Skip to main content

Full text of "The Atharva-Veda Prâtiçâkhya, or Çâunakîyâ Caturâdhyâyikâ"

See other formats


Early Journal Content on JSTOR, Free to Anyone in the World 

This article is one of nearly 500,000 scholarly works digitized and made freely available to everyone in 
the world by JSTOR. 

Known as the Early Journal Content, this set of works include research articles, news, letters, and other 
writings published in more than 200 of the oldest leading academic journals. The works date from the 
mid-seventeenth to the early twentieth centuries. 

We encourage people to read and share the Early Journal Content openly and to tell others that this 
resource exists. People may post this content online or redistribute in any way for non-commercial 

Read more about Early Journal Content at 
journal-content . 

JSTOR is a digital library of academic journals, books, and primary source objects. JSTOR helps people 
discover, use, and build upon a wide range of content through a powerful research and teaching 
platform, and preserves this content for future generations. JSTOR is part of ITHAKA, a not-for-profit 
organization that also includes Ithaka S+R and Portico. For more information about JSTOR, please 









Presented to the Society May 21st, 1863. 

Introductory Note. 

The distinctive title of the work here published is Qdunaktyd catur- 
ddhydyikd, ' Qaunaka's Treatise in Four Chapters.' We have for it, 
however, only the authority of the signatures to the different portions of 
the manuscript containing the treatise ; no reference to the latter by 
name has yet been discovered, so far as I am aware, in any other work 
of the Sanskrit literature. As regards the gender of the word, whether 
feminine or neuter, there is some question. In the signature to the first 
section (pdda) of the first chapter (adhydyd), it is styled calurddhydyikd, 
as also at the close of the first chapter. With this accords, farther, the 
name, caturddhydyi-bhdshya, given to the commentary in the signature 
of chapter IV, section 1, and at the close of the whole work. The 
neuter form, and the ascription to ^annaka, are found only in the final 
signature, which reads as follows (unamended) : iti fdunaktyamcaturd- 
dhydyike caturthah pddah : caturddhydyibhdskya samdptak* The trea- 
tise was first brought to light, and its character determined, by Both 
(see the Preface to his Nirnkta, p. xlvii). It was recognized by him as 
being what is indicated by our title, a Pratic&khya to a text of the 
Atharva-Veda. That it has any inherent right to be called the Pratiga- 
khya to the Atharva-Veda is not, of course, claimed for it ; but, consid- 
ering the extreme improbability that any other like phonetic treatise, 
belonging to any of the other schools of that Veda, will ever be brought 
to light, the title of Atharva-Veda Pratic&khya finds a sufficient justifi- 
cation in its convenience, and in its analogy with the names given to the 

* Weber (Cat. Berl. MSS., p.87; Ind. Literaturgeschicbte, p. 146) calls the trea- 
tise caturadhydyikd • and Miiller (Hist. Anc. Sansk. Lit., p. 139, etc.) styles it cdtura- 
dhydyikd — each by a different emendation of the name given in the manuscript : I 
do not see the necessity of departing from the authority of the latter. 

334 W. I). Whitney, 

other kindred treatises by their respective editors, Regnier, Weber, and 
Miiller.* Any special investigation of the questions of the authorship 
and date of our treatise, its relation to the other Pratigakhyas and to 
the present received text of the Atharva-Veda, and the like, is reserved 
for the commentary and the additional notes : it will be sufficient to say 
here, in a general way, that it concerns itself with that part of the Athar- 
van text which is comprised in its first eighteen books, and with that 
alone, and that it covers the whole ground which the comparison of the 
other treatises shows us to be necessary to the completeness of a Prati- 
(jakhya, differing from any of them not more than they differ from one 

The manuscript authority upon which the present edition is founded 
is a single codex (Chambers collection, No. 143; Weber, No. 361), be- 
longing to the Royal Library of Berlin, a copy of which was made by 
me in the winter of 1852-3 ; it contains, besides the text of the Prati- 
cakhya, a commentary upon it, by an author not named, which styles 
itself simply caturadhy&yi-bhashya, ' Commentary to the Four-chaptered 
Treatise,' as already noticed above. It is briefly described in Weber's 
Catalogue of the Berlin Sanskrit Manuscripts (p. 87-8). The signature 
at the end is as follows (with one or two obvious emendations) : prir 
astu : lekhakapathakayoh fubham bhavatu : p-tcandikdydi namah : qri- 
rdmah : samvat 1714 varshe jydishtkafuddha 9 dine samaptalikhitam 
pustakam. The date corresponds to May, 1656; but it must, as in 
many other cases, be doubtful whether this is the date of the manu- 
script in our possession, or of the one from which this was copied ; in 
the present instance, the latter supposition may be regarded as decidedly 
the more probable. Most unfortunately, considering the extreme rarity 
of the work, the manuscript is a very poor one. Not only is it every 
where excessively incorrect, often beyond the possibility of successful 
emendation ; it is also defective, exhibiting lacunae at several points. 
Some may be of opinion, then, that the publication of the Praticakhya 
upon its authority alone is premature, and should not have been under- 
taken. This would certainly be the case, were any other copies of the 
work known to be in existence : to neglect to procure their collation 
before proceeding to publish would be altogether inexcusable. But, so 
far as is hitherto known, the Berlin codex is unique. No public or pri- 
vate library in Europe, nor any in India accessible to Europeans, has 
been shown to possess a duplicate of it. For assistance in procuring a 
second copy, I made application some years since to Prof. Fitz-Edward 
Hall, then of Benares, whose knowledge, experience, and public and 
private position made him the person of all others most likely to be of 
service in such a way ; and he was kind enough to interest himself zeal- 
ously in my behalf in searching for the work : but entirely without suc- 
cess ; while he collected for me a mass of valuable materials respecting 

* Praticakhya du Rig-V6da. Par M. Ad. Regnier, etc. Published in the Journal 
Asiatique, Ve se>ie, Tomes vii-xii, Paris, 1856-58. — Das Vajasaneyi-Pratic&khyam. 
Published by Prof. Albrecht Weber, in his Indische Studien, Vol. iv, Berlin, 1858. — 
Miiller's edition of the Rig-Veda Priticakhya includes only the first six chapters, 
one third of the whole, and forms part of his text edition of the Rig- Veda itself, 
which also remains a fragment. 

Atharva-Veda Prdtigdkhya. 335 

the other Pratickkhyas, for that of the Atharva-Veda nothing could be 
found. Considering, then, the faintness of the hope that additional 
manuscripts would later be obtainable, and considering the peculiar 
interest of this class of worts — well attested by the triple publications, 
within a few years past, of Regnier, Weber, and Miiller — and the desir- 
ableness of placing as speedily as possible before the eyes of scholars the 
whole material furnished by them, in order to the greater force and con- 
clusiveness of the results which some are already hastening to draw from 
them for the literary history of India, it has seemed best to publish the 
treatise without farther delay. Several circumstances deserve to be 
noted as supporting this decision, by diminishing the disadvantages 
arising from the scantiness and poorness of the manuscript material. In 
the first place, as regards the lacunae, they are, with two exceptions, of 
insignificant importance, and do not either cause the loss of a rule or 
render its interpretation doubtful : while, in the two instances (both 
occurring in chapter III) in which one or more rules are lost, the loss at 
least lies within the limits of a certain definite subject, and, though much 
to be regretted, is of no great extent or essential consequence. As con- 
cerns, again, the corruption of the readings, it is to be observed that the 
commentary is generally full enough to establish the true version of the 
rules, and yet, at the same time, too poor and scanty to render its own 
restoration important. The general method of the commentator is as 
follows : he first states the rule, then restates it in the baldest possible 
paraphrase, merely supplying the lacking copula, and adding the specifi- 
cations, if any, of which the presence is inferrible from previous rales ; 
next follow the illustrative citations ; and finally, the rule is given once 
more, along with the one next following, which is euphonically com- 
bined with it, and of which the paraphrase and illustration then follow 
in their turn. As an example, I cite here in full rule i. 7, with its com- 
mentary, beginning from the final repetition of the next preceding rule : 

STjrT: srswrun =t ^r5nf:i ^r ^cnf. *rcrV irarfrt i ^sttrott: i ^ =aronf : tw- 

Thus we have everywhere (unless, as is sometimes the case, a few 
words have dropped out from the copy) a threefold repetition of each 
rule, and its true form is almost always restorable from their comparison, 
notwithstanding the corruptions of the manuscript. If, now, the com- 
mentary were as full and elaborate as those of the other known Prati- 
cakhyas, it would have been alike trying and unsatisfactory either to 
endeavor to edit it, or to disregard it : while, as the case actually stands, 
it has itself attempted so little that we care comparatively little to 
know precisely what it says. Wherever its usual meagre method is 
followed, accordingly, little attention will be found paid to it in the 
notes. Nor has it seemed to me otherwise than a needless labor to 
notice, except in special cases, the corrupt readings of the manuscript — 
and this the more especially, as my distance from the original renders 
it impossible to test by a renewed collation the accuracy of my copy.* 

* Prof. Weber has had the kindness to verify for me, during the progress of 
publication, sundry passages, of special importance or of doubtful reading, which I 
took the liberty of submitting to him. 
vol. vii. 43 

336 W. D. Whitney, 

The citations from the Atharvan text are also given in their correct 
form, without farther remark ; since, whatever the disguise under which 
the manuscript may present them, it has generally been not difficult for 
one familiar with the Atharvan, and in possession of a verbal index to 
its text, to trace them out and restore their true readings. There are a 
few notable instances in which the commentator abandons his custom- 
ary reticence, and dispreads himself upon the subject with which he is 
dealing : and in such cases the attempt is made to follow him as closely 
as the manuscript will allow. Much more frequently than he ventures 
to speak in his own person, he cites the dicta of other authorities ; 
occasionally referring to them by name ; more often introducing his 
quotations by a simple apara Aha, ' another has said ;' and very fre- 
quently making extracts without any introduction whatever, as if of 
matter which might lawfully be woven in as an integral part of his own 
comment. The work, if it be a single work, from which these anony- 
mous citations are made, is written in the common floka, and is seem- 
ingly of the same general character with our treatise itself, or a kind of 
metrical Praticakhya to the Atharva-Veda ; wearing, however, more 
the aspect of a commentary than does the metrical Praticakhya to the 
Rig- Veda. 

What has here been said of the commentary applies only to that 
part of it which ends with the third section of the fourth chapter : the 
concluding section, on the krama-patha, is of an entirely different char- 
acter, as will be explained at the place. 

While thus but imperfectly aided by the native commentator, I have 
enjoyed one compensating advantage over those who have undertaken 
hitherto the publication of works of this class, in that I have been able 
to avail myself of the results of their labors. Had it not been for their 
efficient help, much in the present treatise might have remained obscure, 
of which the explanation has now been satisfactorily made out; and I 
desire here to make a general acknowledgment of my indebtedness to 
them, which I shall have occasion to repeat hereafter in particular cases. 
I have thought it incumbent upon me to refer, under every rule, or in 
connection with every subject treated of, in the work here published, to 
the corresponding portions of the other Praticakhyas, giving a briefer 
or more detailed statement of the harmonies and discrepancies of doc- 
trine which they contain. To the Rig-Veda Praticakhya reference is 
made primarily by chapter (patala) and verse (floka),* the number of 
the rule cited being then also added, according to the enumeration of 
both Regnier and Muller; the latter (in the first six chapters only) in 
Roman figures, the former in Arabic. The Vajasaneyi Praticakhya is 
cited from Weber's edition, already referred to, and according to his 
enumeration of its rules. For my ability to include in the conspectus 
of phonetic doctrines the Taittiriya Praticakhya of Karttikeya, I have 
to thank Prof. Hall, as above acknowledged ; the excellent manuscripts 
of the text and of the text and commentary (tribhashyaratna) which 
he procured for me will be made, I trust, to help the publication of that 

* In the first chapter, of which the verses are numbered differently by Muller 
and Regnier, the former counting in the ten prefixed introductory verses, the refer- 
ence is according to Regnier : to find the corresponding verse in Muller, add ten to 
the number given. 

Atharva- Veda Prdlicdkhya. 337 

treatise in the course of the next year, either by myself or by some one 
else. The mode of reference to the T&ittiriya Pr&tic&kkya which has 
hitherto been usual I have abandoned. The work is divided into 
twenty-four chapters (adky&ya), which are classed together in two sec- 
tions (pragna), each of twelve chapters : and Roth — as also Weber, fol- 
lowing his example — has cited it by section and chapter, omitting any 
enuaneration and specification of the rules into which each chapter is 
divided. But the pragma division is of as little account as the corres- 
ponding division of the Rik Pr&ticakhya into three sections (adky&ya) ; 
and there appears to be no good reason why this treatise should not be 
cited, like those pertaining to the Rik, the White Yajus, and the Athar- 
van, by chapter and rule simply ; as I have done. To Panini's grammar 
(in Bohtiingk's edition) reference is also frequently made — in all cases, 
it is hoped, where the comparison would be of any particular interest. 
The special relation exhibited by our treatise in many points to the sys- 
tem of general grammar whereof Panini is the authoritative exponent 
would perhaps have justified a more detailed comparison ; but I have 
both feared to be led too far, and distrusted my ability to draw out the 
correspondences of the two in a perfectly satisfactory manner. To 
determine in full the relations of P&nini and the Pratic&khyas, when 
the latter shall have been all made public, will be an important and a 
highly repaying task for some one more versed than I am in the intri- 
cacies of the Paninean system. 

The peculiar method, so commonly adopted in our treatise (e. g. i. 64, 
65, 85), of applying a rule to the series of passages or words to which 
it refers, by mentioning only one of them and including the rest in an 
"etc." (Mi) which is to be filled out elsewhere — or the familiarly 
known c/ana-melhod of PS.nini — and the remissness of the commenta- 
tor, whose duty it was to fill out the ganas, but who has almost always 
failed to do so, have rendered necessary on the part of the editor a 
more careful examination of the Atharvan text, and comparison of it 
with the Praticftkhya, than has been called for or attempted in connec- 
tion with any other of the kindred treatises. It has been necessary to 
construct, as it were, an independent Pratuj&khya upon the text, and to 
compare it with that one which has been handed down to us by the 
Hindu tradition, in order to test the completeness of the latter, fill up 
its deficiencies, and note its redundancies. The results of the compari- 
son, as scattered through the notes upon the rules, will be summed up 
in the additional notes, to which are also relegated other matters which 
would otherwise call for attention in this introduction. In examining 
and excerpting the text, full account has been taken of the nineteenth 
book, and of those parts of the twentieth which are not extracted 
bodily and without variation from the Rig- Veda. References are made, 
of course, to the published text of the Atharva-Veda ;* if a phrase or 
word occurs more than once in the text, the first instance of its occur- 
rence is given, with an " e. g." prefixed. 

Readings of the manuscript which it is thought desirable to give are 
generally referred by numbers to the bottom of the page. 

* Atharva-Veda Sanhita, herausgegeben von E. Eoth und W. D. Whitney 
Erster Band. Text. Berlin, 1856. roy. 8vo. 

838 W. D. Whitney, 

The occurrence, here and there in the notes, of emendations of the 
published text of the Atharvan calls for a few words of explanation here. 
The work of constructing the text was, by the compelling force of cir- 
cumstances, so divided between the two editors that the collation of the 
manuscripts, the writing out of a text, and the preparation of a critical 
apparatus, fell to myself, while Prof. Roth undertook the final revision of 
the text, and the carrying of it through the press after my return to 
this country. Such being the case, and free communication being im- 
possible, occasional misconceptions and errors could not well be avoided. 
Moreover, the condition of the Atharvan as handed down by the tradi- 
tion was such as to impose upon the editors as a duty what in the case 
of any of the other Vedas would have been an almost inexcusable lib- 
erty — namely, the emendation of the text-readings in many places. In 
so treating such a text, it is not easy to hit the precise mean between 
too much and too little ; and while most of the alterations made were 
palpably and imperatively called for, and while many others would have 
to be made in translating, there are also a few cases in which a closer 
adherence to the manuscript authorities might have been preferable. 
Farther, in the matter of modes of orthography, where the usage of the 
manuscripts was varying and inconsistent, our choice was not always 
such as more mature study and reflection justify. Whenever cases of 
any of these kinds are brought up in connection with the rules and illus- 
trations of the Praticakhya, I am free to suggest what appears to me a 
preferable reading or usage. In referring to the manuscripts of the 
Atharvan, I make use of the following abbreviations (which are also 
those employed in the margin of the edited text, in books xix and 
xx) : 1st, sanhita MSS. : "B." is the Berlin MS. (Ch. 115, Weber 338), 
containing books xi-xx ; "P." is the Paris MS. (D. 204, 205), and con- 
tains the whole text, and books vii-x repeated; "M." and "W." are 
manuscripts of the Bodleian library at Oxford, M. in the Mill collection, 
and W. in the Wilson : M. is a copy of the same original, by the same 
hand, and in the same form, as P., and it lacks the part of the text which 
is found double in the other : W. lacks book xviii ; " E." is the East India 
House manuscript, Nos. 682 and 760 ; "H." is in the same library, No. 
1137, and contains only books i-vi; "I." is the Polier MS., in the Brit- 
ish Museum : a copy made from it for Col. Martin is also to be found in 
the East India House library, Nos. (I believe) 901 and 2142. 2nd, pada 
MSS. These are all in the Berlin library. "Bp." is Ch. 8 (Weber 332) 
for books i-ix, and Ch. 108 (Weber 335) for books x-xviii : these are two 
independent manuscripts, but are included under one designation for 
convenience's sake, as complementing one another. "Bp. a " is Ch. 117 
(Weber 331) for book i, and Ch. 109, 107 (Weber 333, 334) for book v, 
and books vi-ix : the two latter are accidentally separated parts of the 
same manuscript, and stand also in very close relationship, as respects 
their original, with Bp. (Ch. 8) : the other is independent. Of book xix 
there is no pada-text to be found, and probably none was ever in exist- 
ence: and the pada MSS. of book xx are only extracts from the Bik 

The mode of transcription of Sanskrit words is the same with that 
which has been hitherto followed in this Journal. 



Contents: — Section I 1-2, introductory, scope of the treatise ; 3-9, sounds which 
may occur as finals; 10-13, aspirates, nasals, surds, and sonants; 14-17, descrip- 
tion of accents ; 18-28, description and classification of sounds according to their 
place and organ of production ; 29-36, do. according to the degree of approxima- 
tion of the organs; 37-39, the r and I vowels; 40-41, diphthongs. 

Section II. 42, visaryaniya ; 43-48, abhinidhdna ; 49-50, conjunction of con- 
sonants; 51-54, quantity of syllables; 55-58, division into syllables; 69-62, 
quantity of vowels. 

Section III. 63-66, abnormal alterations and interchanges of sounds ; 67-72, 
occurrence of nasalized vowels ; 73-81, pragrhya vowels ; 82, treatment in pada- 
text of pragrhya vowels followed by iva ; 83-91, occurrence of long nasalized 
vowels in the interior of a word. 

Section IV. 92, definition of upadhd; 93, what makes a syllable ; 94, only 
an unaspirated consonant allowed before an aspirated ; 95, mode of application 
of rules respecting conversion of sounds; 96, special case of accent; 97, special 
cases of omission of pluti before Hi; 98, conjunction of consonants; 99, yama; 
100, ndailcya; 101-104, svarabhalcti and sphotana and their effect; 105, cases of 

1. Of the four kinds of words — viz. noun, verb, preposition, 
and particle — the qualities exhibited in euphonic combination 
and in the state of disconnected vocables are here made the 
subject of treatment. 

Here is clearly set forth the main object of such a treatise as we are 
accustomed to call a pr&tif&khya : it is to establish the relations of the 
combined and the disjoined forms of the text to which it belongs, or of 
the sanhita-text and the pada-text : sandhyapadyau might have been 
directly translated ' in the sanhitA and pada texts respectively.' The 
ultimate end to be attained is the utterance of the sacred text (f&khA, 
'branch' of the Veda), held and taught by the school, in precisely the 
form in which the school receives and teaches it. The general material 
of the text must, of course, be assumed to be known, before it can be 
made the subject of rules : it is accordingly assumed in its simplest and 
most material-like form, in the state of padas or separate words, each 

840 W. D. Whitney, [i. 1- 

having the form it would wear if uttered alone, compounds being also 
divided into their constituent parts, and many affixes and inflectional 
endings separated from their themes; and the Praticakhya teaches 
how to put together correctly this analyzed text. An essential part of 
such a treatise is also its analysis, description, and classification of the 
sounds of the spoken alphabet, as leading to correctness of utterance, 
and as underlying and explaining the complicated system of phonetic 
changes which the treatise has to inculcate. These two subjects — a 
theoretical system of phonetics, and the rules, general and particular, 
by which pada-text is converted into sanhita — are the only ones which 
are found to be fully treated in all the Pratigakhyas ; although none of 
the treatises confines itself to them alone. Thus, our own work gives 
in its fourth chapter the rules for the construction of the pada-text 
itself, as does also the Vajasaneyi Pratigakhya; and likewise, in the 
final section of that chapter (which is, however, evidently a later ap- 
pendix to the work), a brief statement of the method of forming the 
krama-text, of which it has also taken account in more than one of the 
rules of its earlier portions : and the Praticakhyas of the Rik and the 
Vajasaneyi have corresponding sections. Nor are the instances infre- 
quent in which it more or less arbitrarily oversteps the limits it has 
marked out for itself, and deals with matters which lie properly beyond 
its scope, as will be pointed out in the notes. A summary exhibition of 
these irregularities, and a comparative analysis of the other Pratica- 
khyas, will be presented in an additional note. 

As the Pratigakhya deals with words chiefly as phonetic combina- 
tions, and not as significant parts of speech (as Worter, 'vocables,' not 
Worte, 'words'), their grammatical character is unessential, and the 
distinction of the four classes made in the rule is rather gratuitous : 
the names of the classes do not often occur in the sequel, although our 
treatise is notably more free than any other of its class in availing itself 
of grammatical distinctions in the statement of its rules. For a fuller 
exhibition of the fourfold classification of words as parts of speech, see 
Rik Pr. xii. 5-9, and Vaj. Pr. viii. 52-57. 

In illustration of the term sandhya, the commentator says : " words 
that end thus and thus take such and such forms before words that 
begin so and so." To illustrate padya, he cites rule 8, below — a by no 
means well-chosen example. To show how it is that the treatise has to 
do only with the qualities of words as exhibited in sanhita and pada, 
he cites an instance of what must be done by a general grammarian in 
explanation of a derivative form, as follows : sandhyapadyav iti kirn, 
artham : Udham ity alra ho-dha-tvam : paracaturthatvam : (MS. padaca°) 
shtuna-sht.u-tvam : dho-dhe-lopo dirghatvam iti vaiyakaranena vaktavy- 
am : 'why is it said "the qualities in sanhita and pada"? Because 
the general grammarian must say, in explanation of lidha, "here ap- 
plies the rule ho dhah (Pan. viii. 2. 31), that for the change of the fol- 
lowing letter into its aspirated sonant, the rule shtund shtuh (Pan. viii. 
4. 41), the rule dho dhe lopah (Pan. viii. 3. 13), and that for the length- 
ening of the vowel." ' These rules teach the formation of the partici- 
ple Udha from the root lih, through the following series of changes : 
lih-la, lidh-ta, lidh-dha, lidh-dha, li-dha, Udha; and they are for the 

i. 2.] Atharva-Veda Prdtigdkhya. 341 

most part taken directly from Panini, or at least correspond precisely 
with his rules; only, in the second case, paracaturthalvam takes the 
place of Pan. viii. 2. 40, jhashas tathor dho 'dhah; and, in the last case, 
dtrghatvam stands for dhralope p&rvasya dirgho ' nah (Pan. vi. 3. 111). 
Whether the commentator thus deviates arbitrarily or through careless- 
ness from the letter of the great grammarian's rules, or whether he cites 
from some other authority, anterior to or independent of Panini, and 
with whom the latter agrees only in part, is a question of which the 
solution need not be attempted here : while the former supposition may 
appear the more probable, the other, in the present state of our knowl- 
edge, respecting the relations between Panini and the Pratkjakhyas and 
their commentators, is not to be summarily rejected as impossible. 

2. Farther, that respecting which, general grammar allows 
diversity of usage is made subject of treatment, to the effect of 
determining the usage in this gdkhd. 

This is a broadly periphrastic translation of the rule, which reads more 
literally: '"thus and thus it is here" — to this effect, also, that which 
is allowed to be diversely treated in the general language (is made the 
subject of the rules of the treatise).' The commentator's exposition is 
as follows : evam iha iti ca : asydrh pdkhdydm tat prdtijnam, manyante . 
yaro ' nundsike ' nun&siko ve Hi vibhdshdprdptam sdmdnye : kirh sdma- 
nyam: vydkaranam : vakshyati : uttamd uttameshv iti: '"thus it is 
here :" in these words also : i. c, in this fdkhd they regard this as 
matter of precept : by the rule (Pan. viii. 4. 45) " the letters from y to 
s may or may not be made nasal before a nasal," a choice of usage is 
allowed in general grammar — sdmdnya means vydkarana, 'grammar' — 
but the Pratigakhya is going to say (ii. 5) " mutes other than nasals 
become nasals before nasals." ' The rule is somewhat obscure and dif- 
ficult of construction, and the commentary not unequivocal, substitut- 
ing, as before, an illustration in place of a real exposition of its meaning, 
but I am persuaded that it is fairly rendered by the translation above 
given. Miiller, having occasion to refer to it, gives it somewhat differ- 
ently, as follows (p. xii) : " what by the grammatical text books is left 
free, that is here thus and thus : so says the Prati§akhya." But this 
leaves the ca unexplained, and supposes the iti to be in another place, 
making the rule to read rather evam iha vibhdshdprdptam sdmdnya iti ; 
nor does it accord with the commentator's exposition. It seems neces- 
sary, in order to account for the ca, to bring down prdtijnam as general 
predicate from the preceding rule ; and the iti must be understood as 
pointing out that the Praticakhya says evam iha, ' so and so is proper 
here,' respecting any matter which the rules of grammar leave doubtful. 

The rule is properly neither an addition to, nor a limitation of, the 
one which precedes it, but rather a specification of a particularly im- 
portant matter among those included in the other; for the Praticakhya 
does not overstep the limits of its subject as already laid down, in order 
to determine points of derivation, form, etc., which general grammar 

842 W. D. Whitney, [i. 2- 

may have left unsettled ; nor does it restrict itself within those limits 
to matters respecting which general usage is allowed to vary : it does 
not at all imply or base itself upon the general science of grammar and 
its text book, but is an independent and a complete treatise as regards 
its own subject. 

Of which p&khd of the Atharva-Veda this work is the Pratic&khya, 
it gives us itself no information whatever, nor does it even let us know 
that it belongs to the Atharvan. The name by which it is called, how- 
ever, leads us to suppose that it was produced in the school of the Q&u- 
nokas, which is mentioned in the Caranavyuha among those of the 
Atharvan (see Weber's Indische Studien, iii. 277-8). Its relatiop to 
the only text of the Atharvan known to be now in existence will be 
made the subject of an additional note. 

q^jFRT: TO n^li 

3. A letter capable of occurring at the end of a word is called 

This is simply a definition of the term padya, which, in this sense, is 
peculiar to the present treatise ; it is not found at all in either of the 
Yajur-Veda Praticakhyas, or in Panini, and in the Rik Praticakhya it 
means ' member of a compound word.' The term signifies, by its ety- 
mology, 'belonging to a pada, or disjoined word' (in the technical 
sense), and it is evidently applied specifically to the last letter of such a 
word as being the one which is most especially affected by the resolu- 
tion of sanhita into pada. 

As instances, the commentary cites a series of four words, ending 
respectively in guttural, lingual, dental, and labial mutes, which he gives 
also repeatedly under other rules ; viz. godhuk (p. go-dhuk : e. g. vii. 
73. 6), virat (p. vi-rat : e. g. viii. 9. 8), drshat (ii. 31. 1), trishtup (p. 
tri-slup: e. g. viii. 9. 20). 

*JHhhl(: *5T(: TO iiSh 

4. Any vowel, excepting I, may occur as final. 

The Rik Prati$akhya treats of possible final letters in xii. 1, and ex- 
cepts the long f-vowel, as well as Z, from their number. The latter is 
also excluded by the introductory verse 9 to the first chapter, as given 
by Miiller (p. x). The Vajasaneyi Praticakhya also pays attention to 
the same subject, in i. 85-89, and its rule respecting the vowels (i. 87) 
precisely agrees with ours. It farther specifies, however (i. 88), that r 
is found only at the end of the first member of a compound, which 
is equally true as regards the Atharvan text. 

The illustrations brought forward by the commentator are brahma 
(e. g. i. 19. 4), fdl& (ix. 8. 17), nila (not found in AV.), dadhi (in dadhi- 
-van, xviii. 4. 17), kumari (x. 8. 27), madhu (e. g. i. 34. 2), vayti (only 
in indrav&yu, iii. 20. 6), kartr (no such case in AV., nor any case of 
this word as member of a compound : take instead pitr-bhih, e. g. vi. 
63. 3 ; pitr-lokam, xviii. 4. 64), cakshate (e. g. ix. 10. 26), asy&i (e. g. 
ii. 36. 1), vayo (e. g. ii. 20. 1), tau (e. g. iii. 24. 7). 

i. 7.] Atharva- Veda Prdtiqdkhya. 343 

5. Also I and visarjaniya. 

The instances given by the commentator are b&l (e. g. i. 3. 1), and 
vrkshah (e. g. iv. 7. 5). The word b&l, an onomatopoetic exclamation, 
is the only one in the Atharvan ending in I — excepting the similar 
words pal and phal, in xx. 135. 2, 3, a part of the text of which our 
treatise takes no account. Both the other Prati$akhyas (R. Pr. xii. 1 ; 
V. Pr. i. 86) omit I from the number of possible finals, no word in 
their texts, apparently, ending with it. 

FOT: CmtfFTT: ll \ II 

6. Of the mutes, the first and last of each series. 

That is to say, the unaspirated surds and the nasals, or h, t, t, p, and 
n, n, n, m ; c and n being excepted by the next following rule. In 
speaking of the mutes, our treatise follows the same method with that 
of the other Pratiqakhyas, calling the surd, the surd aspirate, the sonant, 
the sonant aspirate, and the nasal, of each series or varga, the " first," 
" second," " third," " fourth," and " last" of that series respectively. The 
Vaj. Pr. alone also calls the nasal by the name "fifth." 

The commentator gives no instances under this rule : they may be 
added, as follows: pratyak (e. g. iv. 18. 2), vashat (e. g. i. 11. 1), yat 
(e. g. i. 2. 3), tri-slup (e. g. viii. 9. 20) ; arvdri (e. g. iii. 2. 3), brahman- 
-vatim (vi. 108. 2), asm&n (e. g. i. 1. 4), teshdm (e. g. i. 1. 1). The 
guttural nasal, », appears only as final of masculine nominatives singular 
of derivatives of the root anc ; the lingual, n, only in a few instances, 
at the end of the first member of a compound, where, by a specific 
rule (iv. 99), it is left in the pada in its sanhita form (the Vaj. Pr. [i. 88] 
expressly notices this as true of its text) : t is found almost only as 
euphonic substitute of a final e, j, sh, or g (vit-bhyah, iii. 3. 3 : in the 
onomato poetic phat [iv. 18. 3], it doubtless stands for either sk or c; 
bat. [xiii. 2. 29], the only other like case, is doubtful) : k and p are also 
comparatively rare, and especially the latter. 

The Vaj. Pr. (i. 85) gives the same rule, comprising with it also the 
one here next following. The Rik Pr. (xii. 1) forbids only to the aspi- 
rates a place as finals ; but the phonetic rules of its fourth chapter 
imply the occurrence only of surds at the end of a word : see the note 
to rule 8, below. 

7. Excepting the palatal series. 

The commentator mentions all the palatal mutes, c, ch, /, jh, n, as ex- 
cluded from the final position by this rule ; but it properly applies only 
to c and n, the others being disposed of already by rule 6. The Vaj. 
Pr. (i. 85) specifies e and fi : the Rik Pr. (xii. 1) speaks, like our rule, 
of the whole class. 

vol. vii. 44 

344 W. D. Whitney, [i. fr- 

it does not belong to the Praticakhya, of course, to explain into what 
an original palatal is converted when it would occur as a final. 

c[frT: \\z\\ 

8. That the words thus declared to end in first mutes end 
rather in thirds is Qaunaka's precept, but not authorized usage. 

That is to say, Q&unaka prescribes that those words which, as noted 
in rule 6 above, and as implied throughout the rest of the treatise, have 
for their final letters the unaspirated surd, must be pronounced with the 
unaspirated sonant instead : but, although the sage to whom the treatise 
is ascribed, or from whom the school to which it belongs derives its 
name, is thus honored by the citation of his opinion, the binding au- 
thority of the latter is denied. With regard to the question whether 
a final mute is surd or sonant, opinions seem to have been somewhat 
divided among the Hindu grammarians. Panini (viii. 4. 56) does not 
decide the point, but permits either pronunciation. The Rik Pr. (i. 3, 
r. 15, 16, xvi, xvii) cites Gargya as holding the sonant utterance, and 
Qakatayana the surd : it itself declares itself for neither, and at another 
place (xii. 1), as already noted, treats both snrd and sonant as allowable: 
its phonetic rules, however (iv. 1), being constructed to apply only to 
the surd final. If the Rik Pr. were actually, as it claims to be, the work 
of £aunaka, the rule of our treatise now under consideration would lead 
us to expect it to favor unequivocally the sonant pronunciation. The 
Vaj. Pr., as we have seen above (under r. 6), teaches the surd pronun- 
ciation. The Taitt. Pr., liberal as it usually is in citing the varying opin- 
ions of the grammarians on controverted topics, takes no notice what- 
ever of this point ; but its rules (viii. 1 etc.), like those of all the other 
treatises, imply that the final mute, if not nasal, is surd. 

It would seem from this that the sound which a sonant mute assumed 
when final in Sanskrit (for that an original surd, when final, should 
have tended to take on a sonant character is very hard to believe) 
wavered somewhat upon the limit between a surd and a sonant pronun- 
ciation : but that it verged decidedly upon the surd is indicated by the 
great preponderance of authority upon that side, and by the unanimous 
employment of the surd in the written literature. 

In his exposition of this rule, the commentator first gives a bald 
paraphrase of it : prathamantani padani trtiyantani 'ti p&unakasya 
"c&ryasya pratijnanam bltavati: na tu vrttih; adding as instances the 
words already given (see under r. 3), godhuk, vir&t, drshat, trishtup ; 
he then, without any preface, cites two or three lines from his metrical 
authority, which need a good deal of emendation to be brought into a 
translatable shape, but of which the meaning appears to be nearly as 
follows : " mutes other than nasals, standing in pausa, are to be re- 
garded as firsts : a word ending in a first may be considered as ending 
in a third, but must in no case be actually so read (compare Uvata to 

L 9.] Atharva- Veda Pr&tiq&khya. 845 

R. Pr. iii. 8, r. 13, cc), owing to the non-exhibition of authoritative usage 
in its favor" (MS. mdvasdndnikdn spar f dn padydn \ddydnf\ ananund- 
sUc&n: prathaman irltydn [prathamantam trtiydntamf\ vidydt not tu 
pathet leva tit : vrtter ananudarpanat). 

#*rof sr u ^ a 

9. Also adhisparqam. 

The meaning and scope of this rule are exceedingly obscure, and the 
commentator so signally fails to throw any light upon it, that we can 
hardly help concluding that he did not understand it himself. His ex- 
position, without any amendment, is as follows : adhisparfa ca pratijHd 
\Jddin ma 'vasitan spar pan pady&n anunasikan : trtiydn caunakamatdt] 
nam bkavati : na nu vrttih : kim adhisparpd nama : vakshyati : yak&ram- 
vakdrayor Ufavrltir adhisparfam fdkatdyanasya . . .* I have to thank 
Prof. Weber for the highly probable suggestion, made in a private com- 
munication, that the words j&din to matdt, or those enclosed in brack- 
ets, have strayed into the commentary, out of place ; so that the true 
reading is adhisparfam, ca pratijUdnam bhavati : na tu vrttih : ' adhis- 
parcam also is a dictum of Qaunaka, but not authoritative usage.' The 
interpolated words form part of a verse, and are apparently identical or 
akin in signification with the verses cited under the preceding rule : a 
restatement of the same thing, in slightly different terms, and so, we 
may conclude, by a different authority. To explain what adhisparpa 
means here, the commentator simply cites rule ii. 24, in which the same 
word occurs again : a rule which informs us of the opinion of Q&kata- 
yana, that final y and v, the result of euphonic processes, are not omitted 
altogether, but imperfectly uttered as regards the contact (adhisparfam), 
the tongue and lips, in their pronunciation, not making the partial con- 
tact (i. 30) which is characteristic of the semivowels. But how can the 
use of adhisparcam in that rule, as an adverb, give a hint of its mean- 
ing here, where it seems to be treated as a noun ? Are we to under- 
stand that it is taken as the name of that peculiar utterance of y and v, 
and that our rule means to say that the mode of utterance in question 
is also a teaching of £aunaka, but not authoritative ? This is scarcely 
credible : it does not appear hereafter that Qaunaka had anything to do 
with that utterance, which is sufficiently put down by the positive rules 
of the treatise against it, nor would its mention here, in a passage 
treating of padyas, be otherwise than impertinent. Or is adhisparfa 
to be interpreted as the name of a slighted or imperfect utterance, and 
did ^aunaka teach such an utterance as belonging to a final mute, 
which wavered, as it were, between sonant and surd ? This appears 
somewhat more plausible, but not sufficiently so to be accepted as at all 
satisfactory : there is no question of a difference of contact of the 

* Here, as also in the citation of the rule ii. 5, under rule 2 above, the whole 
series of illustrative citations from the Atharvan text, as given by the commentary 
under the rules themselves, are rehearsed : I have omitted them as superfluous. 

346 W. D. Whitney, [i. 9- 

organs (spared) in such a case, and it is one to which the prescription 
of abhintdhana (i. 45) applies.* 

f^termr: m^Ttmiaou 

10. The second and fourth of each series are aspirates. 

The term Ashman, literally ' heat, hot vapor, steam,' is in the gram- 
matical language applied to designate all those sounds which are pro- 
duced by a rush of unintonated breath through an open position of 
the mouth organs, or whose utterance has a certain similarity to the 
escape of steam through a pipe : they are the sibilants and aspirations 
or breathings (see below, i. 31). In the term soshman, ' aspirated mute,' 
and in its correlative anushman, 'unaspirated mute' (i. 94), ushman is 
to be understood not in this specific sense, but in that of ' rush of air, 
expulsion of unintonated breath.' To this rule correspond Rik Pr. i. 3 
(r. 13, xiv) and Vaj. Pr. i. 54, the latter being also verbally coincident 
with it. The Taitt. Pr. has nothing analogous, and does not employ 
the terms soshman and anushman. 

The commentator merely adds the list of surd and sonant aspirates 
to his paraphrase of the rule, citing no examples. For the sonant pala- 
tal aspirate, jh, the Atharvan text affords no example. He next cites 
a verse from his metrical authority : sasthdndir ushmabkih prktds trtiydh 
praihamdf ca ye : caturthdf ca dvitlydp ca sampudyanta iti sthitih ; 
'thirds and firsts, when closely combined with flatus of position corres- 
ponding to their own, become fourths and seconds : that is the way.' 
The most natural rendering of sasthanair ushmabhih would be ' with 
their corresponding ushmans or spirants ;' but this is hardly to be toler- 
ated, since it would give us, for example, ts and ds, instead of th and dh, 
as the dental aspirates. This view is distinctly put forth, however, as 
regards the surd aspirates, by another authority which the commentator 
proceeds to cite at considerable length : the first portion, which alone 
bears upon the subject of our rule, is as follows : " another has said, 
' the fourths are formed with h :' " (now begin the flolcas) " some know- 
ing ones have said that there are five ' first' mutes ; of these, by the suc- 
cessive accretion of secondary qualities (guna), there takes place a con- 
version into others. They are known as 'seconds' when combined with 
the qualities of jihvdmuliya, p, sh, s, and upadhmdniya. The same, 
when uttered with intonation, are known as 'thirds:' and these, with 
the second spirant, are known as 'fourths.' When the 'firsts' are pro- 
nounced with intonation, and through the nose, they are called 'fifth' 
mutes. Thus are noted the qualities of the letters." The remaining 
verses of the quoted passage treat of the combination and doubling of 
consonants, and I am unable in all points to restore and translate them. 

* I add Weber's conjecture : " possibly — ' as regards contact also ' the view of 
Caunaka is only a pratijndnam, and not vrtti ; that is, when the padyas enter into 
sandhi, they are to be converted into trtiyas before nasals (e. g. tad me, not tan me) : 
but this is only pratijndnam, not vrtti." I cannot regard this as the true explana- 
tion, since we have no doctrine of (Jaunaka's, to the effect implied, anywhere stated, 
and'since «parf« is not, so far as I am aware, ever used of the contact or concur- 
rence of one sound with another. 

i. 13.] Atharva-Veda Prdtigdhhya. 347 

3fRT ?FRTf£ro n nu 


11. The last ia each series is nasal. 

The term anun&sika in this treatise means simply ' uttered through 
the nose,' and is applied to any sound in the production of which the 
nose bears a part: see rule 27, below. In ii. 35, it is used of the / 
into which a nasal is converted before an I : in all other cases of its 
occurrence, it designates a nasalized vowel, or what is ordinarily known 
as the independent and necessary anusvara. Our treatise stands alone 
among the Praticj&khyas in ignoring any such constituent of the alpha- 
bet as the anusvara, acknowledging only nasal consonants and nasal 
vowels. For a comprehensive statement of the teachings of the other 
treatises respecting nasal sounds, see Roth, Zur Litteratur und Geschichte 
•des Weda, pp. 68-82. 

The Rik Pr. (i. 3, r. 14, xv) and Vaj. Pr. (i. 89) describe the nasal 
mutes as anundsika ; as does also the Taitt. Pr. (ii. 30), including with 
them the anusvara. 

giHt tfff^«jU<M: 1 5=TT^t sffacFF^ ll <^l^ ll 

12. In the surd consonants, the emission is breath ; 

13. In the sonant consonants and the vowels, it is sound. 

In this case and the one next following, two or three rules are stated 
and explained together by the commentator ; that the division and enu- 
meration is to be made as here given, is attested by the statement at 
the close of the section respecting the number of rules contained in it. 

The Prkticakhya here lays down with entire correctness the distinc- 
tion between surd and sonant sounds, which consists in the different 
nature of the material furnished in the two classes to the mouth organs 
by the lungs and throat : in the one class it is mere breath, simple un- 
intonated air; in the other class, it is breath made sonant by the vocal 
chords on its passage through the throat, and thus converted into sound. 
The same thing is taught by two of the other treatises : see Rik Pr. 
xiii. 2 (r. 4, 5), and T&itt. Pr. ii. 8, 10 : the Vaj. Pr. gives no corres- 
ponding definition, nor does it use the terms aghosha and ghoshavant, 
out adopts instead of them the arbitrary and meaningless designations 
jit and mud for the surds, dki for the sonants (i. 50-53). No one of 
the treatises confuses itself with that false distinction of "hard" or 
"strong,'" and "soft" or "weak," which has been the bane of so much 
of our modern phonology. 

The word anuprad&na means 'a giving along forth, a continuous 
emission,' and hence, 'that which is given forth, emitted material:' 
compare Taitt. Pr. xxiii. 2, where anuprad&na, ' emitted material,' is 
mentioned first among the circumstances which determine the distinc- 
tive character of a sound. The Rik Pr. (xiii. 2) uses instead prakrti, 
1 material.' 

Our commentator gives the full list of the sonant letters : the vowels 
in their three forms, short, long, and protracted (pluta), the sonant 

348 W.D. Whitney, [i.13- 

mutes, the semivowels, h, and, by way of examples of the sonant yamas 
(see below, i. 99), those of g and gh. 1 He then cites again a verse 
from his metrical authority, as follows : vyanjanum ghoshavatsamjnam 
antasthA hah parau yamdu : Irayas trayaf ca varg&niyd aghoshah 
fesha ucyate ; ' the consonants termed sonant are the semivowels, h, 
the two latter yamas, and the three last of each class of mutes : the 
rest are called surd.' There is one striking anomaly in this classifi- 
cation ; namely, the inclusion among the sonants of h, which in our 
pronunciation is a surd of surds. The Sanskrit h is, as is well known, 
the etymological descendant, in almost all cases, of a guttural sonant 
aspirate, gh : are we then to assume that it retained, down to the 
time of establishment of the phonetic system of the language, some- 
thing of its sonant guttural pronunciation, and was rather an Arabic 
ghain than our simple aspiration ? or would it be allowable to sup- 
pose that, while in actual utterance a pure h, it was yet able, by a 
reminiscence of its former value, to exercise the phonetic influence of 
a sonant letter ? The question is not an easy one to decide ; for, while 
the latter supposition is of doubtful admissibility, it is equally hard to 
see how the h should have retained any sonancy without retaining at 
the same time more of a guttural character than it manifests in its 
euphonic combinations. The Praticakhya which treats most fully of 
the h is that belonging to the Taittiriya Sanhita : we read there (ii. 4-6) 
that, while sound is produced in a closed throat, and simple breath in 
an open one, the A-tone is uttered in an intermediate condition ; and 
(ii. 9) that this A-tone is the emitted material in the consonant h, and 
in "fourth" mutes, or sonant aspirates. I confess myself unable to 
derive any distinct idea from this description, knowing no intermedi- 
ate utterance between breath and sound, excepting the stridulous tone 
of the loud whisper, which I cannot bring into any connection with 
an h. The Rik Pr. (xiii. 2, r. 6) declares both breath and sound to be 
present in the sonant aspirates and in h, which could not possibly be 
true of the latter, unless it were composed, like the former, of two 
separate parts, a sonant and a surd : and this is impossible. The Taitt. 
Pr., in another place (ii. 46, 47), after defining A as a throat sound, 
adds that, in the opinion of some, it is uttered in the same position of 
the organs with the following vowel ; which so accurately describes the 
mode of pronunciation of our own h that we cannot but regard it as 
an important indication that the Sanskrit h also was a pure surd aspi- 

14 In a given key, a syllable uttered in a high tone is called 
acute ; 
15. One uttered in a low tone is called grave ; 

1 MS. ITEf, bo that, but for the following verse, it would be very doubtful what 
was meant 

i. 16.] Atharva-Veda Prdtigdkhya. 349 

16. One carried from the high to the low tone is called cir- 

The word samdnayame signifies literally ' on the same pitch :' yama 
has this sense once in the Rik Pr. (xiii. 17), and several times in the 
Taitt. Pr. (xv. 9, xix. 3, etc.). The specification which it conveys is 
omitted in all the other treatises, probably as being too obvious to re- 
quire statement. The meaning evidently is that the acute and grave 
pronunciations are bound to no absolute or fixed tones, but that, wher- 
ever one's voice is pitched, a higher tone of utterance gives the acute, 
a lower the grave. Our treatise, the Vaj. Pr. (i. 108, 109), the Taitt. 
Pr. (i. 38, 39), and Panini (i. 2. 29, 30) precisely accord in their de- 
scription of the uddtta and anuddtta accents: the Rik Pr. (iii. 1) tries 
to be more profound, describing the cause rather than the nature of 
their difference, and succeeds in being obscure : its definition of them, 
as spoken " with tension and relaxation respectively," would teach us 
little about them but for the help of the other authorities. As regards 
the svarita, the definitions virtually correspond, though different in 
form : the Taitt. Pr. (i. 40) and Panini call it a samdhdra, or ' combi- 
nation,' of the other two; the Vaj. Pr. (i. 110) says that a syllable pos- 
sessing both the other tones is svarita ; the Rik Pr. (iii. 2), that a sylla- 
ble is svarita into which the two other tones enter together. The term 
Akshipta, used in the definition of our treatise, is difficult of explanation. 
It corresponds with the term dkshepa, by which in the Rik Pr. (iii. 1) 
the accent in question is characterized, and which Regnier translates 
" addition," Miiller " a clinging to, continuance, persistence (ankalten)" 
and Roth (Preface to Nirukta, p. lvii) nearly the same (aushalten, ' per- 
sistence, perseverance'); while Weber (p. 133) renders our dkshiptam 
"slurred, drawled (geschleift)." Regniers translation is supported by 
the analogy of the corresponding expressions in the other treatises, nor 
would it imply too great an ellipsis in the connection in which it stands 
in his text ; but to understand the participle here in a corresponding 
sense, as meaning 'exhibiting the addition of the other two to each 
other,' could hardly be tolerated. Uvata's commentary explains dkshe- 
pa by tiryaggamana, which would admit of being rendered ' a passing 
through, or across, from one to the other;' and I have accordingly 
translated dkshipta as having the sense of ' thrown, transferred, or car- 
ried from one to the other of the two already mentioned.' 

The words udAtta and anuddtta mean literally 'elevated' and 'not 
elevated ' — that is to say, above the average pitch of the voice. Sva- 
rita is more difficult to understand, and has received many different 
explanations, none of which has been satisfactorily established. I have 
myself formerly (Journ. Am. Or. Soc, v. 204) ventured the suggestion 
that it might come from svara, ' vowel,' and mean ' vocalized, exhibiting 
a conversion of semivowel into vowel,' as would be necessary, in order 
to the full enunciation of the double tone, in the great majority of the 
syllables which exhibit it : but 1 am far from confident that this is the 
true explanation. The accent is once called in the Taitt. Pr. (xix. 3) 
dviyama, ' of double tone or pitch.' The three Sanskrit accents, udatta, 
anuddtta, and svarita, so precisely correspond in phonetic character 

350 W. D. Whitney, [i. 16- 

with what we are accustomed to call acute, grave, and circumflex, that 
it has not seemed to me worth while to avoid the use of these terms in 
treating of them. 

The commentator gives only a paraphrase, and no explanation, of 
these rules, which he states and treats together, as I have done. As 
illustrations of the accents, he cites amdvasyd? (e. g. vii. 79. 2) and 
kanyd> (e. g. i. 14. 2), both circumflex on the final syllable, and the 
words prd' ''mu' ca roha, which are not found in the Atharvan : but the 
reading is probably corrupt, and the phrase meant may be prajd'm ca 
roha (xiii. 1. 34); this would furnish instances of the uddtta and anu- 
ddtta — although, indeed, not better than a thousand other phrases 
which might have been selected. 

fol^HHli^dl HMIMH<lrlHll\oii 

17. Half the measure of a circumflex, at its commencement, 
is acute. 

Our treatise, with which the Vaj. Pr. (i. 126) precisely agrees, con- 
tents itself with this description of the svarila or circumflex, and we 
must commend their moderation. The other two treatises give way 
more or less to the characteristic Hindu predilection for hair-splitting 
in matters unessential, and try to define more particularly the degree of 
elevation of the higher portion, and the degree of depression of the 
lower. Thus the Rik Pr. (iii. 2, 3) describes the higher portion — 
which it allows to be either a half-mora or half the whole quantity of 
the syllable — as higher than uddtta or acute, while the after portion is 
indeed anuddtta or grave, yet has the udatta pitch. The Taitt. Pr. 
(i. 46) notices the doctrine held by our treatise as that of some teach- 
evs, and also remarks (i. 47) that some regard the whole syllable as a 
slide or continuous descent from the higher to the lower pitch. Its 
own doctrine (i. 41-45) is that, when the svarita follows an uddtta, its 
first half-mora only is higher than uddtta, its remaining portion being 
either the same as uddtta, or lower, or the same as anuddtta. 

We have in this part of the work only the general description of the 
accents : a more detailed treatment of them, as they arise and as they 
affect one another in the combinations of the continuous text, is given 
in the third section of the third chapter (iii. 55 etc.). 

The commentator merely cites, as offering instances of the circumflex 
accent, the following words : amdvdsyd> (e. g. vii. 79. 2), kanyd? (e. g. 
i. 14. 2), dhdnyam (e. g. iii. 24. 2), dc&ryah (e. g. xi. 5. 3), rdjanydh 
(e. g. v. 17. 9), nyak (vi. 91. 2), Jeva (e. g. ix. 9. 4), svah (e. g. ii. 5. 2) : 
they all appear again, as instances of the jdtya or original svarita, under 
iii. 57. 

18. In the mouth there are differences of producing organ. 

This rule is simply introductory to those that follow, respecting the 
place and mode of production of the different sounds of the spoken 

i. 19.] Alharva-Veda Prdticdkhya. 351 

alphabet. As regards each of these, two circumstances ate to be con- 
sidered : the sth&na, or ' position,' and the karana, or ' producer.' The 
distinction between the two is laid down by the commentator twice 
over, in identical phrase, under rules 19 and 25 : Mm punafy sth&nam ; 
Mm karanam : . . . yad upakramyate tat sth&nam : yeno 'pakramyate tat 
karanam; 'what, again, is "position," and what "organ"? that is posi- 
tion to which approach is made ; that is organ by which approach is 
made.' The Taitt. Pr. has a similar definition in its text (ii. 31-34) : 
" in case of the vowels, that is position to which there is approximation ; 
that is organ which makes the approximation : in the case of the other 
letters, that is position upon which contact is made; that is organ by 
which one makes the contact." That is to say ; two organs are always 
concerned in the production of a sound, and by their contact or ap- 
proximation the sound receives its character : of these, the more im- 
movable one is called the sth&na, or place of production, and it is from 
this that the sound derives its class designation ; the more movable or 
active one is called the karana, or instrument of production. The 
sth&na does not require to be stated, since it is implied in the very 
name of the sound ; but, lest it should chance to be erroneously imag- 
ined that all the sounds are produced by one and the same organ at the 
places indicated, we are expressly taught the contrary in tikis rule, and 
the treatise goes on to specify the different organs.* 

<=hU<UMIHfe|;(<m: H ^ II 

19. Of the throat-sounds, the lower part of the throat is the 
producing organ. 

That is to say, as the commentator goes on to explain, the upper part 
of the throat, as place of production, is approached by the lower part 
of the throat, as instrument of production. As the sounds constituting 
the class, he mentions a, in its short, long, and protracted values, h, and 
the visarjaniya. The same sounds are defined as kanthya f>y the Rik 
Pr. (i. 8, r. 38-40, xxxix-xli), which also notices that some call h and 
visarjaniya " chest-sounds" (urasya). The Vaj. Pr. (i. 71) declares them 
formed in the throat, but (i. 84) by the middle of the jaw as organ — a 
strange description, and not very creditable to the accuracy of observa- 
tion of its author. The Taitt. Pr. (ii. 46) reckons only h and visarja- 
niya as throat-sounds, and then adds (ii. 47, 48) that some regard h as 
having the same position with the following vowel, and visarjaniya as 
having the same position with the preceding vowel. This latter is the 
most significant hint which any of the Praticakhyas afford us respecting 
the phonetic value of the rather problematical visarjaniya, indicating it 
as a mere uncbaracterized breathing, a final h. There is an obvious 
propriety in detaching these two aspirations and a from the following 
class of " gutturals," k etc, in which the Paninean scheme (under Pan, 

* The meaning i under the title karana in the BontlingkRoth lexicon viz. 

"Aussprache, Articulation"— is accordingly to be struck out: Weber's translation 
of the word, also— "Jfervorbringungswise, ' method of production ' " — is both inac- 
curate and peculiarly cumbersome and unwieldy. 
vol. vii. 45 

352 W. D. Whitney, [1. 10- 

i. 1. 9) ranks them, as they receive no modifying action from any of the 
mouth organs : and the authority who called the aspirations chest- 
sounda may also be commended for his acateness, since in their produc- 
tion it may even be said that the throat has no part : it is only, like the 
mouth, the avenue by which the breath expelled from the chest finds exit. 
The commentator quotes a verse again, of which the general drift is 
clear, although I have not succeeded in restoring its readings so as to 
translate it with closeness. It speaks of the diphthongs as also coi»- 
taining an element of throat-sound, and says that they f as well as- the 
nasal mutes, are declared to have a twofold position. 

20. Of the gutturals,, the base of the jaw 13 the producing 

The name jihvamffltya T by which the class of sounds here spoken of 
is called, means ' formed at the base of the tongue :' I retain for them, 
however, the brief and familiar appellation of "gutturals." They are 
stated by the commentary to be the r vowels, short, long, and pro- 
tracted, the guttural mutes k, kh, g, gh, h, the jihvamuliya spirant, or 
that modification of visarjaniya which is exhibited before the surd gut- 
turals k and kh (intimated by him by means of an illustrative instance, 
purushah khanati : the phrase is a fabricated one, not occurring in the 
Atharvan text), and the vowel I (also intimated by an example, ktptah 
[x. 10. 23j). Precisely the same series of sounds is stated by the Rik 
Pr. (i. 8, r. 41, xlii)' to constitute the class of jihvamulty&s. The Vaj. 
Pr. declares the same, with the exception of the Z-vowel, to be formed 
at the base of the tongue (i. &5) by the base of the jaw (i. 83). The 
Taitt. Pr. (ii. 35, 44) includes in the class only the guttural mutes and 
spirant, and reverses the relation of position and organ, making the jaw 
the former, and the tongue the latter. This is evidently the more natu- 
ral way of defining the mode of production of the class, and the more 
analogous with the method of our own treatise elsewhere, as in the 
cases of the throat-letters, palatals, and labials, the lower and more 
mobile of the two organs concerned being taken as the producer. But 
the usage of naming the class from, the sthana seems to have required 
that the jihvdmula be declared the Mhana, and not the karana, of the 
sounds of which the well established name was jihv&multya. By hanu~ 
mulct, ' root or base of the jaw,' must be here understood, it should seem, 
the posterior edge of the hard palate, which migbt well enough be re- 
garded as the base of the upper jaw, or of the bony structure in which 
the upper teeth are set. It is, in fact, by a contact produced at this 
point between the roof of the mouth and the nearest part of the upper 
surface of the tongue that our own gutturals, k and g, are uttered. That 
the r-vovvel should be included by the Praticakhyas among the guttural 
sounds, instead of among the linguals, where its euphonic value so dis- 
tinctly places it, and where it is arranged in the Paninean scheme, is 
very strange, and would point to a guttural pronunciation of th« r in 
certain localities or among certain classes ; a guttural r is a well recog- 
nized constituent of many modern alphabets. The definition of the 

i/21.] Atharw- Veda Prdtic&kliya. 358 

Z-vowel as a guttural by part of the authorities is probably explainable 
by its occurrence only in the root kip, after a guttural, 'where it might 
naturally enough be so far assimilated as to take on something of a 
guttural character, being removed to a point considerably posterior to 
that in which the common I is uttered. The Vaj. Pr. (i. 69) and the 
Paninean scheme make it dental. The jikv&m&Mya spirant and its 
compeer, the upadhm&niya or labial spirant, are nowhere expressly 
mentioned in our treatise, but are apparently necessarily implied in ii. 
40, and are regarded by the commentator as forming part of the alpha- 
bet which the work contemplates. It does not seem probable that they 
were important modifications of the neutral breathing, the visarjantya. 

The commentator again closes his exposition with a verse, which, 
with some doubtful emendations, reads as follows : jihvamMam rvar- 
nasya kavargasya ca bh&shyate ; yap l cai , va jihv&mMiya Ivarnap ce Hi 
te smrtah s : ' the root of the tongue is declared the organ of the r- 
vowels and the ^-series ; also the spirant which is jihvamuliya, and the 
Z-vowels are so explained.' 

HM«JMI *fclIsl^H II t^ II 

21. Of the palatals, the middle of the tongue is the producing 

The sounds composing this class are stated by the commentator to be 
e, &i, y, g, c, ch, j, jh, n, and the vowel i, in its short, long, and pro- 
tracted values. In this enumeration, he follows the order of the half 
verse which he goes on to quote, as follows : talv aiya$acavargan&m 
ivarnasya ca bh&shyate : ' the palate is explained to be the place of pro- 
duction of &i, y, f, the c-series, and the i-vowels.' The same sounds 
are specified by the Rik Pr. (i. 9, r. 42, xliii) as palatals, and are de- 
scribed by the Vaj. Pr. (i. 66, 79) as formed upon the palate, by the 
middle of the tongue, precisely as by our treatise. The Taitt. Pr. (ii. 
36) furnishes the same definition of the c-series and (ii. 44) of f, but 
holds (ii. 40) that y is formed upon the palate by the middle and end 
of the tongue ; and, as in other cases, it does not include any vowels in 
the class. 

The ancient Sanskrit c and,/ can hardly have been so distinctly com- 
pound sounds as our ch and,;' (in church, judge), or they would have been 
analyzed and described as such by the phonetists. At the same time, 
their inability to stand as finals, the euphonic conversion of t and fol- 
lowing p into ch, the Prakritic origin of c and j from ty and dy, etc., 
are too powerful indications to be overlooked of their close kindred 
with our sounds, and deviation from strict simplicity of nature. That 
the $ was our sh, or something only infinitesimally differing from it, we 
see no good reason to doubt : and certainly, those who hold to the Eng- 
lish ch and j pronunciation for the mutes cannot possibly avoid accept- 
ing the sh pronunciation for the sibilant. 

It has already been noticed above (under r. 10) that one of the palatal 
mutes, jh, does not once occur in the Atharvan text. 

1 yac. s {varnatye Hi ta smxtah. 

354 W. D. Whitney, [i. 22- 

^fernf finpt mIh^^h^ h^h 

22. Of the Unguals, the tip of the tongue, rolled back, is the 
producing organ. 

The sounds composing this class are sh, and the t series, or /, th, d, 
dh, n ; so says the commentator, and fortifies his assertion by adding 
the half verse murdhasthanam, shak&rasya tavargasya tatha matam. 
They are known in all the Pr&ticakhyas by the same name (R. Pr. i. 9, 
r. 43, xliv; V. Pr. i. 67, 78; T. Pr. ii. 37, 44), and the Vaj. Pr. and 
Taitt. Pr. describe them in the same manner with our treatise, even to 
using the same verb to express the action of reverting or rolling back 
the tip of the tongue into the highest part of the mouth cavity. The 
semivowel and vowel r are in the Paninean scheme, and in our custom- 
ary classification of the Sanskrit alphabet, also reckoned as Unguals ; 
and, as the euphonic laws of the language show, with entire propriety, 
since it is in no inconsiderable measure under the assimilating influence 
of the r that the others have come into the alphabet, or won their present 
degree of extension in the spoken system of sounds. The only letter of 
nearly corresponding position in our modern European alphabets is the 
r, which in English, at least, is ordinarily pronounced smoothly over the 
tip of the tongue within the dome of the palate, although not at a 
point so far back as would seem to be indicated by the term miirdhari. 
This word means literally 'head, caput,' 1 and hence an exact translation 
of its derivative murdhanya would be ' capital,' and this would be the 
proper name by which to call the class, if the term had not in English 
another well recognized meaning as applied to letters. Miiller (p. xviii) 
holds murdhan to be used directly in the sense of 'dome of the palate' 
(Gaumendach), and Weber (p. 108) accepts the same meaning for firas, 
but it seems to me exceedingly doubtful whether words which mean so 
distinctly ' head,' as usually employed, can, without limiting addition, be 
taken as signifying a certain region in the mouth : especially when we 
see the Vaj. Pr. (i. 30) once use bhrumadhya, ' the middle of the brows,' 
in a corresponding sense, and the Taitt. Pr. (ii. 3) mention the mouth 
(mukha) along with the "head" (firas) among the organs which give 
form to sound. M&rdhan must be taken to mean ' dome of the palate ' 
indirectly, if at all, in so far as that is the highest point in "the head" 
which the tongue is capable of reaching. Miiller proposes " cacuminal " 
as a name for the class ; a far from unsuitable term, but one which has 
not found acceptance, perhaps as being rather cacophonous. The name 
employed by Bopp and many other later grammarians, "lingual," seems 
as free from objection as any other. "Cerebral" does injustice to the 
Hindu grammarians, and obtrudes offensively a false and absurd theory. 

mA{W ?fl"M^T u ^ II 

23. Of sh, the trough-shaped tongue is the producing organ. 

Our treatise is the only one which singles out sh from among the 
other lingual letters, to make it the subject of a special description. 

i. 25.] Atharva- Veda Praticakhya. 355 

Both the commentator and his metrical authority regard the sh as in- 
cluded in the class which the last rule describes : we are to regard this, 
then, only as a specification which so far modifies the description already 
given. It is very possibly a later interpolation in the text of our 
treatise. The commentary, as usual, offers no explanation of the word 
droniM, which does not occur elsewhere in the grammatical language. 
It is a derivative from drona, ' wooden tub or trough,' and is explained 
in the Bohtlingk-Roth lexicon as " the tongue bent together in the form 
of a trough," which is undoubtedly the true rendering. It can hardly 
be claimed that this rule adds to the distinctness of our apprehension of 
the character of this sibilant, which is clearly enough exhibited by its 
relation to the other lingual sounds : it is not our sh — which is rather, 
as above noticed, the palatal f — but such a sibilant as is formed by re- 
verting the tip of the tongue into the dome of the palate ; much more 
nearly resembling our sh than our s, because uttered at nearly the same 
point with the former, only with the tip, instead of the broad upper 
surface, of the tongue : an s can only be produced pretty close behind 
the upper teeth. 

As an instance of this sibilant, the commentator cites the phrase shad 
dhuh fit&n shad u masah (viii. 9. 17). 

ZytTFU fsT^TTf Htrfluf*i H^ll 

24. Of the dentals, the tip of the tongue thrust forward is 
the producing organ. 

The commentator makes this class include I, s, t, th, d, dh, and n, 
citing again a quarter verse to the same effect : dant& l lasatavargan&m. 
The Vaj. Pr. adds the Z-vowel to the class, which it defines (i. 69, 76) 
as formed at the teeth by the tip of the tongue. The Rik Pr. (i. 9, 10, 
r. 44, 45, xlv, xlvi) composes the class of I, s, and r, besides the ^-series, 
and calls them dantamully&s, ' letters of the roots of the teeth.' The 
Taitt. Pr. (ii. 38, 42, 44) defines the same letters, except r, as formed 
dantam&leshu, 'at the roots of the teeth,' the tf-series and s by the tip 
of the tongue, and I by its middle part. The description of the two 
latter authorities is undoubtedly the more accurate, since the contact 
by which our " dentals " are produced is not upon the teeth themselves, 
but just at their base or behind them : between the tip of the tongue and 
the teeth, where no close contact is possible, are brought forth the Eng- 
lish th sounds. What makes in all cases the peculiar character of an I 
is that in its production the tongue is in contact with the roof of the 
mouth in front, but open at the sides. The Taitt. Pr., then, in defining 
the I as produced by the middle of the tongue, doubtless refers to the 
part where the escape of the breath takes place, while the others are 
thinking only of the part by which the contact is made. 

25. Of the labials, the lower lip is producing organ. 

1 dantyd. 

3 -otkthyam; as also in more than one instance in what follow*. 

866 W. D. Whitney, [i. 25- 

That is to say, as in the case of the throat sounds (r. 19, above) the 
upper surface of the throat was regarded as the passive organ, or posi- 
tion, and the under surface as the active organ, or producer, so here the 
upper lip is passive organ, and the lower lip active : or, as the commen- 
tary phrases it, " the upper lip, the position (stkana), is approached by 
the lower lip, the producer (karana).'" The labials are, according to 
the commentator, the diphthongs o and du, in the normal and the pro- 
tracted form, the p-series, or p, ph, b, bh, m, the upadhmaniya spirant 
(which is not named, but indicated by an example, purushah pibali : 
the phrase is not found in the Atharvan), and the vowel u, short, long, 
and protracted. That the semivowel v is omitted here is doubtless the 
fault of the copyist only, since the sound is not provided with a place 
elsewhere. The verses cited from the metrical treatise are as follows : 
sandhyakshareshu varneshu varnantam oshthyam ucyate : upadhmani- 
yam vah pavargas tatha matah : 1 'in the diphthongal sounds, the 
final sound is called labial ; the upadhmaniya, u, v, and the ^-series are 
also so considered.' The Eik. Pr. (i. 10, r. 47, xlviii) agrees with our 
treatise; the Vaj. Pr. (i. 70, 80, 81) also defines the same sounds as 
produced upon the lip, and by the lip,* but then adds farther that in 
the utterance of v the tips of the teeth are employed : the same speci- 
fication as to the v is made by the Taitt. Pr. (ii. 43 : its commentator ex- 
plaining that in the utterance of that letter the points of the upper teeth 
are placed on the edge of the lower lip) ; and the latter treatise also, 
as in other cases, omits the vowels and diphthongs from the class. The 
descriptions of v given by the two Praticakhyas of the Yajur Veda, as 
well as that offered in the Paninean scheme (which declares its organs 
of utterance to be the teeth and lips), leave no room to doubt that at 
their period the v had already generally lost its original and proper value 
as English w — as which alone it has any right to be called a semivowel, 
and to rank with y — and, doubtless passing through ,the intermediate 
stage of the German w, had acquired the precise pronunciation of the 
English v. Whether the silence of the Rik and Atharvan Praticakhyas 
on this point is due to their prior date, or to a local or scholastic differ- 
ence in their utterance of the v, or to the fact that, in view of the ex- 
clusively labial euphonic character of the sound they were willing to 
overlook the peculiarity of utterance distinguishing it from the other 
labials, I would not undertake to decide : but should consider the first 
supposition the least possible, and the second the most probable, of the 

HlfacftlMI Hlfa^l H t^n 

26. Of the nose-sounds, the nose is producing organ. 

The commentary paraphrases nasikyah by n&sik&sth&na varnah, 

1 pavargaf ca tatha matah. 

* Weber misundertands rule 80, samdnasthdnakarand ndsikydmhthy&h, to sig- 
nify that the nasals and labials have the same sthdna and karana with one another : 
the meaning evidently is that, in each of these two classes of sounds, sthdna and 
karana are the same organ : in the one case, they are both the nose ; in the other, 
both are the lips. 

i. 27.] Aiharva-Veda Prdligdkhya. 857 

' sounds which have the nose as their place of production,' and cites, 
without farther explanation, as instances, brahma (e. g. i. 19. 4), paydnsi 
(e. g. i. 9. 3), qj ijf jf sf, and n, n, n, n, m : that is to say, the ndsikya 
(see below, i. 100), anusvdra, the yamas (see below, i. 99), and the 
nasal mutes. A verse from the metrical authority follows, sustaining 
this exposition : ndsikye ndsikd sthdnam tathd 'nusvdra ucyate : yarned, 
vargottamap cd 'pi yatho 'ktam cdi 'fa te matdh ; ' in the case of ndsikya, 
as likewise of anusvdra, the nose is called the place of production ; the 
yamas, and the finals of the several mute series are also understood to 
be as explained.' But there are grave objections to be made to this 
exposition. In the first place, the nasal mutes have been expressly de- 
clared above (i. 11) to be anundsika, and the anundsikds are the sub- 
ject, not of this rule, but of the next. Again, this treatise, as already 
noticed, acknowledges no anusvdra, and regards such syllables as the 
second of paydnsi to contain nasalized or anundsika vowels, which also 
fall under the next rule. We can hardly doubt that the commentator 
has here allowed himself to be misled by the authority on which he 
relies, and which may have treated the nasals in a manner essentially 
different from that of our treatise. The sounds to which the rule is 
meant to apply must be merely the ndsikya and the yamas. This con- 
clusion is supported by the authority of the Rik Pr., which (i. 10, r. 48, 
xlix) gives the name of nose-sounds (ndsikya) to the ndsikya, yamas, 
and anusvdra ;* and also by that of the Vaj. Pr., which (i. 74) declares 
the same sounds to be formed in the nose, and pronounces (i. 80) therr 
place and organ of production to be the same, only specifying farther 
(i. 82) that the yamas are uttered " with the root of the nose." The 
doctrine of the Taitt. Pr. (ii. 49-51) is less definite and distinct: it 
states that the nose-sounds are uttered with the nose, or else with the 
nose and mouth both, when their organ varies according to the varga 
or mute series to which they belong. 

iMHlfikhMI HisHlRHH U^>ii 

27. Of the nasalized sounds, the mouth and nose together 
are the producing organs. 

The commentator explains anundsikdh by anundsikaslhdnd varndh, 
'sounds which have for their place of production the anundsika.' I 
know of no other case in which anundsika is treated as the name of 
any part or organ in the mouth, and cannot but regard this paraphrase 
as an unintelligent and mechanical continuance of the same mode of 
explication which has been correctly applied to the class appellations in 
the preceding rules. Without any statement of what sounds are to be 
considered as referred to in this rule, the commentary cites the follow- 
ing illustrative instances ; dve ca me vingalip ca (v. 15. 2) ; tisrap ca me 
trinpac ca (v. 15. 3); catasrap ca me catvdrinpac ca (v. 15. 4); pumdn 

* The commentary of one of Muller's manuscripts (see p. xix), by a noteworthy 
agreement in misinterpretation with our own, tries to bring in the nasal mutes also 
as belonging to the class. 

358 W. D. Whitney, [i. 21- 

pumsuh (e. g. iii. 6. 1); tatra pumsuvanam (vi. 11. 1) : they are cases, 
wanting both in brevity and variety, of the nasalized vowels only. But, 
besides the nasal vowels, the rule must be intended to describe the 
character of the nasal semivowel I (ii. 35), and of the nasal mutes 
(i. 11). In the production of all these sounds, the mouth bears a part 
not less essential than the nose : each of them requires a given position 
of the mouth organs, to which the expulsion of the breath, in part or 
in whole, through the nose, then communicates a nasal quality. 

The corresponding definition of the Rik Pr., " a nasal sound is pro- 
duced by the mouth and nose together," does not occur until the latter 
portion of that treatise (xiii. 6, r. 20). The Vaj. Pr. (i. 75) gives an 
equivalent explanation ; the Taitt. Pr. (ii. 52) says, with equal justice, 
" nasal quality is communicated by the unclosing of the nose " — of 
course, in any given position of the mouth organs. 

A verse is again cited by the commentator, as follows : mukhanasike 
ye varnA ucyante te ' nunasikah : samanAsyaprayatnA ye te savarna iti 
smrtah ; ' the sounds uttered in the mouth and nose together are called 
nasalized. Those produced by a like effort of the mouth are styled 
similar.' The term savarna, 'similar,' applied to sounds differing in 
quantity only, and not in quality, is used but once in our treatise (iii. 
42), and is not defined by it : the cited definition is almost the same 
with that of Panini (i. 1. 9) : that of the Vaj. Pr. (i. 43) is more ex- 
plicit : the other treatises, like our own, employ the word without tak- 
ing the trouble to explain it. 

28. Of r, the roots of the teeth are the producing organs. 

By the 'roots of the teeth' must be understood, doubtless, the bases 
of the upper front teeth, at which, according to the Rik Pr. (i. 9-10) 
and the Taitt. Pr. (ii. 38, 42), the whole class called in our treatise 
simply "dentals" (see rule 24, above) is produced. It seems strange to 
find them here called the karana, instead of the sthAna, of r, and we 
are almost ready to assume a break in the anuvrtti of the term karana, 
and supply sthAna in place of it ; and the more especially, as the cited 
verse favors the substitution: rephasya dantamulani pratyag vA tebhya 
ishyate: iti sthanani varnanam ktrtilani yathakramam ; 'of r, the place 
is taught to be the roots of the teeth, or a point close to them : thus 
have the places of the sounds been set forth in order.' The commen- 
tator farther adds : apara Aha : hanumilleshu rephasya dantamuleshu 
vA punah : pratyag vA dantamulebhyo m&rdhanya iti cA 'pare; ' another 
has said : "the place of r is at the roots of the jaw, or, again, at the 
roots of the teeth, or close behind the roots of the teeth : others say 
that it is a lingual."' A considerable difference of opinion among the 
Hindu phonetists respecting the position of the r is indicated by these 
citations and by the teachings of the different phonetic treatises. The 
Rik Pr., as we have seen (under rule 24), includes it with the other 
dentals, as dantamhliya, but adds (i. 10, r. 46, xlvii) that some regard it 
as gingival. The Vaj. Pr. defines it as produced at the roots of the 

i. 30.] Atharva- Veda Pratictikhya. 359 

teeth (i. 68), by the tip of the tongue (i. 11) ; the Taitt. Pr. (ii. 41), 
by the tip and middle of the tongue, at a point close behind the roots 
of the teeth : the Paninean scheme alone reckons it as murdhanyai 
' lingual.' The separation of r and r from one another, and of both 
from the lingual class, is the strangest and least defensible feature in the 
alphabetic classification of the Praticakhyas. By its effect in the eu- 
phonic system of the language, r is clearly a lingual, and can hardly be 
supposed to have been uttered otherwise than as our smooth English r 
is uttered, with the tip of the tongue reverted into the dome of the 
palate, to the lingual position. In this position, however, it cannot be 
vibrated or trilled ; and it is possible that in the laborious and some- 
what artificial pronunciation of the Vedic schools it was, for greater 
distinctness, thrown farther forward in the mouth, to the teeth or nea* 

As instances of the r, the commentator cites faradah pur&cih (ii. 13. 
3), pun& raktaih vasah (not in AV.), pund r&p&ni (i. 24.4), jaghnil ra- 
kshdnsi (iv. 37. 1), agni raksh&nsi (viii. 3. 2&), agnl rakshah, (xii. 3. 43). 

29. In the case of the mutes, the organ forms a contact. 

From this contact (sparfd) of the organ with the place of produc- 
tion, the mutes (sparfa) derive their name. 

The Rik Pr. (xiii. 3, r. 9) gives the same definition, with the addition 
that the organ is also asthitam, ' not stationary.' The Taitt. Pr. (in ii. 
33, 34, cited above, under i. 18) implies a eontact in the case of all 
sounds excepting vowels and spirants (ii. 45), not laying down any dis- 
tinction between the complete contact of the mutes, and the imperfect 
one of the semivowels. 

The commentator cites a verse which establishes a noteworthy ex- 
ception to this rule : svaramadhye dadh&u yatra pidanam tatra varjayet : 
mrduprayatnav ucc&rydv id& mtdkaih nidarganam ; ' where d, and dk 
occur between two vowels, there one must avoid a close contact ; they 
are to be uttered with a gentle effort: instances are id& (v. 12. 8) and 
midham (puru-midkam, iv. 29. 4).' This corresponds, if it does not 
coincide, with the conversion of these letters in a like case into a lin- 
gual I, unaspirated and aspirated, usual in the Rik and in some schools 
of the White Yajus, and taught by the Rik Pr. in i. 11, 12 (r. 51, 52, 
lii, liii), as resting upon the authority of Vedamitra, and by the Vaj. Pr. 
in iv. 143 as the doctrine of some teachers. Our verse does not indeed 
point out that the relaxation of the contact takes place at the sides of 
the tongue, and that the resulting sound is hence of the nature of an I; 
but this is altogether probable. 

30. In the case of the semivowels, it is partially in contact. 

That is to say, the organs are so nearly approximated that their posi- 
tion may be called an imperfect contact. The Rik Pr. (xiii. 3, r. 1 0) 
vol. vn. 4G 

360 W. D. Whitney, [i. 30- 

calls it duhsprshtam, ' imperfectly or hardly in contact.' The Taitt. Pr.,- 
as just remarked, does not distinguish the degree of contact of the semi- 
vowels from that of the mutes. 

The name by which the semivowels y, r, I, v are called — namely 
antahHhd, ' intermediate, standing between ' — is generally explained as 
indicating that the sounds in question, in the arrangement of the alpha- 
bet, stand between the mutes and the spirants. The Bohtlingk-Roth 
lexicon, however (sub verbo), defines it to mean ' occurring only in the 
interior of a sentence, never at its end.' This latter interpretation is 
exceedingly unsatisfactory : in the first place, the definition would be as 
true of the spirants and aspirates as of the semivowels ; in the second 
place, it would not be true of the I ; in the third place, no letter could 
be called antahstha in this sense which could occur at the beginning of 
a sentence, as all the semivowels do. But the other explanation also 
seems too indefinite and indistinctive. Is it not more likely that these 
sounds were named "intermediate" in reference to the mode of their 
formation, as being neither by a complete contact, like the full mutes, 
nor by an open position, like the vowels ? The name antahstha would 
then be virtually accordant with our own " semivowel." 

31. In the case of the spirants, it is also open. 

The final ca of the rule indicates, according to the commentator, that 
ishatsprshtam is also to be inferred from the preceding rule : in the 
formation of the spirants ($, sh, s, and h are specified by the commen- 
tary as constituting the class), the organ is both in partial contact and 
open — a rather awkward way of saying, apparently, that its position is 
neither very close nor very open. The Taitt. Pr. (ii. 44, 45) declares 
that the spirants, in their order, are uttered in the positions of the 
mutes, but with the middle part of the producing organ opened. The 
Rik Pr. (xiii. 3, r. 11) includes the vowels, anusvara, and the spirants 
together, as produced without contact, and with the organ stationary. 

In the absence of a varnasamamnaya, 'list of spoken sounds,' or 
'alphabet,' such as the other Praticakhyas give (Rik Pr., introductory 
verse, and i. 1, 2 ; Vaj. Pr. viii. 1-31 ; Taitt. Pr. i. 1-10), it is not easy 
to assure ourselves how many spirants the treatise acknowledges, and in 
what order it would assume them to stand. As we have already seen, 
the commentary accepts the jihvamuliya and upadhmaniya, which are 
nowhere expressly mentioned in the text, but of which the existence 
seems necessarily implied in ii. 40. The class of spirants is then prob- 
ably composed of h (visarjaniya), h, hk (jihvamuliya), f, sh, s, and hp 
(upadhmaniya). The Rik Pr. (i. 2, r. 10, xi) includes in the class these 
seven, along with anusvara; the Vaj. Pr. (viii. 22), only f, sh, s, h; the 
Taitt. Pr. (i. 9), the seven of our treatise, with the exception of visar- 

32. In the case of the vowels also, it is open. 

i. 36.] Atharva-Veda Prdtigdkhya. 361 

The commentator understands, and doubtless correctly, that vivrtam 
only, and not iskatsprshtam also, is implied in this rule by inference 
from the preceding. He adds the whole list of vowels, both simple 
vowels and diphthongs, in their short, long, and protracted (pluta) form. 

The Rik Praticakhya's doctrine respecting the vowels was cited under 
the last rule. The Taitt. Pr., in its rules ii. 31, 32 (cited above, under 
i. 18), implies that in the utterance of the vowels the organs only ap- 
proximate, and do not touch one another. 

^% FJ^Tll^ll 

33. Some consider it as forming a contact. 

That is, the commentator says, some maintain that in the utterance 
of the vowels the organs are in contact ; others, that they remain open. 
The former opinion is too obviously and grossly incorrect, one would 
think, to be worth quoting. No one of the other treatises favors it in 
any degree. 

^l(Hl(^f%FFFF( n $8 ll 

34. In the case of e and o, it is very widely open. 

The word eke, ' some,' is no longer in force, but this and the two fol- 
lowing rules are more detailed explanations of our treatise itself under 
its own rule 32. For the pronunciation of the Sanskrit e and o, see 
below, under rule 40. 

The commentator cites, as instances of these diphthongs, eke taranti 
(vi. 122. 2), oho asya (v. 22. 5). 

rTrft ^rFfrT{FTnHu 

35. And even more so, in the case of d. 

The a-sound (" Italian a," as in father) is unquestionably the most 
open of all the sounds of the alphabet, the only one in the utterance of 
which all the mouth organs are removed, so far as is possible, from the 
path of the intonated breath, which is thus suffered to stream forth 
wholly unimpeded and unmodified. 

36. The a is obscured. 

The modes of utterance of the short a, of the r-vowel, and of the 
diphthongs e and o, taught by the Praticakhyas, are matters of special 
interest in their phonetical systems, as helping to characterize the period 
in the history of the language represented by these treatises. Neither 
of the sounds in question has fully retained, down to their time, that 
value which general considerations, and the euphonic system of the 
Sanskrit language, show to have been the original and proper one. As 
regards the short a, it was no longer generally spoken with the full 

362 W. D. Whitney, [i. 36- 

openness of a, or as its correspondent short sound. See what Weber 
says upon the subject, under Vaj. Pr. i. 72 — which rule, like the final 
one of Panini's grammar (viii. 4. 68), prescribes that the short a is to 
be treated throughout as if coincident in quality with long a — a pre- 
scription which implies, of course, that in actual pronunciation it was 
different. Whatever degradation from its pure open quality the a had 
suffered must have been, it seems to me, in the direction of the neutral 
vowel (English " short «," in but, son, blood), which has so generally 
taken its place in the modern pronunciation of India, rather than to- 
ward an e or o, as suggested by Weber. The term samvirta, 'covered 
up, enveloped, obscured' (antithesis of vivrta, 'opened'), very well ex- 
presses the quality of this neutral sound, which differs from a only in 
not having the mouth freely opened for its utterance, and which does 
not, like e and o, call for a placing in position of any of the mouth 
organs. The Taitt. Pr. does not separate a from d, but says of both 
(ii. 12) that they are to be spoken " with the lips and jaws not too much 
approximated, and not too widely parted" — a description too indefinite 
to derive any distinct idea from. The Rik Pr. also fails to note any 
difference of quality between the long and short values of this vowel. 
But it is very doubtful whether we are to regard the silence of these 
two treatises upon the point in question as any evidence that they are 
of notably earlier date than the others, as Weber seems inclined to do : 
their peculiarity is much more likely to be due to a local or a scholastic 
difference of pronunciation, or they may have simply disregarded, as of 
little account, the discordance of quality between a and a. 

The commentary gives, as furnishing instances of short a, the words 
ctfvah (e. g. ii, 30. 5), ajah (e. g. iv. 14. 1), and agnih (e, g, i. 7. 4). 

FT li ^ou 

37. The r-vowels are combined with an r. 

In the grammatical language of our treatise and of the Taitt. Pr., 
varna appended to the name of a short vowel causes it to include also 
the long and protracted (pluta) vowels of the same quality : it is a de- 
signation of the quality, without distinction of quantity. The Taitt. Pr. 
(i. 20) gives a special rule establishing the usage. Thus rvarna means 
rkara, rkara, and rzhara. 

The commentator gives no explanation of this rule : he simply re- 
peats it with an added bhavati, and then cites a couple of phrases con- 
taining the r, viz.: idam pitrbhyah pra bhar&mi barhih (xviii. 4. 51), 
and putrair bhralrbhir aditih (vi. 4. 1). But he next proceeds to quote 
from his metrical authority a few verses which are more to the point ; 
they read as follows, with the exception of the first and last lines, which 
are corrupt : . . . . l rvarrie svaramatra yd tasyd madhye 'rdhamatrayd : 
repko bhavati saihsprshto yatha , nyulya nakharh tatha: sutre manir ive 
Uy eke true krimir ive Hi ca : . . . .* ' an r is combined with a half-mora 

1 rvarnasya madhye yugapac ca canoralt. 

2 anena mdtratyadhaydh prafeshe u ubhayar apt. 

i. 38.] Atharva-Veda Prdtigdkhya. 363 

in the middle of the vowel mora in the f-vowel, just as a nail is with 
the finger ; like a pearl on a string, some say; like a worm in grass, say 
others.' With this accords quite nearly the doctrine of the Rik Pr., 
which says (xiii. 14) that r forms part of the r-vowel, and is found in 
the middle of it. Neither treatise attempts to define what constitutes 
the remainder of the vowel. In the analogous rule (iv. 145) of the 
Vaj. Pr., that remainder is (if the rule is in this point correctly inter- 
preted by Weber, which is doubtful ; my own manuscript of the com- 
mentary is too corrupt just here to be made anything of) declared to be 

of the character of a; so that, according to Weber, r = z+2~ r '4' The 
Taitt. Pr. does not, any more than the Rik Pr. in the earlier and more 
genuine part of its text, take any notice of the presence of heterogene- 
ous elements in the r and I vowels; it only says (ii. 18) that in their 
utterance the jaws are somewhat closely approximated, and the tip of 
the tongue brought near to the parts immediately above and behind 
the row of teeth. The etymological and euphonic character of the 
sound in question is simply that of a vocal r, an r which is employed 
with the value of a vowel, as r has been and is employed in other lan- 
guages in different parts of the earth ; and there seems no good reason 
for regarding it as having originally deviated in mode of pronunciation 
from the semivowel r. But it is clear that, at the time of the Pratica- 
khyas, the Hindus had begun to find that difficulty in its utterance and 
use as a vowel which caused its entire disappearance in the later forms 
of the language, and has made of it in the mouth of the modern Brah- 
mans the syllables ri and ri. If I may judge from experiments made 
in my own mouth, the bringing of the r far enough forward in the 
mouth to be trilled would render very natural, and almost unavoidable, 
the slipping in, before and after it, of a fragment of the neutral vowel, 
our « in but, the "obscure (saihvrta) a" of our treatise : of this char- 
acter, it can hardly be doubted, would be what elements the sound con- 
tained which were not r. 

<0<%hm): grfr *n^Tn^H 

.88. Of the long and protracted forms of the vowel, the first 
mora is so combined. 

The commentary paraphrases thus: dirghaplutayos tu phrva, matrh 
sarhsprshtarepham rvarnam bhavati; which is a palpable blunder for 
samsprshtarepha bhavati: i. e. if the vowel is extended so as to occupy 
two or three moras, the r-element which it contains is not prolonged, 
but is found only in the first mora : the whole remainder of the sound 
is .composed of the other element. The Rik Pr. says in like manner 
(xiii. 14) that the r is found only in the former half of long r, and is 
either shorter or of the same length with that which enters into r. 

Two instances of the long r are given by the commentator as illus- 
trations: they are Jcartfn akshasva (x. 1. 14), and pitrnr upe 'mam 
(xviii. 4. 40). 

864 W. D. Whitney, [i. 39- 

39. The Z-vowels are combined with I. 

This doubtless means what is more clearly and unequivocally stated 
by the Rik Pr. (xiii. 14, r. 35) : that when, in such combinations as those 
which have just been described, I takes the place of r, the result is the 
Z-vowel. The other two treatises, as we have seen above, treat the two 
vowels together, in the same rules. The use of the term Ivarna in the 
rule would seem to imply the possible occurrence of the long and pro- 
tracted forms of the vowel, which are, on the other hand, impliedly 
denied in rule 4 above ; they are also ignored by the Taitt. Pr., as they 
are by the Rik Pr. in its proper text (i. 1, r. 1) ; while the prefixed in- 
troductory verses to the latter treatise, and the Vaj. Pr. (viii. 7), ac- 
knowledge them. 

The commentator cites, as instances of this vowel, pancadctfena klpt&h 
(viii. 9. 15), and sinlvaly aciklpat (vi. 11. 3) : the Rik. Pr. (xiii. 14, r. 35) 
notices the fact that the I occurs nowhere excepting in the root kip. 
He then adds a verse from his A metrical authority : rvarne ca rvarne 1 lah 
praplishtap ca yadd tayoh : 1 1 iti tad ichanti prayogam tadvido jandh ; 
the general meaning is clear enough, but the verse needs amending to 
be made translatable. 

40. The diphthongs are composed of combined vowels ; their 
treatment is that of a simple vowel. 

The term sandhyakshara means literally ' syllable of combination ;' it 
is the usual name for a diphthong in all the treatises excepting the Taitt. 
Pr. The correlative samanakshara, ' homogeneous syllable,' is but 
rarely used, as indicating the simple vowels, when it is necessary to dis- 
tinguish them from the diphthongs (in our treatise, only in iii. 42). 
The diphthongs are vowel sounds which, though not simple and homo- 
geneous, yet form but a single syllable, and are treated as if they were 
simple sounds. They are e, o, ai, an. The two former would be more 
properly written ai, au, since the euphonic processes of the language 
clearly show these to have been their original values, each containing a 
short a as its first element, followed by an * or an a respectively. That 
they should be so readily composable of o and i, a and u, in the acci- 
dental and momentary combinations of the phrase, and especially, that 
they should be so regularly resolvable into the same sounds, if they did 
not actually contain those sounds, is not to be credited. The same evi- 
dence proves the other two to be made up of long a, with i or u following. 
The mutual relation of e (ai) and ai must have been nearly that of our 
I and aye. In the Prakrit languages, however, e and o have gained the 
pronunciation of the e in they and o in note ; they have become sounds 
intermediate between, instead of made up of, a and i and a and u; and 
they have acquired short values as well as long. As e and o they are like- 
wise pronounced in the usage of the modern Brahmans. But even at the 

1 \varno. 

i. 41.] Atharva'Veda Pr&lig&lchya. 365 

period of the Pratigakhyas, and in the phonetic systems of the Vedic 
schools, they no longer had uniformly their original value. From the 
present rule, indeed, no such inference could he drawn ; hut the one which 
next follows establishes a distinction in value between them and di, du. 
The Eik Pr. (xiii. 15, r. 38) predicates doubleness of position of all the 
four, and goes on (r. 39) to cite Qakatayana to the effect that a, forms 
half of each, and i and u the remaining half: but it adds (r. 40) that e 
and o f by reason of the fusion of their parts, have not a sound in which 
the separate components are distinct. This might, however, be fairly 
enough said of our own at and au (in pine, house). The Vaj. Pr. (i. 73) 
defines only di and du as composed of two different elements (the com- 
mentary explains them to be ^a+l^c and -Ja+l^o respectively), and 
directs them (iv. 142) to be treated as simple sounds, without seeing any 
reason for giving the same precept as to e and o. The Taitt. Pr. is not 
less explicit ; it says of o (ii. 13, 14) that in its enunciation the jaws are 
to be neither too nearly approached nor too widely sundered, while the 
lips are to be closer than in a; of e (ii. 15-17), that the lips are to be 
somewhat protracted, the jaws pretty closely approached, and the mid- 
dle part and end of the tongue in contact with the upper rows of teeth 
(jambhdn) ; and finally (ii. 23), that in e, as in i, the middle of the tongue 
is brought near the palate. More distinctive descriptions of our e and o 
could hardly be given : there is evidently no thought at all of the com- 
bination of two phonetic elements into one in them. On the other 
hand, di and du are defined with equal clearness (ii. 26-29) as contain- 
ing each the half of an a (which some held to be of closer position 
than the ordinary a), followed by one and a half times i and u in the 
two cases respectively. 

41. Not so, however, with di and du, in a rule of position.- 

The commentator's paraphrase is dikdrdukdrayoh sthdnavidhdne eka- 
varnavad vrttir na bhavati. What the meaning and value of the rale is, 
is not altogether clear ; I can see no other application of it than to for- 
bid the inclusion of di among the palatals only, and of du among the 
labials only, since they are both throat-sounds as well. By implication, 
then, e and o would admit of being ranked as merely palatal and labial ; 
but the commentary to rule 19, above, treated these, as well as the others 
as of double position, and as containing an element of throat-sound. 

A verse is added in the commentary, as follows : dikdrdukdrayap 'cd 
''pi purvd mdird pard ca yd : ardhamdtrd tayor madhye samsprshta Hi 
smrtdh. The \ast pdda is corrupt, and I am too uncertain of the scope 
of the verse to venture to amend it : perhaps the meaning is that, while 
the beginning and end of di, for instance, are clearly a and i, a mora in 
the middle of the sound is of a mixed character. 

This rule ends the first section of the first chapter : the signature is 
calurddhydyikdydm pralhamasyd 'dhy&yasya prathamah pddah: sulra 
41 : ekacatvdrinpat. This is the only case in which the number of rules 
reckoned is assured by being expressed in words as well as in figures. 

366 W. I). Whitney, [i. 42- 

42. Visarjaniya is abhinisht&na. 

The commentator vouchsafes no explanation of the rule, but merely 
paraphrases it, as follows: visarjaniyo varnah : abhinishtano bhavati ; 
and adds, as instances of visarjaniya, agnih (e. g. i. 7. 4) and vrkshah 
(e. g. iv. 7. 3). The term abhinishtano, does not form part of the gram- 
matical language of the Praticakhyas or of Panini : among the former, 
it occurs only in this place : a rule of the latter (viii. 3. 86) determines 
its derivation and orthography, and the instances given in the com- 
mentary show its equivalence with visarjaniya; the Bohtlingk-Roth 
lexicon also refers (sub verbo) to several vocabularies which contain the 
word, giving it the same meaning. More significant is its occurrence 
several times in the grhya-sutras (as cited in the lexica of Bohtlingk- 
Roth and Goldstiicker), also with the signification visarga.* It looks 
as if it bad belonged to an earlier grammatical terminology than that 
of our treatises, and had been retained merely as a reminiscence of 
something formerly current : its introduction into our text is otherwise 
quite unexplained, and, so far as can be seen, without significance. 
Probably it is an ancient name of visarjaniya or visarga, crowded out 
of use by the latter terms. The Bohtlingk-Roth lexicon gives it, with 
reference to this passage, the meaning " an expiring or vanishing sound 
(ein verklingender Laut)," but this is merely a conjecture, and by no 
means so well supported by the etymology of the word (which would 
suggest rather ' a sounding forth, a resonance ') as to be placed beyond 
the reach of question. Panini's rule must be taken as conclusive re- 
specting the derivation and form favored in his time, or by his school ; 
but the analogy of the words abhinidhana, abhinihita, abhinihata, abhi- 
nipata cannot but suggest abhinishthdna as the true form, coming from 
the toot stha with the prefixes abhi and ni. This would not, however, 
relieve the obscurity investing the primitive meaning and application of 
the term ; an obscurity which also attaches, in some measure, to the 
word visarjaniya and its more modern representative visarga. 

*N'UHI^ ll ^ II 

43. The holding apart of a consonant is abhinidhana; it is 
pinched, quite weakened, lacking breath and sound. 

* That the word ever means ' a sound of the alphabet in general,' as stated in 
both the lexicons, seems to me very doubtful : I have not access to all the authorities 
referred to by Bohtlingk-Roth, but the commentary to Panini, abhinishtano varnah, 
does not necessarily imply any thing of the kind, but may rather mean ' an abhi- 
nishtdna letter ;' while, in the citation given by Goldstiicker as an instance of the 
general meaning, it evidently signifies visarga : dirghabhinishtanantam, ' (a name) 
endin" in a long vowel or in visarga.' If the other cases relied on are not less 
equivocal than these, the general definition ' sound' must be rejected. 

i. 43.] Alharva- Veda Praticakhya. 367 

We have here one of those subtleties of phonetic analysis which 
are such marked characteristics of the Hindu science. In order to any 
Satisfactory understanding of it, we must call in to our aid theoretical 
considerations, as the dark and scanty expositions of the grammatical 
treatises and their commentators are insufficient. The phenomenon 
forming the subject of the rule evidently is or includes a defective pro- 
nunciation or indistinctness of utterance, and the two next rules teach' 
us that it affects a mute which is followed by another mute, and one 
which stands as final. In what does the peculiarity of utterance of such 
a letter in such a position consist? A mute is a sound produced by a 
complete closure of the organs of articulation in some defined position, 
entirely cutting off the escape of breath through the mouth ; and it is 
by the breaking of the closure with the utterance of a following open 
sound that the mute is itself made audible. In speaking a p, for in- 
stance, so long as the lips are kept compressed, there is no audible 
sound ; but as soon as the contact is severed with the expulsion of either 
unintonated or intonated breath, in the passing of the voice to the 
utterance of some other sound, the p is clearly heard. A sonant mute, 
as a b, is less absolutely a dumb letter before the breach of the contact, 
because it includes an expulsion of resonant breath from the throat into 
the cavity of the mouth during the closure of the organs, and this re- 
sonance is sufficient to indicate imperfectly the character of the contact. 
A nasal mute, as m, is yet less dependent upon the explosion for its dis- 
tinctness of utterance, since it implies a free flow of sonant breath 
through the nose, and so is continuous and even quasi-vocalic in its 
nature ; yet even the nasals, and still more the sonants, are explosive 
letters, and do not have a perfect utterance unless the contact is broken. 
A following vowel, of course', discovers them most completely ; yet any 
open and continuable letter, as a semivowel or a sibilant, answers the 
same purpose, and in the syllables pya, psa, for instance, we feel that p 
is fairly enunciated. If, however, one mute letter follows another, the 
explosion of the former cannot properly occur ; the organs are supposed 
to pass from one position of complete contact to another, without any 
intervening open sound : the former mute is imperfectly uttered. A 
like thing takes place when a mute is final, or when there is no follow- 
ing open sound to break the contact with : we then have only that very 
imperfect hint of its pronunciation which is given by the formation of 
the contact upon the preceding open sound. We are accustomed, in- 
deed, in order to give distinctness to a final mute, to unclose the organs 
again after making the contact, thus whispering after it, as it were, a bit 
of a vowel ; and the absence of this unclosure is remarked by phonet- 
ists as a peculiarity of the pronunciation of some dialects of spoken 
Chinese, rendering their final mutes almost inaudible : it is hardly pos- 
sible, too, to make one mute follow another so closely that there shall 
not slip out, in the transfer of the organs from one contact to the other, 
a bit of breath or sound, which greatly helps to make the former of the 
two audible : and of both these inorganic or involuntary additions or 
insertions wo shall see hereafter that the Hindu theory takes note: but 
they do not wholly remedy the theoretic imperfection of the utterance. 
That the indistinct pronunciation thus described is the abhinidkana of 
vol. vii. 47 

368 W.D. Whitney, [i.43- 

the Hindu theory, or at least the central and most important fact of 
those comprehended under that name, seems to me tolerably certain, 
although it must be confessed that there are difficulties attending such 
an explanation : none, I think, that may not be done away by supposing 
that the Hindus had not made a complete physical analysis of the phe- 
nomenon, and hence that their descriptions of it partake of vagueness 
and inconsistency ; and also, that they have brought together under the 
name abhinidhdna things not entirely accordant, although analogous, in 
character. The difficulty of the subject is sufficiently attested by the 
doubtful and discordant views taken of it by those who have had occa- 
sion hitherto to examine it, as Miiller, Regnier, Weber, Goldstucker 
(s. v. abhinidhdna). An alternate view to which I have myself been 
somewhat attracted is that by the abhinidhdna is meant the instant of 
silence which intervenes between the closure of the organs for the first 
mute, and their opening for the second : that the Hindu theory regards, 
in the word dpta, for example, the utterance of the p as complete by 
the closure of the lips upon the preceding d, and that of the t as com- 
plete by the unclosure of the tongue before the following o, while the 
brief interval of suspended utterance separating the two acts is abhini- 
dhana. This, better than anything else, would give meaning to the 
first word of our rule, "a holding apart of the consonants," and would 
accord well enough with the rest of the description, translating the last 
term ' deprived of both breath and sound.' Fatal objections, however, 
to this explanation are : the treatment of the phenomenon as something 
affecting the former consonant, not interposed after it; the difficulty of 
assuming any such interval of silence in the case of a concurrence with 
sonant and nasal mutes ; and the non-applicability of the theory to the 
case of a final consonant. The term vyafyanavidhdranam must there- 
fore be understood as used simply in antithesis to the sarhyuktam of 
rule 49 : whereas, in other cases of concurrence of consonants, there is 
actual combination, with partial assimilation of the latter to the former 
(rule 50), here each is held apart from the other as distinct. This, it is 
true, applies only to the concurrence of consonants, and not to a final ; 
but it is allowable to regard as contemplated in a general description or 
designation of a phonetic phenomenon its principal case only, although 
not to adopt an explanation of the phenomenon itself which should 
shut out any of the cases included by it. If I am not mistaken, the 
term abhinidhdna has also a similar meaning. Etymologically, and by 
its use in other than grammatical senses, it should signify, as a neuter 
noun, simply ' a setting down against' the following letter, as distin- 
guished from an actual combination with it. That it is used in our 
treatise as a masculine is somewhat surprising, but cannot be regarded 
as an error of the manuscript. The word seems to be taken almost in 
the sense of abhinihita, as denoting the sound affected by the process 
rather than the process itself, and so to be attracted to the gender of 
varnah or sparfah : the explanations which follow it in the rule, it will be 
noticed, apply rather to the altered letter than to the alteration. The 
Kik Pr. (vi. 5, r. 17, cccxciii) treats the word as neuter, and defines it 
clearly as a process : saihdhdranaih samvaranam ca vdcah, ' a repressing 
and obscuring (holding together and covering up) of the voice.' 

i. 45.] Atharva- Veda Prdtigdkhya. 869 

Our own commentary, as is its wont in difficult cases, leaves us here 
altogether without valuable aid. It simply paraphrases the rule, adds 
the dicta of a couple of other authorities, and closes with a verse ; as 
follows: vyanjanavidh&ranam abhinidk&no bhavati: piditap ca pv&sana- 
d&bhy&m : apara aha : vyanjanavidharanam abhinip&to matro japano 
bhavati piditap ca pvasanadabhyam : apara aha : vyanjanavidharanam 
abhinip&to matro japane guruta bhavati : antahpade padante va ptditah 
sanna eva tu: avakrshtatara sth&n&d avasannataraf ca sah: hinag ca 
fv&san&d&bhyam yo yair&rtho bkidhiyate. I will not attempt to trans- 
late the passage, as I could do so but in part, and as it seems incapable 
of throwing any valuable light upon the subject in hand. The most 
noteworthy circumstance about it is its presentation of abhinip&ta, ' a 
falling down against,' as a synonym of abhinidh&na. 

44. A mute suffers abhinidh&na before a mute. 

The phraseology of the rule would be the same, if abhinidh&na were 
here intended to be taken adjectively, as conjectured above, and if it 
were meant to say that ' a mute before another mute becomes abhini- 
dh&na.'' The commentary merely cites as instances the three words 
brhadbhih, samidbhih, marudbhih, of which only the last is found in 
the Atharvan (p. marut-bhih, e. g. ii. 29. 4). 

The cases in which abhinidh&na alone ensues (only accompanied in 
part by duplication, according to iii. 28 etc.) are those in which a mute 
is followed by another mute (and, if itself non-nasal, then by another 
non-nasal) of the same or a succeeding series. Followed by a mute of 
a preceding series, it suffers also the intervention of sphotana, by ii. 38 ; 
if followed by a nasal, a yatna is interposed, by i. 99. In an additional 
note at the end of the work will be presented a conspectus of all the 
consonantal combinations occurring in the Atharva- Veda, with an exhi- 
bition of the forms assumed by them according to the phonetic rules 
of our treatise. 

The Rik Pr. (vi. 5, r. 11, cccxciii) pronounces not only the mutes, but 
also the semivowels, except r, to suffer abhinidh&na when followed by 
mutes. This would, however, in the Atharvan text, add only the 
groups lis, Ig, Ip, Iph, lb, Im, and vn to those which by our own treatise 
admit the modification, so that the extension of the rule is meant vir- 
tually to include merely the I, a letter which our rule 46 shows to be 
regarded as especially liable to abhinidh&na. The I requires so marked 
a contact of the tongue at its tip that the omission of the breach of that 
contact by a following open letter may well enough have been felt by 
the Hindu phonetists as needing to be looked upon as abhinidh&na. 

c^micW^mI^J ii 8h, ii 

45. Also at the end of a w 

The commentator paraphrases as follows: pad&nte avagrahe ca ipar- 

45. Also at the end of a word, or of the first member of a 

370 W. D. Whitney, [i. 45- 

fasya sparpe paratah: abhinidhdno bhavati: but it is clear that the 
specification sparpe paratah, 'before a following mute,' has no business 
here : that case is included in the preceding rule, and the present pre- 
cept applies to the pronunciation of a final as a final, without any refer- 
ence to what may follow it. This appears partly from the nature of the 
case, partly from the analogy of the corresponding rule in the Rik Pr. 
(vi. 5, r. 18, cccxciv), and partly from the cited illustrations of the com- 
mentator himself: the words given by him under the preceding rule 
would be cases of avagraha in the pada-text, and, of those which he 
presents under this, the last two are instances of avayraha before vowels. 
His citations are tan : vah : yah : devdndm (xi. 1. 5), ap-su (e. g. i. 6. 2), 
sdldnrkdn-iva (ii. 27. 5), and khalvdn-iva (ii. 31. 1). 

The rule of the Rik Pr., already referred to, api c& 'vasdne, ' also in 
pausa? is coincident in meaning with our own. The Taitt. Pr. takes no 
notice whatever of the doctrine of abhinidhdna, nor does the Vaj. Pr. 
directly. The latter, however, presents a couple of rules which are wor- 
thy of remark, as having to do with the same general subject. In i. 90, 
91, it teaches that when a final mute stands either in paum or before a 
following word, there takes place a release or separation of the organs 
of production, the passive and the active organ, or sthdna and karana ; 
that is to say, the contact is dissolved (Weber, and Goldstiicker following 
him, have failed to apprehend the true meaning of the phenomenon de- 
scribed). This dissolution of the contact, in the case of the mute in 
pawsa, is what was referred to above as taking place in our ordinary 
pronunciation after a final contact-letter, in order to make the mute 
more distinctly audible : as occurring before another word, it is analo- 
gous with the sphotana of our treatise (ii. 38), and the dhruva of the 
Rik Pr. (vi. 11), although having a different sphere of occurrence from 
both of them, as they from one another : it is a formal release of the 
organs of articulation from the position belonging to the close of one 
word, before they take up that belonging to the beginning of another, 
in order to the more distinct separation of the two independent mem- 
bers of the sentence. 

46. L suffers abhinidhdna before spirants. 

The only spirants before which I is found actually to occur in the 
Atharva-Veda are q and h : the commentary cites instances of both, 
as follows : fatabalfd vi roha (vi. 30. 2) ; sa gamishyati balhikan (v. 22. 
9) ; vihalho ndma (vi. 16. 2) ; nor are the combinations to be met with 
in the text in any other words than those here quoted. The rule and 
its comment are of particular interest as settling authoritatively the 
reading of the word balhika, 'of Balkh,' which, owing to the customary 
carelessness of the scribes, in not distinguishing Ih from hi (our own 
manuscripts vary between the two), has often been read and explained 
as bahlika. 

L is also noted by the Rik Pr. (vi. 6, r. 20, cccxcvi) as suffering abhi- 
nidhdna before spirants, according to the fakala doctrine, which is not 
that of the treatise itself. By the Vaj. Pr. (iv. 16) it is regarded as to 

i. 48.] Atharva-Veda Prdtigdkhya. 371 

be treated in the same manner as r in a like position. R before a spi- 
rant suffers svarabhakti, or the insertion of a vowel-fragment, according 
to all the other Praticakhyas (see below, rule 101); and the treatment 
of the Vaj. Pr. is virtually, though not formally, the same. The doc- 
trine, then, of the Vaj. Pr., in admitting a svarubkafyi between I and a 
spirant, would differ little from that presented in the Rik Pr. — which 
(by vi. 11) would admit a dhruva, or (by vi. 13, r. 47, ccccxxii) even a 
svarabhakti, after the abhinidhana of the I — except by omission of the 
abhinidhana, of which, as already remarked, it nowhere takes any no- 
tice; but our own treatise, by prescribing abhinidhana, and not allowing 
even sphotana after it, differs quite notably from the others. I must 
confess myself unable to explain why either I before a spirant, or the 
nasals before h, as taught in the next rule, should suffer or be regarded 
as suffering the obscuring process of abhinidhana. 

47. Also the guttural, palatal, and dental nasals before h. 

The instances cited by the commentary, in illustration of this rule, 
are as follows: pralyan, hi (iv. 19. 7); gan hi (a fabricated case: the 
lingual nasal never occurs before h in the Atharvan text) ; krimtn kantu 
(ii. 32. 1); am&n heiih (vi. 29. 1). 

The only consonants ever found to precede h in the Atharva-Veda 
are r, I, n, and n. The first case, rh. is one of svarabhakti (i. 101); the 
second, Ik, falls under the preceding rule; the other two are provided 
for by this rule, which is moreover, like many others in the treatise, 
cast in a theoretical form, or made more general than the requirements 
of the text justify. Since, according to the theory of this Praticakhya 
(see ii. 9), no nasal ever occurs immediately before a sibilant, rules 46 
and 47 might have been cast together into the form : "the nasals and I 
suffer abhimdhdna before the spirants." 

The cases which this rule contemplates are in the Rik Pr. (vi. 7, r. 23, 
cccxcix) included in a much more general precept of the Qakalas, viz., 
that all the mutes except m, when final and followed by initial spirants 
or y, r, and v, suffer abhinidhana. 

48. Abhinidhdna is also called dsihdpita. 

I translate in obedience to the commentator, who says : dstkdpila- 
samjnaf ca bhavati : abhinidhana p ca : el&ny evo 'ddharandrd; 'it both 
receives the name dsthdpila and abhinidhana : the instances are those 
already given.' Unfortunately, this alternative title for the phenomenon 
which we have found so obscure does not notably help our comprehen- 
sion of it : the word admits of being translated, in accordance with the 
explanation of abhinidhana offered above, 'made to stand up to, or 
against ;' but it may also be rendered ' stopped,' that is, ' silenced,' and 
so may favor another theory of the 'phenomenon. 

372 W. D. Whitney, [i. 49- 

49. Any other combination of consonants is conjunct. 

That is to say, all other combinations of consonants than those speci- 
fied in rules 44-47 as accompanied with abhinidhana are simply sam- 
yukta, ' yoked together, conjoined ;' the precise nature of such conjunc- 
tion being defined by the next rule. The commentator says : atah anye 
vyanjanasamdhayuh samyuktd bhavanti : anye abhinidhdndt paddnta- 
sparcdh : ' antahslhoshmanu padadisku 2 ca sarhyvjyante : ' other combi- 
nations of consonants than these are conjunct ; other final mutes than 
abhinidhdna, before semivowels and sibilants commencing a word, are 
conjoined with them ;' and then, instead of citing from the text any 
actual cases, he goes on to put the series of words with which we are 
already acquainted, godhuk, virdt, drshat, trishtup (see rules 3, 8), in 
lengthy and tedious succession, before ydti, vayati, rathe, fete, shande, 
and saye. This by no means exhausts all the possible cases to which 
the name samyulcta applies ; nor has there been any restriction of abhi- 
nidhana to cases of contact between a final and an initial, as the com- 
mentator's language would seem to imply. 

This rule has the appearance of restricting the term sathyopa to such 
combinations of consonants as are not accompanied with abhinidhana. 
But such is not its meaning, at least as regards the general usage of the 
treatise : samyoga is employed everywhere in the more general sense 
expressly attributed to it by a later rule of this chapter (i. 98). 

Nothing is to be found in the other Praticakhyas corresponding to 
this rule and the one next following. 

50. The latter half-measure of the first constituent has the 
same organ of production with the second constituent. 

The term purvarupa is not elsewhere found in our treatise with this 
meaning, although it occurs twice in a like sense in the Rik Pr. (ii. 
12, iii. 7). The construction of the rule is also irregular, and its ellipsis 
of pararupena or parena at the end (parena is added by the commenta- 
tor in his paraphrase) is bolder and more obscure than is usual else- 
where. These anomalies may be owing to the fact that the rule is 
taken in its present form and extent from some other treatise, and a 
metrical one. Weber (p. 127) has noted that it forms a ha\f-floka, and 
it is actually cited as such by the commentator, along with the other 
half-verse, as follows : piirvarupanya mdtrdrdham samdnakaraiiam pa- 
ram : pratyayena bhavet Jcaryam etat samyuktum ishyate ; ' the latter 
half-measure of the first element must be made to have the same organ 
of production with the succeeding element ; such a combination is re- 
garded as conjunct.' We can hardly help, however, both here and in 
the rule, assuming a different meaning for karana from that which it 

1 paddntdt tparfah. 2 pailabhidkishv.. 

i. 52.] Athurva-Veda Pr&liq&lchya. 878 

has elsewhere in our treatise, and usually also in the other kindred 
works, and translating it rather 'mode of production' than 'organ;' 
and this is an additional indication of the foreign origin of the rule 
itself. The only instances given by the commentator are such as do 
not show any difference of organ between the two constituents of the 
conjunction : they are vatsdu virdjah (viii. 9. 1), stomd dsan (xiv. 1. 8), 
and again vaste (xiii. 1. 16). Of the accuracy of the physical observa- 
tions which could discover any actual assimilation of the first element 
of these and other similar combinations, in its final portion, to the lat- 
ter, I find it hard to say much in praise : I am unable to discover that 
any part of the t in vatsdu becomes an s, or any part of the » in vaste 
a t, any more than the s and t respectively become converted in part 
into the following vowels du and e. 

51. A syllable containing a short vowel, excepting before a 
conjunction of consonants, is light. 

The distinction of syllables, as regards their metrical value, is prop- 
erly into light (laghu) and heavy {guru) ; long (dtrgka) and short 
(hrasva) are terms to be used of vowels only. The neuter gender of 
the terms in the rule is to be explained by their agreement with aksha- 
ram, 'syllable,' understood. 

The Rik Pr. (xviii. 19, r. 37) and the Taitt. Pr. (xxii. 15) have rules 
closely agreeing with this. The former also adds (xviii. 20, r. 42, 43) 
that a short vowel with a consonant makes a light syllable, but without 
a consonant one still lighter — an unpractical and useless distinction. 
The Vaj. Pr. has no passage corresponding to our rules 51-54, but re- 
marks, rather out of place, in iv. 105, that vowels which precede a con- 
junction of consonants or a final consonant, or which stand in pausa, 
are of double quantity ; a loose and inaccurate statement, as compared 
with those of the other treatises, since it is the value of the syllable, 
and not the quantity of the vowels, that is increased in the cases men- 

The commentator gives as illustrations the indifferent words dadhi 
and madhu, which we have had already (under i. 4), and shall meet 
with many times more. 

ilcMH II H^ II 

52. Any other is heavy. 

That is, as the commentator goes on to explain, those syllables are 
heavy which contain a short vowel before a group of consonants, or a 
long vowel, or a protracted (pluta) vowel. As instances of the first 
case, he gives takshati (takshati, ix. 10. 21) and rakshati (e. g. viii. 9. 
13); of the second, pdldh (viii. 6.10); of the third, bh&ydi iddzm (ix. 
6. 18). 

The corresponding rules of the other treatises are Rik Pr. i. 4 (r. 20, 
21, xxi, xxii) and xviii. 19 (r. 36, 37), Taitt. Pr. xxii. 14, Vaj. Pr. 

374 W. D. Whitney, [i. 52- 

iv. 105. The Rik Pr. farther adds (xviii. 20, r. 40, 41) that, while a long 
vowel is heavy, it is yet heavier if accompanied by a consonant. 

sFprrf^tr ^ u \% w 

53. Also a syllable containing a nasalized vowel. 

The commentator's illustrative citations are the same which he has 
already once given us, under rule 27 ; it is unnecessary to repeat them 

The other treatises have the same rule (R. Pr. i. 4, r. 21, xxii, and 
xviii. 19, r. 38 ; T. Pr. xxii. 14), but with the difference that the former, 
admitting the anusvara as a separate constituent of the alphabet, de- 
clares a vowel followed by anusvara to be heavy. 

c^t% sru^ii 

54. And at the end of a word. 

The commentator simply paraphrases the rule, and adds one of his 
staple lists of illustrations, viz. godhuJe etc. (see under i. 3). The Vaj. 
Pr. (iv. 105, cited under r. 51, above) holds a like doctrine. The Taitt. 
Pr. (xxii. 14, 15) restricts the heaviness to such final syllables as end 
with a consonant, as our own commentator would seem to do by the 
instances he cites. It is not meant, of course, that in the combinations 
of the phrase the final syllables of words are heavy, but in the disjoined 
or pada-text, where each final is followed by a pause, or at the end of a 
verse or phrase. The Rik Pr. makes no mention of this case. 

55. Consonants belong to the following vowel. 

This and the three succeeding rules concern the division of words into 
syllables, and the assignment of the consonants they contain to the 
proper vowels. It is a matter of pretty pure theory ; the only practical 
bearing it can have must be in determining whether such and such a 
consonant shall receive one or another accent, as being that of the pre^ 
ceding or of the following vowel : and this itself must be almost un- 
mixed theory, since it can hardly be claimed that even sonant conso- 
nants share at all in accentuation : certainly they do not do so con- 
sciously. The teachings of the different Praticakhyas are very nearly 
accordant upon the subject, and this general introductory rule is equiv- 
alently stated by all (R. Pr. i. 5, r. 23, xxiv, and xviii. 17, r. 32 ; V. Pr. 
i. 100; T. Pr. xxi. 2). 

The commentator gives as instances again dadhi and madhu, which 
are to be divided da-dhi and ma-dhu. 

tfjftJTTf^ <J=for»lMA»l 

56. The first consonant of a group belongs to the preceding 

i. 58.] Atharva- Veda Prdliqd&hya. 875 

The commentator here does his work very unsatisfactorily : he fabri- 
cates his illustrations, instead of drawing them from the Atharvan text, 
giving atra sati, adravati, pradravati, and he does not note for us the 
fact that, in the combinations which he presents, the former consonant 
is to be doubled, by iii. 28, and then inform us to which of the two 
products of duplication the precept of the rule applies. In the Rik Pr. 
(i. 5, r. 25, xxvi; also xviii. 18, r. 34), the name samyog&di belongs to 
the second letter, as being the first of the original combination or sam,' 
yoga, while the one preceding it is specifically the product of the dupli- 
cation (kramaja) : and the treatise allows it to be counted either with 
the preceding or following syllable : thus, either at'tra or att'ra. The 
Vaj. Pr. (i. 102) calls the first consonant of the group as it stands after 
duplication samyogddi, and unites it with the former syllable : and in 
the same sense, probably, the term is to be understood in our own 
treatise and in the Taitt. Pr. (xxi. 4) : we are to write and divide at • tra 
sati, dd'dravati, prad'dravati. 

The commentary adds : aparaaha: hasayamam purvasye Hi, of which 
the meaning is obscure and the pertinence questionable. If it has to 
do with the disposition of the yama, it ought to come in under rule 58 
or 104. 

CPJT ST ii Hsu 

57. As does also a final consonant. 

The commentary offers once more godhuk etc, (as under i. 3). 
The equivalent rules of the other treatises are Eik Pr. xviii. 17 (r, 
32), Vaj. Pr. i. 101, and Taitt. Pr. Xxi. 3. 

58. And one generated by krama after r and h. 

The commentator offers no explanation of the rule, merely adding to- 
ft,, in his paraphrase, the words purvasvarasya bhavali, and proceeding at 
once to give his illustrations. These are the same which appear again 
under iii. 31, and also, in part, under i. 100 : they are for the most part 
words which do not occur in the Atharvan text, and, being much cor- 
rupted, are in more than one case of doubtful reading, A comparison 
of the illustrations under some of Pacini's rules (viii. 3. 26, 27 j 4. 46) 
is of important use in restoring their true form. They are arkah, area 
(so under Pan. viii. 4. 46 ; MS. artka, arcco), vartalf (MS. gartte, vartto), 
bhargah (MS. bhagnah, bhagal}: found in AV. only at xix. 37, 1), prah* 
nah, purv&hnah, aparahnak (ix. 6. 46), apa kmalayati (MS. apa brahma 
layati, apa hyalali), vi hmatayati (MS. under iii. 31 vi hyalati), apa 
hnute (omitted under i. 100), vi haute (omitted here), and brahma (e. g. 
i. 19. 4). In all these words, the consonant following the r or the k is 
doubled, by iii. 81, and the former of the two, which is regarded as the 
one that owes its existence to the krama, or duplication, is to be reck- 
oned as belonging to the preceding syllable. Thus we are to read and 

vol. vii. 48 

376 W. D. Whitney, [i. 58- 

divide arh-kah, arced, vart'tah, bharg'gah,prahn-nah,p&rv'vahn'nak, 
aparahn-nah, apahm-malayati, apahn-nute, brahm-ma. 

The rule i. 104 of the Vaj. Pr. corresponds in meaning with this, 
although more general in its form ; the Taitt. Pr. (xxi. 5) teaches that 
a consonant not combined immediately with a vowel belongs to the 
preceding syllable, which would leave only the final member of any 
group to be attached to the following vowel : there are some exceptions 
made, which need not be noticed here. In the Rik Pr., the simple and 
frequent case of a consonant doubled after an r does not seem to be 
provided for at all : its rule (i. 5, r. 26, xxvii) is constructed only for a 
case in which the consonant following the r is itself succeeded by an- 
other : one is tempted there to reject the commentator's interpretation, 
and understand the rule to mean " two consonants are reckoned as be- 
longing to the preceding vowel, when there is duplication of the second 
of a group :" this would make it accord with our own. 



59. A short vowel is of a single mora. 

The commentator gives us again, as instances, dadhi and niadku. 

The word translated 'mora' is matrd, 'measure,' a term commpn in 
this sense to all the Pratigakhyas. It is the fundamental measure, 
which cannot itself be defined by anything else. Only the Rik Pr. 
(xiii. 20) attempts to fix the length of the short, long, and protracted 
vowels, by comparing them with the cries of certain birds. 

The corresponding definitions of the other treatises are Rik Pr. i. 6 
(r. 27, xxviii) ; Vaj. Pr. i. 55, 56 ; Taitt. Pr. i. 33. 

otRRTFr ST || ^o n 

60. The consonants are of the same length. 

The commentator's illustrative instances are again dadhi and madhu. 
All the other treatises (R. Pr. i. 7, r. 34, xxxv; V. Pr. i. 59; T. Pr. 
i. 37) agree in assigning but half a mora as the length of a consonant. 

61. A long vowel has two moras. 

The commentator's instance is gala (ix. 3. 17). 
There is no discordance among the Pratitjakhyas upon this point : 
compare Rik Pr. i. 6 (r. 29, xxx) ; Vaj. Pr. i. 57 ; Taitt. Pr. i. 35. 

f5RT5p SH'. II \\ II 

62. A protracted vowel has three moras. 

The instance cited is id&3m (ix. 6. 18). All the cases of protracted 
vowels which the Atharvan text contains are rehearsed below, in rule 105. 

i. 65.] Alharva- Veda Prdtupkhya. 877 

Compare the accordant rules of the other treatises in Rik Pr. i. 6 
(r. 30, xxxi); Vaj. Pr. i. 58 ; Taitt. Pr. i. 36. 

With this rule ends the second section of the first chapter. The sig- 
nature in the manuscript is pratkamasya dvitiyah p&dah : 62. 

ii ^ ii 

63. The final of shash and puras becomes u before daga and 
ddga respectively, with, substitution of a lingual for the follow- 
ing initial. 

That is to say, shash before dapa becomes sho, and the dapa becomes 
dapa, mating the compound shodapa ; and puras with dapa, in like 
manner, forms puroddpa. The commentator cites from the text the 
words themselves merely, viz. : shodapam (iii. 29. 1), puroddp&u (e. g. 
ix. 6. 12). Neither of the words is analyzed, or restored to its theoreti- 
cally regular form, by the pada-text ; and our treatise, accordingly, ac- 
cording to its own programme, has nothing to do with them : and the 
same is true of the words referred to in the three following rules. 

These two words, with others of somewhat analogous character, are 
treated in the Vaj. Pr., iii. 39-46. 

f^ *^ot ft^? ii \$ n 

64. In the root krp, I is substituted for r. 

The whole commentary upon this rule is lost, and only its repetition 
before the next rule remains. Apparently, the copyist has carelessly 
skipped from the repetition of the rule in the commentator's paraphrase 
to that with which, as usual, the whole exposition closes. The loss is 
of very insignificant consequence : the missing passage would probably 
have afforded us some instances from the Atharvan text of verbal forms 
or derivatives of the root kip or kalp, which are frequent there. The 
rule may be taken as the assertion of an opinion that the original form 
of this root is harp; an opinion rendered plausible by the derivative 
noun krp (see the next rule), and by the analogy of the root kar, of 
which the other seems to be a secondary form. , With it corresponds 
Panini's rule viii. 2.18; none of the other Pratic&khyas offers anything 
equivalent. If our treatise has set itself to note the words in which a 
I appears in the place of a more original r, it should not pass over the 
words in which the root car becomes cal, as avic&cala, punfcali, etc., 
glaha and glahana, which are hardly to be separated from the root grah, 
udumbala (viii. 6. 17), etc. 

:* =Ml{lHI*jNl Vl" 

65. Not, however, in the words krpd etc. 

878 W. D. Whitney, [i. 65- 

Thk is the first instance in our treatise of a rule stated in this form, 
the words or phrases to which the precept contained in the rule refers 
being conceived to form a series, or gana, of which the first only is 
given in the rule, and the others comprehended in an et cetera. The 
form of statement is characteristic of the Atharva Pratiijakhya and of 
Panini, and of them only : the Vaj. Pr. employs it but once (v. 38), 
the others not at all (R. Pr. iv. 39, where, for convenience's sake, a list 
is thus referred to in one verse which is given in full in the next, fur- 
nishes but an accidental and insignificant analogy) . It would seem to 
be the business of a commentator to give the list in full, but the author 
of our commentary evidently does not think so, for he very seldom, if 
the gana have any extent, presents us more than specimens from it. 
Here, he gives krpa p&vaka (xviii. 4. 59), and krp&t svah 1 (vii. 14. 2: 
the reading doubtless is a corrupt one, and should be krp& svah, as is 
read by both the Sama and Yajur-Vedas, in their corresponding verses) ; 
also krpanah (krpanah, xi. 8. 28), and its derivative karpanyam (not 
found in AV.). If these two words, which come from altogether an- 
other root, actually belong to the gana, it should contain also krpama- 
nasya (v. 19. 13) and akrpran (xviii. 3. 23). 

With this and the preceding and following rules are to be compared 
Pan. viii, 2. 18, and the vdrtikas upon it. 

H*I(W t^: MI<H$iHplr*)o|HI<OHI^iiU« 

66. In pddam arigulim etc., r is substituted for I. 

The instances given by the commentary as coming under this rule 
are papre padarn angurim (iv. 18. 6 and v. 31. 11), saham&ran ami 
daha (v. 29. 11), ydhi may&raromabhih (vii. 117. 1), and apvasya v&rah 
paritshasya v&rah (x. 4. 2). The gana should also include pancangurih 
(iv. 6. 4), svangurih (vii. 46. 2), anangureh (viii. 6. 22), and perhaps 
tirya (for tilya, from tila : iv. 7. 3) : angurim also occurs again in xx. 
136. 13. As counter-instances, to show the necessity of constructing a 
gana, of a limited number of instances, the commentator cites anguli- 
bhyo nakhebhyah (ii. 33. 6), and balds te prokshanih santu (x. 9. 3). 

It is not in accordance with the usage of our treatise elsewhere to 
give, in citing a word or phrase in a rule, another form than that which 
it actually has in the text : we should have expected here tfrjjrrjf^fit . 
The form ity evam Adi, instead of simply adi, is found once more, in 
ii. 29. 

H=hl^l(<T i HN ^HkjHlfa*: II \<o ll 

67. In ease of the loss of a n or m, the preceding sound be- 
comes nasalized. 

The cases of elision of n and m are taught below, in ii. 32-34, which 
see for illustrations. The commentator offers here only the words 

1 krpasvih. 

i. 70.] Atharva- Veda Prdticdhhya, 379 

vinpatih (e. g. v. 15. 2) and paydnsi (e. g. i. 9. 3) — which are very ill 
chosen, since, though each offers an example of a nasalized vowel, nei- 
ther exhibits an elision of an original nasal mute, according to any rules 
contained in this treatise. 

Corresponding rules to this and the following one of our treatise are 
offered by the other Praticakhyas : see Rik Pr. iv. 35 (r. 79, ccxcix); 
Vaj. Pr. iii. 129, iv. 3; Taitt. Pr. xv. 1 : there are some differences of 
application, but chiefly dependent upon the different modes of treat- 
ment of the nasal mutes adopted by the different authorities, which 
will be explained in their place. 

68. Also in case of their conversion into y, r, or a spirant. 

The instances given by the commentary are as follows : rathan iva 
(v. 13. 6), salavrkan iva (ii. 27. 5), khalvan iva (e. g. ii. 31. 1) — in all 
these cases, the final n is first, by ii. 27, converted into the spirant visar- 
janiya, the latter then changed, by ii. 41, into y, and this finally, by ii. 
21, dropped altogether; so that we have the successive steps rathan iva, 
rathdnh iva, rathany iva, rathan iva — farther, rtunr rlubhih (not found 
in A V.), rtunr ut srjatc vacA (vi. 36. 2), mo aha paninr abhi (v. 1 1. 7 : 
the commentator repeats the first word in its pada form, mo iti, at the 
end of the citation), and dasyunr uta bodhi (iv. 32. 6) — in these in- 
stances, the final n, by rule ii. 29, becomes r, and, the preceding vowel 
being nasalized, rtun ut is converted into rtunr ut. 

As the n must always be converted into the spirant visarjaniya before 
it becomes y, it seems superfluous to make separate mention of the latter 
in the rule. The commentator apparently feels this objection, and ven- 
tures for once a defence, as follows : ushmano grahanat siddhe punar- 
grahanena kim : nityalvam na sy&t: rtunr ul trjate vapi; 'when the 
matter is made certain by the use of the term us'hman, why any farther 
mention ? it is because this does not apply to all cases, as is shown by 
the instance rtunr tit srjatc vafi.' I do not see the point of this defence : 
it does, indeed, explain the mention of r in the rule, but it has nothing 
to do with that of y. 

sr^prrffr^T ^ ^f^^%nUn 

69. And in case of the combination of a nasalized vowel with 
a preceding vowel. 

The only cases cited by the commentary are those of the combina- 
tion of the initial vowel of ahfa with a preceding final vowel, by simple 
fusion or by the elision of the initial a; they are : ubhdv updnfu (pada 
upa-ahpi) pralhama pibaoa (iv. 32. 7), somasyd 'nfo (vii. 81. 3), and ye 
vrlhayo yava nin/pyante ' nf avail (ix. 6. 14). 

Compare Rik Pr. xiii. 10 (r. 26), Vaj. Pr. iv. 51, Taitt. Pr. x. 11. 

5^T W Sf^rf ^TtT^TR ll u° ll 

70. In the passage purusha a babhuvdri, the vowel is nasal be- 
fore the pause. 

880 W. D. Whitney, [i. 70- 

The passage referred to is x. 2. 28 : sarvA dicah purusha a babhuvan, 
where, in a case of doubt and questioning, the final a of babhura is both 
protracted and nasalized. The pada-text reads simply purushoh: d: 
babhunani: and there would be no call for such a rule as that given 
here, but for the requirements of the kmma-text, in which babhuva, as 
the last word in averse, must suffer parih&ra (iv\ 11 7), or repetition 
with ill interposed, and in which it might bo made a question whether 
the nasality of the vowel should or should not be preserved before the 
it). This rule teaches us that the nasal quality is lost before the it), as 
rule 97, below, teaches also with respect to the protraction ; and the 
same things are taught once more by iv. 120, 121. The three last 
kmmapodas of the verse will be, then: purusha & balhdv&ns: a babhit- 
vAnZ : babhuve 'ti bubhtivati.3. 

*L^UItU 7<Wr^ 3rT II '^\ II 

71. Of the r- vowels, the part following the r receives the 
nasal quality. 

"We have seen above, in rules 37 and 38, that the r-vowel is regarded 
as composed of a piece of a r, with a fragment of vowel sound pre- 
ceding and following it, and that, when it is long or protracted, the r- 
quality is found only in the first mora. Here we learn that, when such 
a vowel is nasalized, the nasal quality does not affect the r, but only the 
part of a vowel which follows it. Any one may perceive, however, 
upon trying the experiment, that there is no physical difficulty in the 
way of nasalizing the r itself, supposing the r-vowel to be properly ac- 
cordant in pronunciation with that letter throughout. 

The commentator cites bhurnidraham acyvUim p&rayishnu (v. 28. 14), 
drnha pratu&n (vi. 136. 2), and janan drnhantam (xii. 2. 9). The in- 
stances, as in many other cases, are wanting in variety and in complete- 
ness : as an example of the long vowel nasalized, we may take pitrnr 
vpe 'mam, already cited under rule 38 : no case of the protracted vowel 
nasalized occurs in the text. 

The other treatises offer nothing corresponding to this rule. 

S^Wn^qWStT ii ©^ ii 

72. 27 is nasalized when standing alone, before ill. 

In the pada-taxt of the Atharvan, as in those of the other Yedas, the 
particle u is always written vn iti. In this rule, its nasality in such a 
situation is noticed : in the rule next succeeding are taught its long 
quantity and its exemption from conversion into a semivowel before the 
following vowel. 

The term aprkta means ' uncombined with any other letter :' it is 
said also of the particles d and o (=« + «) in rules i. 79, iv. 1 13, below. 

^f: m\^m li ^ li 

78. In the same situation it is also long, and pragrhya. 

>. 74.] Atharva- Veda Pr&tkdkhya. 381 

The term pragrhya means, by implication, that the vowel to which 
it applies is not liable to the ordinary changes of sartdhi, viz. fusion 
with, or conversion into a semivowel before, a following vowel. I say, 
by implication : for only in the Taitt. Pr. (which uses, however, not 
pragrhya, but the related term pragraha) does the pronouncing a vowel 
pragrhya exempt it from change ; all the other treatises find it necessary 
to teach by a specific rule (see iii. 33, below, and the quotations there 
given) that the vowels declared to be pragrhya are not subject to eu- 
phonic alteration. The whole proceeding is somewhat analogous with 
that by which the Rik Pr. teaches the conversion of visarjaniya into r ; 
first rehearsing all the cases in which the conversion takes place, and 
pronouncing their visarjaniya to be rephin or riphita, and then finally 
declaring the riphita visarjaniya convertible into r. The word pragrhya 
is explained by Bohtlingk-Roth to mean literally " to be held apart, or 
isolated," i. e., from the combinations of sandhi. 

Any satisfactory reason why the particle u should be treated in this 
peculiar manner by the framers of the ^arfa-text is not readily apparent. 
There are but few cases in our text in which it assumes a long form in 
sanhitd (viz. eight instances : they are given under iii. 4), so that it can 
hardly be said to exhibit any special tendency to protraction ; it nowhere 
assumes a nasal quality in the combined text ; and it has hardly a trace 
of a proper pragrhya character : if, indeed, it be preceded by an un- 
combined vowel and followed by another vowel, it remains uncombined 
with the latter (by iii. 36, which see : only three such cases occur in our 
text) ; but, on the other hand, if preceded by a consonant, it combines 
regularly with a following vowel (of this also there are only four cases 
in AV. : see ii. 37). It seems as if the protraction must have been 
made in order to give the word more substance as an independent pada 
in the disjoined text, it being the only instance of a single short vowel 
possessing such a value ; and as if the nasalization and addition of Hi 
were intended to mark it more distinctly as an exceptional case, requir- 
ing a different treatment in the sanhita-text. Panini (i. 1. 17, 18) allows 
it to be read either « or un. 

The treatise now goes on to detail the other cases of pragrhya final 

74. Final i and & are also pragrhya, in a form having a loca- 
tive sense. 

The instances cited by the commentator are ashtri padam krnute 
agnidhane (vi. 27. 3: the Rig- Veda, in the corresponding passage, has 
the proper locative form, Ashtryam), ato jdlaso dhdrayanta urvi (xviii. 
1. 32|, maht no vdtdh (xviii. 1. 39), and tanu daksham & suvatdm (iv. 
25. 5). This last, however, is a doubtful case, since the word tanu may 
quite as plausibly, or more so, be taken as nominative dual, 'their very 
selves.' A more unequivocal case of u is mdyu in xviii. 4. 4, and it is 
the only one which I have noted in the text. There is also a single 
case of a locative in i not given by the commentary : it is abhihruti, in 

S82 W. D. Whitney, [i. 74- 

vi. 3. 3. As counter examples, of final i and u in other than a locative 
sense, and therefore not pragrhya, the commentator offers dhtti va ye 
(vii. 1. 1), tasya 'm& sarvd (xiii. 4. 28). Of cases analogous with the 
former of these, where the i represents an instrumental case, there are 
several others in the text, as vii. 48. 1, 77. 1 ; ix. 9. 8. 

The ^jarfa-text carefully notes these locatives in t and u as pragrhya, 
in the usual manner, by writing an iti after them : thus, Ashtrt iti, urvi 
iti, tanii iti, etc. The commentator, in citing the several passages, 
under this and the following rules, always repeats at the end of each 
citation the pragrhya word, in its pada form, or with iti appended : I 
have omitted such repetitions, as unnecessary here. 

A corresponding rule in the Rik Pr. is found in i. 18 (r. 72, lxxiii) : 
also in Panini, i. 1. 19. The Vaj. Pr. notes no such cases as those to 
which this rule applies : and the Taitt. Pr., instead of classifying and 
defining the pragrhya terminations according to their grammatical 
values, describes them all in an entirely empirical way (in iv. 1-54), by 
their position and surroundings, whence its rules do not generally admit 
of detailed comparison with those of the other treatises. 

RcNHI-dl II oH, II 

75. The same vowels, i and u, are pragrhya as dual termina- 

The commentator's illustrations are kena p&rshni abhrte (x. 2. 1), 
indrav&yw ubhau (iii. 20. 6), ubhav indr&gni a bharatam (v. 7. 6). 

Corresponding rules are Rik Pr. i. 18 (r. 7l r lxxii) and Vaj. Pr. i. 93 ; 
both of them include also the case* noted by our treatise in the next 
following rule. 

76. As is also e. 

The commentator cites atra dadhele (v. 1. 3), rodhacakre vavrdhete 
(v. 1. 5), sam pilar av rtviye (xiv. 2. 37). 

?TFf js% 51" % ^ *fi<lrW noon 

77. Also the words asrne, yushme, tve, and me, when accented. 

The specification " when accented " is, of course, meant only for the 
two latter of the words named, as the others would never occur other- 
wise than accented. Of the four, yushme and me never occur in the 
Atharvan text : tve is found once, in a Rik passage (AV. v. 2. 3 =^RV. 
x. 120. 3), and also, according to the manuscripts, in viii. 9. 9, twice re- 
peated, and each time written in the pada-text tve iti, as a pragrhya : 
but the accent and the addition of iti are hardly to be regarded other- 
wise than as a blunder of the tradition, since the word is evidently the 
enclitic or accentless tva of the Vedic language : no forms of this en- 
clitic pronoun are found elsewhere in the Atharvan. The fourth, asme, 

i.?9.] Atharva-Veda Prdtigdkkya. 383 

is also hardly an Atharvan word. It is found in three Kik passages, viz. 
iv. 21. 1 (RV. vi. 28. 1), xviii. 1. 3 (RV. x. 10. 3), 42 (RV. x. 17. 8) : in 
another passage (iv. 31.3), where the Rik (x. 84. 3) reads asme, all the 
Atharvan manuscripts have asm&'i, which has been altered to asme in the 
edition, in obedience to the requirement of the sense, and the authority 
of the Rik reading. Another precisely similar case is xix. 40. 4 (RV. i. 
46. 6). The only passage where the Atharvan gives asme independently 
is v. 1. 3, where all the manuscripts except P. and M. (copies of the 
same original, by the same scribe) agree in reading it (pada asme iti) : 
here also, however, the edition reads asm&'i. 

The commentator cites no instances, but says nigame yushmadbhyA 
vibhakter ittvam ishyaii : yushm&kam: asm&kam: tvam aham iti pr&pte : 
asme yushme tve me iti ca vibhakty&degah kriyaie. 

The Rik Pr. (i. 19, r. 73, 74, lxxiv, lxxv) notes asme, yushme, tve, and 
ami as pragrhya : the third, tve, when accented, and not a member of a 
compound word. The Vaj. Pr. (i. 96, 97) notes asme, tve, and me, the 
latter when accented. Asme and tve are dealt with in Taitt. Pr. iv. 
9, 10. 

?pft <M£cHHH " os II 

78. Also ami, as plural. 

The examples cited by the commentator are ami ye yudham (vi. 103. 
3), ami ye vivrat&h (iii. 8. 5), and ami afapre (not found in AV). To 
explain the addition of the specification " as plural," he gives a counter- 
example, pamy aira, which is plainly one of his own fabrication ; nor 
can I find that the text contains anything which should render that ad- 
dition necessary. The Vaj. Pr. says (i. 98) "ami, when a word by 
itself;" the other treatises (R. Pr. i. 19, r. 73, lxxiv; Taitt. Pr. iv. 12) 
see no reason for appending any such limitations. 

fHMIrfl ^pt ^TRT^: no^ll 

79. Also a particle consisting of an uncombined vowel, unless 
it be d. 

This rule is meant to apply solely to the particle o, composed of A and 
«, which is found in two passages of the text, viz. o cit sakh&yam (xviii. 
1. 1) and pr&tam havir o shu (vii. 72. 2), both of which are cited by the 
commentator: the pada-text writes the o in the usual manner of a 
pragrhya, viz. 6 iti. To explain the addition of " unless it be a " to the 
rule, the commentator cites punar e 'hi vacaspate (i. 1. 2), where the pada- 
text reads, of course, a 1 : ihi. 

The form of this rule is not a little strange : why o should thus be 
made an exception from the next rule, and why, when there is no other 
particle, except a, composed of a single vowel, it should be treated as if 
one of a class, it is very difficult to see : we cannot help suspecting here 
the influence of the general grammar : compare Pan. i. 1. 14, the virtual 
correspondence of which with our rule is as close as possible. The Rik 
Pr. (i. 18, r. 69, Ixx) has a similar precept. 
vol. vii. 49 

384 W. D. Whitney, [i. 80- 

80. Also one ending with o. 

That is to say, as we must infer from the preceding rule, and as the 
commentator fills out the ellipsis, a nipata or ' particle,' having o for its 
final. This is a strangely inaccurate description : it was bad enough to 
have the upasarga or preposition A treated as a nipata by the last rule, 
when combined with u: but here we have nouns, verbs, prepositions, 
and particles all confounded together under the same name. The par- 
ticles, it is true, greatly preponderate in number and in frequency : thus 
we have atho (about 130 times in the whole Atharvan text), mo (15 
times), no (12 times), uto (7 times), and iho, yado, ango, evo, dosho (once 
each) ; but of prepositions we have o and upo (twice each), and pro 
(once) ; of verbs, vidmo, datto, atto (once each) ; and of nouns (pro- 
nouns), teno (twice), yo, and m (once each). In the form of the rule is 
perhaps to be seen again the influence of the general grammar : compare 
Pan. i. 1. 15. The other treatises are not open to the same criticism : 
the Rik Pr. (i. 18, r. 70, lxxi) declares pragrhya a final o, except of the 
first member of a compound ; and the Vaj. Pr. (i. 94, iv. 89) constructs 
its rule in very nearly the same manner. 

As regards the actual pragrhya character of these words, there are, 
among the nearly 200 instances of their occurrence, but 11 cases in which 
they stand otherwise than before a consonant or an initial a, and so have 
an opportunity to exhibit that character distinctly. These cases are : be- 
fore a, xx. 127. 13 ; before i, vi. 14. 3, xiv. 2. 4, xx. 130. 17, 18 ; before v, 
xi. 6. 7, xii. 1. 7, 9 ; before e, ii. 9. 1, vii. 56. 5, ix. 8. 7. In xx. 130. 19, and 
only there, an initial a is absorbed by such a final o ; on the other hand, 
in iv. 9. 3, the metre shows that such an absorption of an initial a must 
be made in reading, though it be not so written. In teno (ix. 1. 20) and 
yo (xi. 4. 9), the metre shows that the combined particle u must be sep- 
arated from the final of the original word, and that the two must be 
read Una u and yd' u. 

The examples given by the commentary are dosho gdya (vi. 1.1), 
arigo nv aryaman (vi. 60. 2), atto havinshi (xviii. 3. 44), and datto asma- 
bhyam (xviii. 3. 14). 

81. Also a vocative ending in the same letter, before an iti 
not belonging to the text. 

Literally, ' before an iti not coming from the rshis' or authors of the 
hymns: that is to say, before the iti by which, as already remarked, a 
pragrhya word is followed in the pada-text. The vocatives in o, from 
themes in u, are not in a single instance treated as pragrhyas in the 
sanhitd of the Atharvan, but are always euphonically combined with 
the following vowel.* In the pada-text, however, they are invariably 

* The cases are not numerous in which such a vocative occurs elsewhere than in, 
pausa, before a consonant, or before an a ; they are as follows : before a, v. 1 3. 5 ; 

i. 82.] Alharva- Veda Prdtiqdkhya. 885 

written as if they were pragrhyas, with the usual iti annexed. The 
object of this rule, then, is to teach that they are exempt from euphonic 
combination only in the pada-text, while in other situations they are to 
be treated according to the general euphonic rules (iii. 40, ii. 21). The 
Vaj. Pr. (iv. 89) has a rule corresponding with that of our treatise ; 
it, however, seems to be inconsistent with a previous rule (i. 94), which 
teaches that a final o is pragrhya in general, and not before the iti of 
the pada-text alone. The usage of the sanhitd-text is in accordance 
with the later rule, and not with the earlier, so far as I can judge from 
the passages which correspond with those of the Atharvan referred to in 
the marginal note : I am surprised that Weber has not taken any notice 
of this discordance between the text and the Praticakhya. The Taitt. Pr. 
(iv. 6) says that o is pragraha when it is not the product of sandhi, and 
is followed by a or a consonant — which is a rather absurd way of say- 
ing that it is not pragraha at all in sanhitd ; since before a consonant 
its pragrhya character could not, and before a need not, appear. The 
Rik Pr., after declaring the o of the vocative pragrhya (i. 18, r. 68, lxix), 
is obliged later (ii. 27, r. 52, clvii) to except it from the rule that^ra- 
grhyas are exempt from euphonic change, and to place it under the 
control of rules previously given for its combination with succeeding 
vowels. Finally, Panini (i. 1. 16) gives a rule precisely corresponding 
to ours, but gives it upon the authority of Qakalya. This whole state 
of things is something very peculiar. Why, when the o of vdyo is 
really no more exempt from change than the e of agne, should it be re- 
garded by all the pada-texts as a pragrhya, causing so much trouble to 
the different treatises to explain its treatment ? 

The commentator cites, as examples of the rule, tvayy udite pre "rate 
citrabh&no : citrabh&no iti (iv. 25. 3), yuvaih v&yo savita : v&yo iti (iv. 
25. 3), and manyo vajrin: manyo iti (iv. 32. 6). As counter-examples, 
to show that the vowel is unchangeable only before the iti of the pada- 
text, he gives vdya utaye (iv. 25. 6), manya iditd (iv. 31. 4), and babhra 
& me frnuta (v. 13. 5). 

82. la drtni iva etc., the iti follows the iva. 

This is a rule which concerns only the writing of the pada-text itself, 
and so, as dealing with a matter lying outside of the proper sphere of a 
Pratigakhya, is extra-judicial, and has no correspondent in either of the 
other treatises. It grows out of the difficulty, in a few special cases, of 
combining two methods of writing usual in the pada-text. This text, 
in all the Vedas, always combines the enclitic particle iva, ' as if, like,' 
with the word to which it is attached, as if forming a compound with 
it, giving up often, in favor of this combination, the division which 

before i, vii. 4. 1 ; before i, iv. 81. 4; before «, vi. 68. 1, vii. 26. 8 (bis); before 4, 
iv. 25. 6. In iv. 32. 1, the final o absorbs a following initial a; everywhere else, it 
and the following a both remain unchanged. 
1 "dif^irt cq": 

W. D. Whitney, fi. 82- 

would otherwise be made of a preceding compound : thus, uda-dhim 
(iv. 15. 6), but udadheh-iva (i. 3. 8). When, now, the iva happens to 
follow a pragrhya word, like artni, which ought to be followed in the 
pada-text by iti, in order to bring to light its pragrhya quality, what is 
to be done ? shall we separate the two parts of the compound word — a 
thing unheard of elsewhere — and introduce the iti between them, writ- 
ing Artni iti 'vd "rtni-iva ? or shall we allow the iti to lose its proper 
function, but still be retained at the end of the compound, in order to 
call attention to the pragrhya quality of the first member of the latter, 
and write Artni ive Hy artni-iva ? The second of these two alternatives 
is the one adopted by all the pada-texts, and the one which our rule 
here teaches us to choose. The Atharvan text offers but four such 
cases, which, for once, are all cited by the commentator; they are as 
follows : artni ive Hy Artni-iva (i. 1. 3) ; gharmadughe ive Hi gharmadu- 
ghe-iva (iv. 22. 4) ; nrpati ive Hi nrpati-iva (viii. 4. 6) ; yame ive Hi yame- 
-iva (xviii. 3. 38).* 

4HH|£kTT ^^ ^F*: ll ^ li 

83. A nasalized vowel occurring in the interior of a word is 

Here we have the general fact laid down, and in the following rules, 
to the end of the section, are stated the exceptions to it. The Rik Pr., 
in one of its later books (xiii. 7-10), treats the same subject, and the 
commentator is at much pains (see Regnier's note to r. 22) to explain 
its introduction into the Praticakhya, into whose proper province such 
a matter does not enter. Our own commentator seldom troubles him- 
self about little inconsistencies and redundancies of this kind, which are 
exhibited by all the treatises ; they aid in the general purpose of a 
Praticakhya, which is to preserve the traditional text of the school 
from corruption. Thus, the Taitt. Pr. presents (xvi. 1-31) a complete 
conspectus of all the nasalized vowels, short and long, found in its text 
in the interior of a word, and again (xiii. 8-14), a detailed exhibition 
of all cases of occurrence of the lingual nasal, n. 

The commentator cites a third time the whole series of instances 
given above, under rule 27, and repeated by him under rule 53. 

^f JTJ^TST^R II *$ II 

84. In neuters plural it is long. 

The commentator gives, as examples, parunshi yasya sambhar&h (ix. 
6. 1), yajunshi hotra brumah (xi. 6. 14), atto havinshi (xviii. 3. 44). 

An equivalent rule is found in the Rik Pr. (xiii. 7, r. 22), which far- 
ther specifies that the theme ends in a spirant, and that the long vowel 
precedes the terminations si and shi. The Taitt. Pr., ignoring all help 

* I have given the words here in the full form in which the pada-text presents 
them : our commentator, in his citations, leaves off the repetition of the compound, 
writing Bimply drtni ive 'ti, etc. 

i. 87.] Atharva- Veda Prdtigdkhya. 887 

from grammatical categories in the construction of its rule, as is its 
custom, says (xvi. 14) that a, i, and u are nasal before si and shi at the 
end of a word. 

qrf§ i rmi{lHi i jniT;Mji 

85. Also in pdnsu, mdnsa, etc. 

The commentator cites pdnsun akshebhyah (vii. 109. 2), mdnsam mdn- 
sena (iv. 12. 4), pdnfayenu (vi. 129. 1 [should be $dn$apena? the manu- 
scripts blunder somewhat over the word, but W. E. and H. read dis- 
tinctly fdnfapena]), and fild bhumir afmd pdnsuh (xii. 1. 26). To the 
words thus instanced I have only to add pdnsure (vii. 26. 4), which may 
perhaps be regarded as virtually included in pdnsu. 

The form of this rule is quite peculiar, in that it cites two words, 
instead of one, as heading of the gana. 

#IWf: HftiisMi 

86. Also in a desiderative form from the roots han and gam. 

Of desiderative forms from the root han the text furnishes usjighdn- 
sati (e. g. iv. 18. 3) and jighdnsan (vi. 99. 2). From gam we have no 
such forms, unless, in xii. 4. 29, 30, we are to amend yadd sthdma 
jighdnsati into jigdnsati, which would very much improve the sense, if 
I am not mistaken. Could we trust implicitly to the Praticakhya to 
include in its rules no forms not actually to be found in the Atharvan 
text, this passage would be a sufficient warrant for making the alteration 
suggested : but that is not the case, as the very next rule, for instance, 
notably shows. The reason why these two roots are thus put together 
as the subject of such a grammatical precept may be to be sought in the 
general grammar; compare the equivalent rule in Panini (vi. 4. 16), 
which offers also the same technical term, sari, for a desiderative form. 

The commentary offers as examples the word jighdnsati (e.g. iv. 18. 
3), and the passage spoken of above, yadd sthdma jighdnsati ; and, 
although our manuscript here reads, like those of the text, jighdnsati, 
the absence of any other citation or fabricated illustration of desidera- 
tive forms from gam gives a degree of color to the conjecture that our 
commentator may have meant to give jigdnsati, 

JUKI KM 1^ II *to II 

87. As also from the roots gdn, mdn, and ddn. 

Of these three roots, only man offers in the Atharvan text any forms 
falling under this rule. The commentator cites one of them, mtmdnsa- 
mdndh (ix. 1. 3) : the others are mimdnsamdnasya (ix. 6. 24), mtm&nsi- 
tasya (ix. 6. 24), and amimdnsanta (xii. 4. 42). ' The form mdnsta (xi. 
2.8), as not being of desiderative origin, does not properly belong 
here, but, if genuine, should be included under rule 85 : it may be a 
corrupted reading iovmansta; we have the corresponding second per- 

388 W. D. Whitney, [i. 87- 

son, manslhds, in ix. 5. 4, and there also a part of the manuscripts 
(P. W. I.) read mdnsthdh. For gdn and rfdra the commentator evi- 
dently had no genuine instances at command, and he fabricates pfdn- 
sati, diddnsati. Here also it is a suspicious circumstance that a rule of 
the general grammar (Pan. iii. 1. 6) groups these three roots together: 
although, it is true, for a different purpose from that which calls forth 
our rule. 

cltcHHI CRTTOFr ll ^ ii 

88. Also in a strong case from a theme in vans. 

The strong cases (pancapadi, 'five words or forms') are the mascu- 
line nominatives singular, dual, and plural, and the accusatives singular 
and dual. The suffix vans is that which forms the perfect active parti- 
ciple : it is called in Panini by the same name as here, vasu. The 
commentary cites as instances pareyivdnsam (xviii. 1. 49), pravifivdnsam 
(iv. 23. 1), uttasthivdnsah (vi. 93. 1), and papivdnsah (vii. 97. 3). 

§m% u ^ n 

89. As also from a theme in iyans. 

That is to say, in a strong case of a comparative of the ancient forma- 
tion, or that produced by adding the primary suffix iyans to the, gene- 
rally gunated, root. The commentator gives as examples preydn, pre- 
ydnsdu, preydnsah ; but the only strong case of this word occurring in 
the Atharvan text is preyansam (xv. 10. 2). 

j^sr ii ^o a 

90. As also from the root vid. 

There are two damaging objections to be made to this rule : in the 
first place, it ought to be brought in, if at all, after rule 88, in order 
that vasvantasya as well as pancapadydm may be implied in it by in- 
ference from its predecessor; and in the second place, there is no 
need of any such precept at all, since there is no good reason why 
vidvdn, the word to which it alone applies, should not be considered a 
vasvanta, and therefore regarded as disposed of by rule 88. The Hindu 
theory, indeed, does not regard vidvdn as a perfect participle, and Panini 
(vii. 1. 36) is obliged to teach that in it the perfect participial suffix is 
substituted for that of the present participle ; and probably it is out of 
this circumstance that the introduction of the rule here in question has 
proceeded : yet, the substitution having been made, vidvdn would have 
to be deemed and taken for a vasvanta, one would think, even by the 
Hindu theory itself. 

The commentator gives all the strong forms of vidvdn, of which only 
a part, however, are to be found in the Atharvan, and then winds up 
with an actual citation; as follows: vidvdn (e. g. ii. 1. 2), vidvdnsdu, 
vidvdnsah, vidvdnsam (e. g. ix. 9. 4), vidvdnsdu ; vidvdnsam vrdlyam 
(e.g. xv. 2. 1). 

i. 93.] Aiharva-Veda Prdtigdkhya. 389 

JOT uMu 

91. As also from pumans. 

The commentary instances the five cases of pumans to which the name 
pancapadi belongs ; only one of those to which the rule actually applies 
occurs in our text : puman (e. g. i. 8. 1), pumdnsdu, pumdnsah, pumart' 
sam (e. g. iii. 23. 3), pum&nsau. Then he adds a counter-example, to 
show that the rule is meant for the strong cases alone : pumsi vai reto 
bhavati (vi. 11. 2). 

Here ends the third section of the first chapter : the signature in the 
manuscript is prathamasya trtiyah pddah : 91. 

oiufu^ur^cf OTrrii^ii 

92. A sound preceding a final sound is called its upadhd. 

This is simply a definition of the term upadhd, and, to illustrate the 
rule, the commentator cites the two later rules, ndmyupadhasya rephah 
(ii. 42) and dkdropadhasya lopah (ii. 55), in which the term is employed. 
The Vaj. Pr. (i. 35) has precisely the same definition. In the Rik Pr. 
the word has a more general use, as ' preceding letter or word ' (upa-dhd, 
'a setting against or next to') : it is probably on account of this less 
restricted signification current in some schools that the two treatises 
first spoken of deem it necessary to limit the term by a specific defini- 
tion. The Taitt. Pr. does not employ it at all. 

93. A vowel is a syllable. 

The precise scope of this rule it is not easy to determine ; it seems 
to be rather a general and theoretic doctrine than a precept which 
enters in any active and practical manner into the system of rules of 
our treatise. The Vaj. Pr. and Rik Pr. have similar rules, and that of 
the former (i. 99) is expressed in identical terms with our own ; it re- 
ceives an easier interpretation than ours by being placed at the head of 
the rules for syllabication, which correspond to our rules 55-58, above. 
The Rik Pr. (xviii. 17, r. 31) states more fully that a vowel, whether 
pure, or combined with anusvdra, or combined with consonants, is a 
syllable ; as also (i. 4, r. 19, xx) that both the short and the long vowels 
are syllables; making the former declaration an introduction to the 
rules for syllabication and quantity, and the latter, to the briefer treat- 
ment of the same subjects in the first chapter. We may perhaps regard 
our rule as a virtual precept that the accentuation, which in later rules 
(iii. 55 etc.) is taught especially of the vowels, extends its sway over the 
whole syllable : or, on the other hand, that the accents, which in rules 
14-16 above were declared to belong to syllables, affect especially the 
vowels. "With the subject of accent the commentator seems, at any 

390 W. D. Whitney, [i. 93- 

rate, to bring it into special connection. Omitting his usual explanatory 
paraphrase (a small loss : it would doubtless have been svaro 'ksharam 
bhavati), he proceeds at once to give an exposition, of which a part 
occurs again at the close of the third section of the third chapter; it 
reads, unamended, as follows : kirn aksharasya svaryamdnasya svaryate : 
arddham hrasvasya p&do dirghasye 'ty eke: sarvam iti pdnkliamitrah 
(under ii. 6 and iii. 74, fankhamitrih) : aksharasydi 'shd dhdnam (vi- 
dhand) vidyate yad yad virasvardbhdvah (yad yad vifvaribhava : the 
passage goes no farther under iii. 74) : svaram aksharam ity dhuh : sva- 
rdd anyat vyamjanam sarvam prthak varnasamdnyam dvyaktdmjyate 
budhdih. I translate, in part, as follows, not without some misgivings : 
'what part of a circumflexed syllable is circumflexed ? some say, half a 
short one, quarter of a long one : Qankhamitri says, the whole : here is 
found no rule for a syllable [hrasva and dirgha are said of vowels only : 
see note to r. 51, above]; since, in each case, the vowel alone is con- 
templated [? ?] : now the vowel is declared to be the syllable ....,' and 
hence, perhaps, what is taught of the vowel must be understood to be 
said of the whole syllable. Yet all this would appear to be rendered 
unnecessary by the rules for syllabication, which, as we have seen, hardly 
have a meaning if they do not imply that each consonant shares in the 
accentuation of the vowel to which it is declared to belong. 

#5#T qcfoFpTTli^li 

94. The sound preceding an aspirate becomes a non-aspirate. 

The proper application of this rule, within the sphere of the Pratiga- 
khya, is only to cases of the doubling of the first or second consonant 
of a group, by the varnakrama, as taught in rules iii. 26 etc., yet its 
form of statement is general, and there can be no doubt that it should 
apply to all cases arising in the course of derivation and inflection, and 
that forms such as mrdhdhi, containing a double lingual sonant aspi- 
rate, are strictly excluded by it. Such forms, as is well known, occur 
in almost all the Vedic manuscripts, and those of our own text offer 
several instances of them ;* which, however, we have not hesitated to 
amend in the printed text (except in ii. 5. 4, where the correction has 
been accidentally omitted) to ddh. 

Corresponding rules in the other Prati$akhyas are : Eik Pr. vi. 1 (r. 
2, ccclxxix), and xii. 3 (r. 9) ; Vaj. Pr. iv. 106 ; Taitt. Pr. xiv. 5 : that of 
the Vaj. Pr. and the former one of the Eik Pr. are restricted in terms 
to the cases of duplication arising under the rules of the varnakrama; 
the others are general precepts, like our own. 

The examples given by the commentary are such as illustrate the ap- 
plication of the rule to forms of derivation and inflection, as well as of 
krama ; they are iddhani (sam-iddham, vii. 74. 4), dugdham (e. g. x. 6. 

* The details are as follows : ii. 5. 4, all the MSS. dhdh ; vii. 46. 1, all do. ; 
vii. 97. 7, & I. do., the rest dh; xi. 1. 29, Bp. dh only, all the rest dhdh; xi. 1. 31, 
first time, all dhdh: second time, Bp. dh, the rest dhdh; xii. 2. 19, B. dhdh, all the 
rest dh; xviii! 3. 42, all dhdh; xviii. 4. 1 do.; xviii. 4. 56 do. The true reading, 
ddh, is not given in a single instance by any of the MSS. 

i. 96.] Atharva- Veda Prdtigdkhya. 391 

31), nanv d ruruddhre (iv. 31. 3 : ordinary reading rurudhre; but in 
this instance, as occasionally elsewhere, the pada manuscript obeys the 
rules of the krama, and gives ruruddhre), yo daddhre (xviii. 3. 63, 
dadhre), valagam vd nicakkhnuh (x. 1. 18, nicakhnuh). The commenta- 
tor then once more commences his citations from his metrical authority, 
and gives the verse prathamdf ca dvitiydnam samyoge pratyanantaram : 
trtiyd$ ca caturlhdndm etai sarvatra lakshanam ; 'first mutes are sub- 
stituted for seconds, when directly preceding the latter in a group ; and 
thirds in like manner for fourths : this is a rule of universal application. 1 

4IMMUI ^frT: iiHh 

95. Conversion is according to propinquity. 

That is to say, when any sound is ordered to be changed into another, 
of any class or description, we are to convert it into that one which is 
nearest to it, in situation or in character. The commentator, after his 
customary repetition of the rule, by way of paraphrase, with the bare ad- 
dition of bhavati, proceeds, without any farther explanation, to cite three 
rules in the interpretation of which it needs to be applied. The first 
of these is ii. 31, to the effect that m before a mute is converted into a 
letter of the same position with it : which, by this rule, must be under- 
stood, m being a nasal, to mean the nasal letter of the series, and not 
either of the non-aspirate or aspirate surds or sonants. The second is 
ii. 40, by which the visarjaniya is to be made of like position with a 
following sm'd mute, and, by our rule, still a spirant of like position 
with the latter. The third example is iii. 39, which prescribes the con- 
version of a vowel into a semivowel ; and this semivowel, by our rule, 
must be that of the same class with the vowel : we are not to change i 
into », or « into r, etc. There arc, other rules to which the present 
precept applies ; so, in explaining the one next preceding, it may be 
looked upon as determining the non-aspirate into which the duplication 
of an aspirate is converted to be surd or sonant according as the aspirate 
is surd or sonant (a matter which, in the other treatises, is expressly 
prescribed in the rules themselves, and not left for inference) : and pos- 
sibly its bearing upon that rule is the reason why it is introduced here, 
rather than elsewhere in the treatise. Similar prescriptions are found 
in the Rik Pr. (i. 14, r. 56, lvii) and the Vaj. Pr. (i. 142). 

96. In khanvakhdzi and khdimakhdzi, the i following the d is 

This is a special rule, evidently intended to guard against an appre- 
hended mispronunciation. The two words in question (iv. 15. 15) are 
meant for imitations of the croaking of frogs, and are probably for khan- 
vakhdi, khdimakhdi, with protraction (pluti) of the final syllable: and 
it is feared that, without particular caution, the final i will be made to 

vol. vn. 50 

392 W. D. Whitney, [i. 96- 

share in the irregular accent which falls upon the protracted vowel, 
both words being doubly accented, on the first and third syllables.* 

SfoPtfT ^P£5|T ^rfiHN=hl(l S3rT> II \o II 

97. In avagd and d babhuvdn, with iti, the e is not protracted. 

The commentator, after paraphrasing the rule, adds simply avape 'ti, 
babhuve 'ti. The two passages referred to are found at xii. 4. 42 and 
x. 2. 28, and they read, in the pada and sanhitd texts, as follows : 

p- cr^Ti^TPTi^n srsterri^n^icrii-s gw3 j rg%frn 
p. q^p i ?rr i ^j^t ii 8 1 $ h—s. j^f $rr ^f^tt^ i 

They are not analogous cases, as regards the action of the present 
rule, since one of them actually contains an iti, while, in the other, no 
iti follows the protracted vowel in either the sanhitd or the pada texts, 
but only in the vfo-ama-text (see the note to rule 70). All the sanhitd 
manuscripts, however, observe the precept of our treatise in making the 
combination of avapdz with the succeeding word, and accordingly its 
protraction, which is assured by this rule and by i. 105, and which is 
exhibited by the pada-text, entirely disappears in sanhitd — a strange 
imperfection of the latter text, and one which, if it did not exhibit itself 
in all the manuscripts, we should be very loth to introduce, upon the 
sole authority of this rule of the PraticAkhya. 

By Vaj. Pr. iv. 88, a final pluta vowel retains its pluti before iti, and 
the closing Jcrama-pada of our second verse would be babhuvdn iti ba- 
bhuvdn, instead of babhuve 'ti babh&vdn. Neither of the other Vedic 
texts appears to present any case analogous with the other one which 
forms the subject of our rule. 

98. Consonants not separated by vowels form a conjunction. 

With this definition of a samyoga, a conjunction or group of conso- 
nants — which, as already noticed, is much more comprehensive than 
that which would seem to be implied in the definition of samyukta 
given above, in rule 49 — agree those of the other treatises (R. Pr. i. 7, 
r. 37, xxxviii ; V. Pr. i. 48 : T. Pr. offers nothing corresponding). 

The commentator's paraphrase of the rule, with the accompanying 
examples, forms a verse, as follows : vyanjandny avyavetani svardih 

* E. I. and H. read the first word khanvakhd'3i, with a single accent only, and 
the printed text has — wrongly, as it seems to me — followed their authority instead 
of that of the other manuscripts : and also, by some inexplicable oversight, signs 
of accent have become attached to the pluti figures, as if the preceding «'s were 
circumflex, and the following i's acute. The line ought to read as follows : 

i. 99.] Atharva- Veda Pr&tiq&khya. 398 

samyogo bhavati: agnir indrap ca tushtap ca vrkshah plaksho nidarpa- 
nam. The word tushta is not found in the Atharvan, nor plaksha, ex- 
cepting in the form plakskat (v. 5. 5). 

tHMt|\ ^rWlcWMirft dHM^IU^H II H II 

99. After a non-nasal and before a nasal mute, in the same 
word, is made the insertion of yamas, suited to each case. 

The commentator treats this intricate subject with the utmost possi- 
ble brevity, merely paraphrasing the rule, as follows : sam&napade 'nut- 
tatndt sparp&t: uttame paratah sparpe yam&ir vyavadhdnam bhavati: 
yathdsamkhyam : — and adding as instances sapainam (vii. 109. 3), pra- 
dhndti (not found in AV. ; the word most nearly resembling it is crath- 
ndnah [xiv. 1. 57], for which it may not impossibly be a false reading of 
the manuscript), yajnah (e. g. iv. 11. 4), and grbhn&ti (MS. grhndti, 
which is no example of a yama : the only form in the AV. admitting 
yama is grbhndmi [iii. 8. 6]). Unfortunately, we cannot be permitted 
to dismiss the subject in such an off-hand manner, but must endeavor 
to ascertain, by the aid of the other treatises and of phonetical theory, 
what these yamas are. 

We have already seen (under rule 43) that the euphonic system of 
the Praticakhya does not allow one mute to follow another by a simple 
consonantal conjunction, but regards the former of the two as suffering 
a modification which robs it of part of its distinct quality. Now we have 
the farther direction, which must be taken as to that extent limiting the 
former, that, within the limits of a simple word, if the latter consonant is 
nasal and the former not so, there is interposed between the two a yama, 
or 'twin' to one of the other letters. This is all that our treatise says 
of the yamas : none of its other rules mention them, although one or 
two may be regarded as referring to them, and are so interpreted by the 
commentator — from whose explication of rule 26, above, we have learned 
that they are of nasal character. The Taitt. Pr. is not more explicit : 
it merely says (xxi. 12, 13) : "after a mute not nasal, when followed by 
a nasal, are inserted, in each several case, nose-sounds (ndsikya) : these 
some call yawias." The V&j. Pr., where it teaches the occurrence of the 
yamas (iv. 160), calls them vicheda, ' separation,' a word which it does 
not elsewhere employ ; its doctrine is : " within a word, a non-nasal be- 
fore a nasal suffers separation " — that is, it is to be inferred, a separation 
or division of itself into two parts, which are as twins to one another — 
and the yamas have elsewhere been stated to be nose-sounds (i. 74), and 
formed by the root of the nose (i. 82). The Rik Pr. is decidedly more 
elaborate in its description. After stating (i. 10) that the yamas are 
nose-sounds, it goes on to say (vi. 8-10) that the non-nasal mutes, be- 
fore following nasals, become their own " twins " — that is to say, if we 
rightly understand it, each becomes a pair of twins of its own nature ; 
what is left of the original mute being one of the pair, and its nasal 
counterpart the other ; the latter being especially the yama, or the twin 
which is added to make up the pair. The yama is then declared to be 
similar to its original (pralcrti) ; or, it is said, there is an audible utter.- 

894 W. D. Whitney, [i. 99. 

ance in the mouth, of the same quantity with the yama ; but the office 
of the suffixed sound is not diverse from that of its original. All this 
seems intended to be very explicit, but it is so far from being perspicu- 
ous that it has led both the editors of the Rik Pr., or allowed them to 
fall, into the very serious error of supposing the yama to be something 
prefixed to the non-nasal mute, instead of interposed between it and the 
following nasal. Phonetic analysis does not, as it seems to me, help us 
to recognize the yama of the Hindu grammarians as any necessary ac- 
companiment of the utterance of a mute and nasal, but will lead us to 
a plausible explanation of what they must have called by the name.* A 
nasal is a sound in the production of which there is an expulsion of in- 
tonated breath through the passages of the nose, at the same time that 
the mouth organs are closed in the position in which an ordinary mute 
is uttered ; in any language, then, there will naturally be as many nasals 
as there are classes of mutes, and the unusually complete alphabet of 
the Sanskrit language recognizes and distinguishes them all. If, now, 
we pronounce a t before a following m, as in atma, the /, in the first 
place, suffers abhinidh&na, losing the explosion which is essential to its 
full utterance: the organs pass, without intervening nnclosure, from 
the dental contact to the labial contact, by which latter the m is pro- 
duced, with expulsion of sound through the nose. By taking sufficient 
pains, we can make the nasal utterance so closely simultaneous with the 
labial explosion that nothing shall be audible except the t and the m. 
But we may also commence the nasal sound a perceptible interval be- 
fore the explosion, and we shall even be most likely to do so in a labored 
utterance : if it be made to begin after the labial position is taken up, 
the nasal resonance is merely a preface to the m, and a dwelling upon it 
before the explosion : but if we utter sound through the nose before 
transferring the organs from the dental to the labial contact, we give 
origin to a kind of nasal counterpart to the t, as a transition sound from 
it to the m. If this is not the yama of the Hindu grammarians, I am 
utterly at a loss to conjecture what the latter should be. The theory 
which recognizes it might be compared with that which, in rule 50, 
above, taught a general assimilation of the former consonant of a group, 
in its final portion, to the latter; it is still more nearly analogous with 
the surd which, by ii. 9, is inserted between a final nasal and a following 
sibilant : this arises, like the yama, by an exchange of the emission (the 
anupradana) belonging to the former letter for that belonging to the 
latter before the transfer of the organs from the one position to the 
other ; and the t thus introduced, for example, between a n and a « 
has just as good a right to be called the yama or counterpart of the 
former letter, as has the n inserted after t before m. That the utterance 
of the intermediate sound thus described is not necessary, and can 
readily be avoided, is no objection to our interpretation of the Hindu 
theory : in the studied explicitness of the scholastic utterance, and with 
a phonetical science which delighted itself with subtleties, and of which 
the strong tendency was to grow from descriptive into prescriptive, such 

* That Muller pronounces the theory (p. cxxii) " perfectly clear and physiologi- 
cally comprehensible" must go for nothing, considering his entire misapprehension 
of the situation and character of the yama. 

i. 99.] Atharva- Veda Pr&tic&khya. 395 

transition sounds would naturally enough rise to a distinctness and a 
generality of occurrence much beyond what they were originally en- 
titled to. A much more serious difficulty is, that the theory of the 
yama allows its occurrence between an aspirate mute and a nasal : 
and we should suppose that the unclosure and brief emission of un- 
intonated breath constituting the aspiration would form an impassible 
barrier between the two letters, the nasal utterance being unable to 
precede it, and the position of contact of the former letter to follow 
it, so that no nasal counterpart to the former letter could be uttered. 
I see no way of getting over this difficulty, excepting by supposing 
an inaccuracy in the analysis of the Hindu phonetists : a serious charge, 
it may seem, but one of which I should be glad to see them relieved 
by any other intelligible explanation of the yama. If the whole theory 
of the phenomenon were more solidly founded and more accurately 
worked out by them, I should not think they need have explained it in 
a manner to cause their interpreters so much perplexity. The perplexity, 
indeed, is not confined to the modern expositors : the ancient comment- 
ators themselves (see Muller, p. cxxiii) seem to have been somewhat in 
doubt as to how many different yamas there are, whether twenty, one 
for each of the non-nasal mutes, or a smaller number. The orthodox 
doctrine of the Rik Pr. seems to be that of twenty : but its commenta- 
tor says that there are only four ; one for all the first mutes, one for all 
the seconds, and so on; and the commentary to Taitt. Pr. xxi. 12 sup- 
ports the same view.* This latter view, however, appears to me pecu- 
liarly indefensible : I cannot at all see how the nasal counterparts of 
the of the five mute series should be identical with one another ; 
nor, on the other hand, how they should be physically different from 
the yamas of the following mutes of each series respectively ; although 
it might well enough be loosely said, considering their title of " twins," 
that there are as many of them as of the sounds to which they sustain 
that relation. Physically, it would seem necessary that a nasal transi- 
tion-sound between two mutes should be of the nature either of the 
first or of the second : if of the second, and that second a nasal, it 
would be indistinguishable from it ; if of the first, it would be identical 
with the nasal of that series (except as being abhinikita, or wanting the 
explosion), and so the same for all the mutes of the series. The doc- 
trine of our own treatise upon this point is not entirely clear, since its ex- 
pression, yath&samkhyam, ' according to their number,' might possibly 
be taken as referring either to the non-nasal or to the nasal mutes : yet 
it is, without much doubt, to be understood of the former ; and we are 
to allow theoretically the existence of twenty yamas, although only 
thirteen of them — viz. those of k, kh, g, gh, c, j, t, t, th, d, dh, p, bhj — 
occur in the Atharvan text. 

* Weber (p. 125) suggests that the discordance amoDg the authorities upon this 
point may have grown out of the circumstance that, in speaking of the yamas, 
those of a single series of mutes are sometimes taken as representatives of the 
whole class, and treated as standing for them all. This seems very plausible ; but 
we can hardly acquit the later expositors of having been misled by this usage into 
the belief that there are only four yama; and not twenty. 

f For the details, see the additional note on the consonantal combinations in 

396 W. D. Whitney, [i. 99- 

In the examples which he gives under this rule, the commentator 
does not attempt to write the yamas. Above, under rules 13 and 
26, where the yatnas were instanced, they were — taking those of the 
guttural mutes as representatives of the class — written by the mutes 
with an anusvdra sign above : viz. ^ iipf i jf I ET ! the anusv&ra being evi- 
dently intended here exceptionally to indicate the nasal quality of the 
consonant itself, and not of the following vowel. The method of the 
commentary to the Rik Pr. (see Miiller, p. xix) is the same, or, in other 
manuscripts, ^ | ^ , jj | ri, and this last mode Weber (under i. 80) con- 
jectures, with much plausibility, to have arisen from writing the guttu- 
ral nasal 3 under the other letters, since this would be the most accu- 
rate method which the alphabet renders possible of writing the non-nasal 
and its nasal yama. 

«fc*l{WlRM*lH u ^oo it 

100. After h is inserted in like manner a ndsikya before a 
nasal mute. 

The commentator paraphrases with hak&rat n&sikyena samanapade 
vyavadh&nam bhavati ; and adds as illustrations a part of the words 
already once given, under rule 58 : viz. pr&hnah, purv&hnah, apar&hnah, 
apa hmalayati, vi hmalayati, vi hnute, brahma. 

The Taitt. Pr. (xxi. 14) teaches the insertion of a n&sikya after h and 
before a following nasal in terms nearly equivalent to those of our 
own rule. The Rik Pr. (i. 10, r. 48, xlix) and the Vaj. Pr. (i. 74, 80) 
describe its mode of pronunciation, as a nose-sound ; and the latter, in 
its latest portion (viii. 28), speaks of it again among the constituents of 
the spoken alphabet; but, strangely enough, neither of them gives any 
rule respecting its occurrence. 

What the sound may be which is thus taught to form the step of 
transition from the aspiration to a following nasal, it is hard to say with 
confidence. I can only conjecture it to n>e a brief expulsion of surd 
breath through the nose, as continuation of the h, before the expulsion 
of the sonant breath which constitutes the nasal. The pure aspiration 
A is a corresponding surd to all the sonant vowels, semivowels, and 
nasals of the alphabet : that is to say, it is produced by an expulsion 
of breath through the mouth organs in any of the positions in which 
those letters are uttered ; it has no distinctive position of its own, but 
is determined in its mode of pronunciation by the letter with which it 
is most nearly connected. Thus the A's of ha, of hi, of hu, and those 
heard before the semivowels w and y in the English words when and 
hue, for instance, are all different in position, corresponding in each 
case with the following vowel or semivowel. IT is usually initial in a 
word or syllable, and is governed by the letter which succeeds, and not 
by that which precedes it : but where it occurs before another conso- 
nant in the middle of a word — which is always its position in the Vedas 
before a nasal — the question may arise whether it shall adopt the mode 
of utterance of the letter before or after it : whether in brahma, for ex- 
ample, we divide brak-ma; and pronounce the h in the position of the 

i. 102.] Atharva-Veda Pr&titfkhya. 397 

o, or bra • hma, and in the position of the m, through the nose. Ac- 
oording to the Hindu method of syllabication (see rule 56, above), the 
former is the proper division, and the Hindu phonetists doubtless re- 
garded the h as belonging with and uttered like the a ; and noticing at 
the same time the utterance, scarcely to be avoided, of at least a part 
of the h in the position of the m, they took account of it as a separate 
element, and called it n&sikya. 

101. After a r, and before a spirant which is followed by 
a vowel, is inserted a svarabhakli, half a short a: some say, a 

102. Before any other consonant, the svarabhakli after r is a 
quarter or an eighth of a. 

The two rules are stated and explained separately in the manuscript, 
but I have put them thus together for the convenience of treating the 
whole subject of the svarabhakli at once. 

The term svarabhakti signifies a 'fraction or fragment of a vowel,' 
and the theory evidently is, that a r cannot be pronounced in immediate 
combination with any following consonant : there must always be 
slipped in between them a little bit of a transition-vowel, varying in 
length, according to different authorities, from a half to an eighth of a 
mora, and longer before a sibilant or h, if these be followed in turn by 
a vowel, than before other consonants ; while in quality it coincides 
with the a — that is to say, undoubtedly, with the a samvrta (rule 36, 
above), or the neutral vowel. The theory is this time, at least, perfectly 
intelligible, and any one may readily convince himself by trial how very 
easy it is to introduce such a vowel-fragment after a r, if he pronounce 
the latter far enough forward in the mouth for it to require to be trilled 
— and perhaps especially, if he be one to whom the smoother utterance 
of the r, farther back, is more natural. The reason for distinguishing 
the case of a following spirant — and that, too, only when followed by a 
vowel — as requiring a longer insertion, is not so clear, and I confess 
myself unable to discover the pertinence of the distinction : it is, how- 
ever, a marked and important one to the apprehension of the Hindu 
phonetists, as will appear by a comparison of the teachings of the other 

The V*aj. Pr. (iv. 16) restricts the occurrence of anything like svara- 
bhakti to cases in which a spirant is the second member of a group, 
and is itself followed by a vowel ; but it allows it both after a r and 
a I (see above, under rule 46), and moreover defines it as being the r 
and the 7-vowels respectively. Considering, however, that the same 
authority defines these vowels as ending each with quarter of an a (see 
ahove, under rule 37), its description of the character of the insertion 
cannot be regarded as differing essentially from that of our own treatise. 
The doctrine of the Taitt. Pr. is very nearly the same : it teaches (xxi. 

398 W. D. Whitney, [i. 101- 

15, 16) that when r and a spirant form a group, a r vowel-fragment 
(rephasvarabkaklih) is inserted, except when the spirant is subject to 
duplication (i. e., is not followed by a vowel) or is followed by a "first" 
mute : while the commentary explains that a fragment of that vowel 
which is akin with the r, or the r-vowel, is meant. According to the 
Rik Pr. (vi. 13, 14), the svarabhakti t which is described as being like 
the r-vowel {rkdravarnd), is inserted between a r and a following con- 
sonant when the former is preceded by a vowel : if the following con- 
sonant is a spirant, and itself followed by a vowel, the svarabhakti is 
the longer one, which had before been defined (i. 7, r. 33, xxxiv) as 
being a half-mora in length ; in other cases, the shorter one, of half this 
length (i. 7, r. 35, xxxvi), is interposed. The accordance of this with 
the doctrine of our treatise is as close as possible. But the Eik Pr. also 
allows a svarabhakti between a sonant letter and a following mute or 
spirant ; and it then farther cites the views of different authorities, of 
whom some deny the existence of the svarabhakti altogether, others 
permit it only after a r, and others only before a spirant not duplicated 
(this is very nearly the doctrine of the Vaj. Pr. and Taitt. Pr.), pro- 
nouncing it to agree in character with either the preceding or the fol- 
lowing vowel. 

As we shall see hereafter (under iii. 46), the manuscripts of the 
Atharvan acknowledge the virtual correspondence of the r followed by 
the longer svarabhakti with the r-vowel, by writing the r instead of r, 
where the former comes before a spirant, and should be, by iii. 46, con- 
verted into r after a or d. 

Our commentator gives us, under rule 102, the instances aryamd 
(e. g. i. 11.1), parva (i. 12.2), and dharmand (e. g. vi. 132. 1) — the man- 
uscript not attempting to write the interposed vowel-fragment. Under 
rule 101 he cites no examples, but, after the baldest possible paraphrase 
of the rule, proceeds to quote from other authorities, as follows : apara 
Aha : rk&rasvarabhaktih : tiskmam svarapareshv ardhdkdravarno vyafl- 
janam pesha iti: 'another has said, "a vowel-fragment of the r-vowel;" 
" before spirants followed by vowels is heard half an a-vowel ; the rest is 
consonant." ' These appear to be the dicta of two different teachers. 
Next follow several verses, a part of which are of a character which 
would render their introduction under rule 37, above, more appropriate, 
while one line, the second, belongs rather under rule 98 ; they read : 
rephdd anyad rkdre yat tasyd Wdham purvasasvaram : vacanena vyave* 
tdndrh samyogatvam vihanyale: rvarne 'pi tu rephasya cd 'rdhamdtrd 
pratijnayd: ardliamdtrdm svararh vidydt sa cdi Vam kriyale punah: 
tdn hrasvobhayatah lcurydd yathd mdtra bhaved iti: darpo varsham 
tatha rtavah: 1 barhip cdHramdarpanam: etdm rtim vijdniydt svara- 
bhaktir yadd bhavet ; ' half of what there is in the r-vowel different 
from r is of the same character with the preceding vowel. Of conso- 
nants separated by audible sound, the conjunction is destroyed. In the 
r-vowels there is, by express rule, half a mora of r ; half a mora is to 
be recognized as vowel, and that, again, is thus managed : put the parts 
upon both sides of the short vowel, so as to make out a mora : exam- 

1 ? —MS. tatMtaiah. 

i- 104.] Atharva-Veda Pr&tig&khya. 399 

pies are darfa, varsha, tatha rtavah, barhih: know this to be the way 
when a svarabhakti is to be produced.' I trust that either the commen- 
tator or the manuscript, and not the translator, is responsible for the 
inconcinnity of this passage. 

103. Of the latter value is sphotana. 

That is to say, if I do not misapprehend the meaning of the rule, 
sphotana, like the shorter svarabhakti, has a quarter or an eighth the 
quantity of a short a : or it may be that the emphatic eva would restrict 
the reference to the latter value, the eighth, alone. The commentator, 
as so often, gives not a particle of assistance in comprehending the rule. 
He simply paraphrases, as follows : tad eva sphotana vyanjako bhavati — 
explaining sphotana by its synonym vyanjaka, ' manifester' — and then 
cites the same instances of sphotana which are given later, under ii. 38. 
For the doctrine of sphotana, see the rule last mentioned, and the note 
upon it. The subject is not disposed of here, because the sphotana, 
unlike the other insertions treated of in this part of the work, arises 
only in the combinations of the phrase, when a final mute comes in 
contact with a following initial mute of an earlier series or varga. 

gjfert hmUiiRhih^i iao3n 

104. These belong to the preceding vowel, and do not effect 
the dissolution of a conjunction of consonants. 

There is something wrong with the commentary to this rule ; appa- 
rently we have a repetition of a part of the commentary to rule 102, 
with the loss of what should properly be given here : it reads as fol- 
lows : p&rvap&rvasvaram ca tad bhavati : samyogasya ca vighatah yat 
tat rephdt ak&rasya caturtham v& bhavaty ashtamam va : aryama parva 
dharmand. It furnishes us, it will be seen, no hint as to how far back 
the teachings of the rule apply. I presume, however, that they may 
be properly considered as extending themselves to all the phonetic in- 
sertions taught in rules 99-103 : all these, in the division of the word 
into syllables, are to be reckoned as belonging to the preceding vowel, 
and sharing in its accent ; and whereas it might seem that the insertion 
of the vowel-fragment, and of its kindred sphotana, dissolved the con- 
junction of the consonants between which they were inserted — since 
by rule 98, a conjunction of consonants can only subsist where there is 
no interposition of vowels — the contrary is expressly declared to be 
true. This would regard p&rvasvaram as belonging to some such word 
as angam understood, and used in an indistinctive or collective manner 
of all that precedes. It may be, however, that the specification applies 
only to svarabhakti and sphotana, and that the neuter singular form of 
p&rvasvaram is owing to its agreement with one of the words denoting 
the quantity of those insertions, caturtham, ashtamam, etc. The Eik 
Pr. specifies only of the svarabhakti (i. 1, r. 32, xxxiii) that it belongs 

VOL. VII. 51 

400 W. D. Whitney, p. 104- 

to the previous syllable; the Vaj. Pr. (i. 103) says the same thing of 
the yama alone ; while the Taitt. Pr. (xxi. 6) teaches that the svara- 
bhakti belongs to the preceding syllable, but (xxi. 8) that the yamas and 
ndsikya go with the following one. The Kik Pr. alone, besides our 
treatise, thinks it necessary to say (vi. 10, r. 35, ccccxi) that the svara- 
bhakti does not dissolve the conjunction : in the Vaj. Pr. it is left to be 
pointed out by the commentator (see Weber, p. 217). 

^Uojyi'^ Sw$2[ *m FTjf^ pv- is- is] i ^ 
^TT^^T^^rfrr px.6.i8] i OTT ej ^WT^f^pfe ^ 

q^T^; flcfr f^r: j^r m sr^crft t*. 2. 28 ] ( q^w- 
3rifawMi<Hr$ ^itr &■ 3. 27] i ^w^^irr [x».4. 

42] i Url<IHl^f^ 3 rTT^f^frT OH. s. 50] i ^H ^rllPl 

105 : these are the cases of protracted vowels. 

I have taken the liberty of separating by a stroke the different pas- 
sages rehearsed in this rule ; the manuscript puts them all in sandhi 
together. One or two of the signs of protraction have also been re- 
stored which the manuscript has accidentally omitted. On the other 
hand, I have retained the sign of protraction given by the manuscript 
to the second case in the last passage but one, avafez'ti (the MS. writes 
avafetis), although it is not written by the sanhitd codices of the Athar- 
van text, and is forbidden by rule 97, above. Finally, I have added the 
accent marks which belong to each passage. 

The commentator does not give any paraphrase of the rule, nor does 
he repeat it at the end of his exposition, yet I cannot question that it 
is actually the closing rule of the chapter, and not a gratuitous appendix 
of the commentator's own addition. He discourses respecting it more 
liberally than usual, in this wise : kimarthah paripdthah : ita uttaram 
adhikam: et&vat sv&rtho 'pi: bahuvidh&s trividhdh plutayo bhavanti: 
svarapard abhinishtdnapard vyanjanapardh : tdsdm ydh samdndkshara- 
pards td. itdv aplutavad bhavanti itdv aplutavad bhavanti ; 'for what 
reason is this enumeration made ? because any other instance than these 
is in excess : within these limits the protracted vowel is pointed out by 
its own meaning (?). Protractions are various; namely, of three kinds : 
those which affect a syllable ending in a vowel, in visarjaniya* and in 

* For the use of the term abhinishtdna for visarjaniya — of which this is, I be- 
lieve, the only case which our commentary affords — see rule 42, above, and the 
note upou it. 

i. 105.] Atharva-Veda Prdtigdkhya. 401 

a consonant, respectively ;* among these, those which affect syllables 
ending in simple vowels assume their unprotracted form before iti.'f 
No other reason, it would seem, is to be sought for the rule than that 
here given : it is intended to insure the absence of protraction in any 
other instances in the text than those here given ; in all of which, the 
protraction is due to the requirements of the sense, and is not merely 
euphonic or accentual. A somewhat similar enumeration is made by 
the Vaj. Pr. in ii. 50-53, and, at the same time, directions are given as 
to the somewhat anomalous accentuation of the several cases. In Rik 
Pr. i. 6 (r. 31, xxxii), also, are mentioned the only three instances of 
protraction to be found in the Rig-Veda, all occurring in the latter part 
of its tenth book. 

Our text and commentary say nothing respecting the accentuation of 
these words, except as regards the final i in the two instances con- 
tained in the first passage, for which see rule 70, above. From this we 
may perhaps conclude that the other protracted words offer no anoma- 
lies of accent. There is, however, some discordance among the manu- 
scripts as to their treatment, which it may be well enough to notice 
here. Of the first passage (iv. 15. 15) we have already spoken, in the 
note to rule 96. In ix. 6. 18, all the manuscripts excepting I. read 
bhuyAzh, without accent, and our printed text has followed their au- 
thority : but I cannot consider this reading as anything but an error, 
possibly arising from a blundering confusion of the word with the 
verbal form bhuyAs, from the root bhu: we ought to read, with I., 
bhu'y&zh. In x. 2. 28, Bp. and E. accent the protracted syllable, babhii- 
va'Sn: and this accent is somewhat supported by the analogy of the 
first asi'st in Rig-V. x. 129. 5 : but the case is still more nearly analo- 
gous with Vaj.-S. xxiii. 49, A' vivepAzn, and Rig-V. x. 146. 1, vindati3n, 
and hence the reading of the published text is much the more likely to 
be correct. In xi. 3. 26, all the manuscripts except P. and M. accent 
pratyanca'zm, which is accordingly the best supported reading. In xii. 
5. 50 is only to be noted that the pada manuscript in the second in- 
stance omits the sign of pluti, but doubtless by a clerical error merely. 
The pada-text everywhere writes the vowel in its protracted form, and 
adds the sign of protraction, not immediately after the vowel, but after 
the final consonant of the syllable, and sometimes with a stroke, or even 
a double stroke, interposed. 

Except in the first passage, which contains an imitation of animal 
sounds, we have in all these protractions only cases of doubtful ques- 
tioning as between two alternatives, of hesitating indecision, of mimdnsd, 
as it is called once in the text (xii. 4. 42). 

The signature of the chapter is caturadhy ay ik&y Am ca prathamo 
'dhyAyah samAptah : 13. The figures expressing the number of rules 
contained in it are obviously corrupt, but how they are to be amended, 
unless by simply altering them to 105, I do not know. That they mean 
113, and that any part of the last section is lost, is not at all probable : 
I discover nowhere in the section any signs of a lacuna. 

* These terms I translate rather according to the evident requirement of the 
sense than as they would seem naturally to mean, 
f This is virtually a restatement of rule 97, above. 

402 W. D, Whitney, ("• 1- 


Contents: — Section I. 1, introductory; 2, final mutes before sonants; 8, do. as 
finals ; 4, do. before surds ; 5, do. before nasals ; 6, do. before sibilants ; 7, do. be- 
fore A; 8, t before a; 9, nasals before sibilants; 10, n before p; 11, do. before 
sonant palatals; 12, do. before Unguals; IS, t before f and I; 14, do. before pal- 
atals and Unguals ; 15, dentals after palatals and Unguals; 16, do. after ah; 17, 
p after deDtals; 18, loss of an initial a; 19, do. of r before r; 20, do. of a mute 
after a nasal and before another mute; 21, do. of final y and v after a vowel; 
22-28, exceptions; 24, Cakat^yana's view of this combination; 25, insertion of a 
sibilant after pum ; 26, do. after » before a surd palatal, lingual, and dental ; 27, 
final d» before a vowel ; 28, do. before v, in a special case; 29, insertion of r after 
final in, An, rn ; 30, exceptions ; 31, m before mutes ; 32-33, do. before semi- 
vowels and spirants ; 34, » in like position ; 35, m and n before I ; 36-37, m re- 
tained before semivowels ; 38, aphotana; 89, karshana. 

Section II. 40, visarjaniya before a surd ; 41-42, do. before a vowel ; 48, do. 
before a sonant; 44-50, do. converted into r after a and d; 51-52, exceptions; 
53-54, aa converted to o ; 55-59, loss of final visarjaniya. 

Section III. 60-61, special cases of irregular sandhi of final visarjaniya ; 62, 
conversion of visarjaniya into a sibilant before initial £ and p of the second mem- 
ber of a compound word; 63-80, do. of an independent word. 

Section IV. 81-101, conversion of final or initial » into ah; 102-107, ex- 

fll^HIMI^ ll * ll 

1. The following rules are to be understood as of force in the 
combined text. 

The first chapter of the treatise has disposed of all matters of general 
phonetic theory, and laid down such rules as apply to words in their 
disjoined and independent form, and we now enter upon the considera- 
tion of those changes which may and must occur when the padas of the 
disjoined text are put together into the form of sanhitd. This rule is a 
general heading (adhikdra) belonging to the second and third chapters. 
The other treatises have equivalent or corresponding headings; the 
Bik Pr. at the head of its second chapter, the Vaj. Pr. of its third, the 
Taitt. Pr. of its fifth. "We shall see, however, that our treatise does not 
everywhere strictly limit itself to what concerns the conversion of pada- 
text into sanhitd. 

q«MMMJTHMI rffrn 

2. Finals not nasals become, before sonant consonants and 
tow els, un aspirated sonants. 

Considering that, by i. 6, only the first and last of each series of 
mutes can occur as finals, this rule might have said prathamdndm, ' first 
mutes,' instead of anuttamdndm, ' mutes not nasal ;' both this and the 

»• 5.] Atharva- Veda Prdtigdkhya. 403 

following rules, however, seem constructed in view of the disputed 
character of the final non-nasal mute, and of the doctrine of C&unaka 
himself that it is a media, and not a tenuis (see i. 8). The correspond- 
ing rule of the Vaj. Pr. (iv. 117) is expressed in a precisely equivalent 
manner: those of the Rik Pr. (ii. 4, r. 10, cxiv, and iv. 1, r. 2, ccxxi) 
and Taitt. Pr. (viii. 1, 3) use the term prathama, even although, as already 
noticed (under i. 6), the former work in theory recognizes the medim as 
possible finals. 

The commentator's examples are as follows : yad yatra vipvam (ii. 1. 
1) ; yad yamam cakruh (vi. 116. 1) ; tasm&d var nama (iii. 13. 3) ; vevi- 
shad vishah (v. 17. 5) ; yad rajanah (iii. 29. 1) ; suhasto godhug uta (vii. 
73. 1) ; s& vir&d rshayah (viii. 9. 8) ; and two which are not to be found 
in the Atharvan, and of which the latter, at least, is evidently fabri- 
cated : viz., tad abhutam and trishtub atra. 

q^T% srpftaT: u $ 11 

3. And at the end of a word they are surds. 

This, in view of i. 6, is a superfluous precept, and its introduction is 
only to be accounted for by the considerations adverted to under the 
last rule. 

The commentator cites once more his standard assortment of final 
mutes, viz. godhuk etc. (see under i. 3). 

STsffqsr ^ a 3 ii 

4. As also before surd consonants. 

Also an unnecessary specification ; since final surds do not require to 
become surds before succeeding initial surds, but simply remain un- 
changed. Only the Vaj. Pr. (iv. 118), among the other treatises, gives 
an equivalent precept. 

The commentator instances in illustration v&k ce 'ndriyam ca (xii. 5. 
7), vir&t praj&patih (ix. 10. 24), and trishtup pancadafena (viii. 9. 20). 

3fFrr 3f^r n h ii 

5. Before nasals they become nasals. 

The Praticakhyas are unanimous in this requirement : compare Rik 
Pr. iv. 1 (r. 3, ccxxii), Vaj. Pr. iv. 120, Taitt. Pr. viii. 2. Panini, as has 
already been noticed (under i. 2) allows either the unaspirated sonant 
or the nasal before a nasal, while manuscript usage is almost, if not 
quite, invariably in favor of the nasal. 

The commentator cites in illustration the following passages from the 
Atharvan text: rdhanmantro (p. rdhak-mantrah) yonim (v. 1.1); ya 
udanan ny&yanam (vi. 11.2); arnav&n mahatas pari (i. 10. 4) ; madu- 
gh&n madhumattarah (i. 34. 4) ; madhyan nicaih (iv. 1.3); and ya std- 
yan manyate (iv. 16.1); and finally, as the text affords him no instance 
of a final p before a nasal, he fabricates a case, out of words more than 
once employed by him elsewhere in a similar way, viz. trishtum nayati. 

404 W. D. Whitney, [ii. 6- 

6. Before c, sh, and *, they become aspirated surds. 

On this point there is by no means an agreement of opinion among 
the different Praticakhyas. The doctrine of the Taitt. Pr. (xiv. 1 2) 
accords most nearly with that of our treatise, only omitting its restric- 
tion to the case of a final before an initial ; and the same view is by 
onr commentator mentioned as held by ^ankhamitri, f&katayana, and 
Vatsya : his words are : apadant&ndm api pashaseshu dvitlya bhavanti : 
iti pankhamitri-pakatayana-vatsyah: 1 tasy& agnir vathsah ; 2 ' Qankha- 
mitri, Qakatayana, and Vatsya say that mutes even when not final be- 
come "seconds" before p, sh, and s; as in the instance tasya agnir 
vathsah (iv. 39. 2).' The Taitt. Pr. (xiv. 13) adds that Vadabhikara* 
teaches the conversion of the mute into an aspirate only before a sibi- 
lant not of the same class :f and the doctrine of the Taitt Pr. in this 
form, as modified by Vadabhikara, is by the Vaj. Pr. (iv. 119) ascribed 
to Qaunaka, the putative author of our treatise and of the Rik Pr. 
The Rik Pr., ignoring all these views, and itself holding, like the Vaj. 
Pr., that the mute remains unchanged before the sibilant, remarks only 
(vi. 1 5, r. 54, ccccxxx) that some regard a tenuis before a sibilant as to 
be aspirated, unless it be a final. Finally, a v&rttika to Pan. viii. 4. 48, 
as noticed by Weber (p. 249), ascribes to Paushkarasadi the doctrine 
which our commentator attributes to the three other grammarians men- 
tioned, and which is also taught by the Taitt. Pr. — viz., that a mute in 
any situation becomes aspirated before a sibilant. This comparison of 
conflicting views is exceedingly curious, and it cannot but inspire us 
with some distrust of the accuracy, as well as completeness, with which 
the Hindu grammarians report one another's views. 

The commentator, instead of citing from the text any genuine cases, 
proceeds to repeat a part of the cases which he has already once manu- 
factured (under i. 49), in illustration of a samyukla combination of con- 
sonants, by putting his four words, godhuk etc. (see under i. 3), one after 
another, before pete, shande, and sdye ; and the manuscript uniformly 
fails to write the aspirate, except in the case of drshat. The cases 
which actually occur in the Atharvan text are ks (e. g. iii. 1. 4), tp (ix. 
5. 21), ts (e. g. viii. 9. 9 ; but, by rule ii. 8, it is to be read tts), ts (pas- 
sim), and ps (in avagraha; e. g. ap-su, i. 6. 2) ; ksh and pp are found 
only in the interior of words. The manuscripts of the Atharvan read 
always the simple surd before the sibilant, and in the printed text we 
have of course followed their authority rather than that of the Pratica- 
khya. Weber (p. 250) notices that a single Berlin MS. of the Vaja- 
saneyi-Sanhita writes the surd aspirate before a s not followed by a 

* fdnkhamitifdkatdyanasyavdtsydh. 2 vatsah. 

* My manuscripts vary, as to the reading of this name, between vadabhikara, 
bddabhikdra, and bddavikdra : Weber (p. 18) calls it once vdtabhikdra. 

f Weber says (pp. 246, 250) " only before a sibilant of the same class ;" appa- 
rently misled by an error of his manuscript. 

ii. 9.] Atharva-Veda Pratiqakhya. 405 

FRtf: (JeHrjSfl «£*l^d ii o il 

7. After final non-nasal mutes, h becomes the aspirated sonant 
of the preceding letter. 

The Rik Pr. (iv. 2, r. 5, ccxxiv) and Vaj. Pr. (iv. 121) agree precisely 
with our treatise upon this point ; and the same doctrine is attributed 
by the Taitt. Pr. (v. 38) to Plakshi, Kaundinya, Gautama, and Paush- 
karasadi. The Taitt. Pr. (v. 39-41) goes on to state that in the view 
of some the h remains unchanged ; while the Mimansakas, and Qaitya- 
yana etc. (the " etc." means, according to the commentator, Kauhalipu- 
tra, Bharadvaja, sthavira-Kaundinya, and Paushkarasadi [sthavira-Paush- 
karasadi ?]) hold that an aspirated sonant* is inserted between the final 
surd and the h. Panini's rule (viii. 4. 62), as is well known, allows the 
h either to remain unchanged, or to become the sonant aspirate ; and 
there is but a very trifling phonetical difference between the two modes 
of treatment. 

The illustrative citations of the commentator are ud dharshantdm 
maghavan (iii. 19. 6), ud dharshaya satvandm (v. 20. 8), uddharshinam 
munikefam (viii. 6. 17), had dha nunam (xviii. 1. 4), prthivydm astu yad 
dharah (xviii. 2. 36), tejasvad dharah (xviii. 3. 71). 

^i(iKwi\ rr^nr n z ii 

8. After t is inserted t before s. 

The same phonetic precept is found in the Taitt. Pr. (v. 33), com- 
bined with a part of that contained in our next following rule : t, it is 
said, is to be inserted after t and n, when they are followed by s and sh. 
The Rik Pr. (iv. 6, r. 17, ccxxxvi) also gives it as the view of certain 
teachers that t and re, when followed by s, receive the appendix of a t. 

The commentary quotes from the text virat svarajam (viii. 9. 9), 
prtandshdt suvlrah (xi. 1. 2), and trigatah skat sahasrdh (xi. 5. 2), which 
are the only examples of this combination presented by the Atharvan. 
In the first of the three, P. reads tts, in its second copy of the book, 
and by the emendation of a second hand : the other manuscripts give 
here, as do all of them in the other two cases, simply ts ; and the 
printed text follows their authority. 

3m^r: ^zh\ mm u ^ h 

9. After h, n, and n are inserted k, t, and t before g, sh, and s. 

The form of this rule is a little ambiguous, since we might be left by 
it to query whether, for instance, after n, was to be inserted k before p, 
t before sh, and t before s, or only k before all the three sibilants — in 
other words, whether the transition-sound should adapt itself to the 
character of the following or of the preceding letter. The commentator 

* Weber (p. 261), by a lapsus calami, says " the nnaspirated sonant." 

406 W. D. Whitney, pi. 9- 

either does not notice, or does not deign to relieve, this difficulty ; lie 
offers no explanation of the rule, and, in the instances which he cites, 
the manuscript persistently omits to write the transition-sound. For 
phonetic reasons, however, it cannot be doubted that the latter is deter- 
mined by the preceding letter, and that after n is to be uttered a k, after 
n&t, and after n a t, before all the sibilants. By no means all the cases, 
however, which the rule theoretically contemplates, are found actually 
to occur in practice. The guttural nasal, n, precedes s six times in the 
Atharvan (iv. 11. 8. vi. 51. 1. xiii. 1. 56 ; 2. 3 ; 3. 16. xviii. 1. 29), but 
is never found before ( or sh : the manuscripts do not in a single in- 
stance write the transitional k, nor have we introduced it in the pub- 
lished text. The lingual nasal, n, never occurs as a final, except before 
v, in the cases treated of in rule iv. 99. The case of n before g is pro- 
vided for by rules 10 and 17, below ; n before sh is found three times 
in our text (viii. 9. 17. xiii. 1. 4 ; 3. 6), and nowhere do the manuscripts 
write a t between them (it is done by the edition, however, in the last 
two cases) ; n before * occurs times innumerable, and the usage of the 
manuscripts with respect to the sandhi is exceedingly irregular ; there 
is hardly an instance in which they all agree together either to reject 
the t or to insert it, nor is any one of them consistent with itself in 
its practice. In the edition, therefore, we have followed the authority 
of the Praticakhya, and the sandhi is always made nts (except in one 
instance, viii. 5. 16, where the t has been omitted by an oversight). 

The insertion of these tenues after the nasals is a purely physical phe- 
nomenon, and one which is very natural, and liable to occur in any 
one's pronunciation. There is to be made, in each case, a double tran- 
sition in utterance : from the sonant nasal to the surd oral emission, and 
from the close to the partially open position of the organs. If, then, 
the former is made an instant earlier than the latter, if the nasal reso- 
nance is stopped just before, instead of exactly at the same time with, 
the transfer of the organs to the position of the sibilant, a tenuis of the 
same position with the nasal becomes audible. It is, as already remarked 
under i. 99, the counterpart of the nasal yama, asserted by the Hindu 
phonetists to be heard between a mute and following nasal. It is also 
closely analogous with the conversion of np into rich, as will be pointed 
out below (under rule 17). 

The commentator, by way of examples of the combinations taught 
in the rule, puts pratyan and gan before fete, shande, and s&ye respect- 
ively (the MS., as already noted, always failing to write the transition- 
sound), and then quotes from the text two actual cases : viz. shad dhuh 
ftt&n shad u mdsah (viii. 9. 17), and tdnt satydujdh (iv. 36. 1). 

The Rik Pr. does not itself teach these euphonic insertions, but 
merely records it as the opinion of some authorities (iv. 6, r. 16, 17, 
ccxxxv, ccxxxvi) that k is inserted after n before a sibilant, and t after 
n before s. The Vaj. Pr. so far agrees with our treatise as to prescribe 
(iv. 14) the insertion of k after n and t after n, before *, adding (iv. 15) 
that Dalbhya is of the contrary opinion. The Taitt. Pr. (v. 32, 33) 
inserts k after n, and t after n, before both a and sh, and so precisely 
accords with our own rule, only omitting such cases as are unnecessarily 
and vainly provided for in the latter. 

ii. 12.] Alharva-Veda Prdtiq&khya. 407 

H*I|W 5FRT^ o^: H\ou 

10. Before g, n becomes n. 

This rule is incomplete, except as taken in connection with rule 17, 
below, along with which, accordingly, it will be bere treated. The 
commentator's illustrations are two of those which are given under rule 
17, viz. asmaft chatruyattm abhi (iii. 1. 3), and divi shaft chukrah (xviii. 
4. 59). 

11. As also before a sonant palatal. 

That is to say, before,;'; since jh, as already noticed, never occurs, and 
n is never found as initial. 

This is another rule as to the observance of which the usage of the 
Atharvan manuscripts is quite various ; and it may almost be said here, 
as of the insertion of t between n and s, that there is not a passage in 
which all the codices agree either to make or to neglect the assimilation. 
We find written in such cases either anusv&ra, or ft, or n ; yet the first 
is notably the most frequent, and in the printed text has been made, in 
obedience to the authority of the Praticakhya, the universal usage. It 
might perhaps have been better, in order to avoid ambiguity, to write 
the palatal nasal expressly, instead of intimating it by the employment 
of the nasal sign over the preceding vowel : yet the cases are few in 
which a final ft so written could be mistaken for one which arises from 
the assimilation of a final m. 

The other treatises (R. Pr. iv. 4, r. 9, ccxxviii ; V. Pr. iv. 92 ; T. Pr. 
v. 24) prescribe the conversion of n into ft before any following palatal ; 
and the Rik Pr. and Taitt. Pr. include the palatal sibilant in the same 
prescription, their rules thus corresponding to our 10th and 11th to- 
gether. In the Atharvan, n does not occur anywhere before an original 
ch, and n before c is treated in a later rule (ii. 26). The manuscripts of 
the Rig- Veda (see Muller, p. lxxxvii) show the same irregularity in their 
treatment of final n before a palatal which has been noted just now as 
characterizing those of the Atharva-Veda : but the editor does not ap- 
pear to have attempted to carry out any principle in the readings which 
he has adopted. 

The commentator cites avapapyafi jananam (i. 33. 2), i^nahdft janan% 
(v. 8. 7), praishyaft, janam iva (v. 22. 14), and vivah&ft jft&ttn, (xii. 5. 44). 

12. Before a lingual mute, n becomes 1),. 

As no lingual mute is found at the beginning of any word in the 
Atharvan, any more than in the other Vedas, this rule is as unnecessary 
as is the inclusion of n along with the other nasals in rule 9 of this 
chapter, and as is more than one rule or part of a rule in that which is 

vol. vii. 52 

408 W. I). Whitney, [ii. 12^ 

to follow : such specifications are made merely for the sake of a theo- 
retical completeness. None of the other kindred treatises has a cor- 
responding precept. 

The commentator fabricates, as illustrations of the rule, bhavdn diyate, 
tnahdn diyate. 1 

13. Before f and I, t becomes of like position -with those letters 

There is no discordance among the different treatises with regard to 
the combination of t with either f or I, although there are differences 
in the precise mode of statement of the rules. The corresponding pre- 
cepts are Rik Pr. iv. 4 (r. 10, 11, ccxxix, ccxxx) ; Vaj. Pr. iv. 12, 93 ; 
T&itt. Pr. v. 22, 25. The sandhi of t with g is not complete without 
the addition of rule 17, below, which see. 

The commentator cites one instance for each part of the rule, viz. : 
ucehishte (p. ut-fishte) n&ma (xi. 7. 1), and ghrt&d ulluptam (v. 28. 14). 

There follows a slight lacuna in the manuscript, the copyist heedlessly 
passing, as we may plausibly conclude, from the tak&rasya of the final 
repetition of this rule to that of the paraphrase of the next, thus over- 
leaping the latter altogether, so that it has to be restored from its final 
repetition before rule 15. We may restore as follows, indicating by 
brackets the portion omitted : gkrt&d ulluptam : takarasya [fakarala- 
Jcarayoh parasasth&nap catavargayof ca: catavargayop ca takara$ya\ 
parasaslh&no bhavati. We have had occasion once before (under i. 64) 
to note such an omission, and more than one additional instance will 
appear hereafter. Here, nothing of any consequence is lost. 

^^Fftir ii \& n 

14. As also, before palatal and lingual mutes. 

One part of this rule, again — viz. that relating to the lingual mutes — 
is altogether superfluous ; and it has no correspondent in any of the 
other treatises. The assimilation of t to a following palatal is taught 
by them all (see R. Pr. iv. 4, r. 10, 11, ccxxix, ccxxx; V. Pr. iv. 92 ; 
T. Pr. v. 22, 23). 

For the palatal combination, the commentator instances ue ca tisfyha 
(ii. 6. 2), and yaj j&mayah (xiv. 2. 61) ; and we may add brhacchandah 
(iii. 12. 3). For the lingual combination, he fabricates the examples 
agnicit tikate, somasud diyate : compare those given under the corres- 
ponding rule of Panini (viii. 4. 41). 

rTFnt HHHtlX HcUTftHI u4h^M: hVUi 

15. A dental mute following these in the same word is assimi- 
lated to them. 

1 bhavdr nit/ate, mafidr niyate. 

ii. 17.] Atharva- Veda Prdtigdkhya. 409 

This rule, in its extent as given, is an infringement of the limits laid 
down in i. 1 as those of a Pratic&khya, and also of those laid down in 
ii. 1 as those of the chapter : and a more notable one, as it concerns in 
part the very case which is cited in the commentary to i. 1 as an illus- 
tration of what it does not belong to a Pratic&khya to treat ; the in- 
stances here quoted in the commentary for the assimilation of a dental 
to a preceding lingual — they are rnildha amilr&h (vi. 67. 2), and tesh&m 
vo agnimildh&n&m (vi. 67. 2) — are precisely analogous with the one 
there given, and our rule teaches only one out of the series of changes 
which such a word must undergo, as drawn out in full by the commen- 
tator in his exposition. The only practical application of the precept 
is one which is not recognized, or at least not illustrated, by the com- 
mentator ; namely, to those cases in which an initial s followed by a t 
or th is, by later rules (ii. 90 etc.), converted into sh : the following 
dental then becomes by this rule a lingual. 

In illustrating the other part of the rule, that which prescribes the 
assimilation of the dental to a preceding palatal, the commentator first 
states, varttika-\ike, the restricted form in which alone it applies — cavar- 
ffty&n nakarasya ca, 'following a palatal mute, a n is assimilated' — and 
cites yajnena yajnam (vii. 5. 1), somaya rdjne (ii. 13. 2), and somasya 
rdjitah (vi. 68. 1). He might have added yAcnyaya krnute* (xii. 4. 30), 
the only instance in the Atharvan of a like assimilation after c. 

The other treatises, combining the practical part of this rule with the 
one next following, teach that t and th are everywhere converted into t 
and th after sh (see E. Pr. v. 3, r. 11, cccxxviii; V. Pr. iii. 78; T. Pr. 
vii. 13, 14). 

16. And even in a different word, after sh. 

That is to say, a dental following sh is assimilated to it, and becomes 
lingual, not only when both letters occur within the same word, but also 
when the sh is final, and the dental the initial of an independent word. 
The commentary cites cases of the assimilation in the same and in 
separate words — viz. shashtih (e. g. v. 1 5. 6) and shannavatih — but the 
former belongs under the preceding rule, and the other is such a case 
as never occurs in the Atharvan. The precept was evidently only in- 
tended for such combinations as bahish te (i. 3. 1), in which, by the rules 
contained in the fourth section of this chapter, an original final s be- 
comes lingualized, and the following t is assimilated to it. 

The corresponding rules of the other Praticakhyas have been already 
referred to. 

17. After a dental mute, q becomes ch. 

This rule, taken in connection with rules 10 and 13, above, deter- 

* The reading of the printed text, ydncydya, is an error of the press. 

410 W. D. Whitney, [ii.l?- 

mines the form to be assumed by the combinations t+p and n+p. Ex- 
ception may fairly be taken, however, to the method in which the 
change is taught. By the other rules referred to, t and n are to become 
e and ft before p: and if those rules are first applied, there will be no 
dental mutes for p to follow ; while, if the present rule be first applied, 
the others are rendered wholly or in part superfluous, by the non-occur- 
rence of p after t and n. In the case of t there comes in the still far- 
ther difficulty that rule 6 of this chapter has converted it into th, so 
that a part of rule 13 is thereby also rendered incapable of application. 
These are incongruencies such as the authors of the Pratigakhyas are 
very seldom guilty of. What is the intention of our treatise is, indeed, 
sufficiently clear : the combination of t and p is to produce cch, by the 
conversion of the former into c and the latter into ch; and the combi- 
nation of n and p, in like manner, is to produce nch. The Rik Pr. (iv. 
4, 5, r. 9, 11, 12, ccxxviii, ccxxx, ccxxxi) teaches the same changes, only 
adding (r. 13, ccxxxii), that Qakalya would read instead cp and Up. 
The Vaj. Pr. (iv. 93, 94) also agrees, only exempting the p from con- 
version into ch when it is followed by a mute. The Taitt. Pr. prescribes 
(v. 22, 24) the change of t and « into c and n before p, and (v. 34, 35) 
the conversion of p into ch when preceded by any mute excepting m, 
Valmlki (v. 36) also excepting^?, and Paushkarasadi (v. 37) denying the 
conversion when p is followed by a consonant, and denying in this case 
also the conversion of the preceding n into ft* 

The commentator cites examples only of the combination of n and p ; 
they are dev&ft chlokah (xviii. 1. 33), asm&n chatruyatim nbhi (iii. 1. 3), 
and divi shaft chukrah (xviii. 4. 59) : as an example illustrative of the 
other part of the rule, we may take ar&c charavy&h (i. 19. 1). In the 
orthography of this class of combinations, we have followed in the 
printed text the authority of the manuscripts, which, with hardly an 
exception, write simply ch, instead of cch. This orthography is also, to 
my apprehension, a truer representation of the actual phonetic result of 
combining t with p. That these sounds fuse together into a ch is very 
strong evidence that the utterance of the Sanskrit surd palatals did not 
differ materially from that of our ch (in church etc.); and I conceive 
that the constant duplication of the ch and jh (wherever the latter oc- 
curs) between two vowels is to be looked upon simply as an indication 
of the heaviness of those consonants, and of their effect to make the 
preceding vowel long by position. The c and j, though strictly com- 
pound sounds, are too easy combinations to occasion position : in this 
respect they resemble the aspirate mutes, which are likewise really 
double in their nature : but they are too heavy to bear the farther addi- 
tion of even so light an element as the aspiration without acquiring 
the quantity and phonetic value of double letters. 

The conversion of np into nch, on the supposition of the compound 
nature of the palatal, as made up of a mute and a sibilant element, would 
be almost precisely analogous with that of ns into nis, as taught in rule 
9, above, and would be readily and simply explainable as a phonetic 

* Paushkarasadi would read neither pdpiydn chreyase nor even pdpiyM freyase, 
but pdpiydn p-eyase: this is misunderstood by Weber (p. 238). 

ii. 20.] Atharva- Veda Pr&lic&khya. 411 

Fffar 3^: ^IWKt: U<=hl {HI li \z u 

18. After the preposition ud, the 5 of the roots st/id and stanibh 
is dropped. 

The commentary cites the only cases from the root stJid, occurring in 
the Atharvan text, to which the rule properly applies ; viz. m& ghoshd 
ut thuh (vii. 52. 2), tatas tvo '< thApayAmasi (x. 1. 29), and ut th&paya 
sidatah (xii. 3. 30) ; in each instance, the pada-text reads the *, leaving 
the irregular and mutilated sandhi for the sanhitA to make. Wherever, 
however, the preposition receives the accent, and enters into a more 
intimate combination with the root, as in the participle tilthita, the pada- 
text (by iv. 62) does not separate the compound, or restore the original 
«, but reads the same form which appears in sanhilA. Of this kind is 
also the only example of the root stambh combined with the preposition 
ud which our text presents, viz. salyeno UtabhitA (xiv. 1. 1), where the 
pada reads uttabhitA, and not ut-stabhitA : the passage is cited by the 

The Vaj. Pr. (iv. 95) notices the loss of s from the root stambh, but, 
as Weber remarks with surprise, omits all mention of slhA. The Taitt. 
Pr. (v. 14) includes these cases in a more general rule, that s is dropped 
when preceded by ud and followed by a consonant. 

t<OT 1% H ^ ii 

19. R is dropped before r. 

The corresponding rules in the other treatises are Rik Pr. iv. 9 (r. 28, 
ccxlvii), Vaj. Pr. iv. 34, Taitt. Pr. viii. 16. 

The r which is thus dropped must itself, of course, be the product of 
euphonic processes taught elsewhere (ii. 42, 43). The protraction of a 
preceding short vowel when a r is thus dropped is prescribed in a later 
rule (iii. 20). 

W^lirlHI<HrlHWHrlH II ^o n 

20. After a nasal, a non-nasal mute is dropped before a non- 

This rule, also, is hardly in place as a part of the Praticakhya, unless 
it be meant that in the words to which it applies the non-nasal mute is 
not to be omitted in the pada-text. The most frequent cases occurring 
under the rule are those of forms of conjugation coming from roots ex- 
hibiting a nasal before their final mute, and formed by affixes commenc- 
ing with a consonant : as, from indh, indhe instead of inddhe, for indh- 
-te ; from chind, chintam instead of chintlam, for chind-iam; from anj, 
Antam instead of Anktam, for Arij-tam ; from yuftj, yundhi instead of 
yungdhi, for yuftj-dhi, etc. In all such cases, however, the pada manu- 
scripts, as well as the others, omit the intermediate mute, nor is it at 
all likely that they ought to do otherwise : the rule is one properly of 
supererogation, yet finding a sufficient excuse in the peculiarity of the 

412 W. D. Whitney, [ii. 20- 

mode of utterance ■which it inculcates, and in the desirability that this 
should be noticed in the grammatical text-book of the school. Neither 
of the other known Praticakhyas teaches the same omission, or even 
notices it as prescribed by any authority. 

The citations of the commentator are pantir atra (fabricated: no 
such case in AV.), pdntam chandah (xii. 3. 10), and sapatndn me bhandhi 
(x. 3. 13). As counter-examples, to show that the omission takes place 
only after a nasal and before a non-nasal mute, he instances tasyd vdyur 
vatsah (i. e. valtsah : iv. 39. 4), utso va tatra (i. e. uttso and tattra: vi. 
106. 1), apsarasah sad/iamddam madanti (i. e. appsarasah : xiv. 2. 34), 
and nuddma enam apa rudhmah (i. e. ruddhmah: xii. 3. 43). 

The Atharvan manuscripts are quite consistent in observing this rule, 
although there are cases in which one or another of them preserves the 
mute of which the omission is here directed. In the published text, it 
is uniformly followed — with, I believe, but one accidental exception, viz. 
anuprayunktdm (xii. 1.40): and here, for once, all the manuscripts 
happen to agree in retaining the k. 

^TORTt: q<M*flj ii ^ ii 

21. Final y and v, following a vowel, are dropped. 

This rule applies, on the one hand, to the y and v of the syllables ay, 
av, ay, dv (the latter, however, being excepted by the following rule), 
into which, by iii. 40, e, o, di, and au are converted before a vowel ; 
and, on the other hand, to the y into which, by ii. 41, visarjaniya theo- 
retically passes before an initial vowel. An equivalent rule is found in 
the Vaj. Pr., at iv. 124. The teachings of the Taitt. Pr. upon the sub- 
ject are found at x. 19-23 : that treatise is here, as on so many other 
points, especially liberal in the citation of the opinions of discordant 
authorities. According to it, y and v are dropped when preceded by a 
and d; Ukhya, however, maintaining the contrary; Samkrtya denying 
the loss of v ; Macakiya allowing the elision of both when followed by 
m or o ; Vatsapra holding that they are not lost altogether, but only 
imperfectly pronounced. The treatment of final diphthongs and visar- 
jantya by the Rik Pr. does not include the exhibition of a final semi- 
vowel which requires to be got rid of, and hence it has no precept cor- 
responding with the one now in question. 

The commentator instances ka dsan janydh ke varah (xi. 8. 1), ushne- 
na vaya udakene " 'At (vi. 68. 1), asyA ichann agruvai patim (vi. 60. 1), 
so « eva mahdyamah (xiii. 4. 5), and td imd dpah (xv. 15. 1). In these 
passages, ke, vdyo, and asydi are converted into kay, vdyav, and asydy, 
by iii. 40, prior to the elision of the semivowels: while sah, tdh, and 
imdh are in like manner, by ii. 41, converted into say, tdy, and imdy. 

HI*l^i*l<W II ^ II 

22. But v is not dropped after d. 

That is to say, final dv before a vowel — the result of the change of 
an original aw, by iii. 40 — remains dv, being subject to no farther 

ii. 24.] Atharva- Veda Prdticdkhya. 413 

change. This rule is uniformly observed in the sanhitd of the Atbar- 
van, excepting in a couple of cases in book xix, which book the Prati- 
cakhya does not recognize as forming part of the Atharvan text : these 
are pddd ucyete (xix. 6. 5), and citrd imd vrshabhdu (xix. 13. 1). The 
commentator's examples are dvdv imdu vdtdu vdtah (iv. 13. 2), indra- 
vdyd ubhdv iha (iii. 20. 6), and itbhdv indrdgni d bharatdm (v. 7. 6). 

The Vaj. Pr. teaches the loss of the v of dv as well as of av (iv. 124), 
but adds (iv. 125) that some would retain the v excepting when followed 
by w, o, and du. The doctrines of the Taitt. Pr. have been stated in 
full under the preceding rule. The Rik Pr. (ii. 9, 10, 11, r. 25, 28, 31, 
cxxix, cxxxii, cxxxv) holds the view referred to by the Vaj. Pr. in its 
latter rule : o and &u, according to it, become av and dv before any 
other than a labial vowel ; before a labial, a and d. 

irfoft 3TWIT ^frT STll^ll 

23. Nor in gavishti and gaveshana. 

These are the only words found in the Atharvan in which the diph- 
thong o is the final of the first member of a compound before a follow- 
ing vowel, and in such a case, as we might expect, the fuller pronuncia- 
tion is retained, and the v preserved. The commentator cites ishuman- 
tam gavishtdu (iv. 24. 5 : p. go-isktdu), and gaveskanah sahamdnah (v. 
20. 11 : p. go-eshanah). Other like cases, as gavdfir and garish, occur 
in the twentieth book of the text, but with that book the Pratic&khya 
has nothing to do. 

H^Frl^Hjtf MlhM,IUHH( il $ II 

24. According to Q&katayana, there takes place in these cases 
an attenuated utterance of y and v, as regards the contact. 

The commentator gives us no help whatever as regards the interpre- 
tation of this difficult rule : he simply paraphrases it, as follows : lega- 
vrttir bhavati adkisparpam fdkatdyanasya, and then proceeds to repeat 
all the illustrative citations given above under rule 21. The other 
treatises, however, throw a good deal of light upon its meaning. The 
word iepa, ' diminution, attenuation, mutilation,' occurs in the same con- 
nection in the Taitt. Pr., in a rule already quoted (under ii. 21), which 
states that Vatsapra holds, not the omission, but the Ufa, of final y and 
v after a and d ; and the commentary there explains Ufa by luptavad 
ttccdranam, 'an utterance of them as if they were omitted.' In the 
Rik Pr., too, Ufa is once found, in the chapter treating of faulty pro- 
nunciation (xiv. 5), and is set over against pidanam — Ufena vd vacanam 
ptdanarh vd, which Regnier translates " a pronunciation attenuated or 
pressed (i. e. too forcible)." Panini (viii. 3. 18) attributes to Qakatayana 
the same doctrine as regards the pronunciation of final y and v — vyor 
laghuprayatnatarah, fdkatdyanasya, ' the utterance of y and v, accord- 
ing to Gakatayana, is to be made with slighter effort.' Q&katayana, 
then, is to be understood as holding, like Vatsapra, that the final semi- 
vowels are not to be omitted altogether, but slightingly and imperfectly 

414 W. D. Whitney, [ii. 24- 

uttered, the partial contact (i. 30) which is characteristic of them not 
being completely made. The citation by the commentator of the whole 
body of examples belonging to ii. 21 under this rule shows that he re- 
gards the latter as referring to all the cases included in the former ; and 
its position after rules 22 and 23 would indicate that it applies to the 
combinations treated in those rules also. The scholiasts to Panini restrict 
Qakatayana's doctrine to y and v when preceded by bho, bkago, agho, 
and a: but the Vaj. Pr. (iv. 126) refers to him as exempting only the 
word as&u from the treatment prescribed for y and v in every other 
case— which treatment, however, it does not specify to be attenuation 
instead of omission. 

tfjsrr% ii ^i ii 

25. The m of pum becomes visarjaniya before a surd mute 
not followed by a spirant, except in pungca etc. 

This is a rule very hard to get along with. In the first place, it is 
altogether unnecessary and uncalled for, since, of all the words to which 
it is intended to apply, but a single one, punpcalt, is found in the Athar- 
van text, and that one is written by the pada-text precisely as in sanhitd, 
and so requires no explanation from the Prati§akhya. But we have 
noted, and shall have still to note, many cases in which the treatise deals 
with irregularities of derivation or combination, even though they are 
not reduced to regularity by the pada-text, so that we need not be much 
surprised to find the formation of punpcali taught. Another difficulty 
is that, instead of simply disposing of the case which the text presents, 
the treatise gives to the rule a general form of statement, applicable to 
all possible cases. Yet even this is supported by its usage in several other 
instances, in which it affects a theoretic completeness suited to a general 
rather than to a special grammar ; and the precise virtual accordance 
of our rule, with the exception of its last word, apunpc&dishu, with one 
contained in Panini's grammar (viii. 3. 6), is a sufficient explanation of 
the form of statement adopted. The addition of the word apunpc&di- 
shu remains the last and the worst difficulty, and I must confess myself 
unable to give a satisfactory solution of it. The commentator furnishes 
no help as regards it ; his treatment of the whole rule is as follows : he 
first repeats it, inserting merely the omitted copula bhavati after visar r 
janiyo, and gives as illustrations punsk&ma, punsputra, and punpcalt 
(e. g. xv. 2. 1 : the other words cited, here and hereafter, as already 
remarked, do not occur in AV. ; these are all found, with punskoMla, in 
the scholia to Panini) : he then asks " why does it say ' before a mute ?' " 
and cites in reply pumyanam ; farther, "why 'before a surd mute?'" 
reply, because of pumdana (pumddsa? Pan., pumd&sa and pumgava); 
again, " why ' before one not followed by a spirant V " reply, because 
of purhkshura (Pan., pumkshira and pumkshura) ; and finally, apunpcd- 
dishv iti kim : punpcorah; ' why " excepting in punpca etc. ?" because of 
such cases as punpcora? But punpcora, ' he-thief,' is as regular an in- 

ii. 26.] Atharva- Veda Prdtigdkhya. 415 

stance of the application of the rule as punskdmd or puhpcall ; nor 
does it seem possible to find in apunpcddishu itself any form which 
constitutes an exception to the previous specifications. I can only con- 
jecture that the reading is corrupt, and was corrupt before the comment- 
ator set himself at work upon it, and that his explanation was as unin- 
telligible to himself as it is to us. The specification may have been 
intended for such words as puihkhydna, which constitutes an actual 
exception to the rule, and it is cited as such in Bohtlingk's note to 
Panini viii. 3. 6, as from the Siddhanta-Kaumudi. 

It deserves to be remarked that the introduction of the word visar- 
janiya into the next following rule tends strongly to show that the one 
now under discussion is an interpolation : otherwise the term should be 
understood in the rules which succeed, by implication from this, and 
should not require to be again specified. 

The conversion of the m in pum into visarjaniya of course includes, 
by i. 68, the nasalization of the preceding vowel, and also the adapta- 
tion of the visarjaniya to the following consonant, by ii. 40, 62, etc. 

26. N becomes visarjaniya before surd palatal, lingual, and 
dental mutes not followed by spirants. 

That is to say, virtually, a sibilant is inserted before the mute, of the 
same class with the latter, and the « itself is replaced by the nasaliza- 
tion of the preceding vowel. Here, again, the mention of Unguals is 
superfluous, no cases arising in the text to which this part of the rule 
should apply. The commentator fabricates his whole series of examples 
illustrating the application of the rule, viz. : bhavdnf cinoti, bhavdnp 
chddayati, bhavansh tikate, bhavdns iarati, bhavdns tatra. To explain 
the reason of the specification " surd " contained in the rule, he cites 
two actual cases, brhan dakshinayd (vi. 53. 1), and ndi , nan namasd 
parah (vii. 7. 1), which show that no such conversion is made before a 
sonant or nasal mute. But farther, to explain the addition of the re- 
striction " not followed by spirants," he resorts again to fabricated in- 
stances, bhavdn tsaru, mahdn tsaru : this time with good reason, since 
no such cases occur in our text, and the restriction, so far as concerns 
the Atharva- Veda, is superfluous, and is only inserted, like the specifica- 
tion of the lingual along with the palatal and dental mutes, in order to 
make the rule theoretically more complete. 

The insertion of a sibilant between a final n and an initial c (as in 
parvatdnp ca, i. 12. 3) is made in the Atharvan uniformly, without a 
single exception, and, owing especially to the frequency of the particle 
ca after a nominative or an accusative in n, the cases are very numerous : 
of n before ch the text affords no example. A like insertion of s before t 
(th never occurs as initial) is not rare (the text presents sixty-seven cases), 
but the exceptions — which the treatise notes in rule 30, below — are also 
tolerably numerous : they are all given in a marginal note farther on. 

The doctrine of the Vaj. Pr. on the subject of these insertions cor- 
responds precisely with that of our own treatise; its rules (iii. 133, 134), 
vol. vii. 53 

416 W. D. Whitney, [u\ 26- 

however, omit the unnecessary mention of the lingual mutes, and re- 
striction to mutes not followed by spirants, although they retain the 
equally unnecessary mention of ck and th. The exceptions are given 
in detail (iii. 142-144); among them are to be found no cases of n 
before an initial c. The Taitt. Pr. gives a general rule (v. 20) for the 
insertion of the sibilant before e, and then (v. 21) rehearses all the cases 
(seven in number) in which it does not take place. The insertion before 
t is noted in another place (vi. 14), and all the cases of its occurrence 
(only eighteen in number) are there enumerated. The Rik Pr. (iv. 32 
etc.) catalogues all the words before which the sibilant is added, as well 
before c (iv. 32, r. 74, ccxciii), as before t (iv. 33, r. 76, ccxcv) : such 
words in the Rig- Veda are not very numerous; its usual method ofsandhi 
is ti-c and n-t simply. The Rik presents, on the other hand, a few cases 
(five) in which a sibilant, converted to visarjanlya, is inserted after n 
before an initial p (R. Pr. iv. 34, r. 78, ccxcvii, ccxcviii). 

It is sufficiently evident that this insertion of a sibilant after a final 
n before a surd mute is no proper phonetical process : the combination 
of the nasal and following non-nasal is perfectly natural and easy with- 
out the aid of a transition sound, nor can any physical explanation be 
given of the thrusting in between them of a sibilant, which only en- 
cumbers the conjunction. Some other reason must be sought for the 
phenomenon : nor is such a reason difficult to discover. The historical 
rather than phonetical origin of the r which is appended (see rule 29, 
below) to a few accusatives plural in the Vedic language before a vowel 
has been long since pointed out by Bopp (see his shorter Sanskrit gram- 
mar, § 82 b ) ; and a kindred explanation of the conversion of dn into dn 
before a vowel (see rule 27, below) was added by him in his Compara- 
tive Grammar (see the second edition, i. 468, 478, 479). He has re- 
frained from tracing the insertion of a sibilant before e and I to the 
same cause, doubtless, because of the numerous instances in which the 
insertion is made after a word which is not entitled by origin to a final 
s. But nothing is more natural than that an insertion originally organic, 
but of which the true character was forgotten, and which had come to 
seem merely euphonic, should considerably extend its sphere of occur- 
rence, and should be by degrees, and more and more, applied to cases 
to which it did not historically belong. Now a very large majority of 
the words ending in w are accusatives plural and nominatives singular,* 

* That T might not seem to speak at random upon this point, I have looked 
through half of the Atharvan text, or books i-ix, and have noted the character of 
every word terminating in n which is to be found therein. The result is set forth 
in the following table : 
Accusatives plural in dn 
Nominatives singular in an 

It is thus seen that the forms to which a final < originally belongs outnumber the 
others almost precisely in the proportion of three to one, or constitute three-quarters 
of the whole number of words ending in n. 


Vocatives in 









Locatives in 






Verbal forms in 







ii. 27.] Atharva-Veda Prdligdkhya. 417 

to both of which cases comparative grammar clearly shows that a final 
•s belongs as case-ending; and I can entertain no doubt that the whole 
phenomenon of the insertion of the sibilant arose from its preservation 
in these forms, and from the inorganic extension of the same mode of 
combination, by analogy, to the much smaller classes of vocative, loca- 
tive, and verbal forms. The same conclusion is favored by the aspect 
of the phenomenon of the insertion of s between n and t, as it presents 
itself in the text of the Atharva-Veda. Although the insertion is there 
made after other forms than those originally entitled to a final *, it is 
rare after such forms in a ratio four times greater than that of the rarity 
of the forms themselves : that is to say, while these are in number one- 
third of the others, the insertions after them are only one-twelfth as 
numerous. And, on the other hand, although the insertion is sometimes 
omitted after nominatives singular and accusatives plural, it is omitted 
five times as often, in proportion, after the final n of other forms than 
these. For a detailed and classified statement of all the passages in 
which the sibilant is either inserted or omitted after a final « before an 
initial t, see the appended marginal note.* 

ill*l(ta^Hnt|sHl{iHf ^X II y> H 

27. The final n of upabaddha etc., when preceded by d and 
followed by a vowel, becomes visarjaniya. 

This process includes two additional steps, taught elsewhere in the 

* Cases of the sandhi with insertion of » between n and t: I. Accusatives plural. 
1. in an: i.8.4. ii. 12.7; 25.4. iii.2.5. iv. 19.4; 22.1; 86.3. v.8.8. vi. 112.2. 
vii.97.3. viii.6.7. x.8.15,16. xi I. 6, 10, 26; 5. 2, 4,7; 6. 17; 9. 22,24; 10. 23. 
xii. 3.40, 53. xviii. 1.47; 2.34; 3.68; 4.39,86,87. xix. 27. 4 (ter) ; 28.2; 35 3,5; 
86. 5 ; 49. 3 ; 66. 1. S. in in : iii. 21. 1. S. in An : viii. 8. 8. x. 7. 42. xi. 1. 20. xix. 
6.14. 4. in rn: iv. 27.2. II. Nominatives singular. 1. in an: vii. 18. 1 ; 18. 2. xiii. 
1.32,39. xvii. 10. xviii. 2. 20. 2. in an: ix. 2. 19-24. xii. 1. 18. xiii. 2. 29 ; 4. 44. 
xvii. 16. III. Vocatives singular in an : v. 22. 6. IV. Locatives singular in in : xi. 
8.10. V. Verbal forms in an: xix. 19. 10. xx. 135.6, 7. 

Cases of the sandhi without insertion of « .• I. Accusatives plural. 1. in dn: 
vii. 38. 3. 2. in in: xviii. 2. 15, 18. 3. in An : vii. 84. 8. II. Nominatives singular. 
1. in an: xiii. 2. 83. xiv. 1. 6. xix. 28. 4. 2. in dn: iv. 32. 8. xviii. 2. 82. xx. 128. 
14. III. Vocatives singular in ore: i. 14. 3. vii. 9. 3. IV. Locatives singular in in ; : 
i. 16. 2; 35.3. v. 28.4. vii. 48. 1. X. 3. 17. xiv. 2. 48, 49. V. Verbal forms in an . ; 
i. 11.2. iii. 9. 2; 22.3. iv. 7. 7 ; 14. 1 ; 35. 2. x. 10. 24. xi.5.2. xiv. 2. 14. 
That is, in tabular form : 

with tf. without 8. 

Accusatives plural in dn 40 1 

in 1 2 

An 4 1 

fn 1 46 4 

Nominatives singular in an 6 3 

dn _10 J 6 . _2_? 

Total 62 10 

Vocatives singular in an 1 2 

Locatives singular in in 1 7 

Verbal forms in an 3 9 

Total ~5 "18 

418 W. D. Whitney, [ii. 2?. 

treatise, before the combination is complete, and the final sanhitA form 
reached. The conversion of the n into visarjaniya itself implies, by i. 
68, the nasalization of the preceding vowel, so that upabaddhAn becomes 
vpabaddh&nh ; then the visarjaniya, by ii. 41, becomes y before the fol- 
lowing vowel: upabaddhAny iha; and lastly, by ii. 21, the final y is 
rejected, and we obtain upabaddhAn ilia. This seems a cumbrous and 
artificial process, yet it is in part well-founded and correctly carried out. 
All the cases in which this loss of a final n occurs are accusatives plural 
or nominatives singular, which originally possessed a final * after the n, 
and the loss of the n before the sibilant, with accompanying nasaliza- 
tion of the preceding vowel, and then the disappearance of the sibilant 
itself, as in other cases after A and before a vowel, are unquestionably 
the cause of the sand hi as it finally presents itself. Our treatise, then, by 
bringing in the visarjaniya as a step in the process, and treating of this 
combination in intimate connection with those related ones which form 
the subjects of rules 26 and 29, has a decided theoretic advantage over 
either of the other Pratigakhyas. The Eik Pr. (iv. 26, r. 65, cclxxxiv) 
prescribes simply the omission of the final, excepting at the end of a 
pAda, afterwards (iv. 26, 27) specifying the cases in which the omission 
takes place even at the end of a. p&da, and finally (iv. 30, 31) those in 
which it does not take place even within a p&da (there are only eleven 
such cases). The Vaj. Pr. (iii. 141) and the Taitt. Pr. (ix. 20) come 
one degree nearer to the method of our treatise, by converting the » 
into y before its elision, and both give in detail (V. Pr. iii. 145-149, 
T. Pr. ix. 23-24) the exceptional cases in which the n remains unchanged. 

The commentator cites only the first five instances which the text 
contains, viz. : upabaddh&n ih& " vaha (i. 1. 7), f&sa itth& mah&n asi (i. 
20. 4), yo asm&n abhid&sati (e. g. i. 19. 3), and sarv&n mac chapathan 
adhi (e. g. ii. 7. 1). More than a hundred cases occur in the Atharva- 
Veda, so that the gana upabaddh&dayah must have been a tolerably stout 
one. I add in a marginal note a complete list of the cases, classified.* 

To give with the same detail the exceptions to the rule, or the cases 
in which final &n remains unchanged before a vowel, would be quite 
useless. They are very frequent, by far outnumbering the instances of 
the loss of the n — thus, in the first four books of the text, against thir- 
teen instances of An before a vowel, we have forty-one of An, and twelve 
of these between two p&das — and they are found indifferently in all 
possible situations, so that it is quite impossible to lay down any rule 

* I. Accusatives plural: 1. before a: 119. 3; 21.2. ii. 7. 1. iii. 8. 3. iv. 19.5,7. 
v. 5. 9; 18.11; 20.8. vi. 15. 1, 2; 41. 3; 54. 3; 69.2; 72. 1; 75.3; 76.4; 17. 1; 
113.2; 121.4; 129.2. vii.9.2; 27.1; 57.1; 65.1; 109.4. viii.3.6,20; 4.14; 9. 
24. ix. 1. 19 ; 2. 25 ; 4. 24. x. 2. 22, 23 ; 6. 41 ; 6. 19, 30 ; 7. 7 ; 10. 6. xi. 1. 29 ; 9. 
17 22, 24; 10. 23. xii. 1. 25; 2. 12; 3. 15 (bis), 18 ; 4. 31. xiii. 1. 53; 2. 5, 18, 21. 
xiv. 1. 45, 55 ; 2.5,10. xviii.1.45; 2.11,13,15,18; 4.58,61. xix.6.8; 13.8; 26. 
8; 32.7; 36.4; 50.4. xx. 127.7; 128.4,5; 136.15. 2. before d: ii. 25.4. vi. 
28.2. xviii.3.55. xix. 59. 2. 3. before*.- i.7.7. ii. 27. 5 ; 31.1. v. 8. 1 ; 13.6; 
23.8. vi. 22. 3. vii. 117. 1. 4. before u : iv. 34. 7. vi. 59. 2. viii. 9. 23. ix. 9. 
15, 19 (bis), x. 3. 13, 14, 15. xii. 3. 16, 40. xviii. 2. 21. 5. before r: viii. 8. 7. 
xviii. 1. 18. 6- before e: xi. 1. 4. 

II. Nominatives singular : 1. before a : i. 20. 4. iii. 16.5. vii. 91.1. viii. 5.22. 
xiii. 2. 29 (ter). xviii. i. 24. xx. 128. 4, 8. t. before i: vii. 92. 1. viii. 4. 2. J. be- 
fore u: xviii, 1. 22, 48 (bis). 

ii. 29.] Atharva-Veda Pr&li$khya. 419 

respecting them. The loss of the n with nasalization of the vowel is 
evidently an old-style sandhi, going out of use, and no longer appearing 
except sporadically. It is interesting, as regards this sandhi and that 
taught in the preceding rule — which have both, as explained above, the 
same historical origin — to note the relations of the Rik and the Athar- 
van usage to one another and to the practice of the classical Sanskrit. 
The insertion of the », which has become a necessary proceeding under 
the modern euphonic rules, is almost universal in the Atharvan, and 
comparatively rare in the Rik : the conversion of n into anusvdra, of 
which the general Sanskrit grammar knows nothing, is only infrequently 
observed in the Atharvan, while it is made in the Rik with but few ex- 

apff oMIHlirl ^>f^ II St II 

28. In the passage vrkshdn vandni, n is converted into visar- 
jamya before v. 

The commentator cites the passage, vrkshdn vandni sam cara (vi. 
45. 1), which is the only one of its kind in the text. A few such in- 
stances, of the loss of n before semivowels, with nasalization of the pre- 
ceding vowel, are found in the Rik and White Yajus, and are noticed in 
their Praticakhyas (see R. Pr. iv. 28, r. 68, cclxxxvii, and V. Pr. iii. 
135, 136). 

The commentary, to explain why the rule does not read simply 
vrkshdn iti vaJcdre, says sopapadasya grahanam etdvattvdrtham : iha md 
bhut: vrkshdn vdto vrkshdn vaydh; 'the citation of vrkshdn along 
with its following word is for the purpose of restricting the action of 
the rule to this particular case : the conversion is not to be made in the 
passages vrkshdn vdtah and vrkshdn vaydh? These counter-examples, 
however, are fabricated : no such passages occur in the Atharvan. Nor 
is the citation of vandni in the rule necessary, although excusable enough : 
a v follows vrkshdn in no other passage of the text, except in xii. 1.51, 
where it is separated from it by an avasdna, and so exercises upon it no 
euphonic influence. 

^n^JT^l"^ =Wrj^r^ Q|tf|Ho|Hl{lHI^U^ll 

29. Preceded by an alterant vowel, n becomes r in the pas- 
sages rbdnr ut srjate vagi etc. 

All the vowels except a and d are called ndmin, as tending to produce 
the nati, or conversion, of a following * into sk. The Rik Pr. (e. g. i. 
17, 20) has the same term; see Regnier's note to i. 17 (r. 65, 66) : the 
Vaj. Pr. uses instead bhdvin. 

The Pratigakhya is to be reprehended here for not treating the cases 
to which this rule applies in the same manner as those coming under 
the preceding rules, by prescribing the conversion of n into visarjaniya, 
and leaving it for rule 42, below, to change the latter into r. In fact, 
the first two words of the rule are superfluous, and might advantageously 
be omitted. The origin of this peculiar and rather uncommon sandhi 

420 W. D. Whitney, ["• 29- 

is clearly the same with that of those which form the subject of rules 
26 and 27. Only nine cases of it occur in the Atharvan : of these, 
three are cited by the commentary, viz. : rtunr ut srjate va.fl (vi. 36. 2), 
mo shu paninr abhi (v. 11.1), and dasy&nr uta bodhi (iv. 32. 6) ; the 
others are the word rtun three times before a (vi. 61. 2, 3. vii. 81. 1), 
and pit? n three times before u (xviii. 2. 4, 23 ; 4. 40). 

The Rik Pr. (iv. 29, 30) prescribes the insertion of r after in and tin 
everywhere before a vowel, except at the end of &pAda (and once even 
there), and in a single instance after rn. The same sandhi is also made 
in half a dozen instances before y, v, and h. The Vaj. Pr. specifies (iii. 
140) the few passages in its text where the conversion of n to r occurs 
after i and u ; and the Taitt. Pr. (ix. 20) puts the conversion of An to 
An and of in and tin to inr and tinr into the same rule together. 

=T HH(M-Hl{lHl^il^ou 

30. Exceptions are the passages sam dirayan tdm etc. 

By the position of this rule, the gana samAirayantAdayas ought to 
include exceptions to all the preceding rules, beginning at ii. 26. Since, 
however, the rules 27-29 apply only to certain specified cases, it is dif- 
ficult to see the necessity of specifying any exceptions to them, and we 
cannot help conjecturing that the present • precept belongs to rule 26 
alone, and should properly come in next after it, as rule 27. The first 
passage of the gana, sam Airayan tAm vy tirnuvanlu (i. 11. 2), is the 
first instance which the text presents of a n directly preceding t without 
the interposition of a sibilant, and the commentator goes on to cite the 
two next succeeding cases of the same character, viz. : kulapA rAjan 
tAm u te (i. 14. 3), and asmin tishthatu yd (i. 15. 2) : the three happen 
to be typical examples of the three principal classes of cases — verbal 
forms, vocatives, and locatives — in -which we should not expect to see 
the sibilant inserted, since the forms did not originally end in a sibilant. 
For a complete list of the exceptions to rule 26, see the final marginal 
note to the exposition of that rule. 

31. M, before a mute, becomes of like position with the latter. 

The Bik Pr. (iv. 3, r. 6, ccxxv) adds the restriction visthAne, ' before 
a mute of another class,' which is a matter of course, and does not need 
specification ; and both it and the other treatises (V. Pr. iv. 1 1 ; T. Pr. 
v. 27) state distinctly what is implied in our rule by i. 95, that the sound 
into which the m is converted is the nasal of the same class with the 
following mute. The commentator gives the following instances, writing 
always an anusvAra for the nasal into which the m is converted : san 
kdpayAmi vahatum (xiv. 2. 12), udAgAn jivah (xiv. 2. 44), tan dayamd- 
nam (fabricated : no such case in AV.), san nas tebhih (ii. 35. 2), san 
tAih pafubhih (iv. 36. 5), san nashtena (vii. 9. 4), san tvayAi 'dhishimahi 
/xiv. 2. 17), and md tvA vrkshah sam bAdhishta (xviii. 2. 25). The manu- 

ii. 33.] Atharva- Veda Pr&tic&khya. 421 

script reads farther gdnias tdn nedd , ntahpade tavarge prakrtyd : dur- 
ndmnih sarvdh. The beginning of this is probably an additional citation, 
but, if it be so, it is so corrupted in reading that I am unable to trace 
it out. The rest is a restriction applied by the commentator himself, 
vdrttika-Wko, to the action of the rule : ' in the interior of a word, m re- 
mains unchanged before a dental : e. g. durndmnih sarvdh (iv. 17. 5).' 
It is unnecessary to remark, however, that the Praticakhya has nothing 
to do with explaining the m of such a word, and that the commentator's 
emendation of his text is therefore impertinent; it is also bungling, 
since such a varttika, if constructed at all, should be made to apply, 
not to a dental only, but to a lingual, in such words as aryamnd. 

32. Before semivowels and spirants, it is omitted. 

This omission, by i. 67, carries with it the nasalization of the preceding 
vowel. The commentator's examples are vrksham yad gdvah (i. 2. 3), 
pitaram varunam (i. 3. 3), sarhrddhayaniah sadhurdh (iii. 30. 5), pard 
'dya devd vrjinam prnnntu (viii. 3. 14), ny oshatam hatam (viii. 4. 1), sam 
subh&tyd (iii. 14. 1), and bhavasi sam samrddhyd (xii. 3. 21). 

The Rik Pr. (iv. 5, r. 15, ccxxxiv) converts m into anusvdra before 
the spirants and r, but treats it before y, I, and v (iv. 3, r. 7, ccxxvi) in 
the same manner as our treatise (rule 35, below) before I alone. The 
Vaj. Pr. (iv. 1, 3, 9) teaches precisely the same doctrine, but refers (iv. 
4) to Kagyapa and Qakatayana as holding that the m is dropped. The 
Taitt. Pr. (xiii. 2) declares, like our own treatise, the m to be lost, but 
only before the spirants and r ; before all the other semivowels it con- 
verts it (v. 28) into the nasalized semivowel, agreeing in this with the 
Rik Pr. and Vaj. Pr. ; it also notices, however (xiii. 3), the view of the 
Ath. Pr. as held by some authorities. 

33. In the interior of a word, it is omitted before spirants only. 

As examples of the loss of m in the interior of a word before spi- 
rants, the commentary presents the whole list of examples — dve ca me 
vinfalif ca etc. — already given above, under i. 27, and repeated under 
i. 53 and i. 83. As counter-example, we have patir yah pralikdmyah 
(ii. 36. 8) alone. Instances of m before r in like position would not be 
hard to give — e. g. tdmradh&mrdh (x. 2. 11) — but it is found before I 
only in root syllables, as in malimlucam (viii. 6. 2), and before v only in 
the case which forms the subject of rule 37, below. 

Both this rule and the next concern matters with which the Pratica- 
khya properly has no concern. Accordingly, the Rik Pr. (iv. 3, r. 1, 
ccxxvi) disposes of them simply by specifying that m is altered before 
an initial semivowel, excepting r (in connection with which, in the later 
rule, it omits to repeat the specification), and the Taitt. Pr. says nothing 
upon the subject. But the Vaj. Pr. (iv. 2) gives a precept which in- 
cludes both the rule we are treating of and the one which follows it. 

422 W. D. Whitney, [ii. 34- 

H=hl(W ^ II & II 

34. As is also n. 

The commentator's citations to illustrate this rule are those already 
once given, under i. 84, including the words par&nski, yajunski, and 
havtnski, and, to show that n is not altered in the interior of a word 
before semivowels, he farther quotes tdudt ndmd 'si kanyd (x. 4. 24). 
As instance of n before v within a word, we may take tanvah (e. g. i. 
1. 1) : before r and I it is found only when final. 

The Vaj. Pr. (iv. 2) is the only other treatise which contains a rule 
corresponding with this. 

85. Both m and n, before I, are converted into a nasalized I. 

The commentator quotes from the text, as instances of m before I, 
tanl* lokam (iii. 28. 5), avinl lohena (iii. 29. 3 etc), pratimdnl lolc&h 

ixviii. 4. 5) ; and, as instances of n before I, durgandhlnl lohitdsydn 
viii. 6. 12), and sarvdnl lokdn (e. g. iv. 38. 5). 

It is perhaps to be regretted that the editors of the published text 
did not follow this rule of the Pratic&khya with regard to both m and 
n. The manuscripts, however, are almost unanimous in reading only a 
single I after an original m, with a nasal sign over the preceding vowel 
(there are but two or three cases, if I recollect aright, of a doubled I), 
and their authority has in this respect been followed. Where an origi- 
nal n has disappeared, on the other hand, the manuscripts follow, not 
without some exceptions, the directions of the Pratic&khya, and we 
have done the same, also without absolute uniformity. 

The three other kindred works (see R. Pr. iv. 3, r. 7, ccxxvi ; V. Pr. 
iv. 9 ; T. Pr. v. 28, 29) agree with one another, and disagree with our 
treatise, in converting m before all the three semivowels y, I, and v into 
those semivowels nasalized ; as regards the treatment of the », there 

* The lack of suitable type reuders it necessary to represent the tandhi, in trans- 
cribing the instances, in this imperfect way : properly, no n should be written, and 
the sign of nasality should be set above the first t itself. It will have been noticed, 
also, that (for the same reason) the general method of transcription adopted for the 
nasal sounds is not in accordance with the theory of the Pratic&khya. The latter 
knows no anasvdra, and nothing intermediate between a nasal mute and a nasalized 
semivowel or vowel. We ought, then, in our transcription, to write, on the one hand, 
in every instance a nasal adapted in class to the following mute, as has been done 
in the examples under ii. 31 — only, if we choose, taking the liberty to substitute a 
dotted « and m in case of the assimilation of those letters, according to rules ii. 10, 
11, 31 — and, on the o'ther hand, in cases falling under rules ii. 27, 29, 32, etc., to 
write a vowel with a nasal sign above it. The distinction made in ordinary usage 
between the simple dot and the dotted crescent, as nasal signs, is purely arbitrary, 
founded on nothing in the theory of the Praticakhya, and having but a scanty and 
uncertain support from the Atharvan manuscripts : some of the latter occasionally, 
or even generally, attempt to use the dotted crescent for a nasalized vowel, and the 
dot for a nasal mute, but for the most part they employ the latter indiscriminately 
for both classes of cases. 

ii. 38.] Atharva-Veda Prdtigdhhya, 42S 

is a universal accordance (compare R. Pf. iv. 4> r. 8, ccxxvii ; V. Pr, iv, 
13 ; T. Pr. v. 25, 26). 

36. An exception is the m of sam before the root rdj. 

The only words coming under the action of this rule are samr&j 1 
samr&jni, and s&mr&jya : verbal forms from the root raj with the prefix 
sam are not found in the text. The commentary cites sdmr&d eko vi 
r&jati (vi. 36. 3), samr&jny edhi fvapureshu samr&jny ula devrshu: 
nananduh samr&jny edhi samr&jny ula fvaprv&h (xiv. 1. 44). The de- 
rivative s&mr&jya (p. s&m-r&jya) is found once only (xiv. 1. 43). 

The other treatises duly notice the same exceptional case (see E. Pr, 
iv. 1, r. 23, ccxlii ; V. Pr. iv. 5 ; T. Pr. xiii. 4) ; the Taitt. Pr. alone at- 
tempting to give the rule a more general form, and declaring sam and 
s&m not liable to change when followed by r&: it is strange if the 
treatise do not thus lay itself open to the imputation of an error ; our 
own text, at any rate, has such forms as samr&dhayantalj, (iii. 30. 5). 

37. As also, before a v which is the result of sandhi. 

The passage here referred to, and cited by the commentator, is sam 
v &sn& 'ha asyam (vi. 56. 3), where the particle w, following sam, is con- 
verted into v by iii. 39. There are two closely analogous cases — fam v 
astu — in the nineteenth book (xix, 10. 1, 9), which this rule is not con- 
structed to cover, since the Atharvan text recognized by our treatise 
consists only of the first eighteen books of the present Atharva-Veda. 

5pfMr^ HhUH'J gfftlT %f££PT: ll ^ n 

38. In case of a combination in the inverted order of the 
mute-series, there takes place sphotana, provided the former is a 

Weber (p. 267) regards viparyaya as signifying here simply 'differ- 
ence,' but it does not seem to me possible to give the word so general 
and indefinite a meaning, and the whole treatment of the subject by 
the two Praticakhyas goes to show, at least by negative evidence, that 
the cases contemplated by them are only those in which a mute of one 
series (varga) enters into combination with one of a preceding series, 
so that, in the group, the natural order of the series appears inverted. 
The precept of the Vaj. Pr. (iv. 162) is to the effect that it either is or 
is not an error of pronunciation to utter a guttural after another mute 
with sphotana. This is in appearance a narrowing of the sphere of 
occurrence of the sphotana to no small extent, as compared with our 
treatise : but it is almost only in seeming ; for, allowing the exception 
made in the next following rule, there are but two combinations requir- 
ing sphotana to be found in the Atharvan in which a guttural is not the- 
vol. vii. 54 

424 W. R Whitney, [»• 38- 

second member, and each is represented by but a single case, and one 
of the two is in the nineteenth book of the text : they are pch (trish- 
tupchanddh, vi. 48. 3), and bj (trishtubjagatydu, xix. 21. 1). The com- 
mentator to the Vaj. Pr. (Weber, p. 266) defines sphotana to be "the 
separated utterance of a close combination of consonants" (pindibhii- 
tasya samyogasya prthag uecdranam) ; our commentator gives no such 
explanation, merely paraphrasing the rule as follows : vargdndm vipar- 
yaye sphotanah sandhyo bhavati : p&rvena ced virdmo bhavati : but the 
text itself has already (see i. 103) defined sphotana to be a quarter or 
an eighth of a short a — doubtless a samvrta a, or the neutral vowel. 
There can be no doubt, then, that the sphotana is that very brief un- 
closure of the organs which we often, if not ordinarily, allow to take 
place between two mutes standing in conjunction with one another, and 
of the former of which we desire to make clearer the pronunciation. 
In passing from a t to a k, for instance, while it is possible by an effort 
to make the release of the ^-closure and the formation of the A-clo- 
sure so truly simultaneous that nothing whatever shall escape from 
the mouth during the transfer, it is more natural to let so much breath 
slip out between as shall render audible the unclosure of the den- 
tal position, and so far relieve the imperfect or abhinihita utterance 
of the t, rendering it comparatively clear and distinct (sphuta). This 
insertion is then properly enough called sphotana, ' that which makes 
clear, distinct, or evident:' we have noticed above (under i. 103) that 
the commentator gives it also another kindred name, vyanjaka, 'mani- 
fester.' It is, under other circumstances of occurrence, very nearly the 
same with that release or separation of the passive and active organs of 
production which the Vaj. Pr. (i. 90) prescribes after the pronunciation 
of a final mute in the pada-text, so that the next word may begin with 
a new effort. That the Hindu theory allows sphotana in the combina- 
tion of the phrase only in case two mutes meet in the inverse order of 
the vargas to which they belong has something of arbitrariness in it, 
yet is not without foundation ; for it may be noted, I think, that it is 
perceptibly harder to change from a contact farther forward in the 
mouth to one farther back, than to make a like transfer in the contrary 
direction, without allowing any intervening escape of breath or sound : 
and the order of the vargas follows the advance in the mouth of the 
place of formation. 

The commentator cites, as in'stances of the occurrence of sphotana, 
vashatkdrena (p. vashat-kdrena, e. g. v. 26. 12), avatkam (p. avat-Jeam, 
ii. 3. 1), ejatkdh (p. ejat-kdh, v. 23. 7), trishtub gdyatri (xviii. 2. 6), and 
y ad y ky aire (ix. 10. 1). Of other combinations than these, the text 
presents pk (anushtup katham, viii. 9. 20), tkh (e. g. utkhidan [p. ut- 
-khidanj, iv. 11. 10), and dgh (e. g. padghoshdih [p. pat-ghoshdih], v. 21. 
8). Whether combinations of the dental nasal with a following guttu- 
ral mute are to be regarded as coming under the rule, and admitting 
sphotana, is rendered at least doubtful by our commentator, who goes 
on to say : "why does the rule say p&rvena? because of such cases as 
kramdn ko asydh (viii. 9. 10)." It is evident from this that he would 
understand purva as equivalent here to anultama, ' not last in a mute- 
series,' i. e. ' non-nasal. T This seems to me, however, a very forced in- 

n. 39.] Atharva- Veda Praligakhya. 425 

terpretation, and unsupported by the usage of the word elsewhere, in 
this or in the other Praticakhyas. The Vaj. Pr. makes no such excep- 
tion of the nasals, nor is it noticed in the verse — from the same metrical 
treatise, doubtless, which has often been found cited above — with which 
the commentary on the rule closes: vargdndm viparitdn&th samnipdte 
nibodhata: vyavdyi sphotan&khyas tu yad gdyatre nidarfanam ; 'know 
ye that in the collocation of the series in their inverted order there takes 
place an insertion called sphotana : yad g&yatre is an instance of it.' I 
am disposed, then, to look upon the exclusion of the nasals from the 
operation of the rule as a later gloss, foisted in upon the rule itself. 
There is by no means a lack of reason for making the exception ; since 
the nasals are accompanied throughout their utterance by a free emis- 
sion of intonated breath through the nose, and are by it made so dis- 
tinctly audible that there is felt no impulse to give them additional 
clearness by the insertion of such a sound as sphotana. If the inter- 
pretation of the commentator be rejected, we shall have to add to the 
list of groups given above as coming under the action of the rule not 
only nk, of which an instance has already been cited, but also nkh (e. g. 
kepan khddantah, v. 19.3), ng (e.g.vidvan gandharvah, ii. 1. 2), and 
ngh (e. g. devdn ghrtavatd, iii. 10. 1 1). To the specification of the rule 
that the consonant followed by sphotana must be a final, the commen- 
tary brings up the counter-example venor adgd iva (i. 27. 3), where the 
group dg, although composed of a dental before a guttural, suffers no 
such interposition. The term virdma, which is used once or twice also 
in the technical language of the other Pratigakhyas, has the same sig- 
nification with avasdna, and denotes a pause accompanied with a sus- 
pension of euphonic influences ; such as takes place in the ordinary text 
only where there is a sign of interpunction, or at the end of a sentence 
or paragraph, but in the pada-text is found after every word, and even 
between the two separated parts of a compound word. 

39. But not in the case of a lingual before a palatal ; here 
there takes place a prolongation of the time : and this they call 

That is to say, when a t comes before a c, or ad before a,j (the only 
two cases which can occur under the rule), there is no separation of the 
two consonants by unclosure and reclosure of the organs, but the effect 
of the contact is merely to lengthen out the time employed in uttering 
the group : the name applied to this prolongation, karshana, ' traclio, 
drawing out, extension,' is not elsewhere met with in the grammatical 
literature. The commentator cites as examples shat ce 'mdh, (iv. 20. 2), 
shat ca me shashtif ca (v. 15. 6), and shad jdtd (viii. 9. 16) : they are 
the only instances of these combinations to be met with in the Athar- 
van text, except one in the nineteenth book {shat ca, xix. 47. 4). 

It is easy to see the physical ground of this exception to the rule pre- 

426 W. D. Whitney, [ii. 39- 

scribing sphotana. The same close relationship with respect to place of 
utterance which causes the final palatal to pass often into a lingual, 
instead of reverting to the guttural out of which it originally grew, 
causes the lingual, in coming before the palatal, to virtually double it 
only. The transfer of position of the organs is too slight and easy to 
necessitate the emission of an intervening sound. 

This is the last rule in the first section of the second chapter. The 
manuscript this time omits to specify the number of rules contained in 
the section, and adds simply dvitiyasya praihamah p&dah. 

40. Visarjantya, before a surd consonant, becomes of like po- 
sition with the following sound. 

That is to say, applying rule i. 95, it is converted into the spirant 
(ushman) corresponding in position with the following letter. Thus, 
before c and eh it becomes f ; before t and th, sh; before t and th, s; 
before f, sh, and s, it is changed into each of those letters respectively ; 
before k and Jch it becomes jihvamuliya, and, before p and ph, upa~ 
dhm&niya — these last two spirants being, as already noticed, clearly im- 
plied in this rule, although nowhere referred to by name as belonging 
to the scheme of spoken sounds recognized by the treatise. Visatja- 
niya itself, then, would only stand, in sanhitd, before a pause. The 
theory of the Prati<jakhya, however, is not at all the practice of the 
manuscripts, and the latter, rather than the former, has been followed 
by us in the printed text. In none of the Atbarvan codices is any 
attempt made to distinguish the jihv&m&liya and upadhmaniya from the 
visarjantya — and, as we cannot but think, with much reason : since the 
division of this indistinct and indefinite sound into three different kinds 
of indefiniteness savors strongly of an over-refinement of analysis. Nor 
do the manuscripts — except in a few sporadic cases, and without any 
agreement among one another as regards these — convert visarjantya into 
a sibilant before a sibilant. In the final revision of the edited text, the 
rule of the Prati§akhya in this respect was begun to be followed in the 
interior of a word (see ii. 3. 3, 5 ; iii. 21. 2; iv. 17. 2), but was soon 
neglected again, and the text in general shows visarjantya before a sibi- 
lant in all situations. The rule that the visarjantya is to be dropped 
altogether before a sibilant followed by a surd mute — a rule which is 
laid down by the Eik and Vaj. Pratigakhyas, and not by our own, but 
which is rather more usually, although with very numerous and irregu- 
larly occurring exceptions, followed in the Atharvan manuscripts — has 
been uniformly carried out in the edition ; although many will doubtless 
be inclined to think with me that, considering the varying usage of 
the manuscripts, it would have been better to follow the authority of 
the Praticakhya, and so to avoid the ambiguity occasionally arising from 
the omission of the final spirant. 

The commentator's illustrative citations are as follows : before guttu- 
rals (kavarge : he prefixes in each case such a specification to his classes 
of examples), antahhofam iva (i. 14. 4) ; before palatals, yag ca dvishan 

ii. 40.] Atharva- Veda Prdtigdkhya. 427 

(i. 19. 4) ; before Unguals, vrJcshash Mate (a fabricated instance : no case 
of this conversion is to be found in AV. : the same example occurs 
under Pan. viii. 3. 34); before dentals, mayas tokebhyah (i. 13. 2), avis 
tok&ni (v. 19. 2), balhik&n vd parastardm (v. 22. 7), yujas tujejandh (vi. 
33. 1), yatha pasas tdyddaram (vi. 72. 2), prd 'vantu nas tvjaye (vii. 49. 
1), trayaslrinpena jagati (viii. 9. 20), makhas tavishyate (xviii. 1. 23) ; 
before labials, tatah pari praj&tena (vi. 89. 1) ; before s, vdlabhrajds sta- 
nayan (i. 12.1: edition jd stanayan, and this time with the assent of all 
the MSS. except E.), arussr&nam (ii. 3. 3 : I. aruksrdnam, and, in verse 5, 
I. and H. do. ; all the others, in both cases, arusrdnam ; pada arvhsrd- 
nam), vivratds sthana (iii. 8. 5 : all the MSS. except H., vivratdh sthana), 
ati durgds srotydh (x. 1.16). Instances for f and sh are not given. 

The Vaj. Pr. (iii. 8, 11) gives as taught by Cakatayana the doctrines 
of our treatise — namely, that visarjaniya becomes a sibilant before a 
sibilant, and jihvdmuliya and upadhmdniya before gutturals and labials 
— while it states (iii. 9) that Qakalya leaves visarjaniya unchanged be- 
fore a sibilant, and itself, not deciding that point, maintains the visar- 
janiya before gutturals and labials. Before palatals and dentals, it pre- 
scribes (iii. 6, 7) the conversion into f and s: of the hypothetical case 
of an initial lingual it takes no notice. Before a sibilant followed by 
a surd mute, it rejects the visarjaniya altogether (iii. 12), as already 
noticed. The Rik Pr. likewise treats at considerable length the changes 
which our treatise compresses into a single rule. It first (iv. 10, r. 31, 
32, ccl, ccli) gives rules which agree in all points with our own, only 
excepting the case of a mute followed by a sibilant, and, later, that of a 
sibilant followed by a surd mute, before which (iv. 12, r. 36, cclv) the 
visarjaniya is to be struck out. But it then goes on (iv. 1 1 , r. 33, 34, 
cclii, ccliii) to permit the retention of the spirant unchanged before gut- 
turals, labials, and unaltered sibilants, and even, finally (iv. 12, r. 38, 
cclvii), to pronounce this the approved usage before gutturals and labials. 
TheTaitt. Pr. also (ix. 2, 3) agrees with our treatise, only excepting ksh, 
before which visarjaniya is to remain unaltered. It then rehearses the 
varying opinions of other authorities : Agnivegya and Valmiki (ix. 4) 
hold that the spirant maintains its identity before gutturals and labials, 
while others (ix. 5) who allow it in these cases to become jihvdmuliya 
and upadhmdniya, leave it unchanged before sibilants — it being speci- 
fied, however, that of this number are not Plakshi and Plakshayana. 
The omission before a sibilant followed by a surd is mentioned (ix. 1) 
as the doctrine of Kandamayana. 

The discordance of opinion among the ancient Hindu grammarians 
as to the treatment of visarjaniya before surd letters is thus shown to 
have been very great, the only point upon which they all agree being 
its conversion into s and f before dentals and palatals — or, more prop- 
erly, the retention of the original sibilant in the former position, and its 
conversion into the nearly related palatal sibilant, by assimilation, in 
the latter. The assimilation to a following sibilant must, it should 
seem, be regarded as a more primitive mode of pronunciation than the 
retention of — or, more properly, conversion into — visarjaniya, which 
latter has become exclusively prevalent in the later language. It is pos- 
sible, too, that the so-called guttural and labial spirants may have had 

428 W. D. Whitney, [ii. 40- 

more of the sibilant character than the mere breathing visarjaniya, and 
so have been intermediate steps between the latter and the original 

41. Before a vowel, visarjaniya becomes y. 

This is an intermediate step to the total disappearance of the spirant, 
by ii. 21, and it is very hard to say whether the conversion into y is a 
matter of grammatical theory only, or whether it gives account of an 
actual process of phonetic transition. The rule is limited by the one 
which follows it, and then farther by ii. 53, so that all which remains of 
it is that vuarjaniya is lost when following a and preceding a vowel, 
and when following a and preceding any other vowel than a. The 
commentator's citations, illustrating these two cases, are yasya upastha 
urv antariksham (vii. 6. 4 : p. yasydh : upa-sthah), m,adhyandina ud 
gdyati (ix. 6. 46), abhipagyata eva (x. 8. 24), and sa dpah (not found in 
A V. : perhaps the reading is corrupt, and va dpah [iii. 13. 7] or na dpah 
[xii. 1. 30] is the passage intended). 

The Taitt. Pr. (ix. 1 0) gives its general rule in a form closely corres- 
ponding with that of our own. The Vaj. Pr. (iv. 36) also makes the 
conversion of the spirant to y before elision, but restricts it formally as 
well as virtually to the case of a visarjaniya preceded by a and d. The 
Kik Pr. (ii. 9, 10, r. 24, 27, cxxviii, cxxxi) follows a peculiar method : it 
assumes no conversion into y, nor does it declare the spirant omitted, 
but teaches that when the latter is preceded by a long vowel, not subject 
to conversion into r, and followed by a vowel, it becomes d; and when 
in like circumstances but preceded by a short vowel, it becomes a. 

JTFTTOtT "^fi: II ^^ ii 

42. If preceded by an alterant vowel, it becomes r before a 

The alterant vowels, as already noted (under ii. 29), are the whole 
series excepting a and d. 

The commentator's citations are agnir dsinah (ix. 7. 19), vdyur ami- 
trdndm (xi. 10. 16), °syd 'Aam manyor avajydm iva (v. 13. 6), and tdir 
amitr&h (v. 21. 8). 

The other treatises (R. Pr. i. 20, r. 76, lxxvii, and iv. 9, r. 27, ccxlvi ; 
V. Pr. iv. 35 ; T. Pr. viii. 6) combine into one this rule and the following. 

There is here another lacuna in the manuscript : immediately upon 
the citation tdir amitrdh follow tasyd agnir vatsah and the other illus- 
trations of the conversion of visarjaniya preceded by an alterant vowel 
into r before a sonant consonant, and then follow the words ghoshavati 
ca before the rule dvah etc. It is evident that the copyist has leaped 
over the rule ghoshavati ca, together with its own paraphrase, the final 
repetition of the preceding rule, and perhaps some of the illustrative 
citations belonging to one or both of them. There is no reason what- 
ever to suppose that anything more than this is omitted, or that any 
rule is lost altogether. 

»• 44-] Atharva- Veda Prdtigdkhya. 429 

43. As also before a sonant consonant. 

The remaining citations of the commentator are tasyA agnir vatsah 
(iv. 39. 2), agner bhAgah stha (x. 5. 7), aratiyor bhrAtrvyasya (x. 6. 1), 
and tdir medino angirasah (x. 6. 20). 

It has been already noticed that the other Pratic&khyas unite this 
rule with the preceding. 

44. Also is changed into r before a vowel or sonant consonant 
the visarjaniya of dvah, kah, akah, ca vi vah, and abibhah — ex- 
cept in the case of a pronoun. 

In this and the following rules, as far as the 49th inclusive, are treated 
the words whose final visarjaniya represents an original r, and not s, 
and in which, accordingly, the r is liable to reappear before a sonant 
initial, even though a or it precede. The Rik Pr. and Vaj. Pr. deal 
with this class of words in a somewhat different manner. The former, 
in the concluding part of its first chapter (i. 20-26) rehearses all the 
words of which the final spirant is rephin or riphita, 'liable to pass into 
r,' and then, in a later chapter (iv. 9), prescribes the conversion into r 
of the riphita visarjaniya before sonants. The V&j. Pr. gives a like list 
(i. 160-168), and a like precept for the alteration (iv. 35). The Taitt. 
Pr., like our own treatise, disposes of the whole matter at once (in viii. 
8-15). The words of the class are quite differently combined in the 
different Pratic&khyas, so that any detailed comparison is impracticable : 
thus, for instance, the words treated in this rule of ours are found scat- 
tered through Rik Pr. i. 21, 22, 23, 26, Vaj. Pr. i. 161, 164, 168, and 
Taitt. Pr. viii. 8, 9. 

The cases to which the rule refers are instanced by the commentator, 
as follows : suruco vena Avah: Avar ity Avah (iv. 1.1), sarasvati tarn iha 
dhAtave kah: kar iti kah (vii. 10. 1), apam nirrtyA akah: akar ity akah 
(ii. 25. 1), satag ca yonim asatag ca vi vah: var iti vah (iv. 1. 1), and 
yam parihastam abibhar aditih putrakAmyA (vi. 81. 3). It will be no- 
ticed that the commentator repeats each word to which the rule applies, 
with iti interposed, except in the last case, where the r appears in san- 
hitA. This is in accordance with the usage of the pada-text of the Rig- 
Veda, but not with that of the Atharvan, which in no single instance* 
performs parih&ra of a word ending in a riphita visarjaniya; and we 
must accordingly regard the repetitions as taken from the krama-text, 
which would give such a form to the words in question, as standing at the 
end of a line. In the case of vi vah, we have the preceding word ca also 
extracted, in order to limit the rule to this particular passage, or, as the 
commentator has it, etAvattv&rtham (compare note to ii. 28). The cases 

* Excepting in the twentieth book, whose pada-text is shown by this and other 
peculiarities to be merely a putting together of extracts from that of the Bik. 

430 W.D. Whitney, [ii. 44- 

intended to be thus excluded must be such as vi vo dhamatu (iii. 2. 2), 
but they are sufficiently provided against by the final specification of 
the rule, " when the word is not a pronoun," and I do not see how the 
citation in the text of both the ca and the vi can escape the charge of 
superfluity. As counter-examples, illustrating the necessity of the final 
specification, the commentator cites hiranyavarnd atrpam yadd vah: va 
iti vah (iii. 18. 6), and yatre 'dam vepaydmi vah (iii. 13. 7) : he does not 
choose to notice the fact that these cases are also excluded by their 
lacking a preceding ca vi; and he gives no instances of hah as a pro- 
noun, as it was his duty to do. All the other treatises distinguish the 
hah whose final is riphita by calling it anuddtta, ' unaccented,' which 
compels them then farther to specify the cases in which the verbal form 
hah (tear) happens to be accented. 

The term sarvandman, ' all-name,' used to denote a pronoun, is an 
ingenious and interesting one; it is not found in either of the other 
treatises, but is employed by Panini. Ndman, ' name,' includes sub- 
stantives, adjectives, and pronouns; but while the two former, being 
descriptive of quality, are restricted in their application to certain ob- 
jects or classes of objects, a pronoun may be used of anything indiffer- 
ently ; it is a title of universal applicability. 

ilofl^iH il 8HJI 

45. Also that of dvdh and vdh. 

The commentator cites prathamA dvdh: dvdr iti dvdh (ix. 3. 22), tas- 
mdd vdr ndma (iii. 13. 3), and divyam ghrtam vdh: vdr iti vdh (xviii. 
1. 32) ; repeating, as under the previous rnle, the final words of the 
half-verses, as they would be repeated in the Xranta-text. 

*M«fclH(«fci: II 3^ ii 

46. Also that of ahdh, except it be from the root hd. 

An equivalent and, one would think, preferable form for this rule 
would have been harater ahdh, l ahdh when coming from the root har 

Shr).' The commentator's examples are indras tdn pary ahdr ddmnd 
vi. 103. 2, 3), iha rdshtram d 'hdh (xiii. 1. 4 : the commentator, or the 
copyist, omits to add ahdr ity ahdh), and agnish tad d 'hdh (vii. 53. 3 : 
here is added ahdr ity ahdh, but it is out of place, the word not stand- 
ing in pausa ; perhaps the parihdra has slipped away from its proper 
place after the preceding citation to this : but then the word following 
ahdh should also have been quoted in the last passage, and it should 
read agnish tad d 'hdr nirrteh). As counter-example, to show that 
ahdh from the root hd forms no exception to the general rule respecting 
a visarjantya, the commentator cites ahd ardtim (ii. 10. 7). 

^fTFTfWT (ItecMHMHl ll Soil 

47. Also that of the vocative singular of a noun whose dual 
ends in rdu. 

ii. 50.] Atharva- Veda Prdticdhhya. 431 

The commentator first cites three phrases, of which the first two are 
not to be found in the Atharvan, and the third belongs under the next 
following rule — viz. dhdtar dehi, savitar dehi, punar dehi (xviii. 3. 70 : 
but possibly the three phrases form a single passage together, and are 
a genuine citation from some other text) — and then adds two genuine 
and appropriate instances : bkume mdtar ni dkeki (xii. 1. 63), and tvacam 
etam vifastah: vifaslar iti vi-pastah (ix. 5. 4). He goes on to give 
counter-examples : first, to show that the conversion into r takes place 
only in a singular vocative, he cites a passage — ddivyd koldra urdkvam 
(v. 27. 9) — containing a plural vocative from a theme of the same char- 
acter ; and second, to show that a vocative of such a theme only is 
treated in the manner described, he cites the vocative singular of a 
theme in as, which has sdu and not rdu in the dual, viz. tavishasya pra~ 
cetah: praceta itipracetah (iv. 32. 5) ; adding pracetah s&udvivacan&ntam. 

Here, as in more than one other instance, our treatise shows a greater 
readiness than the others to avail itself of the help of grammatical cate- 
gories in constructing its rules : all the other Prati<jakhyas laboriously 
rehearse in detail, one by one, the words which are here disposed of as 
a class, in one brief rule. 

5PFT: J5T: OTtP H^rT: ^oMMMWNi^h 

48. Also that of antah, punah, prdtah, sanutah, and svah, 
when indeclinables. 

The final specification is intended only for the first and last words 
of the series, antah being possible as nominative singular of aula, and 
svah of sva. The commentator illustrates as follows: antarddve juhutd 
(vi. 32. 1), punar mdi " Hv indriyam (vii. 67. 1), prdtar bhagam p&sha- 
nam (Hi. 16. 1), sanular yuyotu (vii. 92. 1), svar no , pa tod (ii. 5. 2). As 
counter-examples, to show the necessity of the specification "when 
indeclinables," he gives yo nah svo yo aranah sajdtah (i. 19. 3), samagrah 
samantah: samagra iti sam-agrah: samanto bh&ydsam : samanta iti 
sam-antak (vii. 81.4: the commentator thus gives the krama-text for 
the passage samagrah samanto bh&ydsam). 

The other treatises exclude the noun dnta by defining the accent of 
antah, and the Rik Pr. treats svah in the same way — a method which 
renders necessary considerable additional limitation and explanation. 

terror H3N 

49. And that of svah, also in svarshdh. 

The commentator cites the passage, pilsham agriyah svarshdh (v. 2. 8). 
The reason of the word ca in the rule, he says, is that the following 
letter is a surd ; and he adds that the spirant becomes r only in sanhitd, 
the pada form being svaty-sah: this last is rather a gratuitous piece of 

^ H Ho li 

50. Also that of ahah, when neuter. 
vol. vn. 55 

4m W. D. Whitney, pL W- 

Tbe examples quoted from the text are yad aharahar abhigaehdmi 
(xvi. 7. 11), and ahar m& Hy aplparah (xvii. 25). As counter-examples, 
to show that the final is liable to become r only when the word is neu- 
ter, we have samaho vartote (not in AV.), and dv&dap&ho 'pi (xi. 7. 12). 

The passages in which the other Prati§akhyas take note of this irreg- 
ular change of ahas are Rik Pr. i. 26 (r. 103, civ), Vaj. Pr. i. 163, and 
Taitt. Pr. viii. 8, 13. 

The next rule furnishes exceptions to this one. 

jt f^p#^qr^flr^trrtg uru» 

51. But not before a case-ending, or the words riipa, rdtri, 
and rathamtara. 

As examples of ahah before case^endings, the commentator gives us 
ahobhy&m and ahobhih, but the latter of tbem (xviii. 1. 55 : p. ahah-bhih) 
is the only instance of the kind which tbe Atharvan text contains. For 
the compound ahor&tre he cites two cases, ahor&tr&bhy&m nakshatre- 
bhyah (vi. 128. 3), and ahor&tre idam brkrnah (xi. 6. 5) : it is a word of 
frequent occurrence. For the combinations of ahas with following r&pa 
and rathamtara are quoted yad aho rupani drpyante, and yad& 'ho ra- 
tkamtaram s&ma glyate r neither of which passages is to be found in the 
Atharvan. It is a very suspicious circumstance that a v&rttika to a 
rule of Panini's (viii. 2. 68) mentions the same three exceptions which 
our rule gives : and it is very probable that our treatise in this case, as 
in several others, has constructed its rule so as to include all the cases 
noted as occurring in general usage ; and hence, that the two phrases 
quoted are not necessarily to be regarded as having constituted a part 
of the Atharvan text for which the Praticakhya was composed. 

The Rik Pr. (iv. 13, r. 40 r cclix) makes exceptions only of ahobhih 
and ahor&tre; the Vaj. Pr. (i. 163) excepts cases in which ahah is followed 
by bh; the Taitt. Pr. (viii. 13, 14) teaches the conversion of the final 
of ahah into r when it is not the final member of a compound, nor fol- 
lowed by bhih or bhy&m. 

52. Nor is the visarjaniya of tidhah, amnah, and hhuvah con- 
vertible into r. 

This rule is utterly idle in our treatise, since no precept has been 
given which should in any way require or authorize the conversion into 
r of the final of these words. The original form of tidhas, however, is- 
Adhar, as is clearly shown by the comparison of the kindred languages- 
(o59-«9, Enter, udder), and by its treatment in the Rig- Veda ; and the 
Rik Pr., accordingly (i. 22, r. 97, 98, xcviii, xcix) T has to give rules re- 
specting it. Neither of the other treatises takes notice of it or of either 
of the words here associated with it. All three, however, are noted by 
Panini (viii. 2. 70, 71), as words which may or may not, in Vedic use, 
change their final into r ; and the instance there cited for bhuvah, bhuvo 
vipveshu bhuvaneshu, looks as if it were meant to be the same which 

ii. 55.] Atharva- Veda Prdtigdkkya. 433 

our commentator gives, viz. bhuvo vipveshu savaneshu yajniyah, and 
which is not to be found in the Atharvan. The commentator farther 
cites for bhuvah an actual Atharvan passage, bhuvo vivasvdn anvatatdna 
(xviii. 2. 32) ; but even here it would hardly be necessary to understand 
the word as having the same meaning as in the mystic trio bhur bhuvah 
svar, with which the later religious philosophy amuses itself. For frdhah 
and amnah, the examples quoted are yo asyd -udho na veda (xii. 4. 18), 
and ye amnojdtdn marayanti (viiL 6. 19 : the only case). 

^RT^tqwt^Tft ^f^ ii H3 h 

53. When preceded by a, visarjaniya becomes w, before a fol- 
lowing a. 

And this «, of course, combines with the preceding a, so that the ok 
becomes o, as is directly taught by the other treatises (R. Pr. ii. 12, r. 
33, cxxxvii ; V. Pr. iv. 42 ; T. Pr. ix. 7). The instances cited are paro 
*pe 'Ay asamrddke (v. 7. 7), and paro 'pe 'hi tnanasp&pa (vL 45. 1). For 
the treatment of the following initial a, see iii. 53, 54. 

This rule is much mutilated by the copyist, both in its first statement 
(akdro '/care) and in its final repetition (akdropadhasyokdre), so that its 
true form is only restorable from the commentator's paraphrase, which 
reads akdropadkasya visarjaniyasyo 'kdro bhavaii: akdre paratah. Evi- 
dently the triple recurrence of the syllables kdro, kdro, kdre bothered 
the copyist's weak head, and he stumbled from the one to the other of 
them in an utterly helpless manner. 

mI^Ih sthhSii 

54. As also before a sonant consonant 

The commentator's illustrations are tathd saptarshayo viduh (iv. 11. 9), 
tad vdi brahmavido viduh (x. 8. 43), and tasydm naro vapata bijam 
asydm (xiv. 2. 14). 

The corresponding rules of the other treatises are Rik Pr. iv. 8 (r. 25, 
ccxliv), Vaj. Pr. iv. 41, and Taitt. Pr. ix. 8. 

55. When preceded by d, it is dropped. 

That is to say, it is so treated in the position defined in the last rule, 
before a sonant consonant; the case of dh before a vowel was included 
in ii. 41, above. The commentator cites anamivd vivakshavah (ii. 30. 3), 
dhird deveshu (iii. 17. 1), and ekapatam t&janatd yd bhumih'{y. 18. 12), 
which last passage contains three cases of the application of the rule. 

The corresponding rule of the Vaj. Pr. is found at iv. 37, that of the 
Taitt. Pr. at ix. 9. The Rik Pr. (iv. 8, r. 24, ccxliii) teaches that the 
visarjaniya (along with its preceding vowel, of course) in such a case 
passes into d — which is a peculiar way of saying the same thing. 

484 W.D.Whitney, [ii.68- 

56. As also in qepaharshanim and vandane 'va vrhsham. 

The commentator cites the two passages : oshadhim fepaharshantm 
(iv. 4. 1 : p. pepah-harshanim) and vandane 'va vrkshom (vii. 1 1 5. 2 : p. 
vaiidanah-iva). The former is a striking case of arbitrariness in ety- 
mologizing on the part of the authors of the pada-text, for there is 
neither necessity nor plausibility in treating the compound as if made np 
of fepas and harshana : the former member is evidently fepa, which in 
the Atharvan is much the more common of the two forms of this word. 
And as for the other case, of elision of the visarjantya before iva and 
contraction of the two vowels into a diphthong, it is equally surprising 
to find this one singled out to be so written, from among the many 
in the text which are to be so read. For the contraction of a final syl- 
lable, ending either in an original * or m, with the following particle of 
comparison iva, so that the two together form but two syllables, is the 
rule rather than the exception in the Atharva-Veda. Out of 59 in- 
stances in the text, in which a final ah occurs before iva, there are only 
13 in which the metre shows the sandhi to be regular: in 46 cases we 
are to read e 'va ; am iva is contracted in the same manner 25 times 
out of 40 ; Ah iva, only 4 times out of 19 ; im iva, 3 times out of 5 ; 
ih iva, 7 times out of 10 ; uh iva, 6 times, or in every instance the text 
contains ; um iva, only once out of 3 times : and there are single spo- 
radic cases of a similar elision after the terminations eh, an, &n, ad, ud, 
•which would pass without notice, as mere irregularities of metre, were 
it not for their analogy with the others I have mentioned, but which, 
considering these latter, are worth adverting to, as illustrations of the 
same general tendency. 

57. As also in eshah and sah, before a consonant. 

The instances selected by the commentator are esha priyah (ii. 36. 4), 
and sa sendm mohayatu (iii. 1.1). 

All the other treatises (R. Pr. ii. 4, r. 11, 12, cxv, cxvi; V. Pr. iii. 15, 
16 ; T. Pr. v. 15) include syah in the same rules with eshah and sah: 
the word, however, nowhere occurs in the Atharvan text. 

^ FTFT^te II \* II 

58. But not in the passage sas padishta. 

The passage, adharah sas padishta (vii. 31. 1), is cited by the com- 
mentator. It is one which occurs in the Rig-Veda also (iii. 53. 21), and 
is noticed by its Praticakhya (iv. 20, r. 58, cclxxvii). 

59. Visarjaniya is also dropped in dirgh&yutv&ya etc. 

"•■81.] Aiharva-Veda Prdticdkhya. 485 

Besides the word dtrghdyutvdya (e. g. i. 22. 2 : p. dirghdyu-tvdya), 
which is of frequent occurrence in the Atharvan, the commentary cites 
atho sahasracaksho tvam (iv. 20. 5 : p. sahasracalcsho iti sahasra-caksho), 
and barhishadah (xviii. 1. 45, 51 : p. barhi-sadah). In all these eases, it 
■will be noticed, the lopa of the visarjaniya is made in the pada-text, as 
well as in sanhitd, as is directed in a later rule (iv. 100). In the two 
first, there is an actual substitution of the themes dyu and cakshu for 
the usual dyus and cakshus ; in the other we are to assume, rather, that 
the final of barhis is lost in making the sandhi, and we have a right to 
be surprised that the ^)arfa-text does not give us barhih-sadah. 

Here ends the second section of the second chapter : the signature 
is : 59 : dviliyasya dvittyah pddah. 

60. The visarjaniya of duh becomes u before ddga, and the 
following dental is replaced by a lingual. 

That is to say, duh becomes du, and ddpa, ddca, making duddpa. 
The passage containing the word is cited by the commentator, as follows : 
yend duddpe asyasi (i. 13. 1) ; and he adds sanhitdy&m ily eva: durdd- 
pa iti duh-ddpe; this is the form the word would assume in the krama- 
text : the pada reads simply duh-ddpe. The theme of the word is 
rather duddp than duddpa* The commentator says farther : apara 
dha: dhandpaddpadabhayeshu l&pyeta upadhdydp ca dirghah tavargt- 
yasya tnvarglya iti : d&ddpah: dudabhah: dudhyah: ava dudhyo jahi ; 
'another says, before dhd, ndpa, ddpa, and dabha, the visarjaniya is 
dropped, the preceding vowel is lengthened, and a lingual substituted 
for a dental : thus, duddpah, [dundpah], dudabhah, dudhyah ; ava dii- 
dhyo jahi (Rig- Veda i. 94. 9).' Of these words, only the one men- 
tioned in our rule is found in the Atharvan ; the others occur in the 
Rik, and are the subject of a rule of its Praticakhya (v. 24, r. 55, 
ccclxxi). A part of them are also noted by the Vaj. Pr. (iii. 41, 42). 
The Atharvan has duh-napa (v. 11. 6), but treats it according to the 
regular methods of combination, making durnapa in sanhita. 

Sjft rT^: II \\ II 

61. Before gun, it becomes t. 

This is intended merely as an explanation of the mode of formation 
of the word ducchund, which is accordingly to be regarded as passing 
through the stages duh-pund, dut-fund, and hence (by ii. 13, 17) duc- 
chund. Two instances of its occurrence are cited : duchundm grdmam 
(v. 17. 4), and tad rdshtram hanti duchund (v. 19. 8). As the pada-text 
does not analyze the word, but writes it simply duchund, this rule is 
properly superfluous, and we have a right to wonder that it was intro- 

* It is quite probable, however, that dd(a in the rule means, not * the theme 
ddf a,' but * the root ddf.' 

436 W. D. Whitney, [ii. 61- 

duced into our treatise, rather than, with Weber (p. 187, 188) to be 
surprised that nothing like it is found in the Vaj. Pr. 

62. In a compound, it becomes s before k and p, excepting in 
the case of antah, sadyah, qreyah, and chandah. 

The citations given in illustration of the rule are adhaspadam (e, g. 
ii. 7.2: p. adhah-padam), pibasphdkam (iv. 7. 3 : p. ptbah-ph&kam), na- 
maxkdrena (iv. 39. 9 : p. namahkdrena), and yo vifvataspdnir ufa vifva- 
ta&prthah (xiii. 2. 26 : p. vifvatah p&nih ; vifvatah-prthah). With re- 
gard to the second of these, I would remark that its treatment by the 
pada-text, and its citation under this rule, seem to depend upon a false 
etymology, inasmuch as its final member is plainly not phaka, but 
sphdka, a word allied with sphdna and sphdli, repeatedly met with else- 
where : in the combination pibah-sphdka, then, the final of the first 
member has disappeared, according to the rule of the Eik Pr. and Vaj. 
Pr. (spoken of above, under ii. 40) for the loss of a visarjaniya before 
a sibilant followed by a surd mute. The commentator then cites in full 
the passages illustrating the exceptions, viz. antahkofam iva (i. 14. 4), 
antahpdlre rerikattm (xi. 9. 15), sadyahkrih (xi. 7. 10), preyahketah (v. 
20. 10), and chandahpakshe (viii. 9. 12). Next follows what has the 
appearance of being another rule, with its commentary ; parahparah : 
para iti ca "mreditasamase na sakaro bkavati : tvat parahparah ; 'parah- 
parah : the visarjantya otparah also does not become » in a compound 
of repetition: e.g. tvat parahparah (xii. 3. 39).' This is not, however, 
regarded by the commentator as a rule, since after it he gives, as final 

repetition before the next rule, our rule 62, sam&se chandas&m. It 

is also evidently not to be reckoned as a rule, on account of its inter- 
rupting the anuvrlti, or implication by inference from the preceding 
rule, of visarjanlyah sakaro bhavati, which continues to the end of the 
section. It must be looked upon as a vdrttika, or secondary limitation 
by another hand, of the rule under which it is introduced. Eespecting 
the propriety of its introduction, moreover, there may be question. 
The only other case of the kind occurring in the text is parushparuh 
(e. g. i. 12. 3), which, as we shall see, is cited by the commentator under 
rule 80, below : as it ought not to be if it belongs under our present 
rule. It may be, then, that such cases of repeated words are not re- 
garded as distinctly enough compounded to be entitled to the unquali- 
fied name samdsa, ' compound.'* They seem, however, to be treated 
as regular compounds by the other Pratigakhyas, and our own pada- 
text makes no distinction between them and the rest, writing paruh- 
-paruh precisely like vdta-jdh, for example. 

Our manuscripts are not consistent or unanimous in their mode of 
treatment of the class of compounds to which this rule relates, one or 
more of them, in several instances, retaining the final visarjaniya of the 

* The same thing appears from their separate mention in the fourth chapter (iv. 
40), as if not included among regular compounds (iv. 9). 

ii.64.] Atharva-Veda Prdtigdkhya. 437 

first member, instead of changing it into the sibilant : in a single case 
(dyuhprataranah, iv. 10. 4), they all agree in so doing: we have, how- 
ever, regarded this coincidence as accidental, and adopted in the pub- 
lished text the reading which the Praticakhya requires. 

Rule ii. 81, below, directs the conversion into sh, after an alterant 
vowel, of the dental sibilants prescribed by this and the following rules 
of the section. 

The other treatises (R. Pr. iv. 14, r. 41, cclx; V. Pr. iii. 29 ; T. Pr. 
viii. 23) give the same general precept for the conversion of visarjaniya 
into a sibilant before k and p, and detail the exceptions in following 

ftf^rfb|4f^HIH vrjqr H # H 

63. The visarjaniya of nih, duh, dvih, and havih becomes s 
before k and p, even elsewhere than in a compound. 

The particle dvh never appears in the Atharvan text except as com- 
pounded with other words, but it would seem from this rule that the 
compounds which it forms are not entitled to the name sam&sa. The 
commentator cites as examples dushkrlam (e. g. iv. 25. 4) and dushpi- 
tam (not found in AV. ; take instead dmhpratigraha [x. 10. 28], the 
only like case which the text affords) : both are also given in illustra- 
tion of the corresponding rule of Panini (viii. 3. 41). Dushkrt and 
dushkrta are the only words in our text in which duh is followed by k. 
The preposition nih becomes nish before & only, and, excepting a single 
case (nishkritah, ii. 34. 1), only in verbal forms and derivatives of the 
root kar : the commentator cites one of them, viz. kushthas tat sarvarh 
nish karat (v. 4. 10). But it occurs three times before k and p in the 
accidental combinations of the phrase (viz. nih kravy&dam, xii. 2. 16 
[B. I. nish kr°\ ; do. xii. 2. 42 ; nih prthivyah, xvi. 7. 6), and the manu- 
scripts almost without exception read there nih, which has been adopted 
in the edition as undoubtedly the true reading. Avih occurs only in 
connection with forms of the root kar (sometimes not immediately com- 
bined with them) : the commentator instances dvish krnushva rupani 
(e. g. iv. 20. 5), and Avish krnute rupani (not found in AV.). Havih be- 
comes havish but once, in the passage havish krnvantah parivatsarinam 
(iii. 10. 5), which the commentator cites : it evidently forms here a kind 
of half-compound with kar. For this word the commentator is himself 
obliged to instance an exception, as follows : tato 'pa vadati : havih pu- 
roddpam ; ' from this is to be made the exception havih puroddpam 
(xviii. 4. 2).' The rule is evidently not entitled to much credit as a 
clear and complete statement of the phenomena with which it is in- 
tended to deal. 

Hn ii M* ii 

64. As is also that of trih. 

The only case in the text coming under the rule is apdldm indra trish 
putvd (xiv. 1. 41) : the commentator cites it, and adds also trish kratvd, 

438 W. D. Whitney, pL 64- 

trish prakdrena, not found in the Atharva-Veda. The Eik Pr. treats 
the same case in iv. 24 (r. 63, cclxxxii). 

65. As also a visarjaniya, except that of harnayoh, before 
kuru, karam, karat, Irriolu, krti, and krdhi. 

The cases to which this rule is meant to apply are of very different 
frequency of occurrence, and the rule itself is of very different degrees 
of accuracy as concerns the forms mentioned. For kuru, the comment- 
ator cites pitrbhyap ca namas kuru (xiv. 2. 20), the only instance in 
•which this rare form (found elsewhere only at xi. 9. 1, and its repetitions 
in the same hymn) is preceded in the pada-text by visarjaniya. For 
karam we have cited subuddhdm amulas karam (xiv. 1.18): also the 
only instance. For karat, the instance sammanasas karat (vii. 94. 1) is 
given : the text also affords jangidas karat (xix. 34. 2, 1 ; 35. 5), and, 
besides this, only nish karat (e. g. ii. 9. 5). Under krnotu, the com- 
mentator remarks krnotv ity aira tridhdtve Hi vaktavyam, ' under krnotu 
it should have been noted that the change occurs only in three cases :' 
and he proceeds to cite the three cases, viz. dirgliam dyush krnotu (vi. 
78. 3), aguish krnotu bheshajam (vi. 106. 3), and manish krnotu devajah 
(x. 6. 31). In all these passages, however, the manuscripts read, with- 
out dissent, visarjaniya before the k, and the edition has followed their 
authority (except in vi. 106. 3, where, by some oversight, sh has been 
introduced). Other instances in which krnotu has an unaltered visar- 
janiya before it are tolerably numerous,* and nowhere among them does 
a single codex read s or sh. The next word, krti, is one which, for a 
double reason, has no right to a place in the rule : in the first place, it 
occurs nowhere except in compounds ; and secondly, it converts into a 
sibilant only the preceding visarjaniya of nih and havih, and so would be 
sufficiently provided for by rule 63, even if not adjudged to fall under 
rule 62. The commentator cites for it tasya tvam asi nishkrtih (v. 5. 4). 
Finally, krdhi actually shows a marked tendency to preserve the original 
final sibilant of the preceding word, and does it in numerous instances :f 
the commentator cites two of them, viz. anamitram puras krdhi (vi. 40. 
3), and se ''mam, nish krdhi purusham (v. 5. 4 : this belongs rather under 
rule 63) ; the only exception is that noticed in the rule, and cited by 
the commentator, viz. mithunam karnayoh krdhi (vi. 141. 2). Besides 
these, we have, of forms from the root kar, tiras karah (iv. 20. 1), which 
the commentator cites under rule 80, below, but which has as good a 
right to be specially noted in this rule as one or two other of the cases 
the latter contains. Not much can be said in praise of the way in 
which the rule is constructed. 

The other Praticakhyas (R. Pr. iv. 15, r. 43, eclxii ; V. Pr. iii. 22, 28, 
31 ; T. Pr. viii. 25, 26) treat these cases in nearly the same way : whether 
with no greater accuracy, I cannot say. 

» They are vi.40.1,2; 53.3; *73. 3 ; 83.1; 104.3. vii. 32. 1 ; 33.1; 51.1; 91. 
1. viii. 8. 4. ix.2. 7,4. 2. xi. 1. 6. xii. 1.1,48. 

f They are i. 13. 2; 26.4; 84.1. T. 6. 4 (H. I. hlc), vi.40. 3; 14.3; 136.2. vii. 
20. 2 ; 26. 3. viii. 6. 17. six- 8. 6. 

ii. 68.] Atharva- Veda Prdticdkhya. 439 

FTrlFTft sH^lim* 

66. Also the visarfaniya of tatah before pari, when the latter 
is followed by brahma. 

The case referred to, and cited by the commentator, is tatas pari brdh- 
man& (i. 10. 1). As counterexample, to show that it is only before 
brahma that the change takes place, is quoted tatah pari prajatena (vi. 
89. 1) ; one other such case is afforded by the text, but in the nineteenth 
hook (xix. 57. 6). A similar case, in which the suffix lah becomes las 
before pari, is sindhutas pari (iv. 10. 4 and vii. 45/1) ; the Pratisakhya 
takes no notice of it, if it be not intended to be included in the gana 
with which the section concludes (ii. 80). 

r:q>Tl^o|sfH ll fya II 

67. Also the visarjaniya of an ablative before pttri t except in' 
ahgebhyah pari etc. 

The Rik Pr. (iv. 1 5) and "Vaj. Pr. (iii. 30) give the additional specifi- 
cation that the pari must be at the end of Apoda^ or before a pause,' 
and the addition of that restriction to our own rule would have made 
it accurate, and obviated the necessity of the gaita. The instances and 
counter-instances are both tolerably numerous, and are detailed in a 
marginal note.* The commentator cites, as instances,, kushtho himavatas 
pari (xix. 39. 1), jatam himavatas pari (iv. 9. 9 and v. 4. 2), and vidyuto 
jyotishas pari (iv. 10. 1); and, as counter-instance,- the passage heading 
the i/ani, pr&ttim angebhyah pary acardntam (ii. 34. 5). The instance 
first cited is the only one anywhere to be found, in the text or in the 
commentary of our treatise, which seems to come from the nineteenth 
book of the present Atharvan text 5 but I do not regard its occurrence 
as by any means to be taken for an evidence that that book constituted 
a part of the commentator's Atharva- Veda ; it is to be explained rather 
as a slip of his memory in quoting, or as a reading of one of his kushtha 
hymns, differing from that of our text. 

f^TFjfef tWirMsfH II ^11 

68. Also that of divah before prihivi, except when the latter 
is followed by the root sac. 

The commentator cites two of the passages to which the rule relates, 
viz. divas prthivyA antariksh&t samudr&t (ix. I. T), and divas prthivim 
abhi ye srjanti (iv. 21. 4). The text affords three others, viz. vL 100. 3 ; 

* Cases of an ablative in « before pari at the end of a pdda: i. 10. 4". iv. 9. 9; 
10.1; 15.10. v,4. 2. vi. 111. 3. vii. 38. 2 ; 63. 7. viii. 4. 6. x.1.32. xi. 2.4. xiv. 
2.44. xix. 39. 1,5. 

Cases of an ablative iti /f before pari not at the end of a pdda : ii. 34. 5. vi. 125. 
2 (bit), vii, 56. 1. viiii 9 1 . 4C ix.4. 10. x. 7. 25. xiii-. 1.26, xix. 8. 1 ; 35. 4 (bit) ; 44. 
8,9; 48.5; 

vol. vrr. 5©- 

HO W. B. Whitney,. [ii. «8- 

125.2. xix. & I. The only counter-instance is that mentioned in the 
rule, and given in full by the- commentator, via. A yanti divah prthivtm 
sacante (xii. 3. 26)i 

The Rik Pr. (iv. 20, r. 57, ccfasvi) adds the restriction that the con- 
version into the sibilant occurs only when divah begins a p&da. This 
would be correct also, if made a* part of our own rule. The other two 
treatises do not apparently note any such cases as occurring in their texts. 

^ ^ It U H 

69. As also before prshtha. 

The commentator cites a single case, divas pfshthe dhavam&nam su- 
parnam (xiii. 2. 37). The others are iv. 14. 2. xL 5. 10. xii. 2. 12. xiii. 
4. 1. xviii. 1.61. To show that only divah is subject to this change of 
its final before prshtha, the commentator instances bhumydh pfshthe 
vada (v. 20. 6). 

The original sibilant of divas also maintains itself before p in two 
other cases, viz. divas putrau (viii. 7. 20) and divas payah (xix. 44. 5). 
With the latter our treatise has nothing to do : the former it lets fall, 
apparently, into the barathrum of the gana which closes the subject and 
the section (rule 80). 

Compare Rik Pr. iv. 22 (r. 61, cclxxx), and Vaj. Pr. iii. 23. 

«T: qfft ilMlHWI^^y^iitoou 

70. Also that of yah before paH, except when gavdm and 
asy&h precede. 

Nearly all the passages of the text to which the rule relates are cited 
by the commentator; they are divyo gandharvo bhuvanasya yas patih 
(ii. 2. 1), mrdad gandharvo bhuvanasya yas patih (ii. 2. 2), and dhtld 
vidhata bhuvanasya yas patty (v. 3. 9) : to be added are only bhavithyttd 
bhuvanasya yas patih (xiii. 3. 7), and yani caMra bhuvanasya yas patih 
(xix. 20. 2). The exceptions are also cited, as follows : gav&m yah patih 
(ix. 4. 17), and dirghayur asyd yah patih (xiv. 2. 2). One cannot but 
think that a better form for the closing restriction of the rule would 
have been " whenever bhuvanasya precedes." 

These cases are not distinguished by the Rik Pr. and Vaj. Pr. from 
those coming under our next rule. 

71. Also the visarjarAya of a genitive, excepting that of yxcydh, 
becomes s before pad. 

The «ases coming under this rule are almost innumerable, and it would 
be a waste of labor to specify them in full. The commentator cites four, 
Viz. vacas patih (e. g i. 1. 1), ushas patih (xvi. 6.6), brahmanas patih 
(e. g. vi. 4. 1), and jagatas patih (vii. 17. 1). The only exception is that 

ii. 74.] Atharva- Veda Prdticdhhya. 441 

■which the rule mentions; the commentator cites it: ar&tydh facy&h 
patih (xiii. 4. 47). 

The other treatises bring together all the cases of conversion into a 
sibilant before poti, and give their rules a more empirical form (see R. 
Pr. iv. 15, r. 42, cclxi ; V. Pr. iii. 34 ; T. Pr. viii. 27). 

72. Also that of iddydh before pada. 

The passage is idayds padam (iii. 10. 6). The commentator gives a 
counter-instance — hastinydh padena (vi. 70. 2) — to show that only the 
genitive iddydh undergoes this change before pada. 

That the rule is not so framed as to include also the closely analogous 
case idas pade (vi. 63. 4), the only other one in the text where an origi- 
nal final sibilant is preserved before pada, gives reason to suspect that 
the verse containing it was not in the commentator's Atharvau : a sus- 
picion which is supported by the peculiar mode of occurrence of the 
verse, at the end of a hymn with the subject of which it has nothing to 
do. It looks as if it had been thrust in at the end of vi. 63, because in 
the Rik text (x. 191) it preceded the verses of which vi. 64 is composed. 

Compare Rik Pr. iv. 17 (r. 49, cclxviii), and Vaj. Pr. iii. 22, which 
note both the cases. 

fag: fcrrf^ II ©^ II 

73. Also that of pituh before pitar.. 

The commentator cites yas tdni veda sa pitush pild 'sat (ii. 1. 2), and 
a similar phrase is found once more in the text (at. ix. 9. 15). On the 
other hand, we have three cases in the eighteenth book (xviii. 2. 49 ; 3. 
46, 59) in which pituh stands before pitarah, and, by the unanimous 
authority of the manuscripts, maintains its visarjaniya. We must sup- 
pose either that the Praticakhya and its commentator overlooked these 
passages, or that they did not stand in the text contemplated by them, 
or that they stood there with the reading pitush pitarah. 

The Rik Pr. notes a similar case at iv. 25 (r. 64, cclxxxiii). 

^Pter u ©3 n 

74. As also that of dyduh, in the same situation. 

The commentator instances dydush pilar ydvaya (vi. 4. 3*), and 

* The manuscripts vary somewhat in their reading of these words, both as re- 
'gards the sibilant and the accentuation : only I. gives the nh which the 1'raticakhya 
requires; the rest have h. Bp. and I. accent di/d'mh, the others di(^du«/i. Theire 
can lie no question that the latter is the true method : the word must be divided 
into two syllables in reading, and the vocative accent of an initial syllable belongs 
only to the former of the two : thus, di • dush ; which, when written together as a 
single syllable, becomes, of course, dt^dush. One or two other like cases are found 
in the text. 

442 W. D. Whitney, [ii. 74- 

dyaushpitar nyafin adhar&n (not found in AV.). At iii. 9. 1, -where 
the two words occur again in connection, all the MSS. excepting I. and 
H. give dydush, and the edition reads the same. At iii. 23. 6, every 
eodex presents dyduh, which was accordingly received into the edited 
text : and the same reading was adopted at viii. 7. 2, where the same 
line occurs again, although two of the MSS. (W. and I.) there have 
dydush. It might have been better, considering the discordance of the 
manuscripts, and the consequent inferior weight of their authority, to 
allow the Pratigakhya to determine the reading of all' these passages. 

Like cases occurring in the Eik and Vajasaneyi are treated in their 
Praticakhyas (R. Pr. iv. 23, 24 ; V. Pr. iii. 33). 

5TFT: $m It ^»\ II 

75. Also that of dyuh before prathama. 

The passage is cited by the commentator: dyush prathamam prajdm 
posham (iv. 39. 2, 4, 6, 8), but, in every case of its occurrence, the manu- 
scripts read without dissent dyuh. The edition, however, has in this 
instance followed the authority of the Pratigakhya, 


76. As also before pra, when followed by the roots mush or 


The two passages to which the rule relates are md na dyush pra mo- 
shih (viii. 2. 1 7), and dire/ham dyush pra jtvase (xviii. 2. 3). In both 
passages., however, all the manuscripts read dyuh — as does our own 
commentary, in citing them — and the printed text does the same. As 
counter-example, the commentator brings forward jivdndm dyuh pra tira 
(xii. 2. 45) ; other cases are viii. 2, 2 (where P. reads dyush both times) 
and xix. 32. 3. 

vftjfv Mrlldl II oo || 

77. Also that of paridhih before patdti. 

The passage is yathd so asya paridhish patdti (v. 29. 2, 3), and the 
manuscripts are unanimous in supporting the reading prescribed by the 

Ptelrl^Ullrtl II os It 

78. Also that of nivatah before prndti. 

That is, in the passage yd vigvd nivatas pfn&ti (vi. 22. 3), as is read 
by all the MSS. 


79. Also that of manah before pdpa. 

ii. 81.] Atharva-Veda, Prdtigakhya. 443 

The commentary cites the passage : 'pe 'hi < (vi. 45. 
I). The pada-text regards manasp&pa as a compound, writing it ma- 
nah-pdpa. Its separate mention by the Praticftkhya would seem to in- 
dicate that the latter regards the two words as independent; since, as a 
compound, it would fall under rule 62, and would need no special 
notice. The accentuation does not help to settle the question, and the 
sense is nearly as good one way as the other. 

'(IMHltall^J Alison 

80. Also that in rdyasposka etc. 

Under this mle, the commentator instances only three cases, viz. 
rdyas posham (e. g. i. 9. 4), parush paruh (e. g. i. 12. 3), and m& pifdcam 
tiros karah (iv. 20. 7) : the latter has been already remarked upon, as 
not less entitled to be included in rule 65 than some of the cases to 
which that rule relates. There are to be found in the text quite a num- 
ber of other like cases, not provided for in any of the foregoing rules ; 
they are as follows: divas putrdu (viii. 7. 20) and divas payah (xix. 44. 
5), already spoken of under rule 69 ; idas pade (vi. 63. 4), referred to 
under rule 72 ; adhamas padishta (viii. 4. 1 6 : see R. Pr. iv. 20) ; dvisha- 
tas pddaydmi (xi. 1. 12, 21) ; dpas putrdsah (xii. 3. 4) and mahas putrd- 
sah (xviii. 1. 2), of which the latter is a Rik passage, and noted in its 
Praticakhya (iv. 21); rajas prthu (xiii. 2. 22: see R. Pr. iv. 19); and 
firas krtam (xiii. 4. 40). In all these passages, the testimony of the 
manuscripts is unanimous in favor of the sibilant : in iv. 34. 8, a part of 
them (P. M. W. E.) read dhinush kdmadughd ; and in more than one 
instance elsewhere, one or two manuscripts give a s or sh, instead of 
visarjaniya, before an initial p or k. Unfortunately, it is impossible to 
tell what is the teaching of the Prati§akhya with regard to any such 
cases; we do not even know how accurately it or its commentators had 
noticed and noted the instances which their text contained. There is 
no apparent reason why the single cases noted in rules 66, 72, 77, 78, 79 
should not have been left to go into the yana, if a gana was to be estab- 
lished, and neither the accuracy nor the method of the treatise, in dealing 
with this class of phenomena, is worthy of unqualified commendation. 

The eightieth rule ends the section, of which the signature is : 80 : 
dvitiyasya trttyah pddah. 

81. In all these cases, s becomes sh, when preceded by an 
alterant vowel. 

The commentator says: yad etat: samdse sakdrah kapayor ananta 
ityanukrdntah: atra ndmyupadhasya sakdrasya shakdro bhavati: etany 
evo 'ddharanani: n&myupadhasye Hi drashtavyam : ita ultararh yad 
anukramishydmah ; 'the meaning is that, from the rule "in a com- 
pound" etc. [ii. 62] onward, in every case, a * preceded by an alterant 
vowel becomes sh : the illustrative instances are those already given : 

444 Wl D, Whitney, [li. 81- 

the specification " when preceded by an alterant vowel" continues in 
force, as we go on farther from this point' Tbat is to say, the action 
of the present rule is retrospective as well as prospective, and extends 
to all the cases in which the conversion <of visarjaniya to * takes place 
according to the rules of the preceding -section, as also to the cases 
which are now farther to be specified:; and that, whether the « to which 
the rules relate is original, or comes from visarjaniya by rule ii. 40 etc. 
The subject of this whole section is the nali of the dental sibilant, or 
its conversion into the lingual, sh. The same phenomenon is treated in 
Rik Pr. v. 1-19, Vaj. Pr. iii. 55-82, and Taitt. Pr. vi. 1-14. The group- 
ing of the cases is so different in the different treatises that any detailed 
comparison of rule with rule is only occasionally and partially practi- 

82. Also in the root sah, when it takes the form sdt. 

The commentator cites, as cases coming under this rule, turdshdf (ii. 
5. 3), prdshdt (not in AV.), prtan&sh&t (e. g. v. 14. 8). fatr&sh&n nish&t 
(v. 20. 11), abhishdd aumi vipvdshdt (xii. 1.54). As counter-example, 
to show that it is only when tah becomes sdt that the change of sibilant 
takes place, he instances prasahanam, which is not found in our text. 
That sddbh&ta means 'when it becomes s&t,' and not 'when it becomes 
»dA,' appears from the instances amitrasdha and tdtrdsdha (see under 
iii. 23), in which, although the vowel of the root is lengthened, the sibi- 
lant remains unaltered. If ashddha occurred in any earlier book than 
the nineteenth (it is found only at xix. 7. 4), the rule would have to be 
altered, or else sad in sddbk&ta would have to be understood as mean- 
ing both s&d (s&t) and sddh. There is no real necessity for any such 
rule as this, since all the words to which the commentary regards it as 
relating have precisely the same form in pada as in sanhitd. But there 
is a graver objection to the rule than its dispensability : it is quite out 
of place where it stands. In the first place, it treats of an initial *, 
while the treatise otherwise puts off such cases until after all those of a 
final * are disposed of; and, much worse, it interrupts the anuvrtti of 
ndmyupadhasya, which must necessarily take place from rule 81 to 
those that follow ; since, in the majority of the cases which it concerns, 
the sibilant is preceded by long d. Either there is here an unusual de- 
gree of awkwardness and inconsistency of method on the part of the 
author of the treatise, or the rule is an interpolation. 

The later rules iii. 1 and iv. 70, which deal with the same irregularly 
formed words from different aspects, are to be compared. 

88. Also before a taddhita-sv&s. beginning -with t 

As instances, the commentator first gives sarpishtaram, yajushfaram, 
vapushtaram — not one of which words occurs in the Atharvan, while 
the first two are found as examples under the corresponding rule of 

ii. 85.] Alharvw Veda Prdtigdkhya. 445 

Panini(viii. 3. 101) — and then vapushtamam (vapushtame, ■p.vapiih-tome, 
V. 5. 6 : I have noted no other similar case in the text). As counter- 
example, to show the necessity of the restriction " before a taddhita- 
suffix," he cites agnis takmAnam (v. 22. 1) ;■ to show that of "beginning 
with t" he gives tarpissAt, yajitssAt, vapussAt, which are not in the 
Atharvan, while the first of them, again, is an example under Panini's 
rule just referred to. 

The corresponding rule of the Rik Pr. is at v. 16 (r. 31, cccxlviii), and 
Taitt. Pr. vi. 5 includes one or two similar cases. Neither, however, 
gives a grammatical definition of the phenomenon, like that of our 

84. Also before forms of the second personal pronoun, except- 
ing in tats tvam etc. 

Those forms used as " substitutes for yushmat" or as cases of the pro- 
noun of the second person, before which the conversion of final s into sh 
takes place in the Atharvan, are tvam, ivd, and te: the cases are enumer- 
ated below.* The commentator cites, as instances, five of them, via. 
bahinh te axtu (i. 3. 1 etc.), rtubhish tvA vayam (i. 35. 4), dyAush tva pitA 
(ii. 28. 4), tAbhish tvam asmdn (ix. 2. 25), and tAish te rohitah (xiii. 1. 35). 
As examples from the gana of exceptions, he offers tAis tvam putram (iii. 
23. 4), vadhris tvam (iv. 6. 8), and pavastAis tvA (iv. 7. 6). This gana I 
have not thought it worth while to take the trouble to fill up, deeming 
it of more interest to give the complete list of the cases in which the 
change of the sibilant did, rather than of those in which it did not, take 
place. The former are, I believe, the more numerous of the two classes. 

rTrTRSnf^J ^ tl z\ h 

85. Also before tat, tdn agra, etc, 

The forms of the third personal, or demonstrative, pronoun before 
which a final «, after an alterant vowel, is converted into sh, are tarn, tat, 
and tAn. The Cases, which are not numerous, are detailed below.f The 
commentator cites four of them, viz. agnish tad dhotA (e. g. vi. 7 1 . 1 ), 
agnish tAn agte (ii. 34. 3), vAyush tAn agre (ii. 34. 4), and nish tarn bhaja 

* Before tvam: ix 2.25. xi. 9. 6. Before tvd: i. 22. 3; 35. 4. ii. 28. 4. iii. 29. 8. 
T. 28. 5, 18. xiv. 2. 59-62. xviii. 4. 9. xix. 27. 1 (bis) ; 34. 6 ; 37. 4 ; 46. 1. Before te : 
i8. 1-6. ii. 18.4; 29.6. iv. 6. 6. v. 25.9; 28.5. vi. 92. 3; 111.2. vii. 63.2; 74.8. 
viii. 2. 13. ix.4.10. x.9.2. xii.1.18. xiii. 1.35. xiv. 2. 3. xviii ,4. 9, 1 5. xx. 133.2. 
At iii. 10. 10, all the tanhitd MSS., without exception, give rlubhyaxh tvd, which 
was deemed, however, much too anomalous and unsupported a reading to be re- 
ceived into the edited text. 

t Before tat* : iv. 22. 2. Before tat: i. 32. 2. vi. 71. 1, 2. vii. 63. 3. x. 9. 26. 
xviii. 3. 65. xix. 69. 2. Before tan : ii. 34. 3, 4. xviii. 2. 28. There is a single case, 
moreover, in which all the tanhitd MSS. read th after a: viz. nash tebhih, ii. 85. 2 ; 
but this reading has been rejected in the edition, as ah evident blunder, apparently 
growing out of a confusion with the participle nashtebhik. 

446 W. D. Whitney, [ii. 85- 

(iv. 22. 2). He also adds two examples of another class, viz. dushtano 
(p. dustano iti duhtano: iv. 7. 3), and nishlakvaritn (p. nih-talevartm : 
v. 22. 6). The. rule must be meant also to include gobhish tarema (vii. 
50. 7). The word dushtara (vi. 4. 1) would seem to come properly 
under it, but its treatment by the pada-texts (it is written dustara, not 
duh-tara) indicates, I should think, that the Hindus regarded it as an 
irregular compound of duh and stara, from the root star (str) ; hence 
it would fall under the next rule, or else under rule 98. 

86. Also before stria, sva, and the root svap. 

Kilt tIiPqq AQBOd rtrtrtlll* fin/1i»»» + liio vnln mtn -firm nnnl 

But three cases occur under this rule, one for each of the words men- 
tioned in it: they are anishtrtah (p. anih-strtah: vii. 82. 3), nishva (p. 
nih : sva : vi. 1 2 1. 1 and vii. 83. 4), and the frequently occurring dushvap- 
nyam (p. duh-svapnyam : e. g. vi. 46. 3), with which goes its derivative 
d&ushvapnyam (p. dauh-svapnyam : iv. 17. 5). The commentator cites 
them all, excepting the last; the second of them he gives in its connec- 
tion, thus : duritam nishva 'smat. All these words* have in the edited 
text been written in a similar manner, with but a single sibilant, as I 
have here cited them, and as they are given by the commentary in our 
manuscript. This does not, however, appear to be the way in which 
the Praticakhya intends that they should be written, for it prescribes 
that the final visarjanlya or sibilant should be converted into sh before 
the initial s ; not that there should be a dropping out of either sibilant. 
Yet the Praticakhya also is defective as regards its treatment of them ; 
for, while it has directed above (ii. 16) that a dental mute (tavargiya) 
be converted into a lingual after final sh in the preceding word, it has 
given no such rule respecting the dental sibilant. If we follow our 
treatise, then, implicitly, we shall write anishstrtah, nish svd, dushsvap- 
nyam, which are barbarous and impossible forms. The manuscripts 
write, without dissent, anishtrtah, and nishva ; as regards dushvapnya, 
their usage is very irregular ; they vary with the utmost inconsistency 
between that form and duhshvapnya, in no single instance writing dusk- 
skvapnya. While, therefore, we are compelled to look upon anish- 
shtrtah, nish shva, and duskshvapnyam as the forms which the Pratica- 
khya intends to sanction, we cannot but hold the editors justified in 
following for the two former cases the unanimous authority of the 
MSS.. and in making the third conform to them. To dushtara refer- 
ence has been made under the preceding rule, as a case akin with these, 
but differently treated, in that one of the two sibilants has been rejected : 
on account of this irregularity, doubtless, it is treated as an anomalous 
case by the pada-text, and left undivided. 

The Vaj. Pr. (iii. 71) adopts and sanctions the form duhshvapnya: 
the Rik Pr. (v. 1, r. 1, cccxviii) requires, like our treatise, dushshvapnya, 
although the manuscripts, as noticed by Muller, usually read dushvapnya. 

* Except in the passage vi. 121. 1, where the edition, by a reprehensible incon- 
sistency, reads nih shva. 

ii. 90.] Atliarva-Veda Prdtigdhhya. 447 

Hlfa*^lrUr^=hl(HI II to II 

87. Also the sofa suffix, after an alterant vowel, h, or r. 

This rale, like tlie corresponding one in the Vaj. Pr. (iii. 57), is not 
precisely in place in a Pratigakhya, which has nothing to do with the 
processes of derivation in connection with which its application must be 
made : hence the other two treatises present nothing analogous to it. 
The commentator illustrates it by citing phdl&j j&tah Jcarishyati (x. 6. 2), 
iyaksham&na bhrgubhih (iv. 14. 5), and havisha " 'h&rsham enam (iii. 
11.3). He explains lea in the rule by kavargat, ' after a guttural,' prob- 
ably in view of the requirement of some authorities (see under ii. 6) 
that the Jc should become Ich before the sibilant. 


88. Also that of strawhtiya. 

This word occurs but once in the text, in a passage cited by the com- 
mentator : strdishuyam anyatra (vi. 11.3: p. straisuyam). Why, among 
the words mentioned in iv. 83, it should be singled out to be made thus 
the subject of a special rule, is not at all clear. The position of the 
rule, too, thrust in between the two closely related rules 87 and 89, and 
disturbing their connection, is in a high degree awkward, and calculated 
to inspire suspicions of an interpolation. 

R^ffq" tffa ii s* n 

89. Even when a n has been lost. 

This rule attaches itself immediately and closely to rule 87, from 
which it has been blunderingly separated by the intrusion of rule 88. 
It is intended to apply to such cases as yajtinshi, havinshi, where, by 
ii. 34, there has been a loss of n before the ending si, accompanied, by 
i. 67, with nasalization of the preceding vowel, when the ending itself 
is converted into shi after the alterant vowel, although the latter is nasal. 
Among the other treatises, the Vaj. Pr. (iii. 56) alone gives a similar 
precept. The illustrative citations of the commentator are the same 
which he has already twice before given, under i. 84 and ii. 34. 

SOTTTintft: II ^o ii 

90. Also that of a root, after a preposition. 

This is a very general precept, and the instances of its application in 
the text, in forms of conjugation and of derivation, are quite numerous. 
The commentator selects as illustrations but two : vrksham yad gdvah 
parishasvajdnd anusphuram (i. 2. 3 : p. pari-sasvajdndh ; anusphuram 
is an exception to the rule, akin with those detailed in rule 102, below), 
and vishitam te vaslibilam (i. 3. 8 : p. vi-sitam). As counter-examples, 
to show that it is only after a preposition that the initial s of a root is 

vol. vii. 57 

448 W. D. Whitney, [ii. 90- 

regularly converted into sh, the commentator fabricates a couple of 
cases, viz. dadhi sincati, madhu siHcali ; the former of them appears 
also among the counter-examples under the corresponding rule of Panini 
(viii. 3. 65). The proper exceptions to the rule are detailed below, in 
rules 102-107. 

The Taitt. Pr. (vi. 4) has a general rule like this of our own treatise, 
for the change of an initial s after a preposition : the other two Prati- 
c&khyas rehearse the cases in detail (see R. Pr. v. 4-10 ' r V. Pr. iii. 

3PillHl3l iihn 

91. As also after a reduplication. 

The commentator is more than usually liberal in his citations under 
this rule ; he gives us sushudata (i. 26. 4), abhi sishyade (v. 5. 9), A su- 
shvayanti (v. 12. 6 ; 27. 8), sishAsavah sish&satha (vi. 21. 3), tat sixhasati 
(xiii. 2. 14) r and sushuve (xiv. 1. 43). In all these cases, as in all others 
of the same kind which the text affords, the pada-text gives the dental 
sibilant unchanged ; thus, sus&data, sisyade, etc. 

This class of cases is not treated by itself in the other Praticakhyas. 

92. And that of the roots sthd, sah, and sic, even when an a 
is interposed. 

There is, owing to the imperfection of our manuscript, a slight doubt 
as to the precise reading of this rule, the copyist, as in one or two other 
cases which we have already had occasion to notice, having skipped a 
line of his original, or from one kAra to another and later one, whereby 
we have lost part of the original statement of the rule and nearly the 
whole of its paraphrase ; the final repetition of the rule before the one 
following being also slightly defective. I restore the mutilated passages 
conjecturally as follows, noting within brackets the parts supposed to 
be omitted: sthasahisicinAm akAra[vyavAye 'pi: sthasahisicin&m upa- 
sargasthan nimittdd akAravyavAye 'pi dhAtusakArdjsya shakaro bhavati: 
stkasahisicin&m akAravyavA[ye\ 'py abhyasavyav&ye 'pi sthah. 

Of the cases to which the rule relates, the commentator instances 
three ; viz. abhy ashthdm vifvah (x. 5. 36 and xvi. 9. 1), tena devA vy 
askakanta (iii. 10. 12), yenA 'kshA abhyashicyante (xiv. 1. 36). For the 
root sthA the text affords us two other examples, adhy ashth&t (x. 10. 
13) and adhy ashth&rn (xii. 1. 11). These are actually all the cases oc- 
curring in the Atharvan in which the three roots named in the rule are 
preceded by prepositions ending in alterant vowels, with interposition 
of the augment — except the anomalous form vy Asthan (xiii. 1. 5), where 
the irregular initial A is read in pada as well as in sanhitA, and which 
does not come under the present rule, since there is an interposition of 
AkAra, not akara. 

Similar cases are noted by the other treatises (see V. Pr. iii. 64 ; T. 
Pr. vi. 3). 

ii. 95.] Atharva- Veda Prdtigdkhya. 449 

93. And that of the root sthd, even when the reduplication is 

Two cases are cited by the commentator under this rule, viz. sapta 
sindhavo vitashthire (iv. 6. 2), and brahma pururupam vi tashthe (ix. 10. 
19). In the latter passage, the edition gives erroneously vi tasthe: two 
of the manuscripts (W. and E.) read vi cashte ; the others have, cor- 
rectly, vi tashthe (or, by a carelessness very common in the codices, vi 
tashte: E., for instance, never distinguishes shth and sht from one 
another) : the pada-text presents vi : tasthe. We have in two other pas- 
sages (ix. 9. 2. xiv. 2. 9) adhi tasthuh (not tashthuh) ; this apparently 
constitutes an exception to the rule which has escaped the notice both 
of the treatise and of the commentator : possibly, however, the adhi is 
not in these passages regarded as standing in the relation of upasarga 
to the root sthd, since it does, in fact, belong rather, in a prepositional 
relation, to preceding ablative cases, than to the verbal form as its pre- 
fix : and this is the more clearly indicated by its retaining its indepen- 
dent accent before the accented verb. 

qJ7&Tt ^TTTOhi^ii 

94. As also after parama etc. ; but not after dpdlca. 

The commentator's explanatory paraphrase is as follows : paramadi- 
bhyah p&rvapadebhyah andpdke uttarapadasthasya sakarasya shakdro 
bhavali ; 'after parama etc., as former members of a compound, except- 
ing dpdke, the * of the second members of the compound becomes sh? 
This does not, in terms, restrict the rule to cases of the root sthd, in 
composition, implying an anuvrtti of sthah from the preceding rule: 
yet we can hardly doubt that it is properly to be so restricted, and the 
commentator himself gives examples only from the root sthd: they are 
parameshthi (e. g. iv. 11. 1 : p. pararne-sthi), bhuvaneshthdh (ii. 1. 4 : p. 
bhuvane-sthdh), madhyameshthdh (ii. 6. 4 : p. madhyame-sthdh), and 
angeshthdh (vi. 14. 1 : p. ange-sthdh). The exception, as noted by him, 
is dpdkesthdh prahdsinah (viii. 6. 14: p. dpdke-sthdh). These are all 
the words in which sthd is compounded with a locative ending in e; 
and, considering the form of the rule, it is probable that only such com- 
pounds were intended to be included under it : if the other compounds 
in which the s of sthd is lingualized (viz. rayishthd, rayishthdna, giri- 
shthd, pathishlha, goshtha, and gavishthira — of which the last is read in 
pada as in sanhitd, while the others are separated, with restoration of 
the dental sibilant) had also been aimed at, the exception bhuristhdtra 
(iv. 30. 2 : p. bhuri-sthdtra) would likewise have required notice. 

The form of statement of the gana, paramebhyah instead of paramd- 
dibhyah, is quite peculiar. 

95. As also after apa and savya. 

450 W. D. Whitney, ["• 95- 

There is still anuvrtti of sthah : the s of the root sthd becomes sk 
after apa and savya. The cases are separately noticed, because in them 
the change takes place irregularly after a, instead of after an alterant 
vowel. The commentator cites the two passages : apdshthdc chrngdt 
(iv. 6. 5 : p. apdshthdt), and savyashthdg candramah (viii. 8. 23 : p. 

96. Also that of stoma and soma after agni. 

The commentator instances y&vad agnishtomena (ix. 6. 40 : p. ag- 
ni-stomena), and agnishomdv adadhuh (viii. 9. 14 : p. agntshdmdu). As 
was the case with the last rule, the two words to which the precept 
relates are of different character, the one necessarily requiring the rule, 
in order to determine its sanhitd reading, while the other might have 
been safely left unnoticed by the Praticakhya. There is yet another 
case in the Atharvan text, which would seem to call for inclusion in 
this rule : it is agnishvatt&h (xviii. 3. 44 : p. agni-sv&ttah). We can 
hardly suppose that it was intentionally omitted here, to be included in 
the gana of rule 98, below : either it must have been overlooked by the 
maker of the treatise, or the verse which contains it (it is also a Rik 
verse; see RV. x. 15. 11) was not in his Atharva-Veda : that it was, 
however, contained in the text recognized by the commentator, is 
shown by the fact that he several times (under i. 80, 84 and ii. 34) cites 
the phrase atto havinshi, which forms part of its second line. 

The commentator gives, as counter-example, abki somo avivrdhat (i. 
29. 3). 

floT: ii Ion 


97. Also that of su. 

The commentary furnishes the following examples : idam u shu (i. 24. 
4), tad u shu (v. 1. 5), maMm u shu (vii. 6. 2), anya u shu (xviii. 1. 16), 
stusha u shu (xviii. 1. 37), and tyam u shu (vii. 85. 1). Other cases are : 
after u, v. 6. 4 ; after u (the same particle not lengthened), vii. 73. 7 ; 
after mo, v. 11.7. xviii. 3. 62; after evo, vi. 84. 3; after o, vii. 72. 2; 
after vidtno, i. 2. 1. These are all the instances in which su follows in 
the Atharvan the particle u, or a word whose final is combined with the 
latter. On the other hand, in the only passage in the text (prthivt su : 
xviii. 3. 51) in which it follows an alterant vowel of another character, 
it remains unchanged. Our rule, then, is defective, and should have 
been restricted either by the specification of the alterant vowels to 
which the effect upon su is limited, or by a mention of the exception. 
The other treatises are less general, and hence, it may be presumed, 
more accurate, in their treatment of the change of this particle (see 
R. Pr. v. 2, 3 ; V. Pr. iii. 59, 60, 61 ; T. Pr. vi. 2). 

srrft^n u H u 

98. Also after tri etc. 

ii. 100.] Atharm- Veda Prdtigdkhya. 451 

Here is another of those convenient ganas, set as a catch-all for what- 
ever cases may not have been otherwise provided for, and rendering it 
impossible for us to ascertain the precise degree of accuracy with which 
the authors of the treatise examined and excerpted their text. The 
commentator cites three passages, ye trishaptdh (i. 1.1: p. tri-saptdh), 
goskedhdm (i. 18. 4 : p. go-sedh&m), and raghushyado 'dhi (iii. 7. 1 : p. ra- 
ghu-syadah) : he also, in his paraphrase, limits the application of the rule 
to cases of this character, in which an alterant vowel at the end of the 
first member of a compound comes before an initial * of the following 
member. Of such cases, besides those falling under rule 90, above, and 
those which are specially referred to in the rules preceding and follow- 
ing this, or which are mentioned in connection with them, I have 
noted in the Atharvan text the following : dhrshnushena (v. 20. 9 : p. 
dhrshnu-sena), sushtuti (e. g. vi. 1. 3 : p. su-stuti), purushtuta (vi. 2. 3 : 
p. puru-stuta), rshishtuta (vi. 108. 2 : p. rshi-stuta), gabhishak (vii. 7. 1 : 
p. gabki-sak), prthushtuka (vii. 46. 1 : p. prthu-stuka), sushurnan (vii. 46. 
2 : p. su-s&man), anushtup (e. g. viii. 9. 14 : p. anu-stup), trishtup (e. g. 
viii. 9. 14 : p. tri-stup), trdishtubha (ix. 10. 1 : p. tr&istubha), trishandhi 
(e. g. xi. 9. 23 : p. tri-sandhi), rdtishdc (e. g. xviii. 3. 20 : p. rdti-sdc), and 
divishtambha (xix. 32. 7). Exceptions will be given below, under rule 

99. Also that of the root sad, after a word ending in r. 

The commentator instances hotrshadanam (vii. 99. 1 : p. holr-sada- 
nam), and pitrsfiadandh pitrshadane tvd (xviii. 4. 67 : p. pitr-sadandh 
etc.). To these may be added pitrshad (xiv. 2. 33 : p. pitr-sad) ; and in 
ndrshada (iv. 19. 2 : p. ndrsada) is also implied nrshad. 

(Mf^^Rj^fMifH ^ ii %oo ii 

100. As also after barhi, pathi, apsu, divi, and prthivi. 

The instances are, as cited in the commentary, barhishadah pitarah 
j(xviii. 1. 51 : p. barhi-sadah), pathishadi nircakshasd (xviii. 2. 12 : p. 
pathisadi iti pathi-sadi), apsushado 'py agnin (xii. 2. 4 : p. apsu-sadah), 
ye devd divishadah (x. 9. 12 : p. divi-sadah), and prthivishadbhyah (xviii. 
4. 78 : p. prthivisat-bhyah). 

We have reason to be surprised that the root sad is treated in this 
manner, being made the subject of these two separate rules. If the 
compounds into which it enters as final member are to be excepted from 
the general gana of rule 98, we should expect to find it directed that 
the * of sad should always be lingualized after an alterant vowel, as is 
actually the ease. Not only is there, by the method adopted, a loss of 
that brevity which treatises of the siltra class are wont to aim at almost 
as their chief object, but there is also a loss of completeness : the only 
remaining compound of sad of this class, sushad (e. g. ii. 36. 4), is left 
out, to be provided for in the general gana. Or is it possible that su is 
regarded as falling und«r rule 90, as if a proper upamrga or preposi- 

452 W. D. Whitney, pi. 100- 

tion ? If so, the forms into which it enters would be sufficiently pro- 
vided for; since, excepting in the cases noted in the later rules (102 
etc.), it always, lingualizes the initial s of a root, while it has no effect 
upon that of a preposition or adverb, as in susaha (vi. 64. 3 : p. su-saha) 
and the numerous compounds in which it is followed by earn, as susam- 

R.R&^FF#T: n\o\u 

101. Also that of the root as after hi and divi. 

The commentator cites two of the cases falling under this rule, viz. 
apo hi shtha (i. 5. 1), and ye deva divi shtha (i. 30. 3) : to which I add 
divi shan (xvii. 12 and xviii. 4. 59). The commentator then notes an 
exception which had escaped the authors of the treatise — prefacing it 
with tato 'pa vadati, ' from this rule is made the exception ' — vimuco hi 
mnti (vi. 112. 3) : it is, in fact, the only case of the kind which the text 
affords — unless hi saty&h (i. 10. 1) is to be regarded as also falling under 
the rule, satya being an evident and acknowledged derivative from the 
root as. 

Thus ends the Praticakhya's tale of the conversions of final or initial 
s into sh ; what remains of the chapter is only a rehearsal in detail of 
exceptions to the rules already given. But there are still left in the 
Atharvan text a few instances of the same conversion, which can hardly 
be regarded as included in any of the preceding rules, since they are 
analogous with none of the other cases there treated of: unless some- 
thing has been lost from this final section of the chapter — of which 
there are no indications — the treatise-makers and their commentator 
must lie under the imputation of having been careless enough to over- 
look them. The passages referred to are as follows : ad u shtenam (iv. 
3. 4), tarn u shtuhi (vi. 1. 2), ni sha hiyatdm (viii. 4. 10), and cid dhi 
shma (xviii. 1. 33). There would be little plausibility in a claim that 
the verses containing these passages were not included in the Atharva- 
Veda accepted by the school to which the treatise belonged, or that 
the readings of the school were different. And certainly, neither of 
these things could be maintained with regard to svarsh&h (v. 2. 8 : p. 
svah-s&h), since the conversion of the final visarjaniya of its svah into.. 
r has already been made the subject of a special rule, while nevertheless 
the Praticakhya contains no precept directing us to read it otherwise 
than svarsah. The word would fall under the analogy of rule 87, but 
its final syllable could by no means be reckoned as a suffix (pratyaya). 

102. The s of the roots sarp, sarj, sparg, sphurj, svar, and smar 
is not changed into sh according to the preceding rules. 

This is evidently a rule of kindred sphere with rule 106, below, and 
the two might well enough have been combined into one, which should 
teach that a root containing a r, either semivowel or vowel, was not 
liable to the changes prescribed in this section. The corresponding 

ii. 104.] Atharva- Veda Prdti^dhhya. 453 

rules in the other treatises have nearly this form (R. Pr. v. 11, r. 23, 
cccxl; V. Pr. iii. 81 ; T. Pr. vi. 8). As so stated, it would require the 
notice of but a single exception, vi shparat (vi. 56. 1 and x. 4. 8). As 
the rules now stand, they are slightly inexact, for in neither of them are 
included anusphuram (i. 2. 3), atisara (v. 8. 2 etc.), and pratisara (e. g. 
ii. 11. 2), although other forms of the root sar are contemplated in rule 

The commentator's illustrative examples are as follows : vi srpo vi- 
rappin (not in AV. ; take instead parisarpati [v. 23. 3]; there are an 
abundance of such cases in the actual text), atisrshto apam vrshabhah 
(xvi. 1. 1), visprpah (perhaps a corrupted reading for uparisprpah [v. 3. 
lOj, the only case our text contains), vapyamana J bhi spkdrjati (xii. 5. 
20), vidatha 'bkisvaranti (ix. 9. 22), nirrtho yap ca nisvarah (xii. 2. 14), 
and prati smarethdm tujayadbhih (viii. 4. 7). 

103. Nor that of gosani etc. 

The instances which the commentator has chosen to give from this 
gana are gosanim (iii. 20. 10), vi simatah (iv. 1.1), abhi sishyade (v. 5; 
9), and arm sutum savitave (vi. IV. 1-4). We cannot but be surprised 
that the first of these words was chosen to head the ffana; it is no 
proper exception to any of the foregoing rules ; the only one under 
which it would have any claim to come is rule 98, and there its simple 
exclusion from the gana to which alone the precept is calculated to 
apply would be sufficient warrant for its reading. The third instance is 
a fair exception to rule 90, besides being an instance under rule 91. 
The second and fourth are, like the first, no legitimate cases for a rule 
like this, since vi and simatah, arm and s&lum, are disconnected words, 
which only accidentally come together in the arrangement of the sen- 
tence : if all such cases are to be taken note of, the gana will be con- 
siderably and unnecessarily extended. Of actual cases having claim to 
constitute together a gana of exceptions, the text does not furnish a 
great number : I have noted, besides the three words mentioned under 
the last rule, antisumne (vii. 112. 1), pratispapanam (viii. 5. 11), visalyaka 
(e. g. ix. 8. 2), abhisatva (xix. 13. 5), and pr&tisutvanam (xx. 129. 2 : the 
reading is very doubtful). As good a right as gosani to inclusion in 
the gana have also bhuristhatram (iv. 30. 2), and bahusuvari (vii. 46. 2). 

srwferf ^fF^: II \o& || 

104. Nor that of the root skand, after adhi and abhi. 

The cited instances are adhi skanda virayasva (v. 25. 8), and abhi- 
skandam mrgi 'va (v. 14. 11). As counter-instance, the commentator 

brings forward ekapatam vishkandhani (iii. 9. 6), which is a blunder 

unless, indeed, the commentator's grammatical system derives skandha, 
' shoulder,' from the root skand. An actual example of the kind he 
seeks to give is parishkanda (xv. 2. 1 etc.). 

454 W. D. Whitney, [ii. 105- 

cf^": ^UIIH: ii W ii 

105. Nor that of the root star, after pari. 

The commentator cites pari slrnihi pari dhehi vedim (vii. 99. 1), and 
paristaranarn id dhavih (ix. 6. 2). As counter-example, he can find only 
vishtari j&tah (iv. 34. 1). Vishtarin doubtless comes from the root star ; 
yet, as the pada-text does not analyze it, but writes it in its sanhitA form, 
it might have been neglected, and the root star added to those rehearsed 
in rule 102, with which it evidently belongs. 

\w{¥\ srnMii 

106. JSTor one that is followed by r. 

The examples offered in the commentary are sisrat&m nari (i. 11. 1), 
parisrutah kumbhah (iii. 12. V), and pra bh&navah sisrate (xiii. 2. 46) : 
the first and third are exceptions to rule 91, the second to rule 90. 
The text furnishes a number of other cases, which it is unnecessary to 
rehearse here, as there are no counter-exceptions. 

$rfr WFT ^rRTrr: II \o^o u 

107. Nor in. abhi sydma prtanyatah. 

The passage, as cited by the commentator, is vayam abhi sydma prtan- 
yatah (vii. 93. 1*) : as a quite peculiar case, it is properly enough made 
the subject of a special rule. To show the necessity of citing in the 
rule the following word, prtanyatah, the commentator brings forward 
vifvdh prtana abhi shyama (xiii. 1. 22). 

The signature of the chapter is as follows: 16 : iti dvitiyo 'dhyayah 
samaptah. The number which should inform us how many rules are 
counted in the chapter is again unfortunately corrupt, and we are left to 
conjecture as to how it should be amended. I see no reason to suspect 
the loss of a rule or rules in the manuscript. 


Contents : — Section I. 1-25, lengthening of final and other vowels in sanhitd. 
Section II. 26-28, doubling of consonants when final or in combination with 
other consonants ; 29-32, exceptions ; 33-36, vowels exempt from euphonic com- 
bination ; Zl, nasals do. ; 38, method of combination of A between two vowels ; 
39, conversion of final vowels into semivowels ; 40, do. of final diphthongs into 
vowel and semivowel ; 41-52, fusion of final and initial vowels; 53-54, absorp- 
tion of initial a after final e and o. 

* All the MSS. except W. read shydma, which has accordingly been received 
into the edited text. 

iii. 2.] Atharva- Veda Prdtigdkhya. 455 

Section III. 55-64, different kinds of svarita accent; 65-74, occurrence and 
modifications of the several accents in the combinations of the phrase. 

Section IV. 75-85, conversion of the dental nasal n into the lingual in, ; 86-95, 
exceptions; 96, anomalous insertion of a sibilant in a single case. 

fl^T5rn% ^tef: w\w 

1. Before the root sah, when it ends in at, the vowel is 

In this rule, adanta has virtually the same meaning as sAdfihkta in 
rule ii. 82. It would not do to say " before sah when it becomes sah," 
because of the words amitrasdha and abhimdtishdha, in which, though 
the vowel of the root is lengthened, the preceding final remains un- 
changed. The illustrative citations of the commentator are precisely 
those already given, under ii. 82, including the two words, example and 
counter-example, prdshat and prasahanam, which are not found in the 

This rule also belongs in the category of the supererogatory, since in 
none of the words to which it relates does the pada-text afford a differ- 
ent reading from that of the sanhitd. There would be just as much 
reason, so far as we can see, for a rule prescribing the prolongation of 
the a in the root sah itself- — and. that, in fact, is what the Vaj. Pr. (iii. 
121) does, while it takes no notice of the change of the preceding 

The Eik Pr. devotes three chapters (vii-ix) to the subject of the 
irregular prolongation of vowels. In the Vaj. Pr., the same subject 
occupies the sixth section of the third chapter, and one rule in the 
seventh (iii. 95-128) ; in the Taitt. Pr., the third chapter : the latter 
treatise inverts the form of statement adopted by the others, and details 
the cases in which a vowel which is long in sanhitA must be shortened 
in pada. The method in which the different works combine and put 
forth the phenomena of prolongation is so different, and so little would 
be gained by any detailed comparison of their teachings, that I shall 
for the most part content myself with this general reference. 

W M^dUN^UI< f ^5h^l II ^ II 

2. Also is lengthened the a of ashta, before pada, yoga, pah 
sha, parna, danshtra, and cakra. 

Compounds of ashta with parna and danshtra are not to be found in 
the present Atharva-Veda, nor are we necessarily to conclude, from 
their being mentioned in the rule, that any such occurred in the text 
recognized by the makers of our treatise : it is more likely that they 
have here, as in some other cases, detailed all the instances of the pro- 
longation which they had noticed, without being particular as to their 
source. The rule, moreover, is an unnecessary one, since the pada-text 
everywhere offers the same reading with the sanhita, as is expressly di- 
rected by a later rule (iv. 94). 
vol, vii. 58 

456 W. I). Whitney, [fli. 2- 

The citations of the commentary are ashtdpadi eaturakshi (v. 19. 7)^ 
ashtdpa&shdm (ix. 3. 21), ashtdyogdih (vi. 91. 1), ashtdparnah, ashtd- 
danshtram, ashtdcakrd navadvdrd (x. 2. 31), and ashtdcakraih variaie 
(xi. 4.22). 

S. Also a vowel before the root vyadh, wheit it is without a 

The commentator's instances- are pv&vit (v. 13. 9) T and hrdaydvidham 
(viii. 6. 18) ' T to these is- to be added only marmdvidham (xi. 10. 26), 
which, under iv. 68, is also cited in the commentary. By the latter 
rale, the pada in all these cases- reads precisely like the sanhitd, without 
any division of the words, and without restoration of the short vowel. 
As counter-example, to show that, when the root takes a suffix, the 
vowel preceding it is not lengthened, we have a form of verbal inflec- 
tion, pravidhyanto ndma (iii. 26. 4), brought forward. 

4. The particle u is lengthened in idam & shn etc. 

Of the passages in which u is thus made long, the commentator cites 
six, viz. idam u shu (i.24. 4), tad u shu (v. 1.5), mahim u shu (vii. 6. 2), 
any a u shu (xviii. 1. 16), stusha u shu (xviii. 1. 37^ and tyam u shu (vii. 
85. 1) ; the othe* cases afforded by the text are pary u shu (v. 6. 4), 
and para u ie (xviii. 3. 7). "Were it not for this last case, the rule of 
our treatise might have been constructed like that of the Vaj. Pr., 
which says (iii. 109) that w before su is lengthened. 

5. Also the final vowel of oshadhi, except in the strong cases. 

Those of the strong cases of oshadhi which the rule would except 
are, of course, only the nominative and accusative singular, since the 
others would, by the rules of declension, show a gunated or a lengthened 
vowel, or a semivowel, in its place. The commentator cites no actual 
passages in illustration of the rule, but merely catalogues the forms in 
which the long vowel would appear, viz. oshadhibhih (e. g. ii. 10. 2), 
oshadhibhydm (not found in AV.), oshadhibhyah (e. g. vi. 20. 2), osha- 
dhtn&m (e. g. iii. 5. 1), and oshadhtshu (e. g. i. 30. 3) : as counter-exam- 
ple, he adds iyam ha mahyam tvam oshadhih (vii. 38. 5). The pada- 
text, in all these cases, reads precisely like the sanhitd. 

Since the rule does not restrict itself to forms of declension, it is 
guilty of an oversight in taking no account of the compound oshadhija 
(x. 4. 23 : p. oshadhi-ja) as a farther exception. In the only other com- 
pounds which the text affords — viz. oshadhisarhfita (x. 5. 32 : p. osha- 
dhi-sarhfita), and ashadhtmant (xix. 17. 6 ; 18. 6) — the rule of the Pra- 
tic&khya is observed. 

ill- 9.] Atharva- Veda Pr&tigdkhya. 457 

6. And except in the phrase jivantim oshadMm. 

The commentator quotes the passage a little more fully, jivantlm osha- 
dMm aham (viii. 2. 6 ; 7. 6), and adds that the inclusion of the accom- 
panying word in the rule is intended to limit the exception to this par- 
ticular passage ; citing, as counter-examples, imam khandmy oshadkim 
(iii. 18. 1), and oshadkim $epaharshanim (iv. 4. 1). 

Wo'> UCIt 

7. The vowel of sddka is long. 

The commentator cites for us the only passage in which this partici- 
ple occurs, vaca s&dhah parastar&m (v. 30. D). The rule is one of the 
most utterly superfluous presented by our treatise, which, of course, has 
nothing to do with the mode of formation of such words. Moreover, 
if it was inclined to do a work of supererogation as regards them, it 
should not have omitted to notice also r&dhva, rtidha, m&dha, pudha, 
and other like forms which the text contains. Probably the reason why 
this particular one was noticed, and not the others, is that the regular 
form, according to the rules of the general grammar, is sodha. Plnini 
(vi. 3. 113) remarks the occurrence in specific cases of forms like sadha. 
The compound ashadha is found but once in the Atharvan, and that in 
the nineteenth book (xix. 1, 4). 

8. As to the final vowel of rdtri, diversity of usage prevails. 
This is rather a discreditable confession on the part of our treatise, 

whose business it is to settle authoritatively the reading of its school in 
all cases admitting of any doubt, that it does not feel equal to dealing 
with the irregularities of the word in question. Nothing like it has 
hitherto met us, but we shall find several instances in that which follows. 
It is also a very unnecessary acknowledgment; for, in the first place, 
there was no such rule as this absolutely called for, since the pada-text 
everywhere reads all the forms of ratri like the sanhita ; and, in the 
second place, there is no great perplexity in the phenomena. In the 
nominative we always meet with ratri, except in a single case (xiii. 4. 
30), where rdtris occurs ; as accusative, only ratrim is found, and ratrim, 
which the Rig-Veda has (e. g. i. 35. 1), is never read ; as vocative, we 
have only ratri, never ratre ; and the other forms are, with the excep- 
tion of ratraye (viii. 2. 20), such as would come from the theme ratri. 

The commentator, as under rule 5, gives the series of cases ratribhih 
(xviii. 1. 10), r&tribhyam, ratribhyah, ralrfaam, (e. g. iv. 5. 4), and ratri- 
shu, only two of which are to be found in our text. As example of the 
short vowel, again, he cites an actual case, vr&tya ekam, ratrim (xv. 13. 1). 

ferFT *H°iylH>j u * ii 

9. Also as to that of vicva, before nam, vasu, and milra. 

458 W. D. Whitney, [Hi. 9- 

Here are quite heterogeneous cases, mixed together in one rule. The 
noun vifv&nara and its derivative vdipvdnara always show the long 
vowel, and moreover are read in pada precisely as in sanhitd. Vipvd- 
vasu the pada-text divides, restoring the short vowel : thus, vipva-vasu ; 
but the long vowel invariably appears in sanhitd. We also find vipvd- 
tnitra three times in the text (iv. 29. 5. xviii. 3. 15, 16), and it is each 
time written vipvdmitra, without division, by the pada-text. But in a 
half-verse which occurs twice in the eighteenth book (xviii. 3. 63 ; 4. 54), 
we meet with vipvamitrdh, which the pada divides, vipva-mitrdh* It 
is only, then, in regard to this last word that we need in our rule the 
implication of bahulam by inference from the one preceding, and it is 
so very difficult to justify the implication, with no ca in the rule to in- 
dicate it, and with nothing in any following rule to denote its cessation, 
that I am inclined to think that the passage containing vipamitrdh was 
overlooked by the authors of the treatise, or did not occur in the text 
they recognized, and that the interpretation of the rule has been modi- 
fied by the commentator in order to bring it in. 

The commentator cites, as instances of the long vowel, vipvdnare 
akramata (iv. 11. 7), yd vipvdvasum (ii. 2. 4), and vifvdmitra jamadagne 

(xviii. 3. 16) ; as example of the short vowel, tarn areata vipvamitrdh 
xviii. 3. 63 ; 4. 54). 

SpTt Q\ li \o u 

10. Also as to that of gvan before pada. 

The commentator cites all the instances of occurrence of the com- 
pound of these two words which the text contains, three of them as 
examples of the long vowel, and one of the short ; they are as follows : 
atho sarvaih pvdpadam (xi. 9. 10), pvdpado makshikdh (xi. 10. 8), and 
uta v& pvdpadah (xviii. 3. 55) ; and, on the other hand, vydghrah pvapa- 
ddm iva (viii. 5. 11 and xix. 39. 4), All these forms admit of being de- 
rived from a theme pvapad, instead of pvapada, and the last of them 
must necessarily come from pvapad. It would be possible, then, to get 
rid of the necessity of implying an anuvrlti of bahulam in this rule, 
as the commentator does, by regarding it as relating only to the cases 
in which pvan precedes pada, and not to that in which it precedes pad. 
There is the same strong objection to the inference of bahulam here as 
in the preceding rule : that nothing in the rule indicates it, and that 
the next one implies not bahulam, but simply dtrghah, without anything 
to point out the cessation of the former and the resumption of the 
latter. Such ambiguity is quite foreign to the usual method of the 

The pada-text reads, in each of the passages cited, precisely like the 
sanhitd, without division of the compound, so that the rule might be 
omitted altogether without practical loss. 

* The first time, all the manuscripts, without exception, accent vipvdmitrdh, and 
the edition follows their authority : in the repetition of the verse, Bp. and M. and 
the edited text have vipvamitrdh, B. and P. vipvamitrdh ; this part of the eighteenth 
bonk has been lost in E., and is unaccented in I. The word is evidently a vocative, 
and the true reading is vipvamitrdh, without accent, in both passages. 

iii. 12.] Atharva-Veda Pr&tiq&khya. 459 

surfer HifHHl ^ ii nil 

11. A final alterant vowel of a preposition is lengthened be- 
fore t of the root dd. 

The only two words in the text falling under this rule are aprattttam 
(vi. 117. 1) and parittah (vi. 92. 2), in which the final i of prati and 
pari becomes i before the t which is all that remains of the root dd before 
the participial suffix to. The commentator mentions both, and likewise 
nittd, vitta, and parittih, no one of which is to be found in the Athar- 
van ; nor are his counter-examples, prattam and avattam, Atharvan 
words. The rule has a more general form of statement than would be 
required if it were constructed only to fit the cases which our text pre- 
sents : and we may also, indeed, recognize in the fact of its presence at 
all the influence of the general grammar : the words to which it relates 
are read alike in pada and in sanhitd, and none of the other treatises 
has anything corresponding to it: compare Panini, vi. 3. 124. 

12. As also before varta etc. 

What words we are to assume to be implied in this rule, by inference 
from those which precede it, is not entirely clear. The commentator's 
understanding of it we are unable to arrive at, for there is here another 
slight lacuna in the manuscript : the copyist has jumped carelessly from 
the vartddishu of the first statement of the rule to that of the para- 
phrase, so that the greater part of the latter is lost. The instances 
given in the commentary are abhivartena (i. 29. 1 : p. abhi-vartend), and 
vifvam any&m abhivdra (i. 32. 4 : p. abhi-vdra). We may with plausi- 
bility conclude from this that we are to imply in the rule not merely 
the dirghah, ' conversion into a long vowel,' which belongs to the whole 
section, but the specific limitation to a preposition ending in an alterant 
vowel, given in the last mle ; and that the commentary would read, if 
restored, as follows : vart&dishu [: upasargasya ndmyantasya vartddishu] 
paratah dirgho bhavati. The words, then, to which the rule relates, 
would be, besides those already cited : vibarha (ii. 33. 7 : p. vi-barha), 
abhtvarga (e. g. iii. 5. 2 : p. abhi-varga), parigdsa (v. 14. 3 : p. pari-gasa), 
pratibodha (e. g. viii. 1. 13 : p. prati-bodha), prativarta (e. g. viii. 5. 4: 
p. prati-varta), anuvrj (ix. 4. 12 : p. anu-vrj), pratikdfa (ix. 8. 6 : p.prati- 
-kdga), abhivrta (e. g. ix. 10. 7 : p. abhi-vrta), parivrta (e. g. x. 8. 31 : p. 
pari-vrta), samparivrta (x. 2. 33 : p. sam-parivrta), pratihdra (xi. 7. 12 : 
p. prati-hdra), nivid (xi. 7. 19 : p. ni-vid), abhimodamud (e. g. xi. 7. 26 : 
p. abhimoda-mud), abhildpalap (xi. 8. 25 : p. abhildpa-lap), anurddha 
(xix. 15. 2), parinah (xix. 48. 1), and nin&ha (xix. 57. 4) — all which are 
separated by the ^>arfa-text (except in the nineteenth book, which has 
no pada-text), with restoration of the original short vowel of the prepo- 
sition. In the same category would belong anujahire (p. anu-jahire), 
which is the reading of all the Atharvan manuscripts at xviii. 3. 46, but 
which in the published text is altered to anuhire, to accord with the 
Rik reading of the same passage (x. 15. 8), for which the other seems a 

460 W. D. Whitney, [Hi. 12- 

blundering substitution. In a single word, nihdra (e. g. vi. 113. 2), the 
pada does not restore the theoretically correct form, but reads with the 
sanhitd : possibly nindha, were there a pada-text for the passage, would 
be treated in the same way. There are a few cases where a final a of a 
preposition is also lengthened before a root, but in these the pada-text 
attempts no restoration of the regular form, and their omission from the 
rules of the Pratigakhya would therefore be of no practical importance : 
they are prdnaha (ix. 3. 4), prdvrsh (e. g. xii. 1. 46), prdvrta (e. g. xii. 
5. 2) ; and upanah (xx. 133. 4) doubtless belongs to the same class. 

There are, however, still remaining a few compounds in the Athar- 
van text, the final vowel of the first member of which is lengthened in 
sanhitd, and which are not treated of in any of the rules which follow ; 
so that, if the rule now in hand is to be so interpreted as to exclude 
them, no provision would appear to have been made for them in our 
treatise ; they are samdhanajit (e. g. v. 20. 3 : p. samdhana-jit), ukthd- 
mada (v. 26. 3 : p. uktha-mada), tarddpati (vi. 50. 3 : p. tarda-pati), sa- 
hasrdposha (e. g. vi. 79. 3 : p. sahasra-posha ; at vi. 141. 3 we have sahas- 
raposha), dhutivrdh (vii. 32. 1 : p. ahuti-vrdh), svdvasum (vii. 50. 3 : p. 
sva-vasum : EV. reads in the corresponding passage svavasam), naghd- 
risha (e. g. viii. 2.6: p. nagha-rishd), pavinasa (viii. 6.21: p. pavi-nasa), 
puritat (e. g. ix. 7. 11 : p. puri-tat), visMvrt (x. 2. 11 : p. vishu-vrt), pu- 
riivrt (x. 2. 11: p. puru-vrt), puruvasu (xiv. 2. 47: p. puru-vasu), and 
ur&nasa (xviii. 2. 13 : p. uru-nasa). 

13. The a of a reduplication is or is not lengthened. 

As instances of the prolongation of a in a syllable of reduplication, 
the commentator gives us dddhrshuh (i. 27. 3 : p. dadhrshtth), abhivdwr- 
dhe (i. 29. 1 : p. abhi-vavrdhe), vdvrdhete (v. 1. 5 : p. vavrdhete), and 
jitasya vdvrtuh (v. 19. 13: p. vavrtuh). The same prolongation takes 
place in the Atharvan in sdsahe (xix. 34. 5) and sdsahdnah (iii. 6. 4), 
cdklpat (vi. 35. 3), cdkljwh (ix. 10. 19), and cdklpe (e. g. vii. 87. 1), and 
tAtrshuh (xviii. 3. 47) ; and also, in the twentieth book (xx. 127. 3), in 
m&mahe. As examples of the short a in the same position, the com- 
mentator cites anena vifvd sasahe (i. 16. 3) and sasahe fatriln (ii. 5. 3). 
The root sah is the only one in which, while the ^>atfa-text gives always 
the short vowel to the reduplication, the sanhitd sometimes prolongs it 
and sometimes leaves it unchanged. Elsewhere we have in sanhitd. 
either always the long d (and especially often in the forms of vdvrdh), 
or always the short a. 

There are also two or three roots in the reduplications of which other 
vowels than a, short in the pada-text, are lengthened in sanhitd : they 
are rvirupah (iv. 7. 5), ririshah (e. g. v. 3. 8), and f&fucah (xviii. 2. 4). 
The Pratic&khya may intend to include these forms in the gana of rule 
21, below, but they would much more properly have been provided for 
in the present rule. 

sft#TT^[ II \d II 

14. Note jihtdd 'ham. 

iii. 16.] Atharva- Veda Pr&ticakhya. 461 

Or, as the commentator expounds it, there is in this passage a length- 
ened vowel — viz., that of the first syllable of jihida. The passage is 
cited: akratur jihida 'ham* (iv. 32. 5: p. jihida). The commentator 
regards the word bahulam as still in force from the preceding rule, and 
adds, as an instance in which a short vowel appears in the reduplication 
of this root, yad vA pita 'paraddho jihide (vi. 116. 2). But it is evident 
that there is no need here of any implication of bahulam, and that all 
cases but one of the occurrence of jihida are excluded from the rule by 
the citation in the latter of the accompanying word aham. 

Compare farther rule iv. 87, which teaches sam&patti of jihtda in the 

H\^\H H VI II 

15. And scffiydma. 

The commentator quotes from the text ; s&hy&ma dasam aryam (iv. 
32. 1 : p. sahydma). He does not attempt to prove the implication of 
bahulam by anuvftti from rule 13, although there is just as good reason 
for doing it here as under the preceding rule. Compare the converse 
of this rule in the next chapter (iv. 88). 

There are in the Atharvan text a few causative forms from verbal 
roots which show in the first or radical syllable a short a in pada and a 
long d in sanhitd, and which are not specially noted in this section, 
being left, apparently, to fall into the gana of rule 21, below, although 
no reference is there made to them by the commentator. They are 
ydvaya etc. (e. g. i. 20. 3), from yu; cyavayati etc. (e. g. x. 1. 13), from 
cyu; vanayantu (vi. 9. 3), from van; y&maya (vi. 137. 3), from yam; 
and gl&payanta (ix. 9. 10), from gl&. Directions for the shortening of 
their first syllables in pada are given in the next chapter (iv. 91-93). 

R(41l{iHI k\{\\{% II UH 

16. The final vowels of vidma etc. are lengthened before $ara etc. 

This rule, together with two later ones of this section (19 and 25) — 
of which the one relates to the final of an instrumental case, and the 
other to that of the particle adha — makes up the sum of all that our 
treatise has to say respecting the frequent and important phenomenon 
of the prolongation in sanhita of a final vowel. The other Prati$a- 
khyas treat the subject at great length and apparently with exhaustive 
fullness (see R. Pr., chapters vii and viii ; V. Pr. iii. 96-128 ; T. Pr. iii). 
As regards our own treatise, we must confess that we can hardly help 
questioning the actuality of all its ganas, when we find set up in the 
rules such as would have to be of so immense extent as the two here 
presented us, which must contain hundreds of words eacb.f It were 

* The edition reads erroneously jihidd 'ham, with the corresponding Rik passage. 
All the sanhitd manuscripts of the Atharvan give jihida, and in the pada-text the 
word is followed by a figure 4, the usual sign of a samdpanna word, or one which 
the pada has reduced from an irregular to a theoretically regular form. 

f The longest gana belonging to Panini's grammar, ardharcadayah, contains, in 
Bohtlingk's edition, 232 words. 

462 W. D. Whitney, [iii. 16. 

almost as well, we should think, to turn off the matter with a hvaeit, as 
is done below, in rule 54, openly confessing inability to master its intri- 
cacy and vastness, as to dispose of it thus. 

The commentator does not feel called upon to give himself any more 
trouble than the authors of the treatise have taken, and he leaves the 
two ganas to take care of themselves, after citing three passages by 
way of examples, as follows : vidmA parasya pitaram (i. 2. 1), evd rogaih 
cd "srdvam (i. 2. 4), and evd me prdna md bibheh (ii. 15. 1 etc.). In 
filling out the gana, and giving a complete account of the usage of the 
Atharvan text with reference to the prolongation of a final vowel, I 
shall put together all the cases, including those which belong under 
rules 19 and 25, since the same principle evidently governs them all. 

The prolongation is so prevailingly a metrical phenomenon in the 
Atharva-Veda, called out by the exigencies of the verse, that I shall 
arrange my exhibition of it accordingly. There are certain points in 
the pdda, whether of eight, of eleven, or of twelve syllables, at which the 
long vowel is especially liable to be called forth : these are, 1st, toward 
the beginning of the pdda, of whatever length, the second syllable ; 2nd, 
toward the end, the sixth syllable in a pdda of eight syllables, and the 
eighth or the tenth in one of eleven or twelve syllables : the protrac- 
tion evidently tending in the former case to give an iambic movement 
to the commencement of the verse, and, in the latter case, to impress 
upon it an iambic cadence — which, however, is in the 11 -syllable pdda 
made trochaic by the addition of an odd syllable. A long syllable at 
these points, then, I shall regard as regular ; elsewhere, as irregular. 

L The pdda of eight syllables (rarely lengthened to ten), usually 
forming anushtubh verses, but, less often, combined into gdyatrt, pankti, 
or brhatt stanzas : these do not require to be distinguished from one 
another, since the prolongation depends solely upon the interior con- 
struction of the pdda itself, and not upon the mode of combination of 
the latter. 

1 . Protraction of a final vowel in the second syllable of the pdda, as 
in the example vidmd parasya pitaram (i. 2. 1) : 

a. Of verbal forms : vidmd (i. 2. 1 ; 3. 1-5), pibd (ii. 5. 1), and tishthd 

b. Of forms of declension : tend (i. 3. 1-5. iii. 7. 3 ;* 14. 1. iv. 5. 1 ; 
7. 1. vi. 7. 1, 2, 3 ; 12. 1, 2 ; 24. 3; 80. 1, 3; 82. 3 ; 91. 1. vii. 38. 2. 
xviii. 2. 30), and yend (i. 13. 1. iii. 9. 4. vii. 38. 2. ix. 5. 17. xiii. 2. 21). 

c. Of particles: evd (i. 2.4; 3.6-9; 11.6. ii. 15. 1-6; 30.1; 36.4. 
iii. 6. 3, 6. v. 21. 4-6. vi. 8. 1, 2, 3 ; 17. 1-4 ; 46. 3 ; 58. 2 ; 70. 1-3 ; 
85.3; 102.1; 105.1-3; 138.5; 139.4,5; 141.3. vii. 13. 1; 50.5. 
ix. 1. 11, 12, 13, 16, 17. x. 1. 13 ; 3. 13-15, 17-25 ; 6. 33. xi. 4. 19; xii. 
4.34; 5.65. xiv. 1.43. xviii. 4. 55. xix. 50. 4; 57. 1), adhd (i. 28. 4. iii. 
20. 1. vi. 65. 1. x. 4. 25. xiii. 1. 30. xiv. 2. 20. xviii. 2. 23), yatrd (iii. 
28. 6), and ghd (vi. 1. 3). 

2. Protraction in the sixth syllable, or the third from the end of the 
pdda, as in imam vardhayatd girah : 

a. Of verbal forms : vardhayatd (i. 15. 2. xix. 1. 1, 2),janayalh& (i. 5. 

* The edition, following the authority of a part of the MSS., Una. 

iii. 16.] Aiharva-Veda Prdtigdkhya. 463 

3), ilayatA (i. 17. 4), yAvayA (i. 20. 3 ; 21. 4. iv. 19. 7. xii. 1. 32), gamayA 
(i. 21. 2), nApayA (i. 23. 2, 3), krnutA (iii. 9. 1. xii. 2. 34), anadata (iii. 
13. \),jayatA (iii. 19. 7), vardhayA (iii. 20. 1. vi. 5. 3 ; 54. 1), /Away* (iv. 
4. 6. vi. 101. 2), shvApayA (iv. 5. 7), kalpaya (iv. 12. 5), nayathA (iv. 13. 
l),jivayath& (iv. 13. 1), chd.payA (iv. 18. 4), d&shayatd (vi. 100. 2), dyd 
(vi. 103. 2, 3; 104. 2), caifcnrad! (vi. 114. 1 ; 115. 1. x. 3. 8), mAdayA (vi. 
130. 4), shy A (vii. 18. 1), bhavatA (vii. 60. 7), jaghnimA (x. 4. 12), 6Acy'<i 
(xiv. 1. 25), vApayA (xviii. 3. 6), car A (xx. 127. 11). 

b. Of forms of declension : sahasyenA (iv. 5. 1), and bhadrenA (vii. 
60. 7). 

c. Of particles : ivA (vi. 37. 2). 

3. Besides cases of the two kinds already noted, we have a few in 
which the protraction still favors the iambic movement of the verse, 
although not at its two cardinal points : 

a. In the fourth syllable of the pAda, as in harinasyA bhiyam hrdhi : 
of verbal forms, parshA (iv. 33. 8), and prnotd (xx. 127. 7) ; of forms of 
declension, dhAmA (vi. 31. 3), and harinasyA (vi. 67. 3); of particles, 
anga (ii. 3. 2), and achA (iii. 20. 2). 

6. In the final syllable of the pAda: of this class we have but two 
cases, viz. TcalpayA (iv. 12. 5), and anga (ii. 3. 2) ; and in each passage 
the same word occurs in another position, protracted according to the 
usual analogies, so that we may regard the irregular protractions in the 
final syllable as the effect of attraction. 

4. Protraction of a final vowel in such a position that it seems to ob- 
struct, rather than assist, the regular movement of the verse : 

a. Protraction in the third syllable of the pAda, as in prathayA s&- 
shane tvam : of verbal forms, prathayA (i. 11. 3), mrdayA (i. 13. 2 ; 26. 4), 
itd, (iii. 19. 7), sunotA (vi. 2. 3), juluotA (xviii. 2. 2) ; of particles, smA (x. 

b. Protraction in the fifth syllable, or in the fourth from the end : 
only a single case, ivA (iv. 4. 7, third pAda).* 

A summary view of the cases of protraction in the 8-syllable pAda 
(including also the very rare lO-syllable^xirfa) is as follows: 

Regular. Indifferent. Irregular, 

Syllable, 2nd. 6th. 4th. 6th. 3rd. Sth. 

Verbal forms, 8 40 2 1 6 

Forms of declension, 26 2 2 1 

Particles, 77 1 2 1 10 

Sum, 111 43 6 2 7 1 

Total, 154 ~8~ ~8~ 

II. In the pAda of eleven syllables, with trochaic close ; usually form- 
ing regular trishtubh verses, but not unfrequently irregularly combined, 
especially with jagati pAdas, of twelve syllables : 

1. Protraction at the commencement of the pAda, in the second syl- 
lable : e. g. vidmA tarn ■utsam yata AbabkUtha. 

a. Verbal forms: vidmA (iv. 31. 5. x. 1.20. xiii. 3. 21), yuhshvA (xviii. 

* The edition reads, with a part of the MSS., iva. 

464 W. D. Whitney, [iii. IS. 

1. 25), vahd (xviii. 1. 30), mrdd (xviii. 1. 40), dcyd (xviii. 1. 52), hhava 
(xix. 24. 5), and frudhi (ii. 5. 4. xviii. 1. 25). 

b. Forms of declension : tend (iii. 16. 5. vii. 20. 4 ; 79. 1), yend (vii. 
12. 1. xviii. 1. 54 ; 4. 44), and kshdmd (xviii. 3. 21). 

c. Particles : evd (iv. 39. 1, 3, 5, 7. vi. 72. 1 ; 74. 3. xii. 2. 25), arfAd 
(iii. 4. 4. iv. 32. 7. v. 22. 2, 2. vii. 73. 11. viii. 4. 15. x. 2. 7. xii. 3. 2, 9. 
xviii. 1. 16, 51 ; 2. 11 ; 3. 21 ; 4. 48, 70), yatrd (iii. 28. 5. vi. 22. 2 ; 120. 
3. ix. 9. 22. xviii. 1. 50), atrd (v. 1. 5. ix. 10. 12. xii. 2. 26, 27), adyd 
(viii. 4. 15. ix. 10. 9), ghd (xviii. 1. 11), achd (vi. 39. 2), and smd (xii. 
3. 3). 

2. Protraction in the cadence of the pdda. Here we have two classes 
of cases to distinguish, viz., those in which the eighth syllable, or the 
fourth from the end, suffers protraction, as in divam gacha prati iishthd 
farirdih ; and again, those in which the tenth syllable, or the last but 
one, is made long, as in md vah susroc camaso drnhatd tarn. 

a. Protraction in the eighth syllable. 

a. Verbal forms: vivegd (i. 12. 3. xi. 4. 20), iishthd (ii. 34. 5. xviii. 2. 
7), bhajd (iii. 4. 2, 4), avd (iii. 16. 3), medayathd (iv. 21. 6), bhard (iv. 32. 
3), svadayd (v. 12. 2), srjd (v. 27. 11), juhutd (vi. 32. 1), rakshd (viii. 3. 
19), nahyd (x. 9. 1), taratd (xii. 2. 26, 27), nudatd (xii. 2. 30), pacatd 
(xii. 3. 27), bhavatd (xii. 3. 29), srjatd (xii. 3. 46), cakrmd (xviii. 1. 51), 
diyd (xix. 13. 8), and avaid (xix. 13. 11). 

b. Forms of declension: asyd (i. 12. 2), janimd (ii. 28. 2. iv. 1. 3. 
xviii. 3. 22), amrtend (iii. 12. 8), kdvyend (v. 1. 5), and martyend (ix. 10. 
8, 16). 

c. Particles: cd (vii. 4.1. x.8.12), adyd (vii. 82. 6), and utd (vii. 97.1). 
@. Protraction in the tenth syllable. This case is comparatively rare r 

and is nearly confined to verbal forms, of which we have mrdatd (i. 20. 
1. xviii. 3. 16), bhavd (ii. 6. 3. iv. 32. 7), srjd (v. 2. 3), hantand (vii. 77. 
2), and drnhatd (xix. 58. 4) : besides these, we find only ihd (xix. 56. 6). 

3. In the trishtubh pdda, as in the anushtubh, we have in a few odd 
cases the fourth syllable lengthened, as in tirahpur'b eid arnavam jagan- 
van; they are as follows: verbal form, iishthatd (xii. 2. 27) ; form of 
declension, purii (xviii. 1. 1) ; particle, ghd (xviii. 1. 3). 

4. The irregular protractions are found to take place in the third, the 
fifth, or the seventh syllable of the pdda. 

a. Protraction in the third syllable, as in urushyd na urujmann apra- 
yuchan : but two cases occur, viz. urushyd (vi. 4. 3), and vyathayd (xiii. 
1. 31), and in the latter passage the metre is hopelessly irregular. 

b. Protraction in the fifth syllable, as in milra$ cid dhi shmajuhu- 
rdno devdn : of this class, we have only cases of particles, viz. prd (ii. 
5. 5. vii. 26. 1*), and shmd (xviii. 1. 33). 

c. Protraction in the seventh syllable : an example is na yat pur& 
cakrmd kad dha nttnam. The words of which the final is lengthened 
in this position are, with a single exception, verbal forms, viz. krnuthd 
(iv. 21.6), bhard (iv. 22. 6. vii. 73. 9), khidd (iv. 22. 7), crtd (vi.'63. 2; 
84. 3), suvd (vii. 14. 3), nudd (vii. 34. 1), and cakrmd (xviii. 1. 4, 33). 
Of forms of declension, we have janimd (v. 11. 5). 

* In both these passages the printed text reads pro, without any support from 
the manuscripts. 

its. 16.] Atharva-Veda Prdtiqdkhya. 465 

The tabular summary for the 11-sy liable pada is as follows '. 

Verbal forms, 
Forms of declension, 


2nd. Stk. 10th. 

10 23 1 

7 8 

36 4 1 




3rd. 5th. 7th. 
2 10 


63 35 8 


2 3 11 

Total, 96 3 16 

III. In the jagatl pada, or that of twelve syllables. 

1. Protraction at the beginning of the pada, in the second syllable, 
as in piksha no asmin puruhuta y&mani. Here we have, of verbal forms, 
pkshd (xviii. 3. 67) ; of forms of declension, purfo (vi. 49. 3) ; of parti- 
cles, evA (iv. -23. 2. vi. 97. 1), adha (vii. 83. 3. xviii. 4. 63), and yadt 
(xviii. 1. 21). 

2. Protraction in the cadence of the pada, taking place, as in the 
(rishtubh pada, at two different points, either at the eighth syllable or 
at the tenth ; that is, at the fifth or the third from the end. 

«• At the eighth syllable, as in asmakam anpam ud av& bhare-bhare: 
of such cases, the text affords us the verbal forms munch (ii. 35. 3), 
krnuta (vi. 55. 3), ruhema (vii. 6. 3), ava /vii. 50. 4), and icha (xviii. 1. 
16); and the form of declension" dharma {vi. 51. 3). 

§■ At the tenth syllable, as in t&h saptagrdhrA iti pufrumA vayam : 
here we have only the verbal forms sincathA (vi. 22. 2), dhAraya (vii. 
82. 3), pufruma (viii. 9. 18), vepayA (ix. 2. 25), and sidaiA (xi. 1. 12). 

3. Of the protraction in the middle of the verse, or at the fourth syl- 
lable, there is to be found but a single case, yud ejathd maruto rukma- 
vakshamh (vi. 22. 2). 

4. Protraction of a final against the requirements of the metre, in 
the seventh syllable, or the sixth from the end. Only one unquestiona- 
ble case can be found, viz. pary u shu pra dhanvA vAjasAtaye (v. 6. 4) ; 
with which may be classed grnatA, at v. 27. 9, in a verse of irregular 
character, and perhaps rather belonging under the 8-syllable pAda. 

The cases of the 12-syllable pada may be thus summed up : 


Syllable, 2nd. 8th. 10th. 

Verbal forms, 15 5 

Forms of declension, 110 

Particles, 5 

Sum, 7 6 5 









Total, 18 1 2 

There thus appear to be, in the whole text, 268 cases of the protrac- 
tion of a final vowel to help the metrical movement in the opening or 
the cadence of the pAda, and 12 which favor the movement in the 
middle of the pada, against 26 which contravene the orderly flow of 
the metre. To point out the cases in which protraction does not take 
place at the two important points at which it is thus shown usually to 
occur, and to establish the laws, if there be any, which determine in 
each instance the retention or the change of the short vowel, cannot, of 
course, be attempted here. 

466 W. D. Whitney, [iii. 16- 

It will be noticed that the vowel which is lengthened is almost always 
o : the text furnishes us, indeed, but three examples of the protraction 
of a final i (ii. 5. 4. xviii. 1. 21, 25), and two of that of u (vi. 49. 3. 
xviii. 1. 1) — except the particle u, as taught in rule 4 of this chapter. 

3$r*<f Rrlt II \© II 

17. Before the suffix mant a vowel is or is not lengthened. 

The term matu is treated by the commentary as covering both the 
suffix mant and its equivalent vant : in fact, he cites instances only for 
the latter, as follows : f&le 'pv&vati (iii. 12. 2), afv&vattr gomatih (iii. 16. 
7), and agv&vatlm pra tara (xviii. 2. 31) ; and farther, as examples of 
the short vowel before the suffix, vtravatih sadam (iii. 16. 7), and ghrta- 
vatt payasvaty uc chrayasva (iii. 12. 2). 

The complete list of words in our present Atharvan text which ex- 
hibit in sanhita a prolongation of the final vowel of a theme before the 
possessive suffix vant is as follows : apv&vant (e. g. iii. 12. 2), varandvant 
(iv. 7. 1), mad&vant (e. g. iv. 7. 4), sam&vant (iv. 18. 1), priy&vant (iv. 
18. 4), uttardvant (e. g. iv. 22. 5), viry&vant (e. g. iv. 37. 5), vrshnydvant 
'v. 25. 8), bhangurdvant (vii. 71. 1), ufitdvant (ix. 6. 38), rocanavant 
xiii. 3. 10), ishtdvant (xviii. 3. 20), kladtvant (vii. 90. 3), and vishuvant 
ix. 3. 8). In all these words, the pada-text restores the final vowel of 
the theme to its short form, and (by iv. 1 7) separates the theme and the 
suffix by avagraha.* For the suffix mant we have but a single word, 
tvishimant (e. g. iv. 19. 2 : p. tvishi-mant) : and so also for the suffix van 
— viz. satydvan (p. satya-van: iv. 29. 1, 2) — respecting which I do not 
know whether it should be regarded as included under the technical 
designation matu. There are a few words in which the long vowel 
before the possessive suffix is properly regarded as rightfully belonging 
to the theme, and so is left unshortened in the pada-texi : such are 
dalcshin&vant, fitik&vant, hlddikdvant, ulkushimant, and jyotishimant. 

Of the words mentioned above, only two — viz. agvavant (vi. 68. 3. 
xviii. 3. 61) and viryavant (viii. 5. 1. xviii. 4. 38) — ever show in sanhitd 
forms having the vowel short. The former of them is specially noted 
in the Vaj. Pr. (iii. 97) as doing so in a single instance. 

^mt sr n*\{\$\ iivrn 

18. As also before a suffix beginning with y, in a desiderative 

In this rule, the anuvrtti of bahulam is duly indicated, by the inser- 
tion of ca. As examples of desiderative forms exhibiting the length- 
ened vowel before y, the commentator cites adhvariyatam (i. 4. 1 : p. 
adhvari-yat&m), vrshdyamdmh (ii. 5. 7 : p. vrsha-yamdnah), and patru- 

* Oar pada MS. does, indeed, read rocanavant (xiii. 3. 10), without separation : 
but this is, it can hardly be doubted, a mere error of the scribe. 

iii. 19.] Atharva-Veda Prdligdkltya. 467 

yatim abhi (iii. 1.3: p. patru-yatim) : as examples of the short vowel, 
he gives ardtiydt (iv. 36. 1 : p. ardti-ydt), janiyanti (xiv. 2. 72 : p. jani- 
-yanti), putriyanti (xiv. 2. 72 : p. putri-yanti), and mrgayuh (x. 1. 26 : 
p. mrga-yuh) — adding to the last word prabhrtdni ca, which I take to be, 
not a part of this or of any other citation, but (with amendment to 
prabhrtini ca) a simple expression for ' et cetera , although the com- 
mentator does not anywhere else, I believe, give us such an intimation 
that more examples might be had for the seeking. The only ones of 
this class which our text farther presents are sukshetriyd and sugdtuyd 
(both at iv. 33. 2 : p. su-kshetriyd : su-gdtuyd), besides amitrayu, in the 
twentieth book (xx. 127. 13). Of cases of the long vowel, we have also 
aghdyant (e. g. x. 4. 10 : p. agha-yant) and aghdyu (e. g. i. 20. 2 : p. 
agha-yu: in abhyaghdydnti [v. 6. 9 and vii. 70. 3*] the pada-text leaves 
the vowel long, writing abhi-aghdydnti), vay&yd (iv. 33. 2 : p. vasu-yd), 
papathiyant (e. g. v. 14. 5 : p. capathi-yant), janiyant (vi. 82. 3 : p. ja- 
ni-yant), amitrdyant (vii. 84. 2 : p. amitra-yant : the Rik, in the corres- 
ponding passage [x. 180. 3], has the short vowel), kaviyamdna (ix. 9. 
18: p. kavi-yamdna), and nipriydyate (xii. 4. 11 etc.: p. ni-priyayate) : 
satvandyant (v. 20. 1) seems to belong to the same class, but it is written 
by the pada with the long vowel, satvand-yan. From the themes amitra 
and jani, it will be noticed, come desiderative or denominative forms of 
either class, or with both the short and long vowel. 


19. As also the final vowel of an instrumental case. 

In the form of this rule there is nothing which continues the impli- 
cation of bahulam, but such an implication is, of course, unavoidably 
necessary, and is made by the commentator. As instance of the long 
vowel, he cites yend sahasram vahasi yend 'gne sarvavedamm (ix. 5. 17) ; 
as instance of the short vowel, kena protriyam dpnoti (x. 2. 20). All 
the cases occurring under this rule have been detailed above, in the note 
to rule 16 : excepting tend and yend, which are tolerably frequent (tend 
occurs 24 times, and yend 8 times), we have, of instrumental with 
lengthened finals, only a few sporadic instances, viz. sahasyend, bhadrend, 
amrtend, kdvyend, and martyend, each in a single passage. Of other 
forms of declension which undergo a like prolongation, the text "affords 
us two genitives, asyd and harinasyd; and, as accusative singular, puru 
is twice lengthened into purk. Besides these, we find only such forms 
as janimd for janimdni, in which the long vowel may be more properly 
regarded as thematic and not metrical. This latter class of cases, how- 
ever, has been reckoned in with the rest under rule 16, because it is so 
treated by the ^wrcfa-text, which writes the final vowel shortf In a 
single instance (brahma : v. 2. 8) the sanhitd preserves the short vowel. 

* Amended in the published text to abhyaghdydti. 

f Excepting in two cases, viz. varimd (iv. 26. 2), and ndmd (ix. 9. 3). I do not 
at all understand the reason of these exceptions, and they may possibly be mi>!- 
readings of our pada MSS. ; the pada-text reads noma at v. 1. 3, and in the Rik 
passage (i. 164. 3) corresponding with ix. 9. 8, both pada and sanhitd give ndma. 

468 W. D. Whitney, [iii. 2 °- 

20. When r is dropped, the final vowel is lengthened. 

The other treatises (R. Pr. iv. 9, r. 29, ccxlviii ; V. Pr. iv. 34 ; T. Pr. 
viii. 17), with better reason, give this rule in connection with that for 
the omission of the final r, which takes place (by ii. 19) only before a 
following initial r. The commentator repeats once more the same series 
of quotations which he has given already twice before, under i. 28 and 
ii. 19, only excepting the first, faradah puritcl rdyah (ii. 13. 3) : he then, 
as if in surprise at his own audacity, asks himself why he has omitted 
this ; and makes reply that, although it offers an instance of the loss of 
a final r, it shows no protraction of the vowel, which was long already. 1 

hi^i{Ihi wmi u ^ ii 

21. Of n&raka etc., the first vowel is lengthened. 

The commentator cites only three cases under this rule, viz. n&raJcam 
(xii. 4. 36 : p. narakam), sddanam (e. g. ii. 12. 7 : p. sadanam), and Asata 
indra (viii. 4. 8 : p. asatah). The first word occurs only once in the 
Atharvan; the second is found several times, but sadana, with short 
antepenult, still oftener ; for dsat we have also two other passages in 
the same hymn with the one quoted (viii. 4. 12,13). Besides these, 
the word of most frequent occurrence, belonging to the same class, is 
purusha, which usually* becomes purusha at the end of a pdda (as 
noted by the Vaj. Pr., iii. 118), or where its first syllable has that place 
at which a long vowel is especially needed in the cadence of the pdda. 
Moreover, the text offers us s&yamart (iv. 27. 1 : p. su-yam&n\), and s&- 
yavasa (e. g. iv. 21. 7 : p. su-yavasa), with its compound siiyavasad (vii. 
73. 11 : p. suyavasa-at). Causative forms, such as y&vaya (p. yavaya), 
which the treatise perhaps meant to include in this rule, have been 
spoken of above, under rule 15. 

{i<ldl{Ml ferflMHI II *X » 

22. Of diddyat eto., the second vowel. 

The commentator's examples are diddyat (iii. 8. 3 : p. didayat), ushdso 
vtravatih (iii. 16. 7 : p. ushasah), and ushdsdnaktd (e. g. v. 12. 6 : p. 
ushasdnaktd). The latter compound occurs several times in our text : 
from the theme ushas we have also once (xiii. 2. 46) the accusative 
ushdsam (p. ushasam). The only other word of like character which I 
have noted in the Atharvan is frathdya (vii. 83. 3 : p. prathaya). A 
part of the compounds exhibiting protraction in the final syllable of the 
first member, and of which a detailed list has been given above, under 

1 him artham idam noddrddhate : puruci rdyah : yady api ralopo dirgha evdivah. 
The reading is very corrupt, but the thing aimed at is, I believe, clear. 

* In the Atharvan only usually, and not universally : exceptions are v. SO. 2. 
viii. 1.-2. xi. 8. 32. xii. 4. 13. xix. 6. 1, and one or two others. 

f The edition reads, with I. and H, suyamdn. 

lii. 25.] Atharva- Veda Prdticdkliya. 469 

rule 12, might be regarded as falling under this rule; but as a part of 
them also would not, since it is their third syllable that is prolonged, 
and not their second, I preferred to give them all together in that place. 

^NWI^l{iHl^rl^<I^IW li # II 

23. Of the compounds sdtrdsdha etc., the first vowel of the 
second member. 

The instances cited by the commentator are sdtrdsdhasya (v. 13. 6: 
p. sdtrd-sahasya), amitrasdhah (e. g. i. 20. 4 : p. amitra-sahah), and vi- 
shdsahih (e. g. i. 29. 6 : p. vi-sasahih) ; all of them from the root sah. 
The text furnishes another derivative from the same root belonging 
under this rule, viz. abhimdtishdhah (iv. 32. 4 : p. abhimdti-sahah) : the 
only other case which I have found is nyayanam (vi. 17. 2 : p. ni-ayanam). 
The word ukthagdsah, which the pada-texts of the Rik and White 
Yajus write uktha-fasah, and which therefore receives attention from 
their Praticakhyas (R. Pr. ix. 19 ; V. Pr. iii. 122), is read in our pada 
(xviii. 3. 21) uklka-fdsah. 

Wf epjcl(io1H^ II \$ II 

24. Of ria, the final vowel, before vrdh, van, and van. 

The instances cited by the commentator are rldvrdhah (e. g. xi. 6. 19 : 
p. rta-vrdhah), rtdvari yajniye (vi. 62. 1 : p. rla-vari), and rtdvdnam 
vdifv&naram (vi. 36. 1 : p. rta-vdnam) : the same words occur elsewhere, 
in a few passages which it is unnecessary to specify ; they always show 
a long vowel in sanhitd and a short in pada, while the other numerous 
compounds of rta have everywhere the short vowel only. 

^ rq^iq^^T li VI" 

25. Also that of adha, excepting when followed by tyam and 

The commentator's illustrative citations for adha, with lengthened 
final, are adhd yathd nah (xviii. 3. 2 1 ), and adhd pilfnr upa drava (xviii. 
2. 23) : he also quotes the exceptions mentioned in the rule, as follows : 
adha tyam, drapsam (xviii. 1. 21), and adha dhir ajdyata (xviii. ]. 21). 
There is, however, yet another case in the text in which the final vowel 
is left short, and which has been overlooked both by treatise and com- 
mentator: it is adha sydma (xviii. 3. 17). All the passages in which 
adhd occurs have been given above, in the note to rule 16. Except in 
the single case in which it is followed by dhih, the word always stands 
at the beginning of a pdda. The commentator adds from his metrical 
authority a verse containing a virtual repetition of the rule : adhe Hi 
vyanjane dtrgho vatyayitvd tyadhiparam : svardddv api 1 sarvatra vyan- 
jane hrasva eva iu; l adha is long before a consonant, excepting the 
cases in which it is followed by tyam and dhih; but before a consonant 
preceded by a vowel (?) it is everywhere only short.' 

1 t — MS. svaddrdvayya. 

470 W. D. Whitney, [iii. 25- 

Very little can be said in praise of the manner in which our treatise, 
in this section, deals with the intricate and numerous phenomena of the 
irregular prolongation of vowels in the Atharvan text ; its statements 
are greatly wanting in accuracy, in completeness, and in systematic con- 
struction and arrangement. The form of its general rules, which em- 
brace a great number of cases, is objectionable from indefiniteness : the 
cases which are made the subject of its special precepts are chosen 
arbitrarily and with little judgment, and in part are such as required no 
notice ; while, on the other hand, it is doubtful whether one or two 
classes of cases are provided for at all : and finally, in several instances 
it abdicates altogether the office of a Pratic&khya, and, instead of de- 
termining the usage of its school as to the points upon which it touches, 
leaves them as unsettled as it found them. 

As elsewhere in the signature of the separate sections of this chapter, 
no statement is made respecting the number of rules which the section 
contains : the manuscript says simply tfltyasya pratkamah padah. 

c^T% s^R fe> n ^ ii 

26. At the end of a word, a consonant is pronounced double. 

The subject of the duplicated pronunciation of consonants, or of the 
varnakrama, as it is sometimes called, is one of the most peculiar in 
the whole phonetical science of the Hindus. It is also the one, to my 
apprehension, which exhibits most strikingly their characteristic ten- 
dency to arbitrary and artificial theorizing; I have not succeeded in 
discovering the foundation of fact upon which their superstructure of 
rules is based, or explaining to myself what actual phonetic phenomena, 
liable to occur in a natural, or even a strained, mode of utterance, they 
supposed themselves to have noted, and endeavored thus to reduce to 
systematic form. The varnakrama, however, forms a not inconspicuous 
part of the phonetic system of all the Praticakhyas, and is even presented 
by Panini (viii. 4. 46-52), although the latter mercifully allows us our 
option as to whether we will or will not observe its rules. To present 
and compare in full detail the systems of the other authorities in con- 
nection with that of our own treatise would take a great deal of room, 
and, moreover, could be done at best but imperfectly, since our manu- 
script, as will be shown below, exhibits a lacuna of some extent and 
importance in the midst of its treatment of the subject. I shall accord- 
ingly only refer under each of our own rules to those precepts of the 
other treatises which are most nearly related to it. 

If the first rule of the second chapter is still to he strictly applied, 
we must conclude that the makers of the Praticakhya recognized the 
duplicated methods of pronunciation as of force only in the sanhitd- 
patha, and not in the utterance of the disjoined text, or the padap&tha. 
This interpretation is somewhat supported by the fact that both the Kik 
Pr. (vi. 3, r. 14, cccxc) and Panini (viii. 4. 51) attribute to Qakala or 
Qakalya, the teacher to whom the invention of the padap&tha is gene- 
rally ascribed, a denial of all duplicated utterance. 

As regards our first rule, it is directly in contravention with the doc- 

Hi. 28.] Atharva- Veda Pr&tic&khya. 471 

trine of the other treatises (R. Pr. vi. 2, r. 7, ccclxxxiv ; V. Pr. iv. 114 ; 
T. Pr. xiv. 1 5), which unanimously teach that a consonant is not dupli- 
cated in pausa. The commentator's instances are again godhuk, vir&t, 
drshat, trishtup, the same words which we have had adduced so many 
times before (see i. 3 etc) : they are accordingly to be pronounced 
godhukk, virdtt, drskatt, trishtupp; but the manuscript omits, as in 
almost all the examples given under the following rules, to write the 
duplicated letter double. 

3TJFTT ^^taTi ^ II ^ U 

27. As are also n, n, and n before a vowel, when final and 
preceded by a short vowel. 

This is a rule familiar to all students of Sanskrit, as being obligatory 
in the euphony of the later classical dialect, and not in the older lan- 
guage of the Vedas alone. It is equivalently stated by all the treatises 
(R. Pr. vi. 4, r. 15, cccxci ; V. Pr. iv. 104 ; T. Pr. ix. 18, 19), except that 
the others omit the needless mention, along with the other two nasals, 
of n, which never occurs as a final. Panini's praty&h&ra (viii. 3. 32) 
includes all the three. Panini and the Taitt. Pr. very properly treat 
this doubling of a final nasal as something apart from and unconnected 
with the phenomena of the varnahrama, by teaching it in a different 
part of their texts from that which deals with the latter subject ; and 
in the Rik Pr., also, the rule rather follows next after, than is introduced 
among, those which prescribe the other duplications. 

The commentator cites, as examples of a final nasal doubled, pratyann 
rnoti (not in A V.), pratyann ud eshi (xiii. 2. 20), udyann ddityalf (ii. 
32. 1), and sugann dste (no such case in AV. : the instance is also given 
by Panini's scholiasts). To illustrate the necessity of the restrictions 
"when preceded by a short vowel" and "before a vowel," he cites ar- 
v&n dkutyd cara (iii. 2. 3), and udanj&tah (v. 4. 8). Finally, he quotes 
from his metrical authority a verse which restates the rule, with exam- 
ples : nanands tu pad&ntd ye hrasvapurvdh svarodayah : tesh&m dvir- 
bh&vam ichanti pratyann udyann sugann iti. 

28. Also the first consonant of a group, after a vowel. 

In all the other treatises (R. Pr. vi. 1, r. 1, ccclxxviii; V. Pr. iv. 97 ; 
T. Pr. xiv. 1), this is put at the head of the subject of the varnahrama, 
as the fundamental and most important rule. 

The commentator gives as instances the two indifferent words agnih 
(e. g. i. 7. 4) and vrkshah (e. g. iv. 7. 5), which are to be pronounced 
aggnih and vrkkshah. Then follows a lacuna in the manuscript, of some 
extent and importance, since it certainly includes one or more rules. 
After the two quoted examples, namely, follows visarjantyah saslh&ne 
ea, of which the two last words constitute a rule which the commentary 
goes on to expound in full, while the first, visaiyantyah, is the final repe- 
tition of the rule next preceding : and in both of them there is anuvrtti 

VOL. VII. 60 

472 W. D. Whitney, [iii. 28- 

of na, ' not ;' that is to say, the text has passed from giving rules for 
the occurrence of the duplication, to giving those for its non-occurrence. 
As we have found several times before, where there were lesser lacuna, 
that the copyist had skipped carelessly from a word to a like word far- 
ther on, so we may suppose the same to have taken place here: agnih 
and vrkshah were very probably quoted again as examples under the rule 
visarjantyah, and, the copyist's eye falling upon them, he overleaped all 
that intervened. It may be, however, that this conjecture is erroneous, 
and that the lacuna is of greater extent than would fairly admit of such 
an explanation. How many rules it includes, there are no means of 
determining : of this chapter we appear to have remaining ninety-six 
rules, while its signature states a hundred and five to be the number 
contained in it : but, on the one hand, we have not hitherto found reli- 
able the numbers given in the signatures to the chapters ; and, on the 
other hand, there is still another lacuna, of undetermined extent, in the 
fourth section (see under rule 80) ; and yet once more, we cannot be 
absolutely certain that the commentator does not, in one or two cases, 
state and expound two or more rules together, as once or twice in the 
first chapter (rules 12-13, 14-16). The treatment of the general sub- 
ject by the other Praticakhyas varies so much, as regards fullness of 
detail as well as the doctrines held on minor points, that I do not ven- 
ture to attempt to fill up, by their guidance, the gap which the careless- 
ness of our copyist has left, and I continue without break the enumera- 
tion of the rules which still remain to us. 

FT] RlHsMki U ^ n 

29. Visarjarnya is not doubled. 

The other treatises also exempt visarjantya from duplication : see 
Rik Pr. vi. 1 (r. 1, ccclxxviii), Vaj. Pr. iv. 112, and Taitt. Pr. xiv. 15. 

As already explained, in the preceding note, it is probable that the 
commentator gave again, in order to show that visarjaniya is not doubled 
when final (by iii. 20), like any other consonant, the instances agnih and 

TOR STli^o ii 

30. Nor a consonant which is followed by another of the same 

The Rik Pr. makes no such exception as this : the other two treatises, 
however, do so, and each divides the precept into two, the one (V. Pr. 
iv. 108 ; T. Pr. xiv. 23) prohibiting the doubling of a consonant when 
followed by the same consonant, the other (V. Pr. iv. 113 ; T. Pr. xiv. 
23, last half), when followed by a mute of the same series ; the Vaj. 
Pr. farther specifying that the following mute must not be a nasal. 

The commentator cites as instances a number of words, evidently 
without any particular reference to the Atharvan text, although two or 
three of them happen to occur there: they are indrah (e.g. i. 9. 1), 

iii. 32.] Atharva- Veda Prdtigdkhya. 478 

eandrah (e. g. ii. 15. 3), mandrah, (xviii. 1. 30), ushtrah, kroshtrah (krosh- 
tre, xi. 2. 2), bhr&shtrum, neshtram, and r&shtram (e. g. iii. 4. 1). Seve- 
ral of them are found also in the scholia to Panini, as illustrations of 
his rule (viii. 4. 50) that, according to Qakatayana, no duplication takes 
place in a group of more than two consonants. The instances are 
wanting in variety, as illustrating our text, since they all present groups 
of three consonants, while we must suppose our rule to apply no less to 
groups of two, and to forbid duplication in such words as antah, asti, 
ashta, etc. 

31. Nor r and h — but the consonant following those two let- 
ters is doubled. 

The commentator's examples are those which have been already given, 
and in part twice, under i. 58 and i. 100, and they need not, therefore, 
be repeated here. 

The Rik Pr. (vi. 2, r. 4, ccclxxxi) subjects only r to this rule, leaving 
h to meet the same treatment with the other spirants; the Taitt. Pr. 
(xiv. 4) does the same ; the precept of the Vaj. Pr. (iv. 98) is to the same 
effect with ours, and so is also that of Panini (viii. 4. 46), except that 
the latter here, as elsewhere, merely allows, and does not require, the 

32. Not, however, £, sh, and s, before a vowel. 

This is an exception, of course, to the latter part of the preceding 
rule, since the sibilants would not, by any other precept, be subject to 
duplication before a vowel. The illustrations given in the commentary 
are none of them from the Atharvan, although a part of them are to 
be met with in the scholia to the corresponding rule of Panini (viii. 4. 
49) ; they are karshati, adarpah, akshatarfah, and tatarsha 1 purod&fam. 
As counter-example, to show that the sibilant is exempt from duplica- 
tion only before a vowel, we have given us v&rshshyodakena yajeta, 
which is also no citation from the Atharvan text. 

The Rik Pr. (vi. 2, r. 10, ccclxxxvii) exempts from duplication any 
spirant, when followed either by a vowel or by another spirant. The 
Taitt. Pr. (xiv. 16) and Panini (viii. 4. 49) precisely agree with our 
treatise. The Vaj. Pr. fails to make any corresponding exception. 

The manuscripts of the Atharva- Veda, so far as known to me, do 
not, save in very infrequent and entirely sporadic cases, follow any of 
the rules of the varnakrama proper (rule 27, as already remarked, k 
really of another character), excepting the one which directs dupli- 
cation after a r; and even in this case, their practice is as irregular as 
that of the manuscripts of the later literature. See Weber, pp. 246- 
248, for interesting statements respecting the usages of the codices of 
the Vajasaneyi-Sanhita. 

1 ? — MS. tatarfam. 

474 W. D. Whitney, [Hi. 83- 

33. The pragrhya vowels remain unchanged. 

As was already remarked above, under i. "73, the designation of cer- 
tain vowels as pragrhya, made in the first chapter (rules 73-81), is 
not enough to exempt them from the operation of the rules for the 
combination in sanhita of final and initial vowels : it is deemed neces- 
sary to add here that the vowels so designated maintain under all cir- 
cumstances their own proper form. The method of the Rik Pr. and 
Vaj. Pr. is the same : see R. Pr. ii. 27 (r. 50, 51, civ, clvi) ; V. Pr. iv. 84. 

The commentator's examples are kena. pdrshni dbhrte: pdrshni iti 
(x. 2. 1), ind.ravt.yik ubhdu: indravdyh iti (iii. 20. 6), and ubhdv indrdg- 
ni d bharatdm : indrdgni iti (v. 7. 6). The text offers a single case in 
which a final pragrhya vowel is combined with a following initial : it is 
nrpait 'va (viii. 4. 6). The same passage is found in the Rig- Veda (vii. 
104. 6), exhibiting the same anomalous sandhi, and such exceptional 
cases are duly noted by the Rik Pr. (ii. 27, r. 54, clix), as is one of the 
same character by the Vaj. Pr. (iv. 86). That no reference is made to 
the passage in our treatise is possibly to be taken as an indication that 
the true Atharvan reading is nrpati iva, as is actually given by E. and I. 

34. Also end ehdh etc. 

The passage cited in the rule as heading the gana is found at xii. 3. 
33, where both sanhita- and pada read end,' eh&'h pdri etc., the estab- 
lishes of the pada-text evidently regarding end as the instrumental of 
the pronominal root a. To me, however, it seems more plausible to 
take the word as accusative plural feminine of the pronoun ena, which 
is usually enclitic, but in one or two instances (see the Bdhtlingk-Roth 
lexicon, sub verbo) is accented, when standing at the head of a pdda. 
The form then would be end'h, and the sandhi perfectly regular, and its 
treatment as an irregularity would be due only to a misapprehension on 
the part of the author of the pada. The other cases cited by the com- 
mentator, as composing the rest of the gana, are yathd man nd 'pagd 
a*aA (i. 34. 5. ii. 30. 1. vi. 8. 1-3), and prthivi uta dyduh (xviii. 1.5). 
The' latter is found also in the Rig-Veda (x. 10. 5), and is noticed by its 
Pratie&khya (ii. 39). The other is a somewhat intricate case. As re- 
gards, first, the accent of the word — all analogy requires that, as con- 
taining a root for its final member, it should be accented upon the last 
syllable, apagd'. Thus, indeed, the pada-text actually reads where the 
passage first and last occurs ; but at ii. 30. 1 it gives apa-gd, and with 
this accentuation agree all the sanhitd manuscripts in every instance, so 
that this reading has been received into the printed text. Second, as 
regards the form — there is not, so far as I can see, any reason why we 
should not regard apagd as the theme of the word, and not apaga, and 
so consider the found in the text to be apagds, and the sandhi 
to be entirely regular. It is true that most of the Atharvan compounds 

iil 35.] Atharva-Veda Pr&tiqtikhya. 475 

into which the foot gam or ga enters as final member exhibit it in the 
shortened form ga (as durgd, sugd, svargd), but we have once purogA 
(v. 12. 11), and in like derivatives from other similar roots, as jan and 
pa, the short and long forms exchange with one another quite irregu- 
larly (e. g. prathamajA'h, iv. 35. 1, and prathatnajah, iv. 4. 5). I should 
be inclined to accuse the pada-text of a similar misapprehension in this 
passage with that exhibited in the other. There are one or two other 
cases in the Atharvan text which belong more or less clearly under this 
rule. In iv. 16. 1, all the sanhitA MSS., without exception, read adhi- 
shthdtd antikad iva (p. adhi-sihala) : the edition treats this as a blunder, 
and amends to adhishthalA , ntikad iva, but it might be possible to re- 
gard the passage as offering a case of anomalous sandhi. The theory 
of an error in this case, in which all the sanhitA MSS. chance to coin- 
cide, is supported by the analogy of the quite similar passage rocanA' : 
asyA, at vi. 31. 2, where P., M., and I. read rocanA 1 asyd in sanhitA, while 
W., E., and H. give rocanA' 'syd, with the Rik, Saman, and White Yajur 
Vedas. At v. 26. 8, where the sanhitA manuscripts read rtipd' asmin, 
the pada has rupa'h : asmin. Unless authority for r-itpA as a feminine 
noun can be found elsewhere, it will probably seem easier to regard 
rxipA as a briefer form of r&pAni, and the sandhi as an anomalous one. 
A like case is vii. 97. 4, sddanA akarma, where Bp. is amended by a 
second hand from s&danA to sddandh. An evident instance of pretty 
gross blundering is to be seen at xviii. 4. 58, where, instead of pratdrito 
'shasdm, as the sense and metre require, and as is read by the Rik and 
Sama Vedas (RV. ix. 86. 19 ; SV. i. 559), all our sanhitA manuscripts 
present prat&ritA ushasdm, which the pada also supports by giving 

MclHfa hVUi 

35. Where y or v has been dropped, the preceding vowel re- 
mains unchanged. 

That is to say, wherever, by the loss of a final y or v according to 
the phonetic rules of the second chapter (ii. 21), a hiatus takes place, 
and two vowels are brought into juxtaposition with one another, they 
are not combined, but the hiatus remains. Two vowels can be thus 
brought together, according to the rules of our treatise, only by the loss 
of y and v, a final visarjaniya being converted into y, by the theory 
here taught, before its final disappearance. The same prohibition 
against applying the rules of combination twice over to the same case 
is given by the other treatises (R. Pr. ii. 2, r. 5, cix ; V. Pr. iii. 3), in 
the form of a general precept, governing and restricting the application 
of its special rules. 

As instances, the commentator cites again the whole series of passages 
given under ii. 21, and which I refrain from repeating here. 

Although this is a rule of universal application as regards the mode 
of writing the text, the metre shows that it was not always observed by 
the authors of the hymns: see, for example, ix. 4. 19, 23, where we 
have to read brdhma.nebhya rshabham dattvA and asmin goshtho 'pa 
prnca nah, instead of brdhmanebhya rshabham and goshtha upa. ' Such 

476 W. D. Whitney, [iii. 35- 

cases are not infrequent. In a single instance, too (x. 1. 15), our sanhitd 
MSS. make the double combination, reading ayam panthdh krlye Hi 
tvd naydmah, instead of krtya iti (p. krtye : iti) ;* but the edited text 
has restored the latter reading. 

36. Also an u forming a word, when it is preceded by a vowel. 

That is to say, of course, the particle u — whenever, by the operation 
of the euphonic rules, a hiatus is produced before it. There are found 
but three such cases in the Atharva-Veda; two of them are cited by 
the commentator, as follows : sa w eva mahdyamah (xiii. 4. 5), and sa u 
aftn&nam asyali (xiii. 4. 41) ; the third is ayam vd u agnih (xv. 10. 7). 

The corresponding rules of the other treatises are Eik Pr. ii. 28 (r. 
55, clx), Vaj. Pr. iv. 87, and T&itt. Pr. ix. 16, 17 : the Rik Pr. also leaves 
the u unchanged after a y arising from conversion of an original i, one 
such case occurring in its text (praty u adarfi, vii. 81. 1), while the 
Taitt. Pr. inserts a transition v between the u and the following vowel. 

5Rt WKt H$©ll 

37. Also n and m, when the results of euphonic processes. 

The commentator fabricates his examples, instead of deriving them 
from the Atharvan text: they are nadim iarati and trishtum nayati. 
The former is intended to show that a n which is the result of the as- 
similation of a final m, by ii. 31, before an initial t, is not liable to a 
farther insertion of a sibilant before the t, by ii. 26 ; the latter, that a m 
which grows out of the assimilation of a final labial to a following initial 
dental nasal, by ii. 5, is not then, by ii. 31, convertible into n by a sec- 
ond assimilation. 

This rule is of the same character with the last but one, and is re- 
placed, or rendered unnecessary, in the other treatises, by the general 
precepts there referred to. 

*ll*l^: %^<T: cm g^ftrr n ^ n 

38. An d forming an entire word is first combined with the 
preceding vowel. 

An instance will best explain the meaning of this role. In the pas- 
sage which the ^xirfa-text writes dhiyd : d : ihi, if d is first combined with 
the following word, it will form e 'hi, and the combination of this with 
the preceding word will give dhiydi 'hi: but if the combination of d 
with dhiyd be first made, producing dhiyd, the addition of ihi will give, 
as final result, dhiye " 'hi. The latter is the true method of making 
the two successive sandhis, as we are taught by this rule, and by cor- 
responding rules in the other treatises (R. Pr. ii. 2, r. 7, cxi ; T. Pr. v. 3) ; 

* P. only has, by a copyist's error, krtydi Hi in both its copies of the tenth book. 

iii.41.] Atharva- Veda Prdtigdkhya. 477 

which, however, express themselves in a more general manner, declaring 
that all sandhis must be made in the order of their occurrence. 

The commentator is this time conscientious enough to cite all the 
passages illustrating the rule which the text contains : they are dhiye " 
'hi (ii. 5. 4), jushasve " 'ndra (ii. 5. 4), stanayilnune " 'hi (iv. 15. 11), 
kushthe " 'hi (v. 4. 1), udakene " 'hi (vi. 68. 1), and avapctfyate " 'to 
(xviii. 4. 37). 

^ HllHHt ^T^ETTll^li 

39. Before a vowel, an alterant vowel becomes a semivowel. 

Instead of citing actual examples from the text, the commentator 
prefers to fabricate his illustrations, which are dadhy atra, madhv atra, 
m&trartham, pitrartham : a part of them are identical with those given 
in the scholia to the corresponding rule of P&nini (vi. 1. 77). 

The other treatises have corresponding rules : that of the V&j. Pr. (iv. 
45) being precisely like our own ; that of the Rik Pr. (ii. 8, r. 21, 22, cxxv, 
cxxvi) being more elaborately stated; and that of the Taitt. Pr. (x. 15) 
restricting the conversion into a semivowel to i, i, and u — a restriction 
which might with the same propriety have been made by all, since final 
w is always pragrhya, and final r nowhere in the Vedas, so far as I am 
aware, comes to stand before an initial vowel. 

40. The diphthongs, in the same situation, become respect- 
ively ay, av, dy and dv. 

And then farther, by ii. 21, 22, the final semivowels are dropped, ex- 
cept in the case of Av, so that of e and o is left before an initial only a ; 
of Ai, only A — which vowels are then exempt, by iii. 35, from farther 
combination with their successors. The absorption of initial a by final 
e and o, and its retention without change after them in some cases, is 
taught by rules 53 and 54, below. 

The commentator's illustrations are again of his own devising, for the 
most part ; they are agna AsAm, vAya AsAm, agna utaye, vAya ittaye (iv. 
25. 6), asmA ud dhara, asAv Adityah (xv. 10. 7), pvayanam, cAyakah, la- 
vanam, lAyakah, pavanam, pAvakah (e. g. vi. 47. 1). The last six are 
examples of applications of the rule which the Pr&tigakhya does not 
contemplate. A few of the instances are identical with those given 
under the corresponding rules of P&nini (viii. 3. 18, 19. vi. 1. 78). 

The Rik Pr. (ii. 9, 10, 11, r. 25, 28, 31, cxxix, cxxxii, cxxxv) converts 
Ai and Au directly into A, and e and o directly into a, adding that after 
the a and A which come from o and Am a v is inserted except before a 
labial vowel. The Vaj. Pr. (iv. 46) and Taitt. Pr. (ix. 11-15) precisely 
agree with our treatise. 

g^rft^fT: u 8\\\ 

41. In the following rules is taught the fusion of a preceding 
and a following vowel into a single sound. 

478 W. D. Whitney, [iii.41- 

This is a general heading or adhikdra for that which is to follow, or 
a rule governing the interpretation of the remaining rules in the section. 
The technical language of the Praticakhyas has no recognized method 
of indicating the fusion of two sounds into one, and the form of the 
following rules is ambiguous, since rule 44, for instance, literally means, 
according to the usual phraseology of the treatise, that a before i becomes 
«, and not a with i. Hence the necessity of this special rule of inter- 
pretation : it has its correspondents in the Vaj. Pr. (iv. 49), and the 
Taitt. Pr. (x. 1) ; while the Rik Pr. (ii. 6 etc.) attains the desired end 
by a fuller or less technical mode of statement. 

42. A simple vowel, with one of like quality, becomes long. 

The commentator's illustrations are again of his own making : they 
are danddgram, dadhindrah, madhitshnam, hot f$ yah, and pitrpyah. For 
the last case, indeed, the combination of two r's, no Vedic example 
could be found. 

The corresponding rules of the other treatises are Rik Pr. ii. 6 (r. 15, 
cxix), Vaj. Pr. iv. 50, Taitt. Pr. x. 2. 

?ffa% ^T: ll ^ li 

43. In simanta, the resulting vowel is short. 

A most blundering and superfluous rule ! The Atharva-Veda con- 
tains no such theme as simanta. In the passage which the commenta- 
tor cites in illustration of the rule — viz. jinatovajra tvam simantam (vi. 
134. 3) — the theme is evidently simant, from which we find also a plu- 
ral accusative, simatah, at iv. 1. 1 ; it is a parallel form with, and equiv- 
alent to, stman, of which the text affords us an accusative singular, 
simdnam, at ix. 8. 13. The rule evidently assumes simanta as the 
theme, and regards it as composed of siman and anta. Our pada-text, 
as might be expected, makes no attempt to analyze the word. The 
commentator, after his illustrative citation, adds to the rule a restriction 
of his own, quite in the style of a v&rttika to Panini, as follows : siman- 
te kepaveshte Hi vaktavyam : yo hi simno antah simdntah sah ; ' it should 
have been said, " in simanta, when it means the parting of the hair ;" 
for the extremity (anta) of a boundary (siman) is simdnta.'' Since, 
however, simdnta does not occur in the Atharvan, the limitation is just 
as impertinent as the original rule: more so, it could not well be. 
Compare vdrttika 3 to Pan. vi. 1. 94. 

ilciufcfauf ^n^ n 88 ii 

44. A and d, with i and % become e. 

The commentator's illustrations, as given by our manuscript, are 
ravadffom&lendrah, which, however, a comparison of the examples under 
the next rule and under the corresponding rule of Panini (vi. 1. Si) 
shows to require amendment to khatvendrah, mdlendrah. 

iii. 46.] Atharva- Veda Prdtig&khya. 479 

The corresponding rules of the other treatises are Rik Pr. ii. 6 (r. 16, 
cxx), Vaj. Pr. iv. 52, and Taitt. Pr. x. 4. 

45. With u and u, they become o. 

The commentator, as so generally in this portion of his work, makes 
his own illustrations, viz. khatvodakam, m&lodakam: compare under 
Pan. vi. 1. 87. 

See the identical rules in the other treatises: Rik Pr. ii. 7 (r. 17, 
cxxi), Vaj. Pr. iv. 52, and Taitt Pr. x. 5. 

There is a single instance in the text, in which this rule is not ob- 
served, and a + u are not combined into o, but into &u : it is the word 
prAudhah (xv. 15. 4 : p. pra-&dhah). B., indeed, reads prodho, but doubt- 
less only by an error of the copyist. We must suppose, either that the 
authors and commentator overlooked this word, or that its pada as well 
as sanhitA reading in their text would be prAudhah, or that the pas- 
sage containing it was not in their Atharva-Veda — of which suppositions, 
I should regard the first as the most plausible, and the last as the least 

46. With r and f, they become ar. 

The commentator's examples are tasya rshabhasyd 'ngdni (ix. 4. 11), 
yajnartah (viii. 10. 4: p. yajn~a-rtah), kdmartah (not in AV.), and ndi 
, ndn avartih (iv. 34. 3 : the ^xzda-text always writes the word avartih, 
without separation). 

The Taitt. Pr. (x. 8) makes the sandhi in the same manner as our 
treatise: the Rik Pr. (ii. 11, r. 32, cxxxvi) and Vaj. Pr. (iv. 48), how- 
ever, treat it very differently, merely prescribing that both a and A be- 
come a before r, without requiring the conversion of the latter into r. 
The usage of the sanhitd manuscripts of the Atharva-Veda is in con- 
formity with neither rule ; they follow a method of their own, in which 
is to be recognized the influence of a doctrine agreeing with or resem- 
bling that of our Praticakhya respecting the svarabhakti, or fragment 
of vowel sound, assumed to be thrust in between r and a following con- 
sonant (see i. 101, 102). Where the phonetical theory requires the 
insertion of the longer svarabhakti, or where a sibilant follows, there the 
manuscripts usually and regularly give the vowel r instead of r, reading 
iva rshabhah (iii. 6. 4), svasa rshindm (vi. 133. 4), etc. ; before any other 
consonant, or where our treatise and the Rik Pr. interpose the shorter 
svarabhakti after the r, and the other Praticakhyas require no insertion 
at all (see the note to i. 101, 102), there our manuscripts regularly make 
the combination according to the rule now under discussion, writing 
rtasya rtena (vi. 114. 1), iva rbhuh (x. 1. 8), sa rcdm (x. 8. 10), etc. 
These rules are not, however, altogether without exceptions : there are a 
number of passages in which one or more of the manuscripts read the 
semivowel r instead of the vowel r before a sh (viz. iv. 4. 8. ix. 8. 14, 

VOL. VII. 61 

480 W. D. Whitney, [iii. 46- 

15,16. x. 7.14a; 10.10. xi. 1.1,3; 3.38. xii. 1.6. xiii. 1. 55. xv. 2. 4, 
first time), and even a few (viz. x. 7. 14c. xi. 6. 11. xv. 2. 4 (bis) ; 14. 4. 
xix. 9. 13) in which they all agree in so doing; and, on the other hand, 
while in the very great majority of cases the r is changed to r before any 
other consonant (it is found so changed, in the Atharvan text, before Jc, 
g, c,j, ch, t, d, n, dh, and bh), there are a very few instances (viii. 10. 4. 
xviii. 2. 3 1 ; 3. 24) in which one or more manuscripts leave it unaltered, 
and even two (x. 1. 12 ; 5. 30, before g and k) where they unanimously 
read r.* As regards the orthography of the sandhi in the published 
text, accordingly, three courses were open to the editors : first, to follow 
the rule of the Pratigakhya, and to convert the initial vowel every- 
where into the semivowel ; second, to carry out consistently the general 
principle derivable from the practice of the manuscripts, writing always 
r before a sibilant, and r before any other consonant ; and third, to be 
governed everywhere by the authority of the manuscripts where they 
were unanimous, and, where they disagreed, to give due weight to the 
principle just referred to, in choosing between their discordant readings. 
Unfortunately the edition has adopted none of these courses, but, while 
adhering with tolerable closeness to the manuscript readings through 
the early part of the text, gives everywhere only r from the beginning 
of the eighth book onward, thus introducing the rule of the Rik and 
White Yajur Vedas, and following neither the Pratigakhya nor the 
manuscripts of the Atharvan. The details given above, however, will 
show what are the readings of the manuscripts in any 1 given passage. 

It is worthy of remark that the manuscript of our commentary, like 
those of the Atharvan text, writes r in the first instance cited (tasya 
rshabhasya), and r in all the rest. 

There is a single passage of the text requiring special notice, as ex- 
hibiting in the sanhitA manuscripts an entirely irregular sandhi of final 
A with initial r: it is at xviii. 2. 31, where the A of sufevA is left un- 
shortened before rksh&kam, B. writing supevA rksIiAkam, and all the 
others sugevA rkshAkam. Probably the reading is corrupt, or the words 
wrongly analyzed by the pada-text: perhaps we have to correct and 
divide sugevA : ArkshikAm. 

scrferrf^r ^ iidou 

47. Also in uparshanti etc. 

The words and forms contemplated by this rule are exceptions to the 
one next following, according to which we should have upArshanti etc. 
The commentator cites in illustration the passages yA hrdayam upar- 
shanti (ix. 8. 14), yAh pArgve uparshanti (ix. 8. 15), and yAs tiragctr 
uparshanti (ix. 8. 16); and these are the only cases of the kind pre- 

* The lingual sibilant, sh, is the only spirant before which a sandlii of final a or 
d with initial r is made in the Atharvan text : the text does, indeed, contain a single 
passage in which such a combination should properly take place before p — viz. kra- 
masva r^a iva rohitam (iv. 4. 7) — but the tradition has most palpably and grossly mis- 
understood and blundered over the phrase, and the pada writes it krdma : ivdrpah- 
•iva : rohitam, instead of Icrdmasva : fpah-iva : rohitam. 

iii. 51.] Atkarva- Veda Prdticdkhya. 481 

sented by the text, so that the " etc." at the end of the rule is quite 
superfluous as regards the Atharvan. In the passages cited, the pada 
reads upa-rshanti, and the sanhitd manuscripts, as already noticed in 
the last note, vary between uparshanti and uparshanti, W. even giving, 
in the second and third cases, uparishanti. 

48. The a or d of & preposition, with the initial r or f of a 
root, becomes dr. 

The commentator's illustrations are of his own fabrication, and in 
part are to be found under the corresponding rule of Panini (vi. 1. 91) : 
they are updrshdti, prdrshdti, updrchati, prdrchati, updrdhnoti, and 
pr&rdhnoli. The only case arising under the rule in the Atharva-Veda 
is d rchatu, at ii. 12. 6. Our treatise might, then, like the Vaj. Pr. (iv, 
57), have restricted the operation of the rule to the preposition A. The 
Taitt. Pr. (x. 9) states the principle in the same general form in which 
it is here given. 

49. As does also that of the augment. 

This rule, of course, in a treatise whose subject is the sandhyapadyAu, 
gundu of words (i. 1), is out of place and superfluous. It has no cor- 
respondent in any of the other treatises, and its chief interest and value 
to us lies in its presenting a term for 'augment' — bhutakarana, 'maker 
of past meaning' — which is elsewhere unknown, at least in the Pratiga- 
khya literature. 

The commentator cites, in illustration of the rule, sa drdhnot (w. 39. 
1 etc.) ; and also, as counter-example, showing that it applies only to 
the augment, katamd sa rcdm (x. 8. 10). 

^f^=hl{MY^hl{: ll M,o a 

50. With e and di, a and d become di. 

Again the commentator gives us only fabricated illustrations, which, 
with the help of the scholia to the corresponding rule of Panini (vi. 1. 
88), are readable as follows : khatvdirakd, mdldirakd, khatvdiiikayanah, 
mdlAilikdyanah. 1 

The corresponding rules in the other treatises are Kik Pr. ii. 7 (r. 18, 
cxxii), Vaj. Pr. iv. 55, Taitt. Pr. x. 6. 

*n*l(Hl^l(Hl(: II HA H 

51. With o and du, they become du. 

The commentator this time presents us two actual citations from the 

1 Ichadvdiragdh : m&diragah : hhadvebhikdyamdnasali : mdlditikdyamdnah. 

482 W. D. Whitney, [iii. 51- 

text, viz. brahm&udanam pacati (xi. 1. 1 : p. brahma-odanam), and tasy&u 
'danasya (xi. 3. 1) ; but then adds a fictitious case, brahm&upagavah, 
which occurs also in Panini's scholia (to rule vi. 1. 88). 

The corresponding ruies in the other treatises are Rik Pr. ii. 7 (r. 19, 
cxxiii), Vaj. Pr. iv. 55, and Taitt. Pr. x. 7. 

id«W«Nil^5j '^{im^i ii h^ ii 

52. In qakalyeshi etc., the result has the same form with the 
latter constituent. 

The commentator cites under this rule the following cases : fakalyeshi 
yadi va tejanitrarn (i. 25. 2 : p. pakalya-eshi), anamiva upeiana (iii. 14. 
3 : p. upa-etana), arv&ci gaur upe 'shatu (vi. 67. 3), and vpeshantam 
udumbalam (viii. 6. 17: p. upa-eshantam). Of these cases, the first 
would equally admit of being regarded as a case of regular sandhi, and 
analyzed as pakali-eshi : the second is analogous with the combina- 
tions to which rule 38 of this chapter relates, the preposition a being in 
sanhitd combined with upa, and then the resulting upa with itana. Of 
this kind, the text presents one additional instance, in upeyima (x. 1. 
10 : p. upa-eyimd; it is made up of upa-A-iyimd) : it is the only pas- 
sage falling under the rule which the commentator does not give. 

Similar cases are noted by the Rik Pr. at ii. 36, 37, by the Vaj. Pr. 
at iv. 53, 54, and by the Taitt. Pr. at x. 14. 

^hl(kl{l<HIH£of: M<i<(*l(W iiHSu 

53. After a word ending in e or o, an initial a becomes one 
with the preceding vowel. 

The commentator cites a few instances of this exceedingly common 
sandhi (the occurrence of which will be more particularly treated of 
under the next rule), as follows : te 'vadan (v. 17. 1 : the instance, how- 
ever, may perhaps be given as one fabricated, and not as a citation from 
the Atharvan text; it is found in the scholia to Pan. vi. 1. 115), te 'kra- 
van (as is read both here and under iii. 55 ; perhaps we are to amend 
to te 'krnvan, but that also is not to be found in the Atharvan), so 'bravtt 
(xv. 3. 2), yo 'sya daJcshinah karnah (xv. 18. 3), and so 'rajyala (xv. 8. 1). 

The physical explanation of this combination is exceedingly difficult. 
The Rik Pr. (ii. 12, 13, r. 33, 34, cxxxvii, cxxxviii) and Vaj. Pr. (iv. 58), 
as well as Panini (vi. 1. 109), treat it in the same manner as our treatise, 
as a union of a with the preceding diphthong, or its absorption by the 
latter. The Taitt. Pr. alone (xi. 1) teaches an actual loss or omission of 
the a. 

efcNrU^rdl II \8 ll 

54. Sometimes the a remains unabsorbed. 

The commentator quotes from the text, in illustration, the passage 
ye agnayah (iii. 21. 1), and adds sahasrarcam ide atra, which is not found 
in tiie Atharvan text. 

iii. 54.] Atharva-Veda Prdliqdkhya: 483 

With this rule and the preceding our treatise makes short shrift of a 
subject which occupies long passages of the other Praticakhyas (R. Pr. 
ii. 1 3-26 ; V. Pr. iv. 58-82 ; T. Pr. xi. 1-xii. 8), and has cost their au- 
thors a vast deal of labor. The saving is made, however, at the expense 
of its reputation and value as an authority for the readings of its text, 
since it does not determine the sanhitA form of one of the many hund- 
red passages in which initial a comes in contact with final e or o. There 
is not in the whole work another so discreditable confession of unwil- 
lingness or inability to cope with the difficulties of an intricate subject. 

In endeavoring to make some systematic exhibition of the usage of 
the Atharvan text with reference to this sandhi, I have been able, after 
more than one trial, to devise no better scheme of presentation than 
one founded upon a comparison of the actual written usage of the text 
with the requirements of the metre. If there is any rule or system of 
rules, of a more formal character, underlying and determining the very 
various phenomena — which I cannot but seriously doubt — I have been 
unable to discover any trace of it. The tabular conspectus, then, is as 
follows — the lines distinguishing the cases in which the metre appears 
to require the retention of the a unabsorbed, as a separate syllable, 
from those in which its absorption is demanded, in order to make the 
verse of proper length ; and the columns showing how often the a is by 
the manuscripts omitted and retained respectively : 

In written text : 
omitted. retained. 

I. Where metre requires omission, 252 41 298 

II. Where metre requires retention : 

a, at beginning of pdda, 198 39 

b, elsewhere in pdda, 102 965 

300 _ 1004 1304 

Total in metrical part of text, 562 1045 1697 

III. In unmetrical passages, 192 28 220 

Total in whole text, 744 1073 1817 

Upon examining this table, it will be seen, in the first place, that in 
the unmetrical portions of the Atharvan text the greatly prevailing 
method of making the sandhi is that which is followed in the Sanskrit 
literature proper, viz. by omission of the a : the proportion of omissions 
to retentions is that of 7 to 1. In the metrical portions, on the other 
hand, the more common custom is to retain the a, the retentions being 
to the omissions nearly as 2 to 1. "When we come to inquire farther 
what was the usage of the makers of the hymns, we find that the pro- 
portion in favor of the retention of the a as a separate syllable is con- 
siderably greater, or almost as 4-J- to 1. There is, then, considerable 
discordance between the written and spoken usage in the metrical part ; 
and yet this discordance appears in great measure at a single point, or 
where the final e or o of a pdda precedes the initial a of another p&da. 
In this situation — where, if ever, we should expect the traditional and 
written reading to correspond with the original spoken form of the 
verse — the recorded text usually leaves out the a, and mars as much as 
it can the metrical form of the verse : the proportion of omissions to 
retentions is here as 5 to 1 ; and in the Rig-Veda, so far as I have 
observed, it is still larger : there can be but very few instances in the 

484 W. D. Whitney, [iii. 54- 

earlier portions, at least, of that text, where the custom of omission of 
o at the beginning of a pdda is not followed. Making exception of 
this special case, it may be said that the usage of the written text fol- 
lows in the main the requirements of the metre — although with not 
infrequent exceptions, which in the case of the required omissions make 
somewhat less than a sixth of the whole number, and in the cases of 
required retention, considerably less than one-ninth. 

There are, of course, a number of doubtful cases, where the metre is 
irregular and obscure, or where it might be restored either with the 
omission or the retention of the a as a separate syllable : but, rather 
than form of them a separate class, I have determined and ranked each 
case according to my best judgment ; and the general relations and bear- 
ings of the scheme are not, I am sure, perverted by any errors I may 
have committed. 

Here, at the end of the second section of the chapter, the manuscript 
says again simply trtiyasya dvittyah pddah. 

The proper subject of the next section is the calling forth or modifi- 
cation, in connection with the combinations of the phrase, of the accents 
already laid down and defined in the first chapter (i. 14-17). With 
this, however, is also connected a distinction and nomenclature of the 
different kinds of svarila or circumflex accent which the theory of the 
school recognizes : and the latter subject is even allowed in part to take 
precedence of and overlie the former, in such a manner as to render 
their joint presentation confused and imperfect, as will be pointed out 
in detail below. By way of introduction to the section, and before 
stating and explaining its first rule, the commentator gives us the fol- 
lowing four verses :* shad eva svaritajdt&ni Idkshandh pratijdnate : pur- 
vam p&rvarh drdhatarark mradiyo yad yad uttaram : — abhinihitah prd- 
plishto jdtyah kshaipraf ca td ubhdu : tairovyanjanapadavrttdv etat 
svaritamandalam : — sarvatikshno 'bhinihitas tatah pr&flishta ucyate: 
tato mrduiar&u svdrdu jatyah kshaiprap ca tdv ubh&u : — tato mrduta- 
rah svdras tdirovyanjana ucyate : pddavrtto mrdutara iti svdrabaldba- 
lam; 'those skilled in distinctions recognize six kinds of circumflex 
accent, and no more. Of these, each preceding one is harder, each suc- 
ceeding one is softer : viz., the abhinihita, the prd$lishta, and the jdtya 
and kshdipra, these two; farther, the tdirovyanjana and pddavrita: 
this is the series of circumflex accents. The abhinihita is entirely sharp ; 
next is ranked the prdflishta; the next pair, of softer character, are the 
jdtya and kshdipra, these two ; the tdirovyanjana is called softer than 
these; the pddavrtta is still softer: thus are laid down the relative 
strength and weakness of the circumflex accents.' We have here evi- 
dently the schemes of two different authorities, which accord quite 
closely with one another; the only difference being that the former 
seems to rank as equal the two accents last mentioned. Indeed, the 
commentator goes on to add, in accordance with this, apara aha : tdiro- 
vyanjanapddavrttdu tulyavrttt Hi; 'another says, "the tdirovyanjana 

* The last two of these verses are also cited in Uvata's commentary to the Vaj. 
Pr.: see Weber, p. 189. 

Hi. 55.] Atharva-Veda Prdtigdkhya. 485 

and p&davrtta are of like quality." ' Other authorities, however, con- 
struct the scale somewhat differently : thus the Vaj. Pr. (i. 1 25) makes 
the order run as follows : abhinihita, ksh&ipra (with which the com- 
mentator declares the j&tya to agree in rank), praplishta, t&irovyanjana, 
tairovir&ma (see below, under rule 62), and p&davrtta ; while the Taitt 
Pr. (xx. 9-12) declares the effort of enunciation (prayatna) of the ksh&i- 
pra and j&tya, together with the abhinihita, to be harder, that of the 
praplishta and pr&tihata (see under rule 62) to be softer, and that of 
the t&irovyanjana and p&davrtta to be yet less (alpatara). The Kik 
Pr., like our own treatise, takes no notice of these alleged differences 
of utterance among the different kinds of circumflex. It is not, how- 
ever, without good reason that our commentator gives himself the 
trouble to state them, since their recognition can hardly have been 
without its important influence upon the division of the svarita into its 
numerous forms. The three arrangements quoted above agree, it will 
be noticed, in ranking the forms of the independent svarita above those 
of the enclitic, but are discordant as regards the relative position of the 
members of each class; and this may serve to us as a significant indi- 
cation that the differences of secondary rank are of but trifling conse- 
quence. Precisely what is meant by "sharp" (tikshna) and "hard" 
(drdha) on the one hand, and "soft" (mrdu) on the other, is not very 
clear : but that the proper circumflex, which arises upon the combina- 
tion into a single syllable of an original acute and an original grave 
element, is more strongly marked and distinct in its quality of double 
pitch than that circumflex which is only enclitic, need not be doubted. 

After this preliminary exposition, the commentator goes on to add 
nd&ttah p&rvah: paro 'nud&tlah: svariiah sandhih; 'an acute preced- 
ing ; a grave following : their combination circumflex.' I am not alto- 
gether confident that this is not the first rule of the section, since, as 
we shall see, the two rules which follow are defective in form, and need 
some such predecessor. Considering, however, the faulty construction 
of the whole section, the limited applicability of the words in question 
as an adhik&ra or heading for that which follows, their inconsistency 
with rule 66 below, and the absence of the paraphrase and repetition 
which ought to follow them, if they are a rule, I have not ventured to 
regard them as a part of the treatise ; they are more probably an addi- 
tion of the commentator, intended to supply the deficiency of the next 
two rules. 

^i(te>i(l q-^mt q^rft ^nt ?fr 4hIhR.h: i i hh 

55. When an a is absorbed by a preceding final e or o, the 
resulting circumflex is abhinihita. 

This is by no means a close version of the rule as it stands in the 
text; literally translated, it reads as follows: 'final e and o; following 
them, an a : that is abhinihita.' The construction presents a grammati- 
cal difficulty. If ak&ra, ' the sound a,' is not here used as a neuter — 
and such words are elsewhere only masculine — the form ak&ram must 
be understood as an accusative, and can only be construed as dependent 
upon paratah, used prepositionally, and taken as meaning ' before ' in- 

486 W. D. Whitney, [iii. 55- 

stead of ' after,' so that we must translate ' final e and o before an a : 
that is abhinihita :' and to treat paratah thus seems to me hardly ad- 
missible.* The commentator does not help us much, but, if I rightly 
apprehend his meaning, he treats the word as a neuter nominative ; his 
paraphrase reads ek&raukarau padantau udattau paratah ak&rarii 1 pa- 
dadi anudatlam sa abhinikilah svaro bhavati ; ' e and o, final, with acute 
accent; after them, a, initial, unaccented; that becomes an abhinihita 
accent (or vowel).' But passing over this difficulty, as of inferior con- 
sequence, since the virtual meaning of the phraseology is clear, we find 
another and a graver one in the fact that the form of the rule seems to 
imply that the occurrence of this svarita has been already sufficiently 
taught, and that nothing remains but to give it a name ; while never- 
theless the treatise nowhere informs us under what circumstances a cir- 
cumflex accent arises in connection with the meeting of a final e or o 
and an initial a, or even that it arises at all. Its doctrine must be, of 
course, the same which is taught in all the other treatises ; namely, that 
when an initial unaccented or grave a is elided after or absorbed into a 
final diphthong which itself has an acute accent, its own accent is rep- 
resented in that of the resulting diphthong, reducing the latter from 
acute to circumflex. This case of circumflex, as well as that which 
forms the subject of the next rule, is an exception under rule 66 below, 
which teaches that a vowel resulting from the fusion of elements whereof 
one is acute, is itself acute : it ought, accordingly, to be specifically de- 
scribed and taught as such an exception. The admission as a rule of 
the words referred to above as used by the commentator before the 
statement of the precept now under discussion — viz. ' an acute preced- 
ing; a grave following : their combination circumflex' — would not satis- 
factorily relieve the difficulty, because this would be equivalent to setting 
up over against rule 66 another general rule opposed to it, without 
establishing in any way the relation between them. We can hardly 
avoid supposing that the constructors of this part of the treatise have 
not been skilful enough, or careful enough, to combine the two subjects 
of the section in such a manner as to give completeness to both. The 
Vaj. Pr. (iv. 59) and Taitt. Pr. (xii. 9) give rules for the occurrence of 
the abhinihita circumflex in connection with those for the absorption of 
the initial a, and define and name the accent elsewhere (V. Pr. i. 114; 
T. Pr. xx. 4), when treating of the general subject of accent : the Rik 
Pr. deals with both matters together, but with clearness and precision, 
first prescribing the svarita (iii. 7, r. 12, excix), and then later (iii. 10) 
giving it its distinctive appellation. 

The names of the different kinds of svarita are chiefly derived from 
those belonging to the euphonic combinations in connection with which 
they arise. These latter, however, are not expressly given in all the 
treatises. Thus, the Rik Pr. alone (ii. 13, r. 34, exxxviii) calls the ab- 
sorption of initial a into preceding final e and o the abhinihita sandhi 

* The commentator uses paratah very often, in paraphrasing the locative of pre- 
cedence, but always puis it after the locative : thus, in the first rule of this chapter, 
sahdu, ' before the root tah,' is explained by takdu paratah ; and so in scores of 
other cases. 

1 akdtra. 

iii. 56.] Aiharva- Veda Prdtiqdkhya. 487 

while the Vaj. Pr., as well as our own treatise, agrees with it in apply- 
ing to the resulting circumflex the title of abhinihita. The Taitt. Pr. 
(xx. 4) has for the same accent the slightly different name abhinihata, 
which is palpably an artificial variation of the other. 

As examples of the abhinihita circumflex, the commentator gives us 
precisely the same series of phrases as under rule 53 above, viz. te 'va- 
dan, te 'kravan, so 'bravit, yd 'sya dahshvnah karnah, so 'rajyata. In a 
few cases (which are detailed below, in a marginal note*), our Atharvan 
manuscripts fail to give the circumflex to an e or o, originally acute, 
which has absorbed an unaccented a, and leave it acute ; but these are 
evidently to be explained simply as perpetuating original errors of tran- 
scription, and as requiring at our hands the emendation which they 
have not received from the native copyists. 

^TT^ft: srrfe: \\\\\\ 

56. The circumflex arising from the fusion of two short i's is 

Praplishta or praplesha is the name given by the Rik Pr. (ii. 7, r. 20, 
cxxiv) to all those cases of combination in which two vowels, or a vowel 
and diphthong, are fused together into a single vowel or diphthong : 
that is to say, to those of which our treatise treats in the latter half of 
the preceding section (rules 42-51). A svarita accent arising in con- 
nection with such a combination is by all the other treatises called pra- 
flishta. The manuscript of our own treatise, both text and commentary, 
varies between the two forms prdflis/ita and prdkplishta, more often read- 
ing the latter ; which, however, has seemed to me too anomalous, not to 
say too monstrous, a term to be permitted to stand. The Rik Pr. (iii. 
8, r. 1 3, cc) informs us that a single teacher, Mandukeya, held that the 
praplishta svarita arose in all cases of a praplesha combination where 
the former element was acute and the latter grave, and it is well known 
that the Qatapatha Brahmana follows this rule of accentuation through- 
out. Panini (viii. 2. 6) also permits it, whenever the unaccented latter 
element is the initial of a word — that is to say, everywhere in the com- 
binations of the phrase. But the predilection of the Sanskrit for the 
circumflex accent is much too weak to allow of so frequent an occur- 
rence of it as the general acceptance of this rule would condition, and 
all the Praticakhyas lend their authority to the usage prescribed in our 
rule 66 below, that a combination into which enters an ud&tla vowel is 
itself ud&tla, the acute element raising the grave to its own pitch. All, 
however, allow the exception which forms the subject of the last rule, 

♦The passages are x. 10. 10 (bin), xi. 3.49; 10.23. xii. 3. 58 ; 4. 35, 41 (here, 
however, 1. gives so instead of *d, and B. has been amended to the same reading). 
xv. 14. 3 (B. and P. apd). xvi. 4. 3. xviii. 4. 1 5. The edition has generally corrected 
the accentuation in these cases; but in three passages — viz. xi. 10. 23. xii. 3. 58. 
xviii. 4. 15 — the erroneous reading of the manuscripts remains. Once, by a contrary 
error, the codices generally agree in giving the circumflex to a final o, while leaving 
the a unabsorbed after it : thus, yajnd ajdyata (xiii. 1. 48), but I. and E. (the latter 
by emendation) give the consistent reading yajnd 'jayata, which has been received 
into the edited text. 

vot~ vii. 62 

488 W. D. Whitney, [iii. 56- 

and all but the Taitt. Pr. allow also that which is treated of in this rule ; 
if I and i, both short, are fused together into a long vowel, this vowel 
has the prdflishta svarita : thus, P. The illustrations cited by the com- 
mentator are abhi y 'hi manyo (iv. 32. 3), bkindM 1 'ddm (vii. 18. 1), and 
dipP 'tdh (xi. 2. 12 etc.).* But the arising of the circumflex is expressly 
restricted to the case in which both the i's are short : if the former of 
the two is long, it is very natural that, as the more powerful element, it 
should assimilate the weaker grave vowel, and make the whole com- 
pound acute. Thus ent'iva becomes enV 'va, not eni y 'va (v. 14. 11); 
maht' '■: iydm becomes mahi' 'yam (ix. 10. 12), etc. Such cases, especially 
of iva following an acute final t', are not uncommon.f The cases in 
which a long unaccented t, on the other hand, is preceded by a short ac- 
cented i, are exceedingly rare ; there is but a single one in the whole. 
Atharvan text, and it is cited by the commentator as a counter-example 
under the present rule, as follows : ma' vanim md' vd'cam no vi' "rtsth 
(v. 7. 6 : p. vi : irtsih) ; he adds, paro dtrghah: iha asmdt prdflishto na 
bhavati; 'the latter of the two i's is long: hence here the prdflishta 
accent does not arise.' We should, however, expect that in such a 
compound, especially, the circumflex would not fail to appear ; for if, in 
the fusion of i and i, the grave accent of the second element is repre- 
sented in the accentuation of the resulting long vowel, by so much the 
more should this be the case in the fusion of i + i, where the second 
element is the stronger. The teachings of accentual theory are so ob- 
vious and explicit upon this point that it is hardly possible to avoid the 
conclusion that the Hindu grammarians, in establishing their system, 
overlooked or disregarded the combination i + i, on account of its 
rarity, and that the accent of the cases later noted was made to con- 
form to the rule, instead of the rule being amended to fit the.cases. 
The Vaj. Pr. (iv. 133), indeed, makes a special exception of the word 
vfkshita (p. vi-tkshita), allowing it the praplishta circumflex which all 
such compounds palpably ought to have : whether the text of the 
White Yajus contains any other examples of the class, besides the one 
cited under the general rule (iv. 1 32 : hi' "m, from hi : im ; the same 
passage is the only one given under the corresponding rule in the Eik Pr. 
[iii. 7, r. 12, cxcixj), Weber does not inform us. It is not easy to see any 
reason in the nature of things why a combination of two w's should not 
be subject to the same law of accent as that of two i's. This, however, 
is another very rare case ; in the whole Atharvan not a single example 
occurs, nor have I happened to meet with any in the Rig-Veda ; that 
this is the reason why the Praticakbyas generally take no notice of any 
praflishta accent arising from such a combination is altogether proba- 
ble. It is not a little remarkable, then, that the Taitt. Pr. takes no note 
of the fusion of two i's as giving rise to a svarita, but (x. 17) prescribes 
it for the case in which, by a praglesha combination, u is formed, and 
applies (xx. 5) to this alone the name praglishta : the instances cited in 

* The other instances which the text affords are to be found at iii. II. 2. v. 22. 
7. vi. 92. 3 j 123. 2 ; 126. 3. vii. 26. 7 ; 73. 7. xi. 3. 18. xiL8. 84, 41. xv. 2. 3. xviii. 2. 

f In the Atharvan, we have them at iv.88.6. v. 14. 11 (bit); 20.11. vi.6.3. vii. 
62. 1. ix. 2. 6 ; 4. 2 ; 10. 12. x. 1. 14, 32. xi. 6. 1. xiv. 2. 31, 44. 

iii. 57.] Atharva- Veda Pr&tig&khya. 480 

its commentary arc s& x 'nniyam iva, *«£ ''dgAtd', rndsff HOsMhan, and 
dikshft , padadkdti. 

Our commentator discourses upon tbis rule at more than his usual 
length. After the citation of the examples and counter-example, with 
the remark upon the latter, which have already been given, he goes on 
as follows: ati 'i>a yah: ati 've 'ty ati-iva: tAirovyanjana ity eshah: 
ik&rak prdkfliskto yadd syad udattah purvah paro 'nuddttah sa prAJc- 
flisktah svaryala eva nityam mndkijam svaritam nd ''nyad dhuh : divt 
, vacakshuh: divfva jyotih: divi , ve , li divi-iva ; 'in the passage Ati 
, va yAh (ii. 12. 6) — where the ifcrama-text would read dWve Hy dli-iva 
■ — the circumflex of the t is tAirovyanjana' (see rule 62, below). ' When 
an t is the result of praflesha, the former element being acute and the 
latter grave, that result oipraplesha is always made circumflex. No other 
circumflex accent is declared to arise from the sandhi. Instances are divt 
Va cAkshuh (vii. 26. 7), divt Vo jyotih (vi. 92. 3), where the fcrama-text 
reads divt : ve Hi divi-iva' I am not quite sure that I have correctly 
interpreted all of this, but its significance is evidently of the smallest. 

Precisely the same objection lies against the form of this rule as 
against that of the preceding: that, whereas it ought to be a definition 
and a prescription, it is in fact merely nomenclatory. 

sTTrtT: UHon 

57. A circumflex, which follows a conjunction of consonants 
ending with y or v and preceded by an unaccented vowel, or 
which has no predecessor, is the jdtya. 

The meaning of the term jdtya, is ' natural, original, primitive :' the 
circumflex syllables to which it is applied are those which have that 
accent in their own right and always, independently of the combina- 
tions of the sentence. The long, lumbering, and awkward account of 
it which the rule offers may be divided into two parts : that which is 
necessary to its distinction from the enclitic svarita, which, by rule 67 
below, ordinarily follows any acute syllable, and that which is added to 
describe the kind of syllable in which it is invariably found to occur. 
The former part is contained in the words " preceded by an unaccented 
vowel, or having no predecessor," the enclitic circumflex being only 
possible as succeeding an acute. The other part is " after a conjunction 
of consonants ending with y or v :" that is to say, no syllable in Sans- 
krit has an independent circumflex accent except as it results from the 
conversion of an original accented i or u (short or long) into its corres- 
ponding semivowel y or v before a following dissimilar vowel; thus 
kanya> represents an earlier kard-d, svar an earlier sii-ar, and the like. 
The jdtya svarita, then, precisely corresponds in origin and in quality 
with the kshdipra, the one next to be described, and differs from it only 
in period, arising in connection with the combination of syllables into 
words, rather than of words into a sentence. 

The definition or description of the jdtya circumflex given by our 

490 W.D. Whitney, [iii. 57- 

treatise is after all imperfect, since it fails properly to distinguish the 
j&tya from the ksh&ipra. Such ksh&ipra accents as are instanced by 
abhykrcata (vii. 82. 1), nv etena (v. 6. 5), and the like, answer in every 
particular to the defined character of the j&tya. The word pade, 'in 
an independent or uncombined word,' or something equivalent, needs 
to be added to the rule. Such a limitation is not omitted from the de- 
scription of any of the other treatises. The Kik Pr. (iii. 4, r. 7, cxciv) 
defines all the other kinds of svarita first, and describes this as the one 
which remains, and which occurs in an uncombined word ; the Vaj. Pr. 
(i. Ill) gives a definition nearly corresponding to our own, but much 
more concisely expressed, and omitting the specification corresponding 
to our ap&rvam v&, ' or which has no predecessor,' which the comment- 
ator is therefore obliged to supply. Finally, the Taitt. Pr. (xx. 2) agrees 
quite closely with our treatise in the form of its definition, only adding 
the item noticed above as omitted here ; but it calls the accent nitya, 
' constant, persistent,' instead of j&tya. 

The commentator, after a simple repetition of the rule with the ap- 
pendix of svaro bhavati to fill out its ellipsis, adds in illustration the 
same series of words which he has given lis once before, under i. 17 : 
they are, as instances of the j&tya circumflex preceded by an unac- 
cented vowel, amavasya\ kanya' 1 , dh&nyam, Acaryah, and r&janyam; 
of the same in an initial or unpreceded syllable, nyan, kva, svah. 

58. The circumflex arising upon the conversion into a semi- 
vowel of an acute vowel before a grave is the kshdipra. 

The name ksh&ipra is given by the Rik Pr. (ii. 8) to the combinations 
taught in our rule 39, above, or to those in which i, t, u, ti, and f be- 
come respectively y, v, and r before a following dissimilar vowel : the 
term comes from kshipra, ' quick, hasty,' and marks the sandhi as one 
in which there is a gain of time, or a hastened, abbreviated utterance 
of the sernivocalized vowel. All the treatises (R. Pr. iii. 7 ; V. Pr. iv. 
47; T. Pr. x. 16) teach that in such a case, when the former element 
of the compound is acute and the latter grave, the resulting syllable is 
circumflex; and all (R. Pr. iii. 10; V. Pr. i. 115; T. Pr. xx. 1) apply 
to the circumflex thus arising the name of the combination, ksh&ipra. 
Our own rule is this time free from the ambiguities which we have had 
occasion to notice in the definitions of the abhinihita and pr&plishta 
accents, and would admit of being interpreted as a precept as well as a 
definition, as follows : ' in the case of the conversion into a semivowel 
of an acute vowel before a grave, there arises the circumflex called 

The commentator cites two actual cases of the ksh&ipra circumflex, 
viz. abhykrcata (vii. 82. 1 : p. abhi : areata) and vidvangah (vi. 125. 1 : 
p. vidit-anffah), and fabricates two others, viz. m&trkrtham, pitrhrtham 
(i. e. m&tr'-artham etc.). 

59. And even in the interior of a word, in a strong case. 

iii. 60.] Atharva-Veda Pr&tiqdhhya. 491 

It may be questioned whether this rule is merely nomenclatory, or 
also prescriptive; whether its meaning is that the svarita which appears 
in the finals of the strong cases of certain words whose themes are 
oxytone is to be accounted as kshdipra and not j&tya, or that a circum- 
flex of the kshAipra variety arises when the accented final vowel of a 
theme, in the strong cases, is converted into a semivowel before the 
case-ending. The same question comes up also in connection with the 
two following rules. But I presume that they are all to be understood 
as precepts, and to be reckoned along with the other cases in which 
our Praticakhya turns aside to meddle with matters lying without its 
proper sphere. Not one of the other treatises offers anything corres- 
ponding; they would all, apparently, class as j&tya the circumflex 
accents here treated of, not distinguishing them from the others which 
occur within the limits of a word, or in the uncombined text. The 
rules, however, are not without some interest, as showing that the 
authors of our work appreciated the entire analogy which the circum- 
flex accents with which they deal have with the ordinary kshdipra. 
Thus nadyhs is equivalent to nadi'as, as nadyasii would be to nadi' asti, 
while nadyd'i, nadya's represent nadi-d'i, nadi-d's ; the terminations of 
the strong cases showing no trace of that tendency which is exhibited 
by the other case-endings to draw away upon themselves the accent of 
the final vowel of the theme : compare tuddntarn, tuddnlau, tudaritas, 
with tudata', tudatds, iudalas. 

The commentator cites from the text, by way of illustration, nadyo 
na'ma ntha (iii. 13. 1), pippalydh sdm (vi. 109. 2), and rudatyah purushe 
hate (xi. 9. 14) ; and also, as counter-example, to show that the circum- 
flex arises only in a strong case, tayd sahasraparnyd' hr'dayam (vi. 1 39. 
1). Instances of both kinds are of not infrequent occurrence. In a 
small number of cases — viz. karkarydh (iv. 37. 4), prddkvdm (x. 4. 17), 
and viliptydm (xii. 4. 41) — all the manuscripts give an acute accent to 
endings of the class to which the rule refers ; these are, however, evi- 
dent errors, and should properly have been amended in the edited text 

The commentator closes his exposition, as so often elsewhere, with a 
verse which is a virtual re-statement of the rule, but its form is so cor- 
rupt that I refrain from attempting to restore and translate it. 1 

60. Or also throughout the declension, if the final of the 

theme is u. 

The manuscript reads in this rule ukdrasya, ' if the final is «,' but 
the facts seem to require the amendment to &, and the method of writ- 
ing of our copyist is too careless to make his authority of much weight 
against it. The rare termination ■k has power usually to hold its own 
accent, even before the ending of a weak case, and the resulting final 
syllable thus becomes circumflex. The commentator's instances are 
tanvd* (e. g. i. 33. 4), tanve (e. g. i. 3. 1), ultdndyop camvoh (ix. 10. 12), 

1 It reads: anloddttd nadi ny&yd hrasvandmi cat tathd: apancapadydm vaca- 
nam uddttah kshaipra ■ucyate. 

492 W. D. Whitney, [iii. 60- 

and vadkvap ca vdstram (xiv. 2. 41). The only other cases afforded by 
the text are tanvas (gen. or abl. : e. g. ii. 29. 1), tanv&'m (at i. 18. 3, and 
one or two doubtful places in the nineteenth book), and astisvas (x. 
10. 23). But there are also sundry exceptions to be found in the 
Atharvan, which the commentator has not failed to notice and record ; 
he says: lalo 'pavadati, 'from this rule one must make the exceptions,' 
urvdrvd' iva (vi. 14. 2), prddkv&'h (x. 4. 5), pvaprvd'i 1 (xiv. 2. 26), and 
■pvaprvd'h (xiv. 1. 44) ; to which is to be added vadhvd'i (xiv. 2. 9, 73). 
All these exceptional forms, it will be remarked, have a heavy ending, 
while of those which exhibit the circumflex accent the ending is light 
in every instance but one (tanv d K m). The words bdhvds (e. g. vii. 56. 
6) and Arvos (xix. 60. 2) are instances — and, if my search has been 
thorough, the only ones which the text presents — of like forms from 
themes in u, which are not to be regarded as contemplated by the rule. 

61. Also in oi}.yoh. 

That is to say, in the single word onybh the final syllable has a kshdi- 
pra svarita, though the form is a weak case, and not from a theme in #. 
The commentator cites the passage onybh havihratum (vii. 14. 1), the 
only one in which the word occurs. The Atharvan affords one other 
like case, viz. kalydny&H (vi. 107. 3), unless we are to assume there an 
error of the tradition represented by our manuscripts, which seems to 
me more likely. 

^sHGt^HW^sH.' ll ^u 

62. A circumflex between which and the preceding acute 
vowel consonants intervene, is the idirovyanjana. 

Literally, 'one separated by consonants is tdirovyafijana! There is 
here a notable change of subject and of implication. We have passed, 
without any warning, from considering the necessary or independent 
circumflex to treating of that which is enclitic only, arising, according 
to following rules (rules 67-70), in an unaccented syllable which is pre- 
ceded by an acute, and not again immediately followed by an acute or 
circumflex. Our treatise and the Rik Pr. (iii. 9, 10) subdivide the en- 
clitic circumflex into two forms only, the tdirovyafljana, where the cir- 
cumflexed syllable is separated from the acute by one or more conso- 
nants, and the p&davrtta (the Rik Pr. calls it vdivrtta), where a hiatus 
intervenes. The Vaj.'Pr. (i. 117, 119) and the Taitt. Pr. (xx. 6, 7) also 
teach the same accents ; but the former distinguishes under the idiro- 
vyanjana a sub-form, tdirovirdma ('having a pause between'), as occur- 
ring when the acute is parted from its enclitic circumflex not only by 
consonants but by the avagraka, or pause which separates the two parts 
of a compound word : thus, in prajd'-vat, for instance, the enclitic ac- 
cent of vat would be the tdirovirdma, while in pra-jd'ndm that of ndm 

? — MS. fva^urah. 

iii. 64.] Atharva- Veda Pr&tiq&khya. 493 

would be the t&irovyanjana simply. The t&irovir&ma, then, can occur 
only in the pada-text. The Taitt. Pr. takes no notice of this pada 
accent, but allows the name t&irovyanjana only to a circumflex which 
follows an acute in the same word : if the acute syllable is a final, and 
the circumflex an initial, the latter is to be denominated pr&tihata (xx. 3) : 
thus, in tdtra, the enclitic svarita of tra would be t&irovyanjana ; in 
tat te, that of te would be pr&tihata.* The practical importance of 
these numerous and varying subdivisions of an enclitic accent must be, 
as we cannot but think, very insignificant. 

The commentator's examples of the t&irovyanjana accent are id&m 
devah (ii. 12. 2) and id&m indra (ii. 12. 3) : it is, in both instances, the 
initial syllable of the second word which is made circumflex by the in- 
fluence of the preceding acute. According to the Taitt. Pr., both 
would be cases of pr&tihata. 

fezg^t MI<v=Jrt: » ^ <» 

63. Where there is a hiatus, the circumflex is pMavrtta. 

As was noticed in the last note, the Eik Pr. calls this accent v&ivrtta, 
' arising in connection with a hiatus.' The term p&davrtta is evidently 
a mutilated substitute for p&davivrtta or p&dav&ivrtta, ' arising in con- 
nection with a hiatus between two words.'f 

The commentator illustrates from the Atharvan text with y&'h Jcrly&i 
Angirasi'r y&'h krtya 1 asuri'r y&'h krty&'h svayamkrtA y&' u c& 'nyebhir 
&'bhrt&h (viii. 5. 9) : here are three cases of the p&davrtta circumflex, 
viz. in the syllables &hg and &s, following the first and second occurrence 
of krty&', and in the particle «, following the last ya'. 

?T^% tffZW ll $ n 

64. Where there is a division between the two parts of a com- 
pound word, the accent is of the same character. 

This rule is far from possessing all the explicitness that could be de- 
sired ; two or three different interpretations of it seem admissible. In 
the first place, it may be understood to apply only to a class of cases 
falling under the preceding rule, the implication of vivrtt&u being con- 
tinued ; it would mean, then, that when in the division of a compound 
word a hiatus is made between the acute final of the former member 
and the grave initial of the latter member, the enclitic circumflex accent 
of the latter is to be ranked as a p&davrtta, just as if the hiatus caused 
by the avagraha had been due to the operation of the ordinary rules of 
euphony. This interpretation is supported by the character of the 
citations made by the commentator to illustrate the rule ; they are as 
follows: uksha-ann&ya : vag&'-ann&ya (iii. 21. 6), yajna-rtah (viii. 10. 

* Roth (preface to his Nirukta, p. lxv) and Weber (p. 135) regard the tdirovirama 
and prdtihata as identical, but I do not see how this is possible. 

f The definition of the T&itt. Pr. (xx. 6) brings out this derivation more distinctly 
than our own; we read there padavivrttydm pSdavrttah. 

494 W.D.Whitney, [iii. 64- 

4),* patd-odand (e. g. x. 9. 1), patd-dyushd (iii. 11. 3, 4), and dtrghd- 
-dyushd and dlrgkd-dyushah (not to be found in AV. : we have, how- 
ever, the nominative dirghd-dyuh, at xiv. 2. 2, 63) : in all these com- 
pounds, the enclitic svarita of the vowel following the sign of division 
is pddavrtta. But the rule as thus interpreted seems of very little use, 
since such cases might be regarded as falling under the preceding rule 
without any special direction to that effect. The commentator does 
not include the word vivrttdu in his paraphrase, and the fact that his 
citations are all of one class is far from conclusive against the intended 
comprehension of the other classes also under the same precept. If 
the implication of vivrttdu be rejected, and avagrahe be taken to mean 
'in any case of the separation of the two parts of a compound,' whether 
accompanied by a hiatus or not, there will, however, still remain a ques- 
tion as to the signification of savid/iah, ' of the same character.' Does 
it refer only to pddavrltah, and shall we assume that the enclitic cir- 
cumflex of an initial syllable after a pause of separation, or avagraha, 
is always to be reckoned as pddavrtta, the pause having the same effect 
in all cases as a hiatus — so that in patd-vrshnyam (i. 3. 1 etc.), sdm-pru- 
tam (i. 3. 6), d'-bhrtdh (i. 6. 4), and the like, the syllables v r, pru, and bhr 
would be pddavrtta ? Or does it refer also to tdirovyanjana, and shall 
we understand that, notwithstanding an avagraha, the accent of a fol- 
lowing syllable remains just what it would be were there no such pause ; 
a hiatus conditioning a pddavrtta, and the intervention of consonants 
(as in the examples last quoted) a tdirovyanjana? Of these three pos- 
sible interpretations, I am inclined to favor the last; and especially, as 
it is supported by the authority of the Rik Pr., which lays down the 
general principle (iii. 15, r. 23, ccx) that where syllables are separated 
by avagraha, their accentuation is the same as if they were connected 
with one another according to the rules of sandhi. 


65. Of the abhinihita, prdgltshta, jdtya, and hshdipra accents, 
when followed by an acute or a circumflex, a quarter-mora is 
depressed : this the wise call vikampita. 

The reason of this rule cannot be made evident without a somewhat 
detailed exposition of the laws laid down by the Hindu grammarians 
as regulating the rise and fall of the voice in connection with the con- 
secution of the accents. In the first chapter (i. 14-17) we had merely 
a brief definition of the three tones of voice in which a syllable may 
be uttered : the low or grave (anuddtta), belonging to unaccented sylla- 
bles, the high or acute (uddlta), which is the proper tone of an accented 
syllable, and the circumflex (svarita), combining a higher and a lower 

* The pada manuscript reads yajtla-r'tak, but such an accentuation is contrary to 
Jill analogy, and would in itself be in a high degree suspicious ; and it is fully con- 
victed of falsity by the citation of the word as an instance under this rule. 

Hi. 65.] Atharva' Veda Prdtigdkhya. 495 

pitch within the limits of the same syllable, and, as we have seen, always 
resulting, as an independent accent, from the fusion of two originally 
separate elements, of which the one was acute and the other grave. If 
this were the whole story, the subject of accent in Sanskrit would bo 
of no more intricacy and difficulty than in Greek : nor even of so much, 
since in Sanskrit neither of the accents is restricted as regards the place 
which it may occupy in the word ; and we should only have to note 
and learn upon which syllable, and with which accent, each word in the 
language was uttered, and what were the few simple rules which govern 
the combinations of accented and unaccented syllables in the phrase. 
A great complication, however, is introduced into the system, in the>. 
first place, by the rule, taught by all the Hindu authorities (see our rule 
67, below), that an originally grave syllable, when it follows an acute, 
receives an enclitic circumflex : that is to say, that the voice, when once 
raised to the pitch of acute, does not ordinarily come down with a leap 
to the level of the grave, but makes its descent in the course of the next 
following syllable ; or, to illustrate by an instance, that we do not say 

a- mlt 'tran, but a- mtt ' tr &n* To this rule is made the important excep- 
tion (mle 70, below), that the syllable which would otherwise receive 
the enclitic circumflex maintains its character of grave, in case an acute 
or circumflex comes next after it : the theory being, apparently, that 
the voice prepares itself for rising to the acute pitch by sinking before 
it: it must, if possible, mount from the station of a syllable wholly 
grave. Thus we have, as the mode of utterance of amitran p&ri and 

amilrdn leva, a' m,t 'tran- n P ar i, a- mit tran- n kv a . Now comes the farther 
complication, in which all the Pratic&khyas agree (see rule 71, below), 
that the unaccented syllables which follow a circumflex, although re- 
garded as having the value of grave, are yet pronounced at the pitch of 
acute : that is to say, that, in pronouncing prshadajy&pr'anuttanam, we 
&a 7 pr-sha-ddjy a PPran utt&n&m , and not pr'sha-ddjV a PPr a .„ u i. (a . nam , 
This grave accent with the tone of acute is in the Rik Pr. (iii. 11) and 
Taitt. Pr. (xxi. 10) called the pracaya (the word means ' accumulation') : 
its theoretical ground I find it exceedingly difficult to discover. But it 
evidently stands in close relation — whether as cause or as effect, I would 
not attempt to say too confidently — with a somewhat different descrip- 
tion of the character of the circumflex. The first portion of the latter 
accent, namely, is by the Rik Pr. (iii. 2, 3) declared to be uttered, not 
at acute pitch, but with a yet higher tone, and its later portion at acute 
pitch. The Taitt. Pr. (i. 41-42) gives the same account of a circumflex 
that immediately follows an acute, although, as we have seen (in the 
note to i. 17), liberally citing the discordant opinions of other authori- 
ties. These two treatises, then, would require our pronunciation to be 

pr-thad&ry a P Vrar > utt&nAm . Neither our own work nor the Vaj. Pr. 
gives such a definition of the circumflex ; and yet the theory of the 

* In writing these instances, I follow the rules for the division of the syllables 
given in the first chapter [i. 65-68); and also, in order not to misrepresent them, I 
make the duplications of the vartyakrama (iii. 26-32), but omit any attempt to de- 
signate abhinidhdna, yatna, etc. 

vol. vii. 63 

496 W. D. Whitney, [ilu65. 

pracaya accent, to my apprehension, so manifestly recognizes and im- 
plies it, that I cannot believe otherwise than that its statement is omitted 
by tbem, and that it really forms a part of their system. If the voice 
has already, in the utterance of the enclitic svarita, sank to the actual 
grave pitch, it can scarcely be believed that it should be called upon to 
rise again to the level of acute for the utterance of the following unac- 
cented syllables : while, on the other hand, if the circumflex be removed 
bodily to a higher place in the scale, and made to end at acute pitch, 
the following grave syllables might naturally enough be supposed to 
run on at the same level. Hence I regard the form of the word last 
given as representing the mode of its pronunciation which must be 
assumed to be taught by the theory of our treatise, as well as by that 
of the others. If, now, the grave syllables succeeding a circumflex are 
uttered at acute pitch, how shall an actual acnte, occurring after them, 
be distinguished from them ? Provision for this case is made in the 
rule, common to all the treatises (see our rule 74, below), that the grave 
which next precedes an acute or circumflex is not of acute tone, but 
remains grave. Thus, putting an acute syllable after the word which has 
been employed in illustrating the pracaya accent, we should have prsha- 

ddjyapranuttdndm ma 1 pronounced as prsha-ddj-^ a P a 'V ut ' tA 'ndm' . 
Thus is assured to the voice a low syllable from which to rise to the 
following acute, just as where the enclitic svarita, before an acute or cir- 
cumflex, was given up for a grave pronunciation. Analogous with these 
two cases is that which forms the subject of the rule at present under 
discussion. It is constantly happening that an acute syllable follows 
one which has an independent circumflex, which cannot, of course, like 
the enclitic, be converted to grave out of complaisance to its successor. 
If, however, it were left unaltered, the distinction of the following acute 
from a pracaya anuddtta would be endangered. If ye 'sy&'m stha pra- 

ti'cy&m dici should be uttered y ^»''V^ (fc " 3 ' li »'^, it might 
be understood as ye 'sy&ih stha pra° etc. ; while the sinking of the cir- 
cumflex syllable ye to the level of the anud&tta pitch — as e s- s V am etc 
— would mark the following syllable as truly uddtta or acute. But that 
the avoidance of such ambiguities was the sole, or even the principal, 
reason for the depression of tone taught in the rule is rendered improb- 
able by the fact that the same is prescribed also before an independent 
svarita, which latter could not, even without any alteration of tone in 
the preceding syllable, be confounded with an enclitic svarita or with 
any other accent. The depression is more probably owing to the theo- 
retical requirement that the voice should, when possible, always rise to 
the utterance of a real acute or circumflex from the lowest or anud&tta 
pitch : which is satisfied by the retention of the anuddtta quality before 
an uddlla or svarita in a syllable which would otherwise become enclitic 
svarita or pracaya anuddtta, and by the lowering of the final portion of 
an independent svarita in the same position. 

The equivalent rule of the Rik Pr. (iii. 3, r. 5, cxcii) is given in con- 
nection with the first definition of the svarita accent : this is to be of 
the pitch of uddtta in its latter portion, unless an uddtta or svarita fol- 

iii. 65.] Atharva- Veda Prdtfydkhya. 497 

lows, in which case it is anud&lta. The commentary informs us that 
the depression to anuddtta pitch is called kampa, a term connected with 
the vikampita of our rule (both come from the root karnp, ' to tremble'). 
An interpolated verse -at the end of the third chapter of the Rik Pr. 
(iii. 19) restates the same precept, in terms closely corresponding with 
those of our treatise. The Vaj. Pr. (iv. 13V) says, in like manner, that 
before an uddlta or svarita the latter portion of a svarita is farther de- 
pressed. It is upon the authority of the Vaj. Pr. (i. 60) that the term 
an.uma.tra, literally ' minute measure,' used in our text, is interpreted to 
signify the definite quantity of a quarter-mora. The same phenomenon 
of kampa is treated in the Taitt. Pr. (xix. 3 etc.), but with peculiar 
complications of theory which it is not necessary to explain here. 

The commentator offers instances of each of the kinds of circumflex 
mentioned in the rule ; they are as follows : of the abhinihita, yd 'bhiyd- 
tah (xi. 2. 13), ye 'sy&'m (iii. 26. 1 etc.), so 'rtham (the Atharvan con- 
tains no such phrase, nor can it be a genuine instance, as ariha has the 
acute on the first syllable, and the accentuation of the two words com- 
bined would be s6 'rtham; it is altogether probable that the reading is 
corrupt, and that the phrase intended to be cited is so 'bhvam [xiii. 4. 
25] : this is the nearest approach to the other which our text furnishes, 
and is moreover an instance of a circumflex before a circumflex, which 
the commentator would be likely to seek) ; of the prdflishta, bhindhf 
'dam (vii. 18. 1), dip? Hah (xi. 2. 12 etc.) ; of the jatya, am&v&sya'' svah 
(these two words do not occur in juxtaposition in the Atharvan : we 
have amdvdsya> m&'m at vii. V9. 2, and e. g. svar nd at ii. 5. 4) ; and of 
the kshdipra, nadyo nd'ma stha (iii. 13. 1), pippalyah sum (vi. 109. 2), 
and rudatyah p&rushe hate (xi. 9. 14) : they are to be pronounced 

ru-dat- t Va <p .pu- r u-she- ha .te^ bhin' d h- d ^ m i ^oMi-Vd-tah^ etc _ 

Whether the Hindu grammar is much the gainer by this intense 
elaboration of the accentual theory may fairly be questioned : whether, 
indeed, it has not lost more than it has gained by the exaggeration, and 
even the distortion, in more than one particular, of the natural inflec- 
tions of the voice. To me, I must acknowledge, it seems clear that 
those ancient grammarians might better have contented themselves with 
pointing out in each word the principal accent and its character, leaving 
the proclitic and enclitic accents, the claimed involuntary accompani- 
ments of the other, to take care of themselves; or, if they could not 
leave them unnoticed, at least stating them in a brief and general way, 
as matters of nice phonetic theory, without placing them on a level 
with the independent accents, and drawing out a complete scheme of 
rules for their occurrence. The obscurity and false proportion given by 
them to the subject of the Sanskrit accent has availed to confuse or 
mislead many of its modern investigators: and we find, for instance, 
our modern Sanskrit grammarians explaining the independent circum- 
flex as originated by the fusion of an acute with an enclitic circumflex, 
in which fusion the former accent gives way to the latter, the substance 
to its own shadow ! The Praticakhyas, it will be noticed, countenance 
no such explanation, but describe the real circumflex as arising from 
the combination of an acute with a following grave element. 

498 W. D. Whitney, £iii. 65. 

The designation of the accents in the Atharvan manuscripts is some- 
what various, and requires a brief explanation. The Rik method of 
designation is now familiar to all students of the Vedic department of 
the Sanskrit literature, and is perfectly adapted to the theory of the 
accent as above set forth ; all syllables uttered at grave or anud&tta 
pitch have a horizontal mark below them : a svarita or circumflex sylla- 
ble, whether its circumflex be independent or enclitic, has a perpendicu- 
lar stroke above ; all syllables spoken with acute or uddtta tone are left 
unmarked, whether they be proper ud&tta or accented syllables, or only 
pracita anud&tta, grave in value but acute in pitch. An example is the 
verse (xi. 10, 19) from which some of the illustrations given above have 
been drawn: 

trishandhe tdmasd tvdm amitrdn p&ri vdraya, 

4^isdujTi hi hihNi *ftfa wh 

prshadajydpranutt&n&ih md 1 'mt'shdm mod k&p cana. 

The agreement between theory and method of designation here is, 
indeed, so close as reasonably to awaken suspicion whether the latter 
may not have exercised some influence upon the former. This mode 
of marking the accented syllables, now, has been adopted in the edited 
text of the Atharvan, although not followed throughout by a single one 
of our Atharvan manuscripts. In these, the circumflex, whether inde- 
pendent or enclitic, which follows an acute is usually marked within the 
syllable itself, sometimes by a dot, sometimes by a horizontal line ; the 
independent circumflex after a grave, generally by an oblique line drawn 
upward and across the syllable, but often by a convex line below it. 
The grave syllable is marked as in the Rik, or, quite as often, by a dot 
instead of a horizontal line below. The different methods are inter- 
changed in several cases within the limits of a single manuscript, and 
as some of their features are incapable of being represented in printing 
without complete suits of type prepared expressly for the purpose, no 
one can call in question the right of the editors to substitute the Rik 
method.* But there is one other case, namely that which forms the 
subject of the rule to which this note is attached, which gives trouble 
in the designation. A svarita immediately preceding an uddtta cannot 
receive simply the usual svarita sign, lest the following syllable be 
deemed » pracaya instead of uddtta. The method followed in such a 
case by the Rik is to append to the circumflexed vowel a figure 1 or 3 
— 1 if the vowel be short, 3 if it be long — and to attach to this figure 

• Lesser and occasional peculiarities of the manuscript usages are passed over 
without notice : it may be farther remarked, however, that E., throughout a great 
part of its text, marks the acute syllables with the perpendicular line above, in 
addition to all the other usual signs of accent. The peculiar ivarita signs of the 
White Yajus (see Weber, p. 138) nowhere appear. 

iiL 65.] Atkarva- Veda Pr&liq&khya. 499 

the signs both of svarita and of anudatta. Instances, in part from the 
examples already given, are : 

diet Uah nadyb n&'ma rudatyhh puruthe 

.i O _ _ _ — — — *_ 

devy K ushasah gharmb 'bhi y 'ndhe bk&tfd 'pnv antali 

(x. 8.30), ' (xi. 3. 18), (x. 5. 15). 

What is the reason of this style of writing the vikampita syllable, we 
are left to find out for ourselves; the l'ratic&khyas teach ns only 
the modes of utterance. In seeking to explain it, we may first note 
the method pursued in the text of the Sanaa- Veda, which is always to 
protract the vowel of the syllable, lengthening it if short, and adding 
the usual sign of protraction or pluti, the figure 3. This suggests to 
us, as not wanting in plausibility, the conjecture that the two Rik signs 
have a similar meaning, and are marks of protraction, the I indicating 
the mora or part of a mora by which the short vowel is regarded as in- 
creased, and the 3 the protracted or pluta value, to or toward which 
the long vowel is raised in pronunciation ; the cause of the prolonga- 
tion lying in the necessity of carrying the vowels concerned to a lower 
pitch of voice, which leads to an extension of their quantity — even 
though the theory of the Pratic&khyas known to us does not recognize 
any such extension. The attachment to the figure of the signs of both 
svarita and anudatta, tone of course denotes that the syllable, although 
circumflex in its general character, descends, unlike the other circumflex 
syllables, to the full level of anudatta pitch, indicated by the lower 
horizontal mark. This Rik method of notation of vikampita syllables 
has also been followed in the published Atharvan text, along with the 
rest of the system to which it belongs. As to the Atharvan manu- 
scripts, they are not infrequently careless enough to omit the figure 
altogether, and when they write it, it is in almost all cases a 3, whether 
the vowel to which it is appended be long or short.* In about twenty 
passages,f they follow the method of the Sama-Veda, and prolong the 
vowel of the syllable : this has, however, been restored to its short 
form in the edition, except in three instances (x. 1.9. xiii. 1. 15; 3. 
16). It will, I presume, be generally acknowledged that, in this condi- 
tion of the manuscript authorities, the editors followed the safest course 
in accepting and carrying out consistently the Rik method of designa- 
tion of the class of cases under discussion. 

* In the second half of the text, or in books x-xx, I have noted but a single pas- 
sage where all the MSS. read 1 after a short vowel, and but three others in which 
that figure is given by more than a single authority: its occurrence is in general en- 
tirely sporadic; it occasionally appears also, in place of 3, after a long vowel. 

+ The details are as follows: all lengthen the vowel at vi. 109. 1. x.1.9. srii.4.4; 
5.21. xiii. 1.15; 3.16. xix.44. 8; one or more make the same prolongation at ii. 
13.5; 33.5. vi.97. 1; 109.2; 120.3. vlil 4. 18. x.8. 30. xii.5. 53. xvi. 6. 5. xviii. 
1.3; 2.24. xix.3.2. 

500 W. D. Whitney, [iii. 0fl- 

^=hl^l ^<ItM"1<ItI! II \\ II 

66. A vowel produced by combination with an acute is itself 

That is to say, when two simple vowels, or a vowel and a diphthong, 
coalesce and form a single vowel or diphthong (by rules 42-53, above), 
in case either of the two was acute, the resulting syllable is acute. The 
rule is a general one, and suffers only the two exceptions which form 
the subject of rules 55 and 56, above : namely, that e and 6, when they 
absorb a following a, become e and 6, and that i+i become t\ The 
corresponding rules of the other treatises are Rik Pr. iii. 6 (r. 10, cxcvii), 
Vaj. Pr. iv. 131, and Taitt. Pr. x. 10 and xii. 10. 

The series of passages given by the commentator in illustration of 
the rule is the same which was furnished under rule 38, above : viz. 
dhiye " 'hi (dhiyd' : &' : ihi), jushasve " 'ndra (jushasva : &' : indra), sta- 
nayilwune " 'hi (stanayitnund : &' : ihi), kushthe " 'hi (kushtha : d' : ihi), 
vdakeni " 'hi (udakena : a' : ihi), and dva papyate " 'ta (dva :pafyata : 
& : ita). The instances are ill chosen, so far as regards variety of com- 
bination ; but they illustrate sufficiently the fact that, whether the acute 
element is the former or the latter of the two combined, the accent of 
the result of fusion is alike acute. It might have been well to offer 
also examples of the extreme cases in which a final acute d assimilates 
and renders acute a following diphthong, such as pat&'udand (fat&- 
-odand: e. g. x. 9. 1) and ihd'i 'vd (ih& : evd: e. g. i. 1. 3) : such combi- 
nations with di and &u the text does not contain. 

The other treatises give the rule (R. Pr. iii. 6, r. 1 1, exeviii ; V. Pr. iv. 
130 ; T. Pr. x. 12) that when a circumflex is combined with a following 
grave, the result of the combination is circumflex. A like rule is needed 
here also, in order to determine the accentuation of such cases as ianv6> 
'ntdriksham (xvii. 13: tanvd : antdriksham), pathye 'va (xviii. 3. 39: 
pathyd-iva), etc., and its omission must be regarded as an oversight. 

4<lrU«H<lri S5T&7T II \o II 

67. A grave syllable following an acute is circumflesed. 

This is the rule prescribing the enclitic svarita, the position of which in 
the accentual system has been sufficiently explained in the last note bnt 
one. It is, as we have seen above (rules 62, 63), subdivided into the 
two kinds called tdirovyafijana and pddavrtta, according as one or more 
consonants, or only a hiatus, intervene between the acute vowel and its 
successor. The commentator gives here only the two instances of the 
tdirovyaRjana which he had already cited under rule 62. 

The corresponding rules of the other treatises are Rik Pr. iii. 9 (r. 16, 
cciii), Vaj. Pr. iv. 134, and Taitt. Pr. xiv. 29, 30. 

&m 4qr hhh^I hMmi 

68. And even in the disjoined text, within the limits of the 
same word. 

iii. 70.] Aiharva- Veda Prdtiq&khya. 501 

The term vyasa, excepting here and in rule 72, is not met with any- 
where in the Hindu grammatical language. Our commentator, accord- 
ing to his usual custom, spares himself the trouble of giving any ex- 
planation of it, or even of replacing it in his paraphrases by a less 
unusual synonym. The instances adduced, however, in illustration of 
the rules, and the general requirements of the sense, show clearly that 
it means the disjoined or pada text. By their adhik&ra (ii. 1), all rules 
in the second and third chapters should apply only to the sanhild, or 
combined text ; hence it must here be specifically taught that in pada, 
as well as in sanhitd, the syllable following an acute is enclitically cir- 
cumflex, if the latter be in the same word with the former, and so not 
separated from it by a pause or avas&na. The examples given under 
the rule are ay&tam (e. g. viii. 2. 21), amr'tam (e. g. i. 4. 4), and antd- 
rikskam (e. g. ii. 12. 1) ; in these words, the syllables tarn, tarn, and rik 
have the i&irovyanjana variety of the enclitic circumflex, and they are 
accordingly written sajfrPT I «*£ri*jn a *d flirty, and not 3rg?W I ayrtyj 

A corresponding rule is to be found in the Rik Pr. at iii. 4 (r. 6, exciii). 

69. As well as where there is a separation of a compound 
into its constituents. 

That is to say : although, in the pada-text, the pause which separates 
each independent word from the one following it breaks the continuity 
of accentual influence, so that a final acute of the one does not render 
circumflex the initial grave of the other, yet the lesser pause of the 
avagraha, which holds apart the two members of a compound word, 
causes no such interruption ; on the contrary, an acute at the end of 
the former member calls forth the cirenmflexed utterance in the first 
syllable of the latter member. The commentator offers us rather a mo- 
notonous series of illustrations, namely su-sampitah (vi. 105.2), su-yalah 
(vi. 111.1), su-prtam (vii. 72. 3), su-drdham (x. 2. 3), su-bhrlam (e. g. vi. 
39. 1), and su-hutam (e.g. vi. 71. 1) : these are to be pronounced and 
written ^r-s ^•fiftcT: i H-szfcr: etc., and not srssffsnT; t 'CT-sarrT: etc. 

The Rik Pr. (iii. 15, r. 23, ccx) and Vaj. Pr. (i. 148) lay down the 
principle that avagraha makes no difference in the accentuation of the 
syllables which it separates; both, also (R. Pr. iii. 16, V. Pr. i. 148), 
except the rare cases in which each of the parts of a compound has 
an acute accent; such as arc, in the Atharvan, pari-dhdtav&'i (ii. 13. 2) 
and tdnh-ndpdt (v. 27. 1). Jn such words as the latter, the Vaj. Pr. (i. 
120) calls the circumflex of the syllable preceding the pause by a special 
name, tdthdbhdvya. 

sfi<lrMf(rW(H II oo ii 

70. Not, however, when an acute or circumflex syllable suc- 

502 W. D. Whitney, [iii. 70- 

A syllable originally grave remains grave before a following uddtta 
or tvarita, even though preceded by an uddtta, and hence, by the last 
rules, regularly entitled to the enclitic circumflex. The proclitic accent 
thus appears, in the estimation of the Hindu phonetists, to be more pow- 
erful than the enclitic, and the law which requires the voice to plant 
itself upon a low pitch in order to rise to the tone of acute or indepen- 
dent circumflex to be more inviolable than that which prescribes a 
falling tone in the next syllable after an acute. The commentator illus- 
trates by citing path na d'pah (i. 6. 4), yd asyd vipvdjanmanoh (xi. 4. 23), 
and asyd sut&sya svah (ii. 5. 2) : the syllables no, as and vip, and m 
and syas, which by rule 67 would be circumflexed, are by this rule rein- 
stated in their anudatla character, and must be written with the anu- 
ddtla sign below : thus, sf :r sgrt: i nt srer fascisfapT. i $RK smtzr ssh. 

The corresponding rules of the other treatises are Kik Pr. iii. 9 (r. 
.16, cciii, last part), Vaj. Pr. iv. 136, and Taitt. Pr. xiv. 31. Cases of oc- 
currence of the accent called by the Vaj. Pr. tdthdbhdvya (as noticed 
under the last rule) constitute in the other systems an exception, which 
is apparently not admitted by the school to which our Pratic&khya 

*c(f(HI<^j<lTl 4<lrlH}fri: \vo\u 

71. A grave following a circumflex has the tone of acute. 

The position and relations of this rule in the accentual system have 
been sufficiently treated of in the note to rule 65. All the other treat- 
ises (R. Pr. iii. 11, r. 18, ccv; V. Pr. iv. 138, 139; T. Pr. xxi. 10) lay 
down the same principle, stating also distinctly what must be regarded 
as implied in our precept, that not only the single grave syllable which 
immediately follows the circumflex receives the acute utterance, but 
those also which may succeed it, until, by rule 74, the proximity of an 
acute or circumflex causes the voice to sink to the proper anud&tta tone. 
The Rik Pr. and Taitt Pr. use the term pracaya, ' accumulation, indefi- 
nitely extended number or series,' in describing this accent, the latter 
employing it in its ordinary sense, the former giving it as the name of 
the accent. 

The commentator cites from the text, as instances, devtr dvdro brhattr 
vipvaminvdh (v. 12. 5), and md'dhvt dhartdrd vidathasya satpati (vii. 
73. 4) : each passage presents a whole pracaya, or accumulated series, of 
syllables having the accent prescribed by the rule. Such syllables are, 
as has been already pointed out, left unmarked with signs of accentua- 
tion in the written texts, like the proper acute syllables whose tone they 
share : thus, Zc&jft ^Jl fekafiRn : . 

Next follow two rules, identical in form with rules 68 and 69 above, 
and, like them, prescribing the application of this principle in the pada- 
text also. They are covered by the same rules of the other treatises 
which have already been cited. 

iii. H.] Atharva- Veda Prdtigdkhya. 508 

5in% tffq" tWM^ H ^h 

72. And even in the disjoined text, within the limits of the 
same word. 

That is to say, in pada as well as in sanhita, those unaccented sylla- 
bles which follow in the same word a circumflex, whether independent 
or enclitic, are uttered at the pitch of acute ; and they are correspond- 
ingly marked in the written texts. The commentator's examples are 
uru ff&'l&y&h (v. 13. 8) and hika'tjk&m (x. 2. 8) : these are uttered and 
written ^.sn^farr: I 4h=hl(t'<*W; and not 3X-S ijtrilUl: I thohifi'ehUJ. 

73. As well as where there is a separation of a compound 
into its constituents. 

That is to say, again, tbe avagraha, or pause of division, does not in- 
terfere with the influence of a circumflex, any more than (by rule 69) 
with that of an acute, upon the following unaccented syllables. The 
commentator gives us the examples pvan-valih (xi. 9. 15), svah-vaCt 
(xviii. 1. 20), and dpman-vati (xii. 2. 26), which we are to read and write 
sstans cleft: i sf:isi(ft i Atf-l^ =r?ft, and not ssfans 5fi?h i ^oti^arft i Wi.4^ 

^l(rfi<lT* ^H«H*HH<lrlH iio8n 

74. But the syllable immediately preceding a circumflex or 
acute is grave. 

This rule applies only to those originally anud&tta syllables which 
would otherwise, under the action of rule 71, be spoken with the itd&tta 
tone, as following a svarita. It is, as has been already pointed out, 
closely analogous in character with rules 65 and 70, above, and has a 
like theoretic ground. The commentator illustrates it by citing the 
passages ajds tad dadrge koa (x. 8. 41), idam devah crnuta ye (ii. 12. 2), 
and iddm indro prnuhi somapa yat (it. 12. 3), where the syllables pek, 
ta, and pa — which, though properly unaccented, would be pronounced 
at acute pitch, like their predecessors, by rule 71 — are depressed to the 
level of grave, in preparation for the succeeding circumflex or acute. 
We write, accordingly, sn^ ^sr eft I p; £cTT: sjrrjH ir i ^fif^ sjnjf^ 

This rule is common to all the systems : compare Rik Pr. iii. 12 (r. 
20, ccvii), Vaj. Pr. iv. 140, and Taitt. Pr. xxi. 11. 

As, at the beginning of this section, the commentator gave us an in- 
troduction to it, composed mainly of citations from unnamed sources, 
so herc r at the end, he appends a postscript, chiefly made up of the 
dicta, upon points connected with accentuation, of authorities to whom 
vol.. vn. 04 

504 W. D. Whitney, [lii. 74- 

he refers by name. He first says : asvar&ni vyaUjanAni svaravanti , ty 
Anyatareydh ; ^the consonants are destitute of accent: "they are ac- 
cented," says Anyatareya.' Upon this point our treatise nowhere dis- 
tinctly declares itself, but, as already noticed (under i. 55), its rules of 
syllabication may be naturally interpreted as implying that the conso- 
nant which is assigned to a particular syllable shares in the accentuation 
of its vowel. The Vaj. Pr. (i. 107) states this explicitly. Next we 
have: Mm samdheh svaritam bhavati: purvar&pam ity dnyatareyah : 
uttararitpam pAnkftamitrih ; 'what part of a combination is circum- 
flexed (or accented) ? " the former constituent," says Anyatareya ; " the 
latter constituent," says Qankhamitri.' In the absence of any illustra- 
tions, I am at a loss to see to what kind of combinations this question 
and its answers are to be understood as applying. Next follows a pas- 
sage which we have had once before, in the commentary on i. 93 (see 
the note to that rule) ; it reads here as follows : Mm aksharasya svarya- 
m&nasya svaryate: ardham hrasvasya pAdo dtrghasye 'ty eke sarvam iti 
fAnkhamitrir aksharasydi , sha vidhA net vidyate : yad yad vifvaribhava. 
A renewed consideration affords me no new light upon this passage. 
Finally, we read rgardharcapadAntandvagrahavivrttishu mAtrAkAlakAlah, 
which appears to me to have to do with the determination of the length 
of the different pauses occurring in the recitation of the Veda ; namely, 
the pauses at the end of a half-verse, between two words in the pada- 
text, between two parts of a divisible compound (also in pada-text), and 
where a hiatus occurs in sanhitA ; but I have not succeeded in restoring 
any intelligible and consistent reading of the passage. The Vaj. Pr. 
(v. 1) and Rik Pr. (i. 6, r. 28, xxix) declare the pause of division of a 
compound, or the avagraha, to have the length of a mora ; the Taitt. 
Pr. (xxii. 13) recognizes four pauses (virAma) : that at the end of a 
verse, that at the end of a pada or disjoined word, that of a hiatus, and 
that of a hiatus within a word (as praugam), declaring them to have 
respectively the length of three moras, two moras, one mora, and a 

The signature of thepAda or section is trttyasya trtiyah pAdah. 

75. After r and f, r, and sh, within the limits of the same 
word, n is changed to n. 

This rule is an adhikAra, or heading, indicating the subject of the 
section, and making known what is to be implied in the following rules. 
It might properly, then, have been left by the commentator without 
illustration, like ii. 1. He prefers, however, to cite from the text the 
passages pari strnihi pari dhehi vedim (vii. 99. 1) and paristaranam id 
dhavih (ix. 6. 2), which offer instances of the lingualization of the nasal ; 
and he also adds koshanam and toshanam, which are not Atharvan 
words. As counter-examples, to show that the nasal is converted only 
if the preceding cerebral is found within the same word, he gives svar 
nayati, prAtar nayati, which are also fabricated cases ; compare the 
similar ones in the scholia to Pan. viii. 4. 1. 

iii. 77.] Atharva-Veda Prdtiqdkhya. 505 

The Eik Pr. introduces the same subject with a precept (v. 20, r. 40, 
ccclvii) which also includes our next rule and a part of 89 below. The 
leading rules of the Vaj. Pr. (iii. 83) and Taitt. Pr. (xiii. 6) are in close 
agreement with ours. 


76. In drugha,7ia etc., the same effect takes place after a cause 
which is contained in the former member of a compound. 

The commentator paraphrases p&rvapad&t by the expression which, 
for the sake of clearness, has been substituted for it in translating, viz. 
purvapadasthdn nimittdt. His cited illustrations are drughanah (vii. 
28. 1 : p. dru-ghanah), s&td gramanydh (iii. 5.1 : p. gr&ma-nyah), ra- 
kshohanam vdjinam (viii. 3. 1 : p. rakshah-hanam), brhaspatipranuttdndm 
(viii. 8. 19: p. brhaspati-pranuttdndm), prshaddjyapranuttdndm (xi. 10. 
19 : p. prshaddjya-pranuttdndm) , and durnihitdishinim (xi. 9. 15 : p. 
durnihila-eshinim). The latter case, we should think, is one of some- 
what ambiguous quality, since in the form of the word, as given by the 
pada-text, there is nothing to show that dur stands in the relation of 
purvapada, or former member of a compound, to nihita, they being un- 
separated by avagraha. The same objection lies against the two pre- 
ceding instances ; but also the much more serious one that they are 
examples properly belonging under rule 79, below, the converted n 
being that of a root after a preposition. 

The other examples of the action of this rule afforded by the Athar- 
van text are pardyana (e. g. i. 34. 3 : p. pard-ayana), vrshdyamdna (ii. 
5.1 : p. vrsha-yamdna ; the uttarapada is here a suffix of derivative 
conjugation), paripdna (e. g. ii. 17. 7 : p. pari-pdna), vrtrahanam etc. 
(e. g. iv. 28. 3 : p. vrtra-hanam), durnaga (v. 11. 6 : p. duh-napa), trina- 
man (vi. 74. 3 : p. tri-ndman), purundman (vi. 99. 1 : p. puru-ndman), 
ur&nasa (xviii. 2. 13 : p. uru-nasa), and sahasranttha (xviii. 2. 18 : p. m- 
hasra-nltha). We have also a few cases of a class analogous with the 
last one given by the commentator, where the principle is precisely the 
same, although, in the actual division of the words, the avagraha falls 
elsewhere than between the converting lingual and the nasal : they are 
suprapdna (iv. 21. 7 : p. su-prapdna; our ^ada-manuscript, to be sure, 
reads here, but doubtless only by an error of the copyist, su-prapdna^, 
suprdyana (v. 12. 5 : p. su-prdyana), and anupraydna (vii. 73. 6 : p. anu- 
-praydna). Some other words, which would otherwise have to be noted 
under this rule, are made the subject of special precepts later in the 
section (rules 82-85). 

The general rule of the Rik Pr., as already noticed, includes this of 
ours as well as the preceding. By the Vaj. Pr. and Taitt. Pr., the cases 
are stated in detail. 

77. The n of ahan is changed after a former member of a 
compound ending in a. 

506 W. D. Whitney, [iii. 77- 

Pa-nini's rule, viii. 4. 7, is precisely coincident with this, and the illus- 
trative citations of its scholiasts are in good part those which our com- 
mentator gives us, and which are all strange to the Atharvan : they are, 
as examples of the rule, prdknah, purvdhnah, and, as counter-examples, 
nirahnah, paryahnah, and duraknah. Our text has only the two exam- 
ples apardhnah (is. 6. 46 : p. apara-ahnah) and sahasrdhnyam (e. a. x. 
8. 18: p. sahasra-ahnyam), and furnishes no counter-examples at all, so 
that the rule evidently finds its justification in the observed phenomena 
of the general language, and not in those of the Atharva-Veda. 

faWrfiWMHlirHli^hMW ll oc II 

78. Also is liable to be changed the n of a case- ending, that 
of an dgama, and the final n of a pratipadika. 

This rule is the exact counterpart in form of Panini's rule viii. 4, 11, 
and the technical terms which it contains are undoubtedly identical in 
meaning with those there given : dgama is an augment or inorganic 
insertion ;* prdtipadika is a theme of regular derivation, ending in a 
fart or iaddhita suffix. The commentator's instances are as follows ; 
for the case-ending ina, svargena lohena. (not in AV.), vardhena prlhivi 
samviddnd (xii. 1. 48) ; for the dgama n, ati durgdni vipvd (vii. 63. 1 : 
p. duh-gdni) ; for the final of a theme, ndi 'nam ghnanti parydyinah (vi. 
76. 4 : p. pari-dyinah). 

The other treatises have nothing corresponding to this rule, which is, 
indeed, an unnecessary one, as a Pratigakhya usually takes the words 
of its text in their ^>arfo-form, without going farther back to enquire 
how they came to assume that form. Such a word as vardhena, where 
pada and sanhild read alike, is not regarded by the others as calling 
for any notice : those analogous with the other instances cited might 
have been ranked as falling under the preceding rule. Parydyinah is, 
it may be remarked, the only word of its class which the text presents, 
excepting pravdhinafy, in book xx (xx. 127. 2), and durgdni also stands 
alone in its class ; nor have I noted a single instance of such a compound 
form as would be svah-gena or duh-gena, where the alteration of the 
case-ending in the second word would be mads in the reduction of pada 
to tankitd. 

SqWlJFfhTRFT*^ m ll ©* \\ 

79. Also that of a root after a preposition, even in a separate 

That is to say, even when the words are not connected together as 
parts of a, compound. The commentator gives us a single instance 
where the cause of conversion stands ndndpade, and two in which it 
stands p&rvapade; they are apah pra nayati (ix. 6. 4), yd eva yajila dpah 
praniyante (ix. 6. 5 : p. pra-ntyante), and jtvdm rtebhyah parmtyamdndm 

* The Bohtlingk-Roth lexicon, upon th« authority of this rule, erroneously makes 
it signify ' suffix.' 

iii. 60.] Atharva- Veda Prdtiodhhya. 507 

(xviii. 3. 3 : p. pari-ntyamdndm). Instances of a somewhat different 
character, where the preposition lingualizes the initial n of the root in 
sanhita and not in pada, although in the latter text no avagraha sepa- 
rates the two words, have been already given above, under rule 76. 
The text presents us pranutta in like combinations also after vdibddka 
(iii. 6. 7) and sdyaka (ix. 2. 12) ; and we have farther, in pada, su-pra- 
niti (e. g. v. 11. 5), and pardnaydt* (xviii. 4. 50). The initial n of a 
root is almost always cerebraJized by the preceding preposition in the 
Atharvan, even when (as at ii. 7. 1, ix. 2. 17, x. 4. 26, etc.) the augment 
intervenes : the only exceptions are the combinations of nabh with pra 
(vii. 1 8. 1, 2), and of nart with pari (e. g. iv. 38. 3) and pra (e. g. viii. 6. 
11). It is unnecessary to detail, therefore, all the rather numerous in- 
stances of the change. 

The cases forming the subject of this rule are not classified together 
by the other Pratigakhyas. 

80. Also that of ena, after pra and para. 

The commentator cites nearly all the examples which the text affords : 
they are prdi ''nan chrnihi (x. 3. 2), prdi 'nan vrkshasya (iii. 6. 8 : the 
same verse has, in its first pada, prdi ''nan nude), and pardi 'ndn devah 
(viii. 3. 16). As counter-example, to show that it is only after the two 
prepositions mentioned in the rule that ena exhibits the lingualization 
of its nasal, is farther cited pary endn prdnah (ix. 2. 5). 

Here follows a lacuna, apparently of considerable extent. Where 
this rule should be repeated, before the one next succeeding, we read 
prapardbhydrntrnaydmasi navatec ca ; and the sequel shows that navatec 
ca is a rule by itself, while what precedes is the mutilated remnant of 
another. This, aided by the implication in navatec ca, and by the text, 
which offers us the passage to which the rule was evidently intended to 
refer, we are enabled to restore with tolerable certainty to its true form, 
as punar nay&masi: what has become lost in the interval, we can only 
conjecture. Perhaps the treatise next took note of another case which 
the text affords of the lingualization of the nasal of ena, viz. djabhdrdi 
'ndm (v. 31. 10). Possibly there followed also a mention of the pas- 
sage asrjan nir enasah (ii. 10. 8) ; but this is very questionable, as the 
reading itself is doubtful.f But it is beyond question that a part of 
the omitted passage had reference to the not infrequent change of the 
» of nah in sanhita into n after a lingual near the end of the preceding 
word : this is much too common to have been passed over without notice, 
and the class of cases is too large and distinct to have been thrown 

* This is a case of entirely anomalous division and accentuation. We should 
expect npa-sampdrdnaydt, like anu-*amprdydhi (xi. 1. 36) etc. ; but the pada reads 
upa-sdm : pardnaydt, and all the sanhita MSS. agree with it as regards the accent ; 
it can hardly be otherwise than an error of the tradition : see below, under iv. 2. 

t It is given by M. W. H. and I. ; but P. (if I have not been careless enough to 
omit to note its reading) and E. have enasah, and the printed text has followed 
their authority. 

508 W. D. Whitney, [iii. 80- 

under any of the other rules of the section. The statement might be 
put into the form of two rules, as follows : 

' The n of nah is lingualized after pra and pari. As also, after Affr, 
urushya, grheshu, and pkska! 

Under the first rule, the cases would be i. 2. 2 ; 7. 5. ii. 4. 2, 6 ; 7. 3, 
4. iii. 16.3; 20.2.3. iv. 10. 6. v. 7. 3. vi. 37.2. vii. 2. 1 ; 5.5. xi.2.20. 
xii. 2. 13; 3. 55-60. xiv. 2. 67. xix. 34. 4, 5. Under the second, the 
passages are Afir na iirjam (ii. 29. 3), urushyA na urvjman (vi. 4. 3 : 
E. na), asam&tiih grheshu nah (vi. 79. 1 : E. nah), upahuto grheshu nah 
(vii. 60. 5 : E. nah), and fiksha no asmin (xviii. 3. 67). Where nah fol- 
lows grheshu, the edition reads both times nah, without conversion, 
although the manuscript authority for the lingual nasal is precisely 
what it is for the same after urushya, E. alone dissenting. At iv. 31. 2, 
a part of the manuscripts (P. M. W.) read senanir nah, but the edition 
properly follows the prevailing authority of the others (E. I. H.), and 
gives, with the Rig- Veda (x. 84. 2), nah. I have noted a single case 
where the Atharvan reads nah, while the Rik, in the parallel passage, 
has nah: it is vii. 97. 2 (RV. v. 42. 4). 

It is altogether probable that a whole leaf, or a whole page, of the 
original of our copy of the text and commentary (or possibly, of the 
original of its original), is lost here, with as many as three or four rules. 
Fortunately, it is in the midst of the rehearsal of cases of a certain 
kind, which rehearsal can be made complete without the aid of the 
treatise : so that the loss is not of essential consequence. 


81. Also that of naydmasi, s&expunah. 

The passage referred to is tarn tvA punar nay&masi (v. 14. 7) : all our 
sanhitA manuscripts agree in giving the lingual nasal. "Whether I have 
given the form of the rule correctly is not quite certain, a portion of it 
being lost altogether, as was pointed out in the last note. 

82. As also that of the root nu. 

The word to which alone the rule relates is punarnava{-p. punah-nava), 
for which the commentator cites three passages, viz. candramAf ea pu- 
narnavah (x. 7. 33), yA rohanti punarnavAh (viii. 7. 8), and punar A 'gAh 
punarnavah (viii. 1. 20). The authors of our treatise, then, must have 
derived nava, ' new,' from the verbal root »u, ' to praise,' instead of from 
the pronominal word nu, ' now.' 

^nrr: u ^ u 

83. Also in p&rydna. 

iii. 87.] Atharva- Veda Prdtigdkhya. 509 

The commentator instances but a single phrase in illustration of the 
rule, viz. pathibhih pury&ndih (e. g. xviii. 1. 54 : p. p&h-ydndih). There 
is small reason to be seen for singling out this word in" order to make it 
the subject of a special rule, and the same is true of those treated in 
the two following rules : they might all have been as well left to fall 
into the gana of rule 76. 

iUliy: II cd u 

84. Also that of durndman. 

The commentator extracts from tbe text three of the passages in 
which this word occurs, viz. durndmnth sarv&h (iv. 17. 5), durndmd 
tatra md grdhat (viii. 6. ]), and durndmd ca sundmd ca (viii. 6. 4). 
The pada writes duh-naman. 


85. Also after an r at the end of tbe former member of a 

The commentator's examples are devaydndh pitrydndh (vi. 117. 3: 
the pada form of the word is everywhere pitr-ydna), pitryanaih sam va 
a rokaydmi (xviii. 4. 1), and nrmand ndma (xvi. 3. 5 : p. nrmandh). 
The text furnishes but a single other word falling under the rule, and 
that in the nineteenth book ; viz. nrpdna (xix. 58. 4). 

It is worthy of note that Panini has a rule (viii. 4. 26) which pre- 
cisely corresponds with this, and that his scholiasts quote the same two 
words in illustration of it. 

=7 fsRlirr II *\\\ 

86. But not tbat of the root m%. 

The commentator brings forward all the illustrations of the rule 
which the text contains — they are pra minij janitrim (vi. 110. 3), pra 
minanti vratdni (xviii. 1. 5), and pra mindti sarkgirdh (xviii. 4. 60) — 
excepting one in the nineteenth book, pramindma vratdni (xix. 59. 2). 
The rule itself is to be understood, it may be presumed, as giving ex- 
ceptions to rule 79 above : yet the latter would seem to apply only to 
conversions of the nasal of a root itself, and not of the appended con- 
jugational syllable. 

The manuscript reads minanti instead of mindti, and the final repeti- 
tion of the rule before its successor is wanting. 

HFftir n so ii 

87. Nor that of bhdnu. 

This rule is entirely superfluous. Of the two cases cited under it by 
the commentator, the first, citrabhdno (iv. 25. 3), could fall under no 
rule for lingualizing the dental nasal excepting 76, and from that it 

510 W. D. Whitney, [iii. 87- 

would be excluded by absence from the gana to which alone the precept 
applies ; the other, pra hhanavah sisrale (xiii. 2. 46), cannot be forced 
under any role that has been laid down. 

The Rik Pr. (v. 22, r. 49, ccclxvi) and Vaj. Pr. '(iii. 91) also note 
bkanu as a word whose nasal is not subject to be changed to a lingual. 

^[^Hlri: ll ecu 

88. Nor that of the root hi after pari. 

The Atharvan text furnishes but one such case, which the commenta- 
tor quotes : it is parihinomi medhaya (viii. 4. 6). As counter-example, 
the commentator brings up pra hinomi d&ram (xii. 2. 8) ; but here, as 
well as in the other eases where the forms of the same verb exhibit a 
lingnalized nasal after pra, the petrfa-text also (by iv. 95) shows the same. 
A strict application of rule 79, then, to the nasal only of a root itself, 
would render this rule also unnecessary. 

The Rik Pr. (v. 22, r. 50, ccclxvi b ) has a corresponding precept. 

89. Nor a final n, nor one conjoined with a mute. 

The commentator's illustrative instances are pushan tava vrate (vii. 9. 
3), samkrandanah (v. 20. 9), and pdfo granihih (ix. 3. 2). To the first 
part of the rule would need to be made the exceptions noted at iv. 99, 
but that, by the operation of that precept, they are made to read in pada 
as in sanhita, and so are withdrawn from the ken of the Pratigakhya. 

The first part of this rule is included in the general precept for the 
conversion of n as given by the Rik Pr. (v. 20, r. 40 f ccclvii) ; which 
adds later (v. 22, r. 47, ccclxiv) that the n is not altered if combined 
with y or a mute. The Vaj. Pr. also divides the two parts of our pre- 
cept (iii. 88, 92), but specifies only a dental mute as preventing the 
conversion of the nasal by combination with it. The Taitt. Pr. (xiii. 
15) includes in one rule this of ours and also 91, 93, and 94 below. 

R$J: <4MW ii $o it 

90. Nor that of the root nag, when it ends in sh. 

This rule is precisely the same with one of Panini (viii. 4. 36), and 
it belongs rather to the general grammar than to a Prati§akhya of the 
Atharvan, since our text does not furnish a single case to which it should 
apply. The examples which the commentator gives are in part those 
which are found also in Panini's scholia : they are pranashtah, pari- 
nashtah, nirnashtah, and durnashtah. A counter-example, durnapam 
cid arvak (v. 11. 6), he is able to bring up from the Atharvan. 

F^yffq- ^%: nhii 

91. Nor that of the root han, when its vowel has been omitted. 

iii. 94.] Aiharva- Veda Prdtigdkhya. 511 

The commentator offers the two following examples : vrtraghnah sto- 
mdh (iv. 24. 1), and indrena vrtraghnd medi (iii. 6. 2) ; and the counter- 
example rakshohanam, v&jinam (viii. 3. 1), which has already figured as 
example under rule 76 above. Unless sparpayukta, in rule 89, meant 
only ' combined with a following mute,' which is very unlikely, the 
present precept is superfluous, as merely specifying a case already suffi- 
ciently provided for elsewhere. 

Panini (viii. 4. 22) looks at this matter from the opposite point of 
view, and teaches that the n of han is lingualized whenever it is pre- 
ceded by a. The Taitt. Pr. (xiii. 15) mentions the case along with 
others, in a comprehensive rule. 

92. Nor that of the root Jcshubh etc. 

This, again, is coincident with one of Panini's rules (viii. 4. 39), and, 
so far as tcshubh is itself concerned, is out of place in the Praticakhya 
of the Atharva-Veda : for not only does this Veda contain no deriva- 
tives from the root to which it should apply (we find only vicukshubhe 
[vii. 57. 1] and kshobhana [xix. 13. 2]), but also, if such forms as kshu- 
bhnali (which the commentator gives as an example under the rule), kshu- 
bhnitah, etc., did occur, their reading would be sufficiently determined 
by rule 89, above. The commentator fills up the gana with parinrt- 
yantyor iva (x. 7. 43), madhund prapindh (xii. 3. 41), and pari nrtyanti 
kepinlh (xii. 5. 48) : to these are to be added other forms of the root 
nart after pari and pra, and forms of nabh after pra — as already noted, 
under rule 79. 

*m% m&> » ^ ii 

93. Nor 'when there is intervention of g, s, or I. 

The instances cited in the commentary are kah prpnim dhenum (vii. 
104. 1), garbhe antar adrpyamanah (x. 8. 13), and savitd prasavdndm 
(y. 24. 1). For the intervention of /, no case is brought up, nor do the 
Eik Pr. and Taitt. Pr. make any account of this semivowel as ever com- 
ing in to hinder the conversion of the nasal. The Vftj. Pr., however, 
does so, and cites an instance, nirjarjalpena (in which, however, there is 
intervention also of a palatal). The latter treatise (iii. 94) and the Taitt. 
Pr. (xiii. 15) throw together into one this rule of ours and the one next 
following : the Rik Pr. gives a separate precept answering to each (R. 
Pr. v. 21, r. 44, 42, ccclxi, ccclix). 

^H^fel il \% ii 

94. Or of palatal, lingual, or dental mutes. 

The commentator gives the examples upe 'Ao 'paparcand 'smin gosh- 
tha upaprnca nah («ix. 4. 23), reshaydi 'ndn (xi. 1. 20 : this, however, is 
no example under the rule), yathd 'so mitravardhanah (iv. 8. 6), and tam 
vartanih (vii. 21.1). 

vol. vii. 65 

512 W. D. Whitney, [Hi. 94- 

The corresponding rules of the other Praticakhyas have been referred 
to above (under rule 93). 

The physical explanation of the effect of the sounds mentioned in 
these two rules to prevent the ' lingual ization of the nasal is obvious : 
they are all of them such as cause the tongue to change its position. 
When this organ is once bent back in the mouth to the position in 
which the lingual sibilant, semivowel, and vowels are uttered, it tends 
to remain there, and produce the next following nasal at that point, 
instead of at the point of dental utterance ; and it does so, unless thrown 
out of adjustment, as it were, by the occurrence of a letter which calls 
it into action in another quarter. 

95. Or of a word, unless it be d. 

As an example of prevention of the lingualization of n by the inter- 
position of a word or words, even though composed only of such sounds 
as would not in the same word produce such an effect, the commentator 
gives us pari 'me gam aneshata (vi. 28. 2). As an illustration of the 
specified exception, that A may so intervene, and the conversion yet 
take place, he gives paryanaddham (xiv. 2. 12: p. pari-anaddham), 
which is, I believe, the only case of the kind afforded by the text. 

The construction of the other rules relating to this general subject in 
the remaining Praticakhyas is not such as to require them to make the 
restriction here taught. Panini, however, takes due note of it (viii. 4. 
38), but omits to except the preposition d, so that (unless he makes the 
exception by some other rule which I have not observed) he would read 

rjf^OT llKll 

96. Note tuvishtamah. 

This word occurs only once in the Atharvan, in a passage cited by 
the commentator, as follows : indrah path tuvishtamah (vi. 33. 3). The 
pada-text reads tuvi-lamah, so that there takes place, as the comment- 
ator says, an inorganic insertion of s (sakar&gamah). The same word 
is found more than once in the Rig- Veda, but is written by the pada- 
text tuvih-tama, so that there is nothing irregular in the sanhita form, 
and it requires and receives no notice from the Eik Praticakhya. It is 
a legitimate matter for surprise to find the rule thrust in in this place, 
in connection with a subject to which it stands in no relation whatever : 
we should expect to meet it in the second chapter, along with rules 25 
and 26 of that chapter, or after rule 30, or elsewhere. Its intrusion 
here, and the indefiniteness of its form, cannot but suggest the suspi- 
cion of its being an interpolation, made for the purpose of supplying 
an observed deficiency in the treatise. 

The commentator, after citing the passage containing the word, goes 
on to say : sanhil&y&m ity eva : tuvitama iti tuviiamah : 1 anadhik&re 

1 The MS. writes, in every case, tuvittama in the commentary, but doubtless 
only by a copyist's error. 

iv.] Atkarva- Veda Prdliqdkhya. 513 

s&lranirdepah : s&trarthas tardyagap ca carcdparih&rayor vacanam m& 
bhud Hi ; ' this is the form only in sanhitd ; the lemma reads tuvitama 
ili tnvi-tamah : in the absence of an explanatory heading, the rule sim- 
ply points out the form ; and the significance of the rule is . . . (?) : in 
cared and parihdra [repetitions of krarna] the reading is not to be so 
made : that is the meaning.' 

The signature of the chapter is as follows: 105: iti trtiyo' dhydyah. 
If its enumeration is to be trusted, our two lacunce (unless some parts of 
the introduction and postscript of section iii are to be accounted as rules) 
have cost us nine rules. 


Contents: — Section L 1-2, combination of prepositions in pack-text with follow- 
ing verb; 8-6, exceptions; 7, separation by avagraha of such combinations; 
8-12, do. of the constituents of compound words; 13-80, do. of suffixes of de- 
rivation from primitive words ; 81-34, do. of case-endings from themes; 35-40, 
do. of other suffixes and constituents of compounds; 41, do. of iva from the pre- 
ceding word ; 42-46, do. of the constituents of words doubly compounded. 

Section II. 47-48, absence of avagraha before suffix mant etc. in certain 
cases ; 49-50, absence of division by avagraha of certain copulative compounds ; 
51-54, do. of other compounds; 55, do. of vrddhi derivatives from compound 
words; 56, do. of compounds with the negative prefix; 57-72, do. of other com- 
pounds and derivatives. 

Section III. 73-77, general rules for restoration in ^Wa-text of original or 
normal forms of words; 78, their application in krama-text also; 79-93, rules 
for restoration in certain cases and classes of cases; 94-100, exceptions. 

Section IV. 101-109, necessity of the krama-text and of its study; 110-126, 
rules for construction of krama-text. 

Our three preceding chapters have covered the whole ground which 
a comparison of the other treatises shows it to have been the bounden 
duty of a Praticakhya to occupy, and in this final chapter are brought 
up matters which might have been left unhandled without detriment to 
the character of the work as a complete and comprehensive phonetic 
manual for the school to which it belonged. Its first three sections, 
namely, teach the construction of the pada-text: whereas we have 
hitherto assumed this text as established, and have been taught how to 
construct the sanhitd upon its basis, we now look upon the body of tra- 
ditional scripture from just the opposite point of view, and, assuming 
the sanhitd, receive directions for forming the pada from it. No other 
of the kindred treatises thus includes in its plan the construction of the 
pada-text ; the Vaj. Pr. is the only one which takes up the matter at 
all : devoting, indeed, the whole of one of its chapters, the fifth, to an 
exposition of the rules determining the use or omission of the avagraha, 
or pause of separation between the two parts of a compound word, 

514 W. D. Whitney, [iv. 1- 

which is the subject of our first two sections, but leaving untouched the 
subject of our third section, or the restoration of words to their normal 
form in pada, which is not less indispensable than the other to the 
formation of the text. As regards the fourth section and its theme, 
the krama-text, their correspondences and relations will be set forth 
when we arrive at the place. 

As was the case with the third section of the foregoing chapter, our 
commentary offers us here a long introduction, containing about twenty 
plokas, to the subject of the chapter. It is prefaced with the following 
words : samds&vagrahavigrahAn pade yatho 'vdea chandasi pakat&yanah : 
tathd va&shy&mi catushtayam padam n&m&khy&topasarganip&tAn&m ;* 
' as Qakatayana has set forth for the Veda the combination, division, 
and disjunction of words in pada, so will I set forth the quadruple word 
— noun, verb, preposition, and particle.' Next follows a definition of 
each of these four classes of words, and then an exposition of the rules 
according to which they are to be regarded as compounded with one 
another, as they occur in the connection of continuous text, illustrations 
of the principles stated being drawn from the Atharvan text. Finally is 
given a list of the twenty upasargas or prepositions, and a designation of 
their accentuation, which is also followed by a list of the upasargavrttird, 
or words which are treated as if they were prepositions — such as acha, 
tirah, purah, hin, etc. The verses are in one or two places very cor- 
rupt, and their precise meaning is sometimes doubtful : hence, in order 
not to interrupt the progress of our treatise by too long an interlude, I 
defer to an additional note, at the end of the work, a presentation of 
their text and an attempt to translate them. 

3cmrf illi^HHlrM wm& » \ 11 

1. A preposition is compounded with a verb when the latter 
is accented. 

This is the well-known usage of all the Vedic jMtefa-texts, at least so 
far as they have been brought to general knowledge. With a true ap- 
preciation of the slightness of the bond which connects a verb with its 
prefix, the constructors of the disjoined text have ordinarily treated the 
two as independent words : unless, indeed, by the laws of accentuation 
of the sentence, the usually enclitic verb retains its accent, in which case 
the preceding preposition in turn loses its accent and becomes proclitic ; 
and the two are then written together as a compound. The commentator 
cites, as instances of verbs thus compounded, pari-y&nli (i. 1. 1), sam- 
-abharah (i. 9. 3), and sam-sravanti (i. 15. 3). He adds, ttpasargavrtti- 
bhif ca, ' the same is the case with the words which are to be treated 
as if prepositions,' and cites yam. : ar&te : purah-dkatse (v. "7. 2), and 
yam : ami' iti : purah-dadhire (v. 8. 5). That our treatise itself, in mak- 

* With the slight alteration of yatho 'vdea to yathd "ha, we should have here 
three equal successive pddas of a metrical verse : but the impossibility of forcing 
the last compound into any such metrical form, as a fourth pdda, renders it very 
doubtful whether this is anything more than a curious coincidence, and whether 
the words are not meant for simple prose. 

iv. 3.] Alharva- Veda Prdtitj&khya. 515 

ing no special reference to such ambiguous words, means to regard 
them as to all intents and purposes actual upasargas, is altogether prob- 
able. As counter-example, to show that the preposition is not com- 
pounded with the verb when the latter is unaccented, the commentary 
offers ydtu-dhd'ndn : vi : l&paya (i. 7. 2). 

The rule of the Vaj. Pr. (v. 16) is to the effect that avagraha is used 
between a verb and a preceding preposition, when the latter is unac- 
cented. This is in some degree a better form of statement, since the 
loss of accent in the preposition is rather the cause of its combination 
with the verb than the latter's acquisition of a tone. 

3R3Tt sM<lrlHlft ll \ li 

2. If there be more than one of them, they are compounded 
even with an unaccented verb. 

If more prepositions than one are compounded with an unaccented 
verb, only one of them, the one next the verb, is accented, the others 
becoming proclitic. In such a case, the constructors of the ^otrfa-text 
have very properly combined all with the verb, instead of simply putting 
the prepositions together, since it is not the relation of the former to 
the latter preposition that costs the former its accent, but rather their 
common relation to the verb : we have not a compound preposition, 
but a duplicate verbal compound. A later rule (rule 7, below) teaches 
us that in such a combination the first of the prepositions is separated 
by avagraha from the rest. 

The commentator gives as illustrations the compounds ut-dvasyati (ix. 
6.54), mm-d'cinushva : anu-sampraydhi (xi. 1. 36), and upa-sampardna- 
ydt* (xviii. 4. 50). He adds, as under the last rule, upasargavrttibhif 
ca, and illustrates with acfia-d'vaddmasi (e. g. vii. 38. 3) and abhi-Mn- 
krnota (xii. 3. 87). 

I do not find any corresponding precept in the Vaj. Pr., although the 
cases which it concerns are not such as would properly fall under the 
rule of that treatise already referred to. 

^H^hHU oMHld M'-jfitfsfa^t 4 

3. Disjoined from the verb, however, are such as are used 
without significance, or to set forth the object, or such as are 
otherwise connected — namely, in the cases ablii vi tanu etc. 

The Sanskrit language, much more than any other of the Indo-Euro- 
pean family, has, throughout its whole history, maintained the upasargas 
or prepositions in their original and proper character of adverbial pre- 
fixes, directing the action of the verb itself, and not belonging especially 
to the object of that action, or immediately governing its case-relation. 
Even in the Vedic dialect, where the preposition admits of being widely 

* The citation of this word here and under rule 1 Bhows that its true pada-ioim 
is as here given, and not as it has been noticed above (under iii. 19) that our pada- 
MS. actually reads. 

516 W. D. Whitney, [iv. 3. 

separated from its verb, it yet, in the great majority of cases, belongs 
distinctly to the verb, and not to any noun. But, while this is true as 
a general rule, there are yet cases, in the Vedic as well as in the more 
modern classical Sanskrit, where the preposition has detached itself 
from the verb, and is to be construed more intimately with the object 
of the action. Even here, it more often follows the noun, as auxiliary 
to its case-ending, and so occupies an intermediate position between 
adverb and preposition, something like the German adverbs of direction 
in such phrases as aus dem Walde heraus, in den Wald hinein, or like 
the German an in ieh blicke ihn an, as compared with the at in our 
nearly equivalent expression " I look at him." "Whenever it happens 
that a preposition thus attached to a noun cornea, in the construction 
of the sentence, to stand before an accented verb, or before another 
preposition which is connected with a verb, there arises an exception to 
the two foregoing rules, and such exceptions are made the subject of 
this and the two following rules: the first of them deals with such cases 
as grhd'n : tipa : pra : sidami (iii. 12.9), where the preposition is fol- 
lowed by another accented preposition and an unaccented verb, so that, 
by iv. 2, 7, we ought to have upa-prastdami ; the second, with such as 
yah : vlfvA : abhi : vi-pdfyali (vi. 34. 4), where the following preposition 
loses its accent before the verb, and the same rules would require abhi- 
-vipdfyati; the third, with passages where a single preposition comes 
before an accented verb, and so ought to lose its own accent and enter 
into combination with it, so that we should have, instead of ye : dsatah : 
pdri •.jajniri (x. 7. 25), pari-jajnire. The description which our rule gives 
of such uses of the prepositions is in near accordance with that of Panini, 
who also (i. 4. 83-98) calls by the name karmapravacantya ('concerned 
with the setting forth of the object of the action') such prepositions as 
are used otherwise than in immediate connection with a verb. He like- 
wise uses in the same connection (i. 4. 93) the term anarthaka, 'non- 
significant,' applying it, according to the scholiast's illustrations, to 
adhi and pari when used after an ablative adverb — thus, kutah pari, 
'whence' — where they are unessential to the completeness of the sense. 
Whether the same term would be applied to the same prepositions when 
following a case, as an ablative or locative, I do not know ; nor pre- 
cisely how it is to be understood in our rule : the commentator gives 
no explanation of it, nor of anyayukta, nor does he assign his illustra- 
tions to the several items of specification which his text furnishes him. 
The term anyayukta probably means ' belonging to another verb,' but 
such cases are quite rare in the text : an instance of the kind intended 
is perhaps nih stuv&ndsya pataya (i. 8. 3), where the preposition nih be- 
longs to the verb pataya, and not to the intervening participle, to which 
it would otherwise be attached, with loss of its own accent. Panini's 
karmapravacantya is comprehensive enough to include all the cases to 
which our rules apply. 

In filling up the ganas of our treatise, and giving all the cases of a 
preposition preceding a verb, but not placed in accentual and composi- 
tional relations with it, I shall include together all that would fall under 
this and the two following rules, since they evidently form a single class, 
and are only formally distinct from one another : 

iv. 3.] Atharva-Veda Prdiigdkhya. 517 

Of adhi, with an ablative case (or, rarely, an ablative adverb), ii. 7. 3. 
viii. 9. 4 (bis), ix. 5.6; 9.18. xiii. 1.42; 3.2. xix. 56. 1 ; with a locative 
case (or locative adverb), i. 3. 6 ; 32. 4. ii. 1. 5. viii. 9. 19. ix. 5. 4, 8 ; 
9.2; 10.7. x. 7. 1,12; 8.41. xi. 7. 8, 9, 14. xii. 3. 36. xiii. 1.37; 3. 
6, 18. xiv. 1.1; 2. 48. xviii. 4. 3 : anu with an accusative, ii. 34. 3. vi. 
97.3; 122.1. viii. 9. 12. x. 5. 25-35. xi. 8. 11, 19-22, 24, 27. xii. 2. 21. 
xiii. 2.40; 3. 1. xv. 6. 1-9 ; 9.1; 14.1-12. xviii. 4. 28. xix. 13.6; 44. 
10 : antar with a locative, i. 13. 3 : abhi, with an accusative, iii. 21. 5. 
iv. 1. 3. v. 19.4. vi. 34. 4. viii. 2. 4 ; 3.9. ix. 10. 6. xii. 1.29, 33; 3.8, 
12, 30, 52. xviii. 3. 2 ;* with an ablative, viii. 6. 22 ;f with a locative, 
xviii. 3. 40 :| a with a locative, xviii. 1. 59 ; upa with an accusative, iii. 
21. 1, 9. xix. 56. 3 ; pari with an ablative, ii. 34. 5. x. 7. 25. xii. 3. 53. 

But these prepositional or quasi-prepositional uses of the upasargas 
are of less importance to give in detail, because they are liable to occur 
in any part of the sentence, and their treatment as exceptions to the 
first rules of our chapter is a result of their accidental position in con- 
tact with a verb. There are other passages, considerably less numerous, 
where the prefixes, although evidently belonging to the verb, have an 
adverbial signification which is so far independent that they maintain 
their separate accent before an unaccented verb, or before another verbal 
prefix. The one most often thus treated is abhi, which is found before 
pra at iii. 1.2; 2. 5. iv. 8. 2 ; 32. 7. xviii. 3. 73, and before other upa- 
sargas at i. 1. 3. viii. 4. 21. ix. 9. 3 : upa occurs only before pra,§ at i. 
28. 1. iv. 31. 1. vi. 37. 1. xviii. 2'. 53 : &, also before pra alone, at iii. 4. 
5. vi. 35. 1. xviii. 4. 49 : apa, at iv. 31. 7 ; 32. 5 : pari, at iii. 2. 4 : ava, 
at vi. 65. 1 (p. ava : a'-yalA : ava etc.) : anu, at xiii. 4. 26 : sam (perhaps 
to be connected with the following instrumentals), at xviii. 2. 58 : and 
ud, at xii. 1. 39 — where, however, it is difficult not to believe the manu- 
script reading erroneous, and requiring amendment to ut-dnrcuh. 

In a single phrase, sam sam sravantu (i. 15. 1. ii. 26. 3. xix. 1.1), the 
preposition sam, being repeated for emphasis before the verb, is both 
times accented, and so is left uncombined. At vi. 63. 4, in a like repe- 
tition, there takes place a combination, with accentuation only of the 
former word — thus, sdm-sam : it, etc. ; and, at vii. 26. 3, the repetition 
of pra is treated in the same manner. 

A briefer, and, we should have supposed, a more acceptable manner 
of disposing of all the cases to which these three rules relate, would 
have been to prescribe that when an upasarga maintained its own accent 
before an accented verb or another upasarga, it did not undergo combi- 
nation with them. Such exceptions to the general rules for combination 
are treated by the Vaj. Pr. in rules 5 to 10 of its sixth chapter. 

The commentator's cited illustrations are iha : evd : abhi : vi : tanu (i. 
1.3), sam : sam : sravantu (e. g. i. 15. 1), su : pra : sadhaya (i. 24. 4), 

* In some of these passages — viz. iii. 21. 5. xii. 3. 12, SO — the special connection 
of abhi with the noun is but the faintest, and the cases are hardly to be reckoned as 
belonging in this class. 

f Here, too, the preposition belongs rather with the verb than with the noun. 

\ This is a combination unsupported by other passages, and hardly to be borne : 
for abhi is doubtless to be read adhi, with the Rig-Veda (x, 18. 8). 

§ Except in the anomalous combination upa : vandydh, at xviii. 4. 65. 

518 W. D. Whitney, [iv. 3- 

and &pa : pr& : ag&t (i. 28. 1). The third is a case having no proper 
analogy with the others, since su is not at all a verbal prefix ; I have 
taken no account of it in drawing out the above statement. 

y«funPiRwJwi^iR$j ii 8 ii 

4. In abhi vipagydmi etc., the former preposition is disjoined. 

The cases falling under this rule — those of a preposition retaining 
its accent and independence before another preposition which is itself 
made proclitic and combined with a following accented verb — have been 
detailed in the preceding note. The commentator quotes four of them, 
viz. ya'vat : te : abhi : vi-p&py&mi (xii. 1. 33), mdnasd : abhi : sam-vidtih 
(iii. 21. 5), ya'vat : s&' : abhi : vi-jdngahe (v. 19. 4), and ydh : vifvd : abhi : 
vi-pdfyati (vi. 34. 4). The examples, as in sundry cases elsewhere, are 
wanting in variety. 

ulHwQ^-Hii^i STllHH 

5. In yondv adhy dirayanta etc., the preposition is also dis- 
joined from the verb. 

This rule applies to such of the cases detailed in the note to rule 3, 
above, as show an accented and independent preposition immediately 
before an accented verb. The commentator instances samane : ydndu : 
ddhi : d'irayanta (ii. 1. 5), adhi : tasthuh (ix. 9. 2), ye : asatah : pari : 
jajniri (x. 7. 25), samttdra't : Adhi :jajiiishe (iv. 10. 2), and pari : bh&'- 
ma :ja'yase (xiii. 2. 3). The citation of the last passage seems to imply 
that the commentator regarded bh&ma as a verbal form, from the root 
bhA; but he can hardly, except in the forgetfulness of a moment, have 
been guilty of so gross a blunder. 

imflsj^fri MH^W fteTTrTllMl 

6. Agih and babhUva are disjoined, owing to the determination 
of the protracted vowel. 

That is to say, if I do not misapprehend the meaning of the rule, 
owing to the recognition of the final syllable of each word as a pro- 
tracted one. To what end the precept is given, unless the words re- 
ferred to (x. 2. 28 and xi. 3. 26, 27) have an irregular accent on the 
protracted syllable, I do not see. If accented, they would have a right, 
by the first rule of this chapter, to combination with the preceding 
preposition ; but, the present rule virtually says, they are seen to lose 
this right upon a recognition of the fact that the accented vowel is pro- 
tracted, and that its accent is therefore of an anomalous character. It 
has been already noticed (under i. 105) that a part of our manuscripts 
accent babMtva'Zn in the latter of the two passages referred to: not 
one gives an accent to agt2h, in either case of its occurrence. 

ydfmicw«£: ho ii 

7. The former preposition is separated by avagraha. 

iv. 10.] Atharva-Veda PrdiifdJchya, 519 

The avagraha, or pause of separation between the two parts of a 
compound word, is defined by two of the Pratigakhyas (as has been 
already remarked, in the note following iii. 74) as having the length of 
a mora. From here to the end of section II (rule 72), the treatise is 
occupied with rules for its employment or omission. And, in the first 
place, with however many prepositions a verbal form may be com- 
pounded, it is always the first of them that is separated from the rest 
of the compound by avagraha. The commentator gives us as examples 
the same series of words which we had under rule 2", above, only pre- 
fixing to them upa-avditi (ix. 6. 53). 

8. In ydtumdvat, also, the former constituent is separated by 

I interpret this rule according to the explanation of the commentator, 
who regards purvena as implied in it, by inference from the preceding 
rule. It comes in rather awkwardly here, as only prepositions have 
been contemplated, thus far in the chapter, as former constituents of 
compounds. The commentator adds an exposition of the matter, which 
is, however, too much corrupted and mutilated for me to restore and 
translate it : it reads matvarthe : vdyam mdvacchabdo manupo makdra- 
sya vakarah : dkdrdgamah. The word is apparently regarded as being 
ydtu-mant, with an added suffix vant, and the rule is designed to teach 
us that, instead of being divided yatumd-vat, as we might expect, it is 
to read ydtu-mdvat — as our pada manuscripts do in fact present it to us. 
The word occurs only at viii. 4. 23. 

FFfTFf ^ II \ II 

9. Separation by avagraha takes place also in a compound. 

The ca in the rule evidently continues the implication simply of ava- 
grahah from rule 7, and the connection of the text casts upon ydtumd- 
vat the suspicion of being an interpolation. A corresponding precept, 
samdse 'vagrahah, is by the Vaj. Pr. set at the head of its chapter upon 
the use of avagraha (v. 1). 

The commentator cites, as examples of separable compounds, upa- 
-hulah (e. g. i. 1. 4), bhkri-dhayasam (i. 2. 1), bh&ri-varpasam (i. 2. I), 
and bhuri-dhandh (vii. 60. 4). Whether such words as the first of these, 
or the participles of compounded verbs, should be regarded as falling 
under the designation samdsa, appears to me doubtful, as they present 
the closest analogies with the verbs from which they come. I have in- 
cluded them with verbal forms in filling up the ganas of rules 3 to 5. 

33sTH q^TTT II \o ll 

10. When a compound is farther compounded with an ap- 
pended member, the latter constituent is separated. 

vol. vii. 66 

520 W. D. Whitney, [iv. 10- 

The instance which the commentator selects for illustrating this rule 
is prajdpatisrshtah (x. 6. 19). The word prajd is itself divisible as a 
compound — thus, pra-ja (e. g. vii. 35. 3) ; upon farther adding pati, the 
former division is given up in favor of that between the old compound 
and its added member, and we have prajd-pati (e. g. ii. 34. 4) ; and a 
similar addition and removal of the pause of separation gives us prajd- 
patisrshtah: while we might have, did the words occur, the yet farther 
change prajdpatinrshtah-iva, or prajdpatisrshti-bhih, and prajdpathrshti- 
bhih-iya. In no compound is the separation by avagraha made at more 
than one point, and it is always the member last appended which is 
entitled to separation. 

The form of the corresponding rule in the Vaj. Pr. (v. *7) is vastly 
preferable to that of our own. That treatise says " in the case of a 
compound composed of several members, separation by avagraha is 
made of the member last added." This puts the matter upon its true 
basis, and accounts for the usage of the £>arfa-text-makers both where 
they separate the latter member, and where they separate the former 
member, from the rest of the compound. We shall see below (rule 12) 
that the treatment by our Fraticakhya of the separation of a former 
member is very obscure and imperfect. 

tjUMI ^ ii ^ n 

11. Also in suprdvyd. 

That is to say, in the word specified the last member is separated 
from the rest of the compound, and not the first, as would be more in 
accordance with the general analogies of the system of separation. 
The passage, the only one in the Atharvan where the word occurs, is 
quoted by the commentator : supra-anya' : yajamandya (iv. 30. 6). Su- 
prdvi is divided in the same manner by the pada-text of the Kig-Veda 
(e. g. i. 34. 4) : it is of obscure derivation and meaning, and whether 
the etymology of it which is implied in its analysis as made by the 
Hindu grammarians is correct, is at least very doubtful. But whether 
composed of su-pra-avi or su-pra-vi, we should expect the ^cwfa-text to 
write it su-pravi or su-pravi, and it is the recognition of its anomalous 
division which has caused it to be made the subject of a special rule. 
The commentator adds a verse about it, which, however, appears to deal 
rather with its accentuation than its division : avateh prapurvasya [su- 
prapurvasya?\ tdpabdah [yapabdah?] svarilah parah : suprdvi 'ti trtiya- 
ydh kshaiprah chandasi svaryate. 

The Atharvan reading suprdvyd is at any rate established by this 
rule, against suprdvye, which the Rik (x. 125. 2) offers, and which the 
connection also appears to require. Possibly the form implies an ex- 
planation of the word which seems suggested by the first line of the 
verse just quoted, as if from the root av, with the prefixes su and pra, 
and with the added desiderative suffix yd, ' with desire to show propi- 
tious favor.' Such an explanation, of course, would be futile, being 
sufficiently disproved by the accent alone. 

iv. 12.] Alharva-Veda Prdtigdkhya. 521 

12. If the appended member is indivisible, the former mem- 
ber is separated. 

This rule is very obscure, and I am far from feeling confident that my 
translation rightly expresses its meaning. The manuscript readings of 
its first word, in text and in comment, vary between anirigena, atihgena, 
and amigena. If the word really means 'immovable,' i. e. 'indivisible, 
inseparable,' we should expect rather aningyena (compare ingya in rule 
76, below) ; and possibly the latter may be the true reading : but as 
inga also is an acknowledged word, having the meaning ' movable,' I 
have not ventured to alter the form presented by the manuscript. As 
I have rendered it, the rule would appear to mean only that when to a 
word already compounded an inseparable appendix, for instance a suffix, 
was added, the division would remain as before — as in su-viratayai, 
pra-paddbhyam, and the like — but this is a perfectly simple case, and 
one which hardly calls for especial attention and determination. The 
commentator's exposition is as follows : aningeno , pajate : aningena vife- 
shalakshanena avikrshitena : dvayoh samfaye jdte purvena 'vagraho bha- 
vati ; 'when a word is farther compounded with an indivisible — that is 
to say, with a modificatory appendage which is not taken apart — and 
there arises a doubt between two, separation is made of the former.' 
My translation of the rule is founded upon my (somewhat questionable) 
interpretation of this paraphrase. The cited illustrations, however, do 
not at all support it; they are m-kshetriya : su-gdluyd (iv. 33. 2), saha- 
-suktavakah (vii. 97. 6), m-antardefdh (ix. 5. 37), and su-praj&h (iv. 11. 
3). The last three of these are plain cases of separation of the con- 
stituent last added from the rest of the compound : the first two are 
less unequivocal, since we should rather regard the suffix yd as added 
to sukshetra and sugdtu, and the more natural division as being sukshe- 
tri-yd, sugdtu-yd ;* but neither is yd an aningya suffix, as is shown by 
the next following word, vasu-yd, and the others detailed in rule 30, 
below. The commentator adds a verse of farther exposition, but this 
also throws no additional light upon the matter in hand : it is dve yalrd 
, vagraliaathane purvene Hi parena va : purvena 'vagrahas talra sukshetri- 
yd sdntardefdh suprajdp ca nidarfanam ; 'where there are two places 
for separation by avagraha, either of the former or of the latter mem- 
ber of a compound, separation is there to be made of the first mem- 
ber : instances are svkshetriyd, sdntardefdh, and suprajdh? I do not 
see how this statement can be accepted as a correct one ; for, of the 
compounds consisting of more than two members, the last is even more 
often separated from the first two than the first from the last two : the 
point of division being, except in a very few cases of which the treatise 
takes special note, determined by the history of the double or triple 
compound, upon the principle distinctly laid down in the Vaj. l'r., that 
the member last added is the one which must be separated. Thus, in 

* The Rik pada (i. 97. 2) actually reads sngdtu-yd, although it divides su-kshetriyd 
like our own text. 

522 W. D. Whitney, [iv. 12- 

the first four books of the text, we have, upon the one hand, rla-prajAta, 
madku-samdrfa, vi-AyAma, prati-abhicarana, mwhka-Abarha, brahma- 
•sariifita, nu-samnaia, svapna-abhikarana, A-uts&ryam, A-vyusham, Ayvh- 
-pratarana, fiva-abhimargana, aksha-parajaya, and afva-abhidhAni ; and, 
on the other hand, sarupa-krt, swriipam-karani, sapatna-kshayana, viskan- 
dha-d&shana, sapatna-han, abhim&ti-jit, abhifasti-pA, samkalpa-kulmala, 
agnikotra-hut, Ak&ii-pra, ApAra-eshin, vijesha-krt, abhibhfitiqjas, and 
abhimAti-sahas. And in the rare cases where three independent words 
are fused into a compound by a single process, the last one is separated 
from the other two : such are nagha-risha and iliha-Asa (xv. 6. 4). It 
will be difficult, I think, to find any interpretation for our rule which 
shall make it other than a bungling and inaccurate account of the phe- 
nomena with which it professes to deal. To connect aningena in con- 
struction with purvena, and translate ' the former member is separated 
from the rest when it is an indivisible word,' although it would satisfy 
well enough the requirements of the instances given by the comment- 
ator, would only throw us into new difficulties, for it would require us 
to read sa-rupakrt as well as sa-antardefa, su-parnasuvana (v. 4. 2) as 
well as su-kshetriyA, and the like.* 

FTfl^r ^Tll^ll 

13. Dhd is separated, when a taddhita-sxx&x. 

As instances of the separation of the secondary suffix dhA from the 
themes to which it is attached, the commentor cites catuh-dhA : retah (x. 
10. 29), ashta-dhA : yuktoh (xiii. 3. 19), nava-dhA : hitah (xiii. 4. 10), and 
dvAdafa-dhA (vi. 113.3): in the same manner is treated mitra-dhd, at 
ii. 6. 4, for which the Vajasaneyi-Sanhita, in the corresponding passage 
(xxvii. 5), has mitra-dheye. On the other hand, the text offers a single 
exception to the rule, vigvAdhA (vi. 85. 3), which neither the Pratica- 
khya nor its commentary notices : it is accented on the penult, while 
all the other compounds with dhA accent the suffix itself. 

The commentator adds; vyatyayasvafraddirghebhyo dhApraiyaye na 
avagraho bhavati; 'the suffix dhA does not suffer separation after . . . (?), 
sva, prat, and a long vowel.' The words svadhA and praddhA, into 
which the root dhA enters as last member, are here referred to, and 

* The best way, it seems to me, of saving the credit of our PraticSkliya as re- 
gards its treatment of the subject of double composition in these two rules, will be 
to regard rule 10 as equivalent to Vaj. Pr. v. 7 ; understanding upajdte as meaning 
simply 'added to,' whether by prefixion or suffixion, and parena as signifying 'the 
later (i. e. Uie last added) constituent:' and then farther, taking rule 12 to be added 
in limitation of 10, and to mean: 'when, however, the added constituent is incapa- 
ble of being separated, the division remains as before.' It might well enough be 
thought that, in such cases of prefixion as avira-han (from vira-han), or in such 
cases of suffixion as su-viratd, vddhu-ya, parivatmrbna, vi-irtsd, etc., the addition 
of another element virtually fused the prior compound into one word, and would 
be understood as annulling its division by avagraha, unless some direction was 
given to the contrary. This interpretation, however, would be contrary to the 
authority of the commentator, would require us to understand parena and p&rvena 
in a different from their usual sense, and would convict rule 11 of being an interpo- 
lation, made since the misinterpretation of rule 10. 

iv. 16.] Atharva-Veda Prdticdkkya. 523 

perhaps godha (iv. 3. 6) ; but to what the first item in the enumeration 
refers, I have not succeeded in discovering. 

The kindred suffix dd remains always attached to the theme to which 
it belongs. 

The Vaj. Pr. (v. 27) forbids the separation of dhd from a numeral, 
but allows it in other cases : it would read caturdhd, ashtadhd, etc., in 
pada. The usage of the Rik pada-text also does not entirely corres- 
pond with that of the Atharvan in regard to the same suffix: thus the 
former has bahudhd, while the latter separates bahu-dhd, 

5TT^T% II \& ii 

14. Also trd, when it ends in d. 

The commentator's examples are deva-trd : ca : krnuhi (v. 1 2. 2), and 
puru-trA : te : vanvat&m (vi. 126. 1); and his counter-examples, of tra 
inseparable, are yatra : devdh : amrtam (ii. 1. 5), and tatra : amrtasya : 
cakshanam (v. 4. 3). 

The Vaj. Pr. (v. 9) declares tr& separable, and, as our treatise might 
just as well have done, regards tra as excluded by the designation of 
the form of the separable suffix as trd. The usage of the Rik pada is 
also the same with that taught in our rule. Doubtless it is the charac- 
ter of the forms to which tra is attached, as being pronominal roots, 
that prevents its separation from them, rather than anything in the 
suffix itself. The ablative suffix tas is not separated, even when it fol- 
lows a word having an independent status in the language, as in abkitas. 

^H*M*(UI ll V< ll 

15. Also thd, when it is combined with a polysyllable. 

The commentator's examples are rtu-tha : vi : cakshate (ix. 10. 26), 
ndma-thd : sa : manyeta (xi. 8. V) ; and his counter-examples are adka : 
yatha, : nah (xviii. 3. 21), and tatha : tat : agne (v. 29. 2). Here, again, 
it is evidently not the length of the words to which the suffix is ap- 
pended, as monosyllables or polysyllables, that determines its separa- 
bility, but the character of the former as pronominal roots and of the 
latter as nominal themes. 

The Vaj. Pr. (v. 12) details the words after which th& is separable. 

16. Also tara and tama. 

The commentator cites, as instances of these suffixes with separable 
character, mddhoh : asmi : madhu-tarah : madughat : madhumat-tarah 
(i. 34. 4), ut-tamah : asi : dshadhindm (e. g. vi. 15. 1), aham : asmi : ya- 
cdh-tamah (vi. 39. 3), and nrnd'm : ca : bhdgavat-tamah (ii. 9. 2). He 
adds, according to his usual method of introducing counter-examples : 
taratamayoh iti Mm artkam, ' why does the rule say tara and lama P 
and gives the counter-examples afvatardsya (iv. 4. 8) and afvataryah 
(viii. 8. 22). The citation of these words in this manner, as if they were 

524 W. D. Whitney, [iv. 16- 

excluded by the very form of the rule itself from falling under its ope- 
ration, suggests as the true reading and interpretation of the rule tara- 
tamapoh, ' tara and tama when unaccented,' giving to the indicatory 
letter p the same force which it has as appended to the same suffixes by 
Panini (e. g. v. 3. 65, 57). My copy of the manuscript, indeed, gives 
me everywhere taratamayoh ; but, considering the small difference be- 
tween y and p when hastily formed, I cannot be confident that the 
Hindu scribe did not mean to write the latter letter. But, in the first 
place, I find it very hard to admit that our Praticakhya would thus use 
an indicatory letter as an essential and determinative part of one of its 
rules, without giving anywhere any explanation of its value. Other 
such appendages to a suffix, corresponding with those which Panini 
employs, are, it is true, elsewhere found in the treatise: thus we have 
(iv. 20) tdtil, where the I indicates that the syllable preceding the suffix 
has the accent ; also matu, vaiu, and vasu, whose appended u is intended 
to show that the suffix receives an augment (dgama) n in the strong 
cases, and takes the feminine ending i ; and the particles u and su are 
called, with Panini, un and sun, to distinguish them from the exclama- 
tion wand the case-ending su : but the appended letters are nowhere 
used as significant. Again, we should expect that the p, if used at all, 
would be applied to each suffix, and that the rule would read tarapta- 
mapoh (compare Pan. i. 1. 22), which is certainly not its form in the 
manuscript. Finally, the admission of the indicatory letter, with its 
Paninean signification, would not make the rule complete and accurate, 
as stating the usage of our pada-iexX with regard to the suffixes in ques- 
tion; for, on the one hand, we have ratham-tarh (e.g. viii. 10. 13), and 
the prepositional comparatives pra-tardm (e. g. v. 1. 4), vi-tardm (v. 12. 
4), and sam-tardm (vii. 16. 1) ; and, on the other hand, there are cases 
in which the unaccented endings are not separated from the themes to 
which they are attached, and one of these cases is even noted and 
specified by the commentator. The latter's discussion of the rule is 
fuller than usual, as he seems, for once, to feel the necessity of doing 
something to supply the deficiencies of his text; but his effort is only 
partially successful, and moreover, his language is so mutilated by the 
manuscript that I can make it out but imperfectly. He first asks why 
the separation of tara and tama is not made also in katard and katamd, 
yatard and yatamd (to which we may add tiara and dntara), and an- 
swers netduche [ndi Hdu stas f\ taraiamdv anydv etdv akdrddi pratya- 
ydu : angasya 'trd "dimdtram u fishyate lupyate param : svarddi pra- 
tyay&v etdu padatvam n&Hra fishyate; 'these are not tara and tama, 
but other suffixes, commencing with a [viz. atara, atama~\ : in the de- 
rivatives in question, the first portion of the theme remains, but the 
last is dropped [i. e. yatara=y + atara etc.] : these are two suffixes with 
initial vowel : capability of standing as separate padas is not taught of 
them.' This distinction of the suffixes as applied to pronominal roots 
from those applied to other themes is evidently artificial and false : the 
difference is that the roots themselves are not, like derivative themes, 
detachable from the suffixes appended to them — as we have seen to be 
the case under the two preceding rules. The commentator, changing 
his subject, then goes onto say gotamah pratishedho vaitavyah ; 'the 

iv. 18.] Aiharva-Veda Prdtigdkhya. 525 

word gotama must be noted as an exception to the rule ;' and he cites 
the passage in which it occurs: yd'u : gotamam : dvathah (iv. 29. 6) ; 
adding, by way of explanation, samjndydm rudhipabdo dhd tamd J tr& 
'nalipdyane : asamdnah samudradis tasm&n netyati golamah. I have 
not succeeded in restoring this verse so as to translate it, but so much 
of it as applies to golama may be pretty clearly understood : the word, 
as a proper name, is one of common currency, a samjna, or specific ap- 
pellation employed without reference to its etymology, and its suffix 
tama has no superlative signification ; hence there is a propriety in 
treating it otherwise than as an ordinary superlative. 

There yet remains one word in the Atharvan, viz. uttara, whose treat- 
ment by the pada-t&xt requires an exposition which the Praticakhya 
omits. While ullama is always divided — thus, ut-tama — the compara- 
tive is variously treated, being sometimes divided and sometimes left 
without avagraha. The Vaj. Pr. (v. 2) takes special note of this irregu- 
larity, declaring a word formed with lara or tama not separable when 
opposed in meaning to 'south;' that is, when meaning 'north' — this 
being, as we may suppose, another case of samjna. This principle is 
so far followed in our text that the word is never divided when it has 
the meaning ' north :' but it is also left without avagraha in quite a 
number of passages* where it has its own proper signification ' upper,' 
or the naturally and regularly derived ones 'superior,' 'remoter,' or 
' later.' I can only suppose that the true principle which should deter- 
mine the separation or non-separation was misunderstood and misapplied 
by the constructors of our pada-text. 

I TrTt II \o II 

17. Also mant. 

The illustrations chosen by the commentator from among the very 
numerous examples of this rule presented in the Atharvan text are 
madhu-mat (e. g. i. 34. 3) and go-mat (xviii. 3. 61). Exceptions are 
noted farther on, in rule 47. 

The Vaj. Pr. (v. 8) states the principle more broadly, including to- 
gether all possessive suffixes ; among them, most of those which form 
the subject of our next rule. 

18. Also a taddhita beginning with v. 

The commentator cites as examples atri-vat : vah : krima.yah : hanmi : 
kanva-vat : jamadagni-vat (ii. 32. 3), rta-vdnam (vi. 36. 1), satya-vdnam 
(iv. 29. 1, 2), apva-v&n (vi. 68. 3), aftji-vam (viii. 6. 9), kega-vdh (viii. 6. 
23),f mdrta-vatsam (viii. 6. 26), d-vayam (viii. 6. 26), and vddhii-yam 

* They are as follows: iii.5.5. iv.22.6. v. 28. 10. vi. 16.4; 118.3; 134.2. viii. 
2.15. xi. 8. 18. xii. 1.54. 

f The MS. next gives rdnvatah, which I have not succeeded in identifying with 
any word in the Atharvan text. 

526 W. D. Whitney, [iv. 18- 

(xiv. 1. 29). He must have been nodding when he added the last three 
words, of which the third has no suffix beginning with v, and the two 
others are not formed by suffixes, but by composition. He goes on to 
say vijno pratishedho vaklavyah, but what is to be understood in, or 
from, vijno, I do not see : the cases of irregular absence of separation 
by avagraha which he adduces are druvayah : vi-baddhah (v. 20. 2), 
ubhay&vinam (v. 25. 9), and the three words, of kindred character with 
the latter, amaydvt, mekhaldvt, medhdvt, no one of which is found in 
the Atharvan, although we once have an accusative, of the last of them, 
medh&vinam (vi. 108. 4). 

5Tf% °fNlMR n \$ ii 

19. Also gas, with distributive meaning. 

The instances given in the commentary are paru-fah : kalpaya : enam 
(ix. 5. 4), and dhdma-fah : sth&tre : rejanle : vi-krtdni : rtipa-fah (ix. 9. 
16) ; the counter-examples, showing that fas is only separable when dis- 
tributive, are ankufah (vi. 82. 8) and kindfdh (e. g. iii. 1 1. 5) — cases 
which it was very unnecessary to cite, since their suffix, if they have 
one, is fa, not fas. 

The Vaj. Pr. (v. 9) marks fas, along with other secondary suffixes, as 

HlIrlH II J^o ii 

20. Also tdti. 

Panini (e. g. iv. 4. 142) calls the suffix tdti by the same name, tdtil, 
the appended I signifying, as has been already noticed, that the syllable 
preceding the suffix is accented. The Vaj. Pr., in its corresponding 
rule (v. 9), calls it simply tdti. 

The commentator illustrates with the two passages mahydi : arishta- 
-tdtaye (iii. 5. 5), and havishmanlam : md : vardhaya : jyeshtha-tdtaye (vi. 
39. ]'). 

The related suffix td is never separated from the theme to which it is 
appended : tva forms the subject of rule 26, below. 

3>WTf 5lf>T II ^1 II 

21. Also dyu, after ubhaya. 

The commentator cites the examples ubhaya-dyuh : abhi-eti (i. 25. 4) 
and ubhaya-dyu h : upa : haranli (viii. 10. 21), and the counter-example 
yah : anyedyuh (i. 25. 4) ; which are the only cases (except a repetition 
of the phrase in i. 25. 4 at vii. 116. 2) presented in our text of deriva- 
tives — or, more properly, of compounds — formed with dyu or dyus. 
The name dyubh or dyubhi, which our treatise gives to the latter, is 
a strange one, and not supported by anything in Panini : indeed, the 
latter seems never to use bh as an indicatory letter ; the general grammar 
forms ubhayadyus with the suffix dyus, and anyedyus and its numerous 
kindred (see the Bohtlingk-Eoth lexicon, under dyus, or Pan. v. 3. 22) 

iv. 25.] Atharva- Veda Prdtifdkhya. 527 

with the suffix edyusac : one is tempted to conjecture that the authors 
of our system may have regarded dyus in these compounds as a con- 
traction for the instrumental plural dyubhis. The reading of the word 
is well assured, not only by the text and comment, but also by a cited 
verse with which the commentator closes his exposition : it reads anyasyd 
dyubhi tv etvarh sydd anyo vd dyahbhir ishyate : lopa edyubhi cd 'ntyasya 
siddho vd ''yam, punar dyubhi. This evidently has to do with the forma- 
tion of anyedyus, accounting for the e which precedes the suffix : but I 
can offer no entirely satisfactory restoration of the text. 

*TT% ^T ll ^ II 

22. Also mdira. 

This is most palpably a rule which has its ground in the observed 
phenomena of the general language, and not in those of the Atharva- 
Veda ; for although, in the later language, mdtra came to be used in 
such a mode and sense as to give some ground for its treatment as a suf- 
fix, it is in the Atharvan nothing but a noun, and even enters into com- 
position only with ati, forming the adjective atimdtra, ' above measure, 
excessive.' The commentator cites, in illustration of the rule, the two 
passages in which this compound occurs, namely ati-mdtram : avardhanta 
(v. 19. 1) and ye : dtmanam : ati-mdtram (viii. 6. 13). 

n^KhIIh ii ^ ii 

23. Also ddnim, after vigva. 

The commentator cites, as example and counter-example, vifva-ddnim 
(e. g. vii. 73. 11) and taddntm (e. g. x. 8. 39), the only two Atharvan 
words which are formed with this suffix. Here, again, is an instance of 
a suffix remaining attached in pada to a pronominal root, while it is 
separated from a nominal theme (compare under rules 14-16). 

*# **H)ltlr(H^3ll 

24. Also maya, excepting after s, 

A single example of the separation of the suffix maya is cited in the 
commentary, viz. faka-mayam : dhkmam (ix. 10. 25) ; and also a single 
example of its non-separation, when following a theme ending in *, viz. 
anah : manasmayam (xiv. 1. 12). 

% okJsMlrJJlVUl 

25. Also Tea, after a consonant. 

Words in which the suffix Jca is appended to a consonant are not at 
all frequent in the Atharvan : the commentator brings up two cases, viz. 
avat-kam (ii. 3. 1) and ejat-kd?} (v. 23. 7), and I have noted but two 
others, viz. manalp-kam (vi. 18. 3) and aniyah-kam (x. 8. 25). As coun- 
ter-examples, where the same suffix, following a vowel, is left attached 

vol. vn. 67 

528 W. D. Whitney, [ir. 25- 

in pada to its theme, he gives us tundikah (viii. 6. 5) and fipavitnukdh 
(v. 23. 7). Such formations occur by scores in our text.* 

26. Also tva, in a word accented on the final syllable. 

The instances cited by the commentator of the occurrence of this com- 
mon suffix are mahi-tvd 1 : kdsmdi (iv. 2. 4), amrta-tvam : dnapuh (ix. 10. 
1), and ficcayah : cuci-tvdm (xii. 3. 28). As counter-example, to show that 
the tva is separated only when it receives the accent, is cited aditih : 
janitvam (vii. 6. 1 : the Rik pada, in the corresponding passage [i. 89. 
10], has jdni-tvam) ; and this is the only word of the kind which the 
text contains, for at ii. 28. 3, where the edition reads jdnitvdh, all the 
manuscripts have janltrdh. The commentator adds a verse respect- 
ing his counter-example, as follows : janitvam aditeh param ne 'ngyate 
krjjanaparatah : itvo vd sydj janas tena padatvam nd Hra fishyate: this 
informs us that janitva following aditi is not to be treated as separable, 
as being formed from the root^'aw by the suffix itva, which is not taught 
to be an ending capable of constituting an independent pada. He then 
proceeds to ask the question iha kasmdt samdso na bhavati : ekaja : tvam : 
mahi : tvam; ' why is there no combination in the passages ekaja tvam 
(iv. 31.3: p. eka-ja : tvdm) and mahi tvam (not found in AV.) ; and he 
answers by a verse, from the authority, doubtless, which had suggested 
to him the query : ekaja tvam mahi tvam ca tad ubhayam samasyate [tti 
, bhayam na samasyate f\: dmantritam layoh piirvam yushmada[s f\ tvam 
param padam; 'in neither of the phrases ekaja tvam and mahi tvam is 
there a combination ; the former word is there a vocative, and the latter 
is tvam from y^hm^d^ It is clearly a work of supererogation on the 
part of the commentator to explain such self-evident matters. But he 
is not content even with this ; he continues " why is not agrepitva sepa- 
rated ?" — that is to- say, I presume, why is it not divided agrepi-tva, in- 
stead of agre-pilva ? — and he again cites a verse : taddhite 'vagrahah 
fishtah padatvam nd 'tra fishyate : pibates tarn nibodhata itvam chdnda- 
sam ishyate ; 'separability is taught only of taddhiias; division into 
separate padas is not taught of this case ; note that the word comes 
from the root pd, 'to drink,' with the Vedic suffix itva.'' But, even 
were this exposition in itself worth giving, the word to which it relates 
does not occur in the Atharvan, nor — judging from its non-appearance 
in the Bohtlingk-Roth lexicon — in any other of the known Vedic texts. 
We could wish that our commentator had reserved bis strength for 
points where its exertion would have done us some service. 

The suffix tra, which appears in the single word sanskrtatrd to fill 
the office of tva, is in our pada-text (iv. 21.4) left inseparable, while the 
Rik, in the corresponding passage (vi. 28. 4) r interposes the avagraha 
before it. Taya, in catushtaya (x. 2. 3), is not separated from its theme. 
Of td, notice has been taken under rule 20. 

* For example, in the first eight books, from which alone I have excerpted them, 
at i. 2. 2 ; 3. 6-9, 9 ; 11. 5 ; 25. 4; 84. 2. ii. 8. 1 ; 24. 1, 2 ; 26. 5. iii. 11.2; 23. 4. T. 
18. 8. vi. 29. 3 ; 43. 1 ; 83. 8 ; 121. 4. va 56. 6. viii. 6. 19, 21 (6»«). 

iv. 29.] Atharva-Veda Pr&tv$kh,ya. 529 

27. Krtva is combined or not, according to its appearance as 
an independent word or otherwise. 

Krtvas occurs but three times in the Atharvan : twice it is treated as 
an independent word, both it and the preceding numeral having an ac- 
cent; once it is combined with the numeral, the latter losing its separate 
accent, and it is then separated by avagraha from it. The commentator 
cites all the passages, as follows : calur : n&moh : ashta-kr'tvah : bhavA'- 
ya : dapa : kftvah (xi. 2. 9), and trih : saptA : kftvah (xii. 2. 29). After 
this, having not yet recovered from the impulse which made him so 
fertile of exposition under the foregoing rule, he continues: nanv evam: 
katham, : vyavasthitena vikalpena vapabdena pratipaditatvdt ; ' now then, 
how is it ? since a diversity of usage is taught respecting the word, by 
the use of the term vA? and he makes reply in a lengthy citation from 
his metrical authority, which wanders at the end far beyond the limits of 
the subject in hand : karoter dapasaptAbhyAm, tvapabdah krd vidhiyate : 
sankhyAyA anuddltAyA ashtapabdAt samasyate : udAttAd dapampte , ty 
evam purvena vigrahah : dhAparyantas taddhitA ye te 'shte vA 'vagraho 
bhavet : ato 'nyena padalve 'pi yuvatyAdishu taddhitam : dhAtrAthalAtili- 
pasidAnimtaratamomatup : vamAtradyubhi ketvApi mayakrlveshv avagra- 
hah. In the last verse we have an enumeration of all the suffixes thus 
far treated of as separable. 

a ca 

sTTrTraTT^J ^ II \* It 

28. Jdtiya etc. are also separable. 

As instances of the use of j&tiya, the commentator gives v& patu-j&- 
tiya, mrdu-jAtiya, pandila-j&tiya, and pobhanA-jAtiya: none of these 
words, however, nor any other compounds with the same final member, 
are to be met with in the At&arvan text. The general grammar also 
treats jAliya as a suffix, and Pacini's scholiasts (under v. 3. 09) give, as 
an example of its use, the first of the instances of our commentator. 
The latter farther cites, to fill out the gana of the rule, some of the 
compounds of dheya, viz. bhAga-dheyam (e.g. vi. 111. 1), rtipa-dheyam 
(r&pa-dfteyani: ii. 26. 1), and nAma-dheyam (vii. 109. 6). "What other 
frequently occurring final members of compounds it may have pleased 
the authors of our treatise to regard as suffixes, and to include in this 
gana, I do not know : I have noted no actual suffixes as needing to be 
comprehended in it. ° 

«JI<lRl$!Uf ^Ir^^HH ^lPS^jg n ^ II 

29. Also a suffix commencing witb y and preceded by a vowel 
in a desiderative form; namely, in participles, denominatives! 
and desiderative adjectives. 

Not one of the technical terms used in this rule is known to me to 
occur elsewhere than in the grammatical language of our treatise. One 

530 W. B. Whitney, [iv.29- 

of them, ichd, we have had at another place (iii. 18), and in such a 
connection as to show that it is employed to designate the whole class 
of words to which this rule applies ; the others, then, are specifications 
under it, or a classified statement of the cases which it includes. The 
same thing is indicated by the commentator, who paraphrases as follows : 
ydddv ichdydm svardt avagraho bhavati karma" etc. Karmandma, 
then, I have without much hesitation rendered by 'participle;' and 
tanmdnin seems to me to mean ' implying the making or doing of that 
which the theme indicates,' and so to be applicable to such words as 
gatriiyanti, agh&yanii, where the signification is not simply desidera- 
tive : but of this I do not feel altogether confident, and I have at one 
time sought in the word a designation of the middle participles having 
the termination m&na ; prepsu I think must belong to such derivative 
adjectives as devayu, fravasyu. The commentator, as usual, fails to 
give us any light upon these points : he only cites, as instances of the 
separable suffixes to which the rule relates, adhvari-yatdm (i. 4. 1), 
agha-yuh (e. g. iv. 3. 2), vrsha-yamdnah (ii. 5. 7), and fatru-yatim : abhi 
(iii. 1. 3) : and farther, as counter-examples, tat : sis&sati (xiii. 2. 14), to 
show that no desiderative suffix is separable unless beginning with y ; 
and yena : pravasyavah (iii. 9. 4), to show that the suffix beginning with 
y must not be preceded by a consonant. 

The Vaj. Pr. has a corresponding rule (v. 10), but more briefly ex- 

c(MciHMytwigiw n $o n 

30. Also yd, when combined with vasu, ava, svapna, sumna, 
and sddhu. 

Why this rule should be necessary, after the one which precedes it, 
and which would include all the cases to which it is intended to apply, 
I find it difficult to see. It can hardly be'that it was meant to exclude 
such words as su-kshetriyd,, su-gdtuyd, since these have been otherwise 
provided for (see under rule 12, above): more probably, forms like 
mithuya (e. g. iv. 29. 7) and urviyd (e. g. v. 12. 5) are to be regarded as 
its counter-examples. The Vaj. Pr. (v. 20) also does a like work of 
supererogation in reference to sundry words of its text. The com- 
mentator repeats the words, but gives nothing of the context of the 
passages in which they occur : they are vasu-ya (iv. 33. 2), ava-yd (ii. 35. 
1), svapna-yd (v. 7. 8), sumna-yd (yii. 55. 1), and sddhu-y& (x. 4. 21). 
The second of them is classed with the rest only by a blunder, since it 
is evidently avayds, the irregular nominative singular of ava-y&j, and 
ought to be written by the pada-text ava-ydh, instead of ava-yd. 

The comment closes with another verse : pancdi '»A 'vagrahdn dha 
ydpabde pdkatdyanah : antoddltah padatvaih ca vibhaktyarthe bhavet tu 
yd ; ' Qakatayana mentions five cases in which yd suffers separation by 
avagraha : it is accented as final, and stands as an independent element, 
when used as a case-ending.' I am by no means confident that I have 
correctly interpreted the last line. 

iv. 34.] Atkarva- Veda Prdtigdkhya. 531 

Ppsffattg ll^\ll 

31. Also bhis, bhydm, and bhyas. 

As illustrations, the commentary furnishes panca-bhih : anguli-bhih 
(iv. 14. 1), uru-bhydm : te : ashthivad-bhydm : pdrshni-bhydm : pra.-padd- 
bhydm (ii. 33. 5), and astki-bkyah : te : majja-bhyah : sn&va-bkyah : dha- 
mani-bhyah (ii. 33. 6). The case-ending bhyam, as in tubhyam and as- 
mabhyam, is not treated as separable. 

The Vaj. Pr. (v. 13) puts our rules 31 and 33 into one, declaring a 
case-ending beginning with bh separable, when following a short vowel 
or a consonant. This would teach the division tu-bhyam, asma-bhyam, 
also ; but the latter is expressly declared indivisible in another rule (v. 
35), and the former was perhaps overlooked. 

82. Also su. 

The commentator's instances are anka-su (vi. 35. 2), ap-su (e. g. i. 4. 
4), 1 and vayam : rdja-su (vii. 50. 7). 

The Vaj. Pr. (v. 14) combines this rule with rule 34, and says that su 
is separated when its s is not changed to a lingual. 

33. But not after a long vowel. 

This restriction applies to both the two preceding rules : no case-end- 
ing is separable after a long final vowel, whether this be an original long 
final of the theme, or the result of a prolongation according to the rules 
of declension. The commentator instances yabhili : satyam : bhavali 
(ix. 2. 25), tdbhih : tvam : asmdn, (ix. 2. 25), akshibhy&m : te : n&siM- 
bhydm : karndbhydm (ii. 33. 1), gobhyah : apvebhyah (iii. 28. 3), and 
dsu : itardsu (iii. 10. 4). 

Compare Vaj. Pr. v. 13, as quoted under rule 31, above. 

34. Nor where conversion into a lingual takes place. 

This is an exception under rule 32, applying only to the termination 
su. The two, as was already remarked, are by the Vaj. Pr. combined 
into a single rule (v. 14). Our commentator cites prali : tishtha : dik- 
shu (iv. 14. 9), namasyah : vikshu : idyah (ii. 2. 1), mdnushishu : dikshu 
(v. 11. 8), marutah : vikshu (viii. 4. 18), yam : ca : vikshu (ix. 5. 19), 
pari : pa$ya : vikshu (viii. 3. 10), and su-vrjandsu : dikshu (xviii. 1. 46). 

1 The MS. adds dup-su, which I have not been able to identify with any Athar- 
van word. Po9sibly hrt-m is intended. 

532 W. D. Whitney, [iv. 35- 

35. Vans is separated after a short vowel. 

The commentator quotes from the text the examples colcr-v&n (ii. 35. 
3) and papi-vdn (xiv. 1. 3), and the counter-example vidv&n (e.g. ii. 1. 
2), and then goes on with a long citation from his metrical authority, 
as follows : apade 'vagrakah fishta ikdrena padddind : dhdlvantdc ca 
vasdu hrasvdc cakrvdn papivdn id : upasargamm&se 'pi vas&v eva 'va- 
grhyate: kikin&uvdoifeshcna bkkle tdbhydm vidhiyate : vasusvdrlhe tayo 
lubdham bahulam chandasl Hi ca : v& tayoh krtoh samdsdc cd 'pajdyeta 
vastuttth: avakdre padatvark na p&rvenai 'vd 'vagrhyate: ahrasve 'pi 
padatvam sydt prdptas tatrd 'py avagrahah: ahrasve chdndasatvdt tu 
rutvam dhur manishinah. The meaning of some of these lines is very 
clear, and they are seen to cover the ground of our present rule and of 
the two following : others are obscure, and need emendation before they 
can be intelligently rendered. 

The Vaj. Pr. gives but one rule (v. 11) respecting the separation of 
the suffix of the perfect participle, combining together the specifications 
and counter-specifications of the three rules of our treatise. 

36. And that, even when the form is combined with a prepo- 

The cited illustrations are pareyi-vdnsam (xviii. 1.49), pravipi-v&nsam 

Siv. 23. 1), jakshi-vdnsah : papi-vdnsah (vii. 97.3), and uttasthi-vdnsah 
vi. 93. 1). The same words (excepting jakshivdnsah, perhaps omitted 
by the carelessness of the copyist) were found cited under i. 88, and it 
is probably their association there which has caused the inclusion among 
them here of the two forms from simple roots, which are of no value 
as regards the matter now under treatment. But for this rule, we might 
expect pra-vifivdnsam, ut-tasthiv&nsam, and so on, like pra-visktam, 
ut-tixhthatah, etc. And yet, the separation as here taught is not dis- 
cordant with the general principle that the last added member shall be 
the one which receives avagraka, since we may more properly regard 
the participial suffix as combined with the root after the latter's compo- 
sition with its prefix than before : were ta a separable suffix, we should 
doubtless also have pravif-ta, utthi-ta, and the like. 

37. But the preposition is separated, when the suffix shows 
no v. 

That is to say, when the suffix is contracted into ush, in the weak 
forms of declension, it is no longer separable, and the avagraha remains 
where it was before, between the preposition and the verb ; as in the 
forms cited by the commentator, d-jagmushah : arm-mate (ii. 26. 2), and 
vaf& : pra-dadushe : duke (xii. 4. 35). He adds a verse : yada prasdra- 

iv. 39.] Atharva- Veda Pr&ii<fikhya. 583 

nam tasya padatvam ne 'shyate tadd : purvend 'vagrahah siddho yatas 
taj jiyate padam ; 'when the suffix suffers contraction, its capability of 
standing as an independent pada is not taught : the former constituent 
then maintains the avagraha, as having a superior right to it (?).' 

38. Samanta is divided, when it has the sense of completion. 

The commentator gives us, as instances of the separable compound, 
pushkarinlh : sdmantdh (iv. 34. 5 etc.), and sdm-agrah : sdm-antah : bhu- 
ydsam (vii. 81. 4) ; and, as instance of the separable compound, ydthd : 
vrkshdm : libujd : samantdm (vi. 8. 1). But how the word has the sense 
of completion any more in the two former cases than in the latter, I 
quite fail to perceive. The commentator adds a farther exposition, 
which puts the distinction upon a safer, though still an arbitrary, ground : 
samanlam sarvalo'rthe 'ntoddttam nd 'vagrhyate, ddyuddltam avagrhyate : 
purandrtham pushkarinlh samantdh; ' samantam, having the sense of 
sarvatas, ' wholly,' and accented on the final, does not suffer avagraha ; 
when accented on the first, and having the sense of completion, it suffers 
avagraha, as in pushkarinlh samantdh^ 

39. The prepositions vi and sam are separated from the root 
an, when the word formed is a name of the breath. 

We should have expected this rule to be stated the other way; 
namely, that the root an was not separated from pra and apa (in the 
compounds prdna and apdna, which are always thus written in pada, 
without division). This would, on the one hand, be theoretically pref- 
erable, since the general rules for division would lead us to expect the 
paefa-readings pra-dna, apa-dna, vi-dna, and sam-dna, and we therefore 
ought to have the first two denied, rather than the last two ratified, by 
a special rule : and, on the other hand, it would be practically more 
accurate, since uddna, which occurs in the combination vydna-uddndu, 
is doubtless a separable compound, and is in fact so regarded by the 
commentator, under rule 42 below. Why prdna and apdna should not 
also be divided, it is far from easy to see. 

The commentator gives us the examples vi-dnah : dyuh (xviii. 2. 46), 
and sam-dnam : asmin : kah (x. 2. 13). To show that it is only after vi 
and sam that the avagraha takes place before an, he brings up the coun- 
ter-example prdnah : apdnah (xviii. 2. 46) ; and, to show that the com- 
pound must be a name of the breath, he cites samdnam : astu : vo : ma- 
nah (vi. 64. 3). The specification prdndkhyd cet, however, is after all 
pleonastic, since the adjective samdna, ' resembling, like, accordant,' is 
from sa + mdna, not sam + dna. 

The Rik and White Yajus treat the word prdna in the same manner 
as our text : apdna does not appear to occur in the former Veda, and 
in the latter it is (Vaj. Pr. v. 33, comm.) separable. Compare also Vaj. 
Pr. v. 36, which deals with samdna. 

534 W. D. Whitney, [iv. 40- 

40. Also are separated kdmya and a repeated word. 

This is a strange rule. In the first place, the Atharva-Veda furnishes 
no ground whatever for the treatment of k&mya as a sufEx, even though 
it be regarded as such in certain combinations in the general grammati- 
cal system (see Pan. iii. 1. 9 etc.). We find it only in such compounds 
as the commentator instances by citing apraddhaJ} : dhana-k&myA (xii. 
2.51) and anrtam : vitta-kdmyA (xii. 3. 52), which would fall under 
rule 9 of this chapter without occasioning any difficulty or hesitation. 
In the second place, I can discover no possible reason for combining 
together in one rule things so utterly unconnected and incongruous as 
the occurrence of this suffix and that of words repeated in an emphatic 
or a distinctive sense. The dual termination, however, is our warrant 
that we have not here, as in the case of rules 12 and 13 of the first 
chapter, two rules written and explained together by the commentator. 
The latter cites a single passage containing two words which are Amre- 
dita, viz. bh&yah-bhAyaJi : fvah-fvah (x. 6. 5 etc.). 

The Vaj. Pr. (v. 18) has the same rule respecting repeated words, 
and calls them (i. 146) by the same name. Panini also employs the 
term dmredita (e. g. vi. 1. 99) in a kindred sense. 

3% sr \\8\u 

41. Also iva. 

The commentator cites but a single instance — s&ldvfMn-iva (ii. 27. 
5) — of this exceedingly frequent case of combination. The Vaj. Pr. 
notes it at v. 18. 

ii 3^n 

42. Separation is made between two words which are each of 
them separable. 

Or, as the commentator paraphrases it, when two words, themselves 
separable, are combined into a single word, separation of the middle 
member (parvan) is made. His instances are yat : Anjana-abhyafijanam 
(ix. 6. 11), prajA-amrtatvam : uta : dtryham : ayuh (xi. 1. 34), and vyAnar 
-udan&u : vdk (xi. 8. 4) ; to which we might add indefinitely, not only 
copulative compounds, but possessives (e. g. Ahrta-yajnakratuh, ix. 6. 27) 
and others. 

The Vaj. Pr. finds no need of such a rule as this, nor does it seem 
imperatively called for, all possible cases being already disposed of by 
rules 10 and 12, above. Still less is to be seen the necessity of adding 
to it the two which next follow, and which it obviously includes. 

spmwtsr ii 8$ n 

43. As also, between two compounds. 

iv. 45.] Atharva- Veda Pr&ti$&khya. 535 

The sole example furnished in the commentary is aghapansa-duhpan- 
sdbhydm : karena (xii. 2. 2). 

fi^#j -cjio|j|tjj n gg ii 

44. As also, between a separable word and its repetition. 

The commentator cites instances of repeated words occurring in 
five successive verses of the Atharvan text, without troubling himself 
about the fact that two of them are not separable : they are kurvatim- 
-kurvalim : eva (ix. 5. 32), samyatim-samyattm : eva (ix. 5. 33), pinvatim- 
-pinvatim : eva (ix. 5. 34), udyalim-udyatim : eva (ix. 5. 35), and abhibha- 
vantlm-abhibhavantlm : eva (ix. 5. 36). He adds a verse : prthagingyasa- 
mdse ca madhye kurydd avagraham : samyaltmsamyaltm c&i Vo vydno- 
ddndu nidarfanam ; 'separation by avagraha must be made in the 
middle of a compound made up of two severally separable words : 
instances are samyalim-samyatim and vydnoddndu? 

45. In vasudhdtara and sahasrasdtama, separation is made after 
vasu and sahasra. 

The passages are cited by the commentator : vasu-dhdtarah : ca (v. 
27. 6), and sahasra-sdtamd : bhava (iii. 28. 4). It is not without reason 
that the Prati§akhya takes note of these cases ; for, since the suffixes 
tara and tama are separable (by iv. ] 6), and are plainly the last added 
members, the words they form should read, in pada, vasudhd-tara and 
sahasrasa-tama. Comparatives and superlatives of this particular class, 
however, where the suffixes are appended to root words which directly 
govern the preceding member of the compound, are treated in the same 
manner by the pada-texts also of the Rik (e. g. ratna-dhdtamam, i. 1. 1) 
and White Yajus, and the latter's Pratigakhya (V. Pr. v. 3), makes spe- 
cial mention of them. The commentator adds : vasudhdtara iti : vasij,- 
ndrh dhdtrtarah: shashthyantena \arthenaf\ samdsah : samdse avagraho 
bhavati : vasimi vd dadhati : vasu-dhdtarah : samdse avagraho bhavali ; 
'vasudhdtara: that is, one who is in a high degree a giver (dhdtrtara) 
of good things; composition is made with a form having a genitive 
sense ; the compound suffers avagraha : or, vasudhdtaras, ' they bestow 
good things;' the- compound suffers avagraha.' 1 The only item of value 
derivable from this exposition is that some authorities regarded vasu- 
dhd'tarah as the plural of vasu-dhdtar. It would be, in fact, in its 
Atharvan connection, much more easily interpretable in this manner, 
but that the accent speaks strongly for the other mode of derivation. 
The passage in which it occurs is shown by comparison with the White 
Yajus (xxvii. 15) to be curiously misunderstood and corrupted, and the 
Atharvan vasudhd'tarah corresponds to vasudhd'lamah of the other 
text : we may suppose that the former means to give the plural of vasu- 
dhdtar, but gives it the accent which belongs to vasudhd'tarna and its 
corresponding comparative vasudhd'tara. The commentator closes his 
treatment of the subject with a verse: sanibhyam [sddhdbhydm?~\ ca 
vol. vii. 68 

536 W. D. Whitney, [ir. 45- 

krdantdbhydm vihitdu taddkitdu pardu: tdbhydm shashthisamdse ea 
pilrvend 'vagrahah smrtah : ' after id and dh& (?), as /fcrtf-endings, taddhi- 
tas are declared to follow : in a compound with these having a genitive 
meaning, the former member is separated by avagraha.'' 

46. Subhishahtama suffers separation by avagraha before tama. 

The commentator cites »ubhishak-tamdh (vi. 24. 2) ; we have also the 
nominative singular masculine at ii. 9. 5. He adds fobhanah bhishak : 
subkishak ; ' subhishak means propitious physician ;' and then again gives 
a verse: bhishajd hi supabdo 'yam pumlingena samasyate: upajdtas 
tamas tasmdt p&rvend [parend ?] 'vagrahah smrtah ; ' here su is com- 
pounded with the masculine bhishaj, and tama is farther appended : 
separation by avagraha is made of the latter/ 

I can see no reason at all for any such rule as this : the case specified 
is simply one in which the separation by avagraha takes place normally, 
according to the general rules, and a score more of precisely similar 
cases might easily be quoted from the Atharvan text : instances are 
bh&gavut-lama (ii. 9. 2) and bhagaval-tara (iv. 13. 6), sphalirn&t-tama 
(iii. 24. 6), mrtamanah-tara (vi. 18. 2), and vrtrahan-tama (vii. 110. 1). 

The signature of the first section, which closes here, is as follows : 
47 : caturthasya praihamah padah : caturddhydyibhdshye caturthasya 
prathamah padah samdptah. We have found but forty-six rules in the 
section, but have remarked one (rale 40) which ought to have been 
divided and counted as two. Possibly two may have been fused to- 
gether in it, in order to allow the commentator's introduction to the 
chapter to count as a rule, without altering the received number in the 
section : but I have neither been willing to allow the rank of a rule to 
anything in that introduction, nor ventured to divide rule 40 into two 

*T H*l^=hl(l^rt Randan 

47. The suffix mant and its equivalents are not separable 
after t and s. 

The commentator cites in illustration datvati (e. g. iv. 3. 2), garutmdn 
(e. g. iv. 6. 3), marutvdn (e. g. vi. 104. 3), Arjasvdn (tirjasvantah, vii. 60. 
2), payasvdn (e. g. vii. 73. 5), itrjasvati (e. g. iii. 12. 2), and payasvati 
(e. g. iii. 10. 1). The only consonants other than t and s which are 
found to occur before the suffix vant are n and n, which allow separa- 
tion ; instances are dsan-vat (vi. 12. 2), asthan-vantam (ix. 9. 4), brah- 
man-vatim (vi. 108. 2), etc. The rule is an exception under rule 17 
above ; by the Vaj. Pr. (v. 8) it is included with the general rule in one 

UrlXd^ft ^r?t II £tll 

48. Nor vant, after ya, ta, and eta. 

iv. 50.] Atharva- Veda Prdtigdkhya. 587 

The commentator's examples are ydvat : te : abhi : vi-papy&mi (xii. I. 
33), tdvat : sam-ditu : indriyam (iii. 22. 5), and etdvat : asya : prdctnam 
(iv. 11. 8); and his counter-examples, which are hardly called for, are 
itinrtd-vat (e. g. v. 20. 6), and apdshtha-vat (xiv. 1. 29). This rule, also, 
is included in Vaj. Pr. v. 8, since in each of the words to which it refers 
there is an irregular prolongation of the final vowel of the theme before 
the suffix. 

X^MH^ ^T II &\\l 

49. Nor a copulative compound made up of the names of 

The commentator gives pretty nearly the whole series of such com- 
pounds which the text furnishes: they are indragni (e.g. i. 35. 4), in- 
dravdyA (iii. 20. 6), bhavdrudrdu (xi. 2. 14), bhav&parvdu (e. g. iv. 28. 1), 
vdtdparjanyd (x. 4. 16), agntshomd (e.g. i. 8. 2), mitrdvarund (e. g. iii. 
4. 4), indrdvarund (e. g. vii. 58. 1), and indrdsomd (e. g. viii. 4. 1) : to be 
added are only somdrudrau (e. g. v. 6. 5), indrdpushand (vi. 3. 1), and 
agndvishnu (vii. 29. 1,2). A number of verses follow in the comment- 
ary, in the usual corrupt condition of text : devatdndm iha dvandve dtr- 
ghatvam yadi 1 drpyate: aningyam tat 2 padam, v&cyam agnishomdu 
nidarpanam : thus much is clear, and is a virtual repetition of our rule, 
but with a restriction to cases in which a long vowel appears at the end 
of the first member of the compound, which requires a specification 
farther on of the single exception indravdyu ; what follows is more ob- 
scure, and I have not been able, with what time I have given to it, to 
restore the text to an intelligible form ; it reads : vdsurdndrh dvandve 'py 
avagrhyam katham padam: fdkalyasye ''ngite nityam yathd satydnrte 
[i. 33. 2] tathd : brahma prajdpatis [xix. 9. 12] tv aha nd 'vagrhyarh kadd 
oana: Ananah pratishedhaf ca vdyof co , bhayatah param: indravdyv 
[iii. 20. 6] Adishu katham dirgho yatra tad [wa. ? ] drpyate: dvandvamd- 
trenashedastvam ahordtre nidarfanam. 

The rule of the Vaj. Pr. (v. 28), which includes also our rules 50 and 
52, is to the effect that dual copulative compounds whose first members 
end in a vowel are not separable. 

50. Nor one which shows a long vowel before an initial con- 
sonant of the latter member of the compound. 

The instances furnished by the commentator are ishtdp&rtam (e. g. ii. 
12. 4), pitdputr&u (vi. 112. 2), hasdmuddu (xiv. 2. 43), dydvdprthivi 
(e. g. ii. 1. 4), dydvdbhumA (xviii. 1. 31), and ushdsdnaktd (e. g. v. 12. 6). 
To these I add fundsird (iii. IV. 5), s&ryamdsdu (iii. 29. 5*), surydcan- 
dramasdu (vi. 28. 3), and yajndyajriiyam (viii. 10. 13). To the same 
class, of dvandvas exempt from division, belongs prdndpdndu (e. g. ii. 16. 
1), although it does not show the peculiarities of form which this rule 

1 yad. 3 anityat. * In the edition, mryam° is a misprint. 

538 W. D. Whitney, [iv. 60- 

demands, and therefore ought to be made the subject of a special pre- 
cept : our treatise-makers and their commentator, apparently, have over- 
looked it. A single counter-example the commentator gives, viz. sa- 
tydnrte iti satya-anrte (i. 33. 2). He adds a verse : irdmivdpurobhyaf 1 
ca prakrtyd dirgha eva sah: hrasvasya yatra dirghatvam sa dvandvo 
nd ''vagrhyale ; 'after ird, amivd. and parah [avagraha comes in], for 
in those cases the vowel is long by nature ; but where a short vowel is 
lengthened, there no division by avagraha takes place.' This is a very 
blundering statement, so far as concerns the instances given in the 
first pdda: purah, of course, could form no copulative compound; 
amtva forms none such in the Atbarvan, and it also, though a femi- 
nine with a long final vowel, as a separate word, always shortens its 
final in composition (amiva-cdtano, e. g. i. 28. 1 ; amtva-han, e. g. RV. 
i. 18. 2); ird, too. is found only in the compound ird-kshird (x. 10. 6), 
which is not copulative. 

The implication of dvandva from the preceding rule seems clearly 
made by the particle ca in this rule, and is supported by the connection 
as shown by the two following rules : the commentator, also, inserts 
dvandvasya after yasya in his paraphrase. Such compounds, then, as 
vifvdnara, svdvrk, vlrudh, sukara, etc., which are left undivided in the 
pada-text on account of the irregularly protracted final of their first 
members, must be left to fall into the general gana of rule 54. 

51. Nor shodaqin, on account of the interfusion of the two 
members of the compound. 

Or, it may be, 'on account of doubt' — that is to say, of doubt as to 
the form to which the constituents should be restored, their mode of 
combination being an entirely anomalous one. It is to be observed, 
however, that our treatise has itself (at i. 63) given special directions 
as to how shot and following dafa are combined together, so that to the 
student of the Praticakhya the ^arfa-reading shat-dafa ought to occa- 
sion no difficulty. That the rule reads instead of shodafa is 
surprising, since both words (each in but a single passage) occur in the 
text: the commentator cites them, as follows : ishtdpurtasya : shodafam 
(iii. 29. 1), shodafi : sapta-r&lrah (xi. 1. 11). 

The Vaj. Pr. mentions shodafa in a rule (v. 37) containing a long 
list of indivisible words. 

^\{\^ II H^ II 

52. Nor ahordire. 

The commentator's illustrations are ahordtrdbhy&m : nakshatrebhyah 
(vi. 128. 3), and ahordtre idam brumah (xi. 6. 5). The Vaj. Pr. includes 
the word in the same rule with shodafa (v. 37). 

There is nothing in the character of either ahordtre or shddafa, so 
far as I can discover, which should withdraw them from the action of 
rule 50, and render their separate mention necessary. 

1 MS. idd". 

iv. 54.] Atharva- Veda Prdti^dkhya. 539 

53. Nor the root anc: nor former constituents of a compound. 

It is, if possible, even more surprising here than at rule 40 to find 
two so utterly heterogeneous matters put together in the same rule. 
We cannot suppose that the commentator would combine them, in 
statement and in exposition, unless he regarded them as composing a 
single precept ; but, on the other hand, we have not the same warrant 
here as in the former case that his division is a correct one : there is 
nothing in the form of the rule which would absolutely forbid its simple 
division into two parts, without further change — although we should, in 
that case, expect rather aiicatdu than ancati. 

As illustrations of the inseparability of the root anc, we have given 
us prdclh (e. g. v. 28. 11), pratici (e. g. iii. 27. 3), and udtcih (e.g. xii. 1. 
31). All the compounds with this root are treated as indivisible by our 
pada-text : the Vaj. Pr. also (v. 30) declares the root inseparable, with 
exception (v. 19) of a single derivative. 

To show that, when new members are added to a compound, the 
existing division by avagraha of their former members is given up, the 
commentator instances pamtdti-bhih : arishtatdti-bhih (iv. 13. 5) : com- 
pare arishta-tdtaye, cited above, under rule 20. The principle has been 
already sufficiently illustrated in these notes, under rule 10. The Vaj. 
Pr. has nothing corresponding to this part of our rule, which is, in fact, 
virtually superfluous, since the directions already given for the separa- 
tion of a newly-added member might be understood as involving the 
suspension of the ancient division. 

The commentator ends with a verse which seems to say precisely the 
opposite of the rule of his text: yatro 'bhe pratividhye te upajdtamja- 
rath, ca yat, jaratA ''vagrahah kdrya rksdmdbhydm nidarfanam; 'when 
both members are severally separable, both the newly-added and the 
ancient, separation by avagraha is to be made of the ancient one : an 
instance is rk-gamabhyam.'' But this is mere nonsense, as it stands, the 
word cited being a case where the last-appended element is inseparable, 
as following a long vowel (see rule 33, above), and where, therefore, the 
division must be suffered to remain between the two original constitu- 
ents of the compound. If the theme of declension had been rk-sdman, 
instead of rksdma, we should have an instrumental dual rksdma-bhydm, 
which would be a true illustration of the rule. One may conjecture 
that the last line originally read jare nd 'vagra/iah kdrya rksdmabhydm 
nidarpanam, and that it was amended to its present form by some 
copyist who knew that the Atharvan read, not rksdma-bhydm, but rk- 
sdmdbhydm, but who was careless enough to overlook the discordance 
which he thus introduced between the text and its comment. 

^'Rl ^tiih^u 

54. Nor samudra etc. 

The whole comment upon this rule is wanting in our manuscript : the 
copyist has again carelessly skipped from its first statement to its final 

540 W. D. Whitney, [iv. 54. 

repetition before the rule next following. This, however, gives us reason 
to believe that the commentator had performed his work in his usual 
brief and unsatisfactory style, and had done very little toward filling 
up the gana. It would have been particularly curious and interesting 
to see how many and which of the words contained in the Atharvan 
the makers of the Prati§akhya looked upon as fairly entitled to a di- 
vision which the constructors of the pada-text had not admitted. The 
Vaj. Pr. (at v. 37) gives a list of such words for its text, but Weber 
finds it, as was to have been expected, both deficient and redundant. 
It is by no means easy to draw up a list which shall include all that 
ought to be received, and exclude all that ought to be left out ; but I 
have looked through the Atharvan text with some care for this purpose, 
and trust that my filling up of the gana of the text will be found to 
comprehend all or nearly all of the matter to which the rule ought to 

There is, in the first place, in this as in the other Vedic texts, a con- 
siderable class of compound words exhibiting an irregular prolongation 
of the final vowel of the former member, and which the constructors of 
the ^>arfa-text have chosen to leave unchanged, instead of separating 
them by avagraha and restoring the normal quantity of the altered 
vowel. Why they should be thus treated, however, in distinction from 
the words with which our treatise deals in the first section of its third 
chapter, no sufficient reason appears. They are as follows : ap&mdrga 
(e. g. iv. 17. 6 : the word, by V. Pr. v. 21, is divisible), apdshtha (iv. 6. 
5 : see above, ii. 95), ashthivant (e. g. ii. 33. 5), idavatsara (vi. 55. 3 : cf. 
V. Pr. v. 32), ubhayadant (e. g. v. 31. 3 : divisible by V. Pr. v. 21), ubha- 
ydvin (e.g. v. 25. 9 : see above, under iv. 18), ekddafa (v. 16. 11 : cf. V.Pr. 
v. 37), kakshivant (e. g. iv. 29. 5 : cf. V. Pr. v. 37), taldmaha (e. g. v. 24. 
17), dv&dafa (e. g. iv. 1 1. 11 : cf. V. Pr. v. 15), nardfama (v. 27. 3 : cf. V. 
Pr. v. 37) and ndrdfansi (e. g. xiv. 1. 7), nihara (e. g. vi. 113. 2 : cf. V. 
Pr. v. 37), pr&naha (ix. 3. 4), pr&vrta (e. g. xii. 5. 2 : cf. V. Pr. v. 37), 
prdvrsh (e. g. xii. 1. 46), marmdvidh (xi. 10. 26 : cf. iii. 3, iv. 68), vipvd- 
mitra (e. g. iv. 29. 5 : cf. iii. 9 and V. Pr. v. 37), vigvdnara (e. g. iv. 11. 
7 : cf. iii. 9 and V. Pr. v. 37) and vdipvdnara (e. g. i. 10. 4), vtrudh (e. g. 
i. 32. 1), fvdpad (e. g. viii. 5. 11 : cf. iii. 10), fvdvidk (v. 13. 9 : cf. iii. 3, 
iv. 68), saranga (e. g. ii. 32. 2), sdrathi (e. g. viii. 8. 23), s&kara (e. g. ii. 
27. 2) s&nrta (e. g. iii. 20. 3), svdvrk (xviii. 1. 32), and hrdaydvidh (viii. 
6. 3, iv. 68). 

Another smaller class is composed of certain words which have as 
their first member a real or an apparent case of declension : such are 
anyedyus (i. 25. 4 : cf. iv. 21), dfuihga (vi. 14. 3), gavishthira (iv. 29. 5 : 
cf. V. Pr. v. 37), narishtd (e. g. vii. 1 2. 2 : cf. V. Pr. v. 37), patamga (e. g. 
vi. 31. 3), pildmaha (e. g. v. 5. 1), madhyamdina (e. g. iv. 11. 12), mdta- 
rifvan* (e. g. v. 10. 8), and fitimga (xi. 5. 12). 

The number is by no means an inconsiderable one of words whose 
division seems so naturally suggested by an etymology which is either 
incontestable or at least very plausible, that we are reasonably surprised 

* At v. 2. 9, all the manuscripts have mdtaribhvari, which the edition, hardly 
with sufficient reason, has amended to mdtarifvari : it is, like the latter, left mi- 

iv. 54.] Atharva- Veda Prdtigdkhya. 541 

that they were not divided by the Hindu grammarians. Of these, I 
name anavAya (viii. 4. 2), anushtku (xii. 4. 45 : probably regarded as 
formed by an inseparable suffix), apAna (e. g. ii. 28. 3 : cf. under iv. 39), 
abhishti (e. g. i. 6. 1), avarti (e. g. iv. 34. 3), avaskava (ii. 31. 4), apvatara 
(e. g. viii. 8. 22 : cf. under iv. 16), asvaga (xii. 5. 45), aghrni (vii. 9. 2), 
AdM (vi. 131. 1 etc.: Rik pada, A-dhl), Anushak (iv. 32. 1), apri (xi. 7. 
19), Ayudha (e. g. iii. 19. 5 : cf. V. Pr. v. 37), arli (e. g. iii. 31. 2), Arpita 
(e. g. vi. 112. 3 : at viii. 9. 19 only, we have A-ArpitAni), Aptviska (xii. 5. 
34), asakti (xiv. 1. 26), Asikta (xii. 3. 25 : probably the absence of divis- 
ion is here only an error of the manuscript ; we have A'-siktam at iv. 7. 
1), uttAna (e. g. ix. 9. 14), rtvij (e. g. vi. 2. 1), oshadhi (e. g. i. 23. 1 : cf. 
V. Pr. v. 35), ffopd (e. g. iii'. 8. 4 : cf. V. Pr. v. 37), gopiiha (e. g. v. 9. 7), 
candramas (e. g. v. 24. 10 : cf. V. Pr. v. 37), dAyada (v. 18. 6, 14), nyag- 
rodha (e. g. iv. 37.4 : cf. V.Pr. v.37), purodApa (e. g. ix.6. 12 : see i. 63 : 
cf. V. Pr. v. 37), prAna (e. g. ii. 12. 7 : cf. under iv. 39), prAyapcitti (xiv. 1. 
30), vivasvant (e. g. xi. 6. 2), vishtap (e. g. x. 10. 31 : cf. V. Pr. v. 41), vish- 
tambha (xiii. 4. 10 : cf. V. Pr. v. 41), vishtArin (iv. 34. 1 etc.) pinpumAra 
(xi. 2. 25), praddhA (e. g. v. 7. 5), sabhA (e. g. iv. 2 1. 6), samantdm (vi. 8. 1 : 
cf. iv. 38), samudra (e. g. i. 3. 8 : cf. V. Pr. v. 37), surabhi (e. g. vi. 124. 
3 : cf. V. Pr. v. 37), sthapati (ii. 32. 4), svadhA (e. g. ii. 29. 7), svapati 
(viii. 6. 16), svasti (e. g. i. 30. 2), and hAridrava (i. 22. 4). It is not hard 
to conjecture, in the case of some of these words, reasons which may have 
led to their being treated as exceptional cases, but in many of them no 
such reason is apparent, and in a part, at least, we are compelled to 
suppose that the composition was fully recognized, and the division 
neglected for some arbitrary and unexplained cause. That the four 
compounds of pronominal elements cana, nahi, nakis, and mAkis were 
left by the pada in their sanhitA form is not to be wondered at : three 
of them are noted also by the Vaj. Pr. (v. 35, 37) as indivisible. 

_ There yet remains quite a list of compounds and derivatives, the di- 
vision of which may plausibly be supposed to have been neglected from 
uncertainty of etymology, anomaly of form, difficulty of restoring the 
original constituents, or the like : while yet, in most cases, we should 
not have been surprised to see the constructors of the pada making an 
attempt at their analysis. In drawing up this part of the list, especially, 
I may very possibly have omitted to note down some words of the text 
which to another would seem not less worthy of mention than those 
given : the series, as collected by me, is akupAra (v. 17. 1), ajagara (e. g. 
iv. 15. 7: cf. V. Pr. v. 37), adomada (vi. 63. 1) and adomadha (viii. 2. 
18), anadvAh (e.g. iii. 11. 5 : cf. V. Pr. v. 37), anrkshara (e. g. xiv. 1. 
34), abhipu (e. g. vi. 137. 2), abhra (e. g. iv. 15. 1 : cf. V. Pr. v. 34), ava- 
dya (e. g. ii. 10. 6), Atura (vi. 101. 2), AmikshA (e. g. ix. 4. 4), Ahanas 
(e. g. iv. 30. 6), udArathi (iv. 7. 3), urvapi (xviii. 3. 23), karmAra (iii. 5. 
6: cf. V. Pr. v. 37), karpapha (iii. 9. 1), kasarnlla (x. 4. 5, 17), kucara 
(e. g. vii. 26. 2 : cf. V. Pr. v. 37), krkavAku (v.'31.2), godhA (iv. 3. 6 : 
cf. V. Pr. v. 37), jAshkamada (xi. 9. 9), daponasi (x. 4. 17), duchunA (e.g. 
v. 17. 4 : cf. ii. 61), duradabhna (xii. 4. 4, 19), durAhA (viii. 8. 24), dru- 
vaya (e. g. v. 20. 2 : cf. under iv. 18), dhivan (iii. 5. 6), padbtpa (e. g. vi. 
96. 2), pandaga (viii. 6. 16), prAnada (iv. 35. 5), maryAda\e. g. v. 1. 6), 
mahiluka (x. 10. 6), ralharvl (x. 4. 5), vansaga (xviii. 3. 36), valaga (e. g. 

642 W. D. Whitney, [ir. 64- 

v. 31. 4 : cf. V. Pr. v. 35), vifvdhd (e. g. vii. 50. 1 : cf. V. Pr. v. 37), vyd- 
ghra (e. g. iv. 3. 1 : cf. V. Pr. v. 37), fdndad&rva (xviii. 3. 6), sadyas (e. g. 
viii. 10. 21), and svdhd (e. g. ii. 16. 1).' ' 

The Vaj. Pr. (v. 37) notes a couple of words as indivisible which are 
found divided in the Atharvan : they are upa-sti (e. g. iii. 5. 6) and pa- 
vira-vant (iii. 17. 3). In like manner, samvatsara, which the Rikpada 
does not analyze, is in our text uniformly written sam-vatsara (e. g. iii. 
10. 9). One or two other such cases of discordance among the several 
pada-texts are pointed out in the notes to the different rules : but there 
is, in general, such close agreement among them as to show conclusively 
that the pada method of text-analysis, in its details as well as in its 
main plan, is the production of a single teacher, or of a single school. 

It may be well to add here, per contra, a few of the cases in which 
the pada-text makes unintelligible or palpably erroneous divisions of 
words : I have noted, as the most striking instances of this kind, anam- 
-ffureh (viii. 6. 22), anapa-dyatdm (iv. 17. 6), jighat'Svam (ii. 14. 1), go- 
-pana- (xii. 4. 10), pam-opya (i. 14. 3), hr-dyota (i. 22. 1) and hr-dyotana 
(v. 20. 12). The peculiar form, accentuation, and division of two pas- 
sages in the fourteenth book — fUbham : yatih (xiv. 1. 32) and pdtim : 
yatih* (xiv. 2. 52) — is also worthy of remark in connection with this 

cj^klt^UI H{MH ll W 

55. Nor is a member which, has suffered vrddhi separable, if it 
be monosyllabic and end in a vowel. 

The commentator's examples of an inseparable vriddhied initial sylla- 
ble are sdpatnah (ii. 7. 2), sdumanasah (e. g. iii. 30. 7), sdumanasam (e. g. 
xiii. 1. 19), s&udhanvanah (vi. 47. 3), trdistubham (ix. 10. 1), ndubhagam 
(e. g. ii. 36. 1), and saubhdgyam (e. g. xiv. 1. 42). His counter-exam- 
ple, brought forward to show that the inseparable member must have 
suffered vrddhi, is su-parnah (e. g. i. 24. 1) ; to show that it must be 
monosyllabic, they are dird-vatah (viii. 10. 29), mdrta-vatsam (viii. 6. 
26), and vddhil-yam (e. g. xiv. 1. 29) ; to show that it must end in a 
vowel, they are ndih-bddkyena : havishd (vi. 75. 1) and dduh-svapnyam : 
dduh-jtvityam (iv, 17. 5). I add, in the farther illustration of the in- 
separable class, vdimanasya (v. 21. 1) and prdhrddi (viii. 10. 22) ; of the 
separable class, sdm-itya (viii. 10. 6), sdm-rdjya (xiv. 1. 43), pdurna-mdst 
(yii. 80. 1), sdurya-varcasa (viii. 10. 27), and avdira-hatya (vi. 29. 3). 
The rule is, I believe, carefully observed throughout the whole of the 
Atharvan text, and the Vaj. Pr. (v. 29) has one precisely correspond- 
ing ; nor have I noted any cases in which the usage of the Rik pada- 
text was not in accordance with it. Its somewhat arbitrary character, 
however, is patent. 

The commentator again closes his exposition with a verse : avagrhydt 
paddd yamtu taddhito vrddhim&n bhavet: ekdt vrddhisvaranteshu na 
cdi 'vd 'vagraho bhavet: dirdvato marlavatsam vddh&yam ca nidar$andt. 

* In this passage, the printed text reads pdtim yati'h, but without any support 
flrom the manuscripts. 

iv. 56.] Aiharva-Veda Prdti^dkhya. 543 

A little amendment makes this mean, in restatement of our rule, ' where 
a taddhita suffix requiring vrddhi is appended to a divisible word, sepa- 
ration is not made of a member which is monosyllabic and ends in a 

itelufkCUl^UI MirlRi^HIMMNli^clsi^ii^H 

56. Nor a monosyllabic member ending in a or d and nega- 
tived — excepting in the case of apraydvan etc. 

The form of statement which our treatise has adopted for its rule re- 
specting the separability of negative compounds is not particularly well 
chosen. The general usage of the Atharvan ^arfa-text, as regards such 
compounds, is as follows. The negative prefix a or an is not itself ever 
separated by avagraha from the word to which it is attached : we have 
atsat, anrta, etc. If, however, the word negatived is already a compound, 
the negative prefix is in almost all cases treated like any other added 
inseparable element, and leaves the prior division of the compound un- 
affected : we have parA-jila and apard-jita, vtra-han and avira-han, etc. 
A few exceptions to this mode of treatment occur, and with them it is 
the province of our rule to deal. The preposition A, with an prefixed, 
is always made inseparable : thus we find d-srdva, but andsrdva (e. g. ii. 
3. 2, 3), and, in like manner, andjdnant, anddhrshya, andrambhana, and- 
dishta, and andvrtta. The same analogy is followed by the negative 
forms of compounds with sa, and by a single one of those with pra — 
viz. aprajasam (e. g. vii. 35. 3) — and by these alone. The Vaj. Pr., then, 
which declares (v. 24, 25) the negative prefix inseparable when alone 
and when followed by d, leaving other rare and exceptional cases to be 
provided for as such, expresses more truly the usage of the text. Our 
commentator gives us, first, as illustrations of the rule, the only two 
cases of negative compounds with sa which are found to occur in our 
text : they are asabandhuh (vi. 15. 2) and asapalnah (e. g. i. 19. 4) : the 
latter is mentioned by the Vaj. Pr. (v. 37) in its list of indivisible words, 
along with asajdta; asabandhu, according to "Weber (p. 305, marginal 
note), is treated as divisible in the White Yajus.* The commentator 
adds aprajdh, aprajdtdh, but neither of the words is to be found in the 
Atharvan. As counter-examples, he gives first avi-dvesham : krnomi : vah 
(iii. 30. 1), to show that the negatived member must end in a or A in 
order to be inseparable ; secondly, to show that, if ending in a or A, it 
must also be monosyllabic, he gives agne : akravya-at (xii. 2. 3) ; and 
thirdly, as evidence that a monosyllabic member ending in the vowels 
specified is not separable unless negatived, he cites yah : sa-patnah (i. 19. 
4). Finally, he partly fills up the gana, with apra-ydvan (iii. 5. 1 ), apra- 
•mddam (e. g. xii. 1. 7), apra-hildu (vi. 29. 2), and apra-cankap&h (viii. 6. 
16): I have noted in addition only apra-yuchan (e. g. ii. 6. 3). To 
close up the exposition, is added the verse ekaksharasavarndntam yad 
bhavet padam utiaram : tat padam nd , vagrkntydd aprayAvAdivarjilam ; 

* In one of the two cases where it occurs in our own text (vi. 54. 8), the pads. 
divides it, an-bandttuh : this, however, is probably a copyist's error. 
vol. vii. 69 

544 W. D. Whitney, [iv. 56- 

' if the word following the negative particle be monosyllabic and end in 
a vowel homogeneous with it, it is not to be separated, except in the 
cases apraydvan etc.' 

57. Nor are prdnati and prdnanti divided. 

The commentator cites yah : prdndti (iv. 30. 4), ydt : ca : prdndti (xi. 
4. 10), yina : prdndnti (i. 32. 1), and ydsmdt : prdnanti (xiii. 3. 3). But 
the rule is an exceedingly insufficient exposition of the treatment by 
the pada-text of the forms of the root an with the prefix pra. Division 
is, in fact, omitted only when the verb, and not the preposition, has the 
accent ; but then, not in the two forms specified only, but also in the 
participles — as prdnat (e. g. x. 8. 2), prdnnte (xi. 4. 8), prdnatds (iv. 2. 2), 
prdnatd'm (iii. 31.9), and prdnatt'ndm. (viii. 9. 9) — and in the causative, 
as prdndyati (xiii. 3. 3). On the other hand, if the prefix takes the 
accent, it is disjoined from the verb, according to the general usage in 
such cases, and we read pra : ana (iii. 31. 9), and pra : anati (x. 8. 19. 
xi. 4. 14). If the root is compounded with apa, also, the same usage is 
followed, and we have apdnate (xi. 4. 8) and dpa. : anati (xi. 4. 14). 

The Vaj. Pr. (v. 33), as acutely amended by Weber, gives a nearly 
corresponding precept, although it appears (Weber, p. 303, marginal 
note) that the text to which it belongs contains no verbal forms in which 
the division requires to be made. 

Wtf$*ti H*l^ 3u[trTtllHEn 

58. Nor are sam and pari separated from the root kar, if the 
latter begins with s. 

The commentator cites the only words occurring in our text in which 
the root kar has the sibilant prefixed to it, in composition with the two 
specified prefixes: they are xanxkrtatram (iv. 21.4), sanslcrtom (xi. 1. 
35), and pariMrtd (e. g ix. 3. 10). 

The doctrine of the Vaj. Pr> (v. 43) is the same, so far as concerns 
the compounds of sarn and kar; but it apparently allows the division 
of parixhkrla (which also occurs in its text : see iii. 52). 

59. Nor is division made in any case where a s is inserted— 
except in tuvushtama. 

The instances which the commentator gives of the insertion of * as 
an augment, (dgamn) between the two members of a compound word, 
and of the consequent unresolvability of the compound, are (ilask'irdm 
(xii. 1 47), Idtikiriih (e.g. iv. 3. 2). vdmmpdtih (e.g. iv. 3. 1), and hr'hus- 
pdtih (e.g. ii. 13. 2). Their citation under such a precept implies the 
acceptance of some such etymological theories of their derivation and 
form as are given by the Vaj. Pr. (iii. 49, 51), which explains fcw- 

iv. 62.] Atharva- Veda Pr&ti$dkhya. 545 

kara and brhaspati as from tat-kara and brhat-pati respectively, with, 
loss of t and insertion of s, and vanaspati as from vana-pati, with inser- 
tion of *; but it is unnecessary to remark that such explanations are 
futile : taskara is obscure, and the other two are without much doubt 
compounds of pati with the preceding genitive of an obsolete noun, 
being analogous with brahmonas p&ti, v&cds p&ti, pubh&s p&ti, etc. ; and 
they would doubtless be separated by the/xida-text into two independent 
words, like these, but for their frequency of occurrence, and, yet more, 
the irregularity of the accent of their former members as genitives of 
a monosyllabic theme. The counter-example, which the commentator 
also cites — indrahpatis tuvuhtamah (vi. 33. 3 : p. tuvi-tamah) — has been 
made the special subject of one of our previous rules (iii. 96). 

Rwldl^wrfl II \o H 

60. Nor in vigpati and vi$patnt 

The commentator cites instances of the occurrence of each of these 
-words — viz. svaptu : vifpatih (ix. 5. 6) and ya : vifpatnt (vii. 46. 3) — 
and adds a verse in explanation of their etymology, as follows : vif patir 
vifpatnt yasya patir vifvasya vifpatih : vafabdo lupyale paty&u vif Am 
va patir vifpatih. This gives us our option as to whether we will take 
vif pati to represent vifvapati or vif&m pati : we shall not be slow to 
choose the latter. The indivisibility of the compound is doubtless 
owing to the rarity of the consonantal conjunction fp, and the embar- 
rassment which would accompany the restoration of the sanhita form 
from a ^arfa-readiug vit-pati. 

S^lrTt ST d*m«^ llUll 

61. Nor is the root dd separated when it begins with L 

We have given us once more, under this rule, the whole series of de- 
rivatives presenting the root da reduced to the form of a simple / which 
the commentary to iii. 11, above, presented, and of which aprattttam 
(vi. 117. 1) and partttah (Vi. 92. 2) are the only ones fonnd to occur in 
the Atharvan. The difficulty of making out an acceptable analysis of 
them for the pnda-tcxt is reason enough for their being treated in that 
text as indivisibles. 

The Vaj. Pr. marks partita as indivisible at v. 45. 

3^r ^Ph<^(h^iw[^ii^h 

62. Nor the roots han, har, sthd, and stambh, after the preposi- 
tion ud. 

For the combination of han with ud, the commentator cites uddha- 
tah; no such word, however, is to be found in the Atharvan, nor does 
any other combination of these elements occur there (except at xiv. 2. 
16, where the preposition is separated from the root by the intervention 

546 W. D. Whitney, [iv. 62- 

of other words). For ud + har, the selected instances are uddhrtd and 
uddhriyamdnd (both xii. 5. 34), and uddhrteshu* (xv. 12. 1). The pada- 
text, however, appears to treat the combinations as inseparable only 
where there is actual composition, as in the participles, and as would 
also be the case if the unaccented preposition preceded the accented 
verb, for we find ut : hara in three passages (iv. 14. 7. ix. 6. 19. xii. 3. 
36). For sthd with ud, two cases are cited, viz. utthdtuh (ix. 4. 14) and 
utthitah (e. g. vi. 43. 2) : it has already been noticed (under ii. 18) that 
•where the preposition would be, by the general rules of combination, 
disjoined from the verb, it is actually so disjoined, and that the pada 
accordingly has ut : stkuh, ut : sth&paya, etc., where the sanhita has ut 
thuh, ut thdpaya, etc. For ud+stambh is quoted the only example 
which the text affords, viz. uttabhitd (xiv. 1. 1). 

The Vaj. Pr. takes note of this class of cases at v. 38, but says noth- 
ing of the roots kan and har ; nor is any reason apparent why their 
compounds should be treated in this peculiar manner. One would have 
thought it especially desirable that the pada-text should separate ut-hrta 
etc., in order to mark the forms as coming from the root har, and not 
from dhar. 

^riffi" ^r «£*hm{i ii ^ n 

63. Nor the root dkd, in a form beginning with h. 

The commentator illustrates with ye : dagdhdh : ye : ca : uddhitdh 
(xviii. 2. 34), and we have also uddhita at ix. 3. 6 : no finite verbal forms 
of this root as compounded with the preposition ud are found in the 
Atharvan. We meet, however, with uddhi once (viii. 8. 22), and our 
2>arfa-text leaves it undivided, although it does not fall under this rule, 
being composed of ud and dhi. 

The same rule in the Vaj. Pr. (v. 38) might cover both this and the 
preceding one of our treatise ; but no such forms as uddhita are there 
cited by the commentator. 

slIKJrMH « $ II 

64. Nor is jdspatyam divided. 

The commentator cites the only passage in which the word in ques- 
tion occurs: sam :jahpatyam (vii. 73. 10). This rule and one in the 
next section (iv. 83), taken together, show that the true pada reading 
recognized by our treatise is jahpatyam; our pada manuscript, how- 
ever, gives j&h-patyam, with avagraha. The commentator adds an 
attempt at an etymological explanation of the form : jay&patyam : yd- 
fabdo lupyate: patydu: asanturuskm&fu dvyakxharo jdy&h vd jdbhdvah. 
Although much corrupted, it is evident that this teaches the same ety- 
mology with that given by the Vaj. Pr. (at iv. 39) : jdspatya for jdyds- 

* Oar pada manuscript writes all these words with simple dh, instead of ddh : 
thus, udhrtd etc. 

iv. 67.] Atharva-Veda Prdtiqdkhya. 547 


65. Nor manushyat. 

The commentator cites the passage containing the word — ida ma- 
nushyat (v. 12. 8) — and adds an explanation, of its form, as follows: 
manushyavan manushyat : yapabdo lupyate vakarasya ca yakarah ; 
1 manushyat is properly mannshyavat: ya is dropped, and v converted 
into y.' It is unfortunate that, the Atharvan form of the word being 
thus fully established, and its treatment having been prescribed by the 
Praticakhya with so much care, it should have been altered in the 
edited text to manushvat, even though the latter is theoretically de- 
cidedly the preferable reading, and is presented by the Rig-Veda in the 
corresponding passage (x. 110. 8). 

66. Nor tredhd. 

This word, which our pada-text, like that of the Rig- Veda (and, I 
presume, the other Vedas also), always leaves undivided, is an exception 
under rule 1 3 of this chapter. 

The manuscript has a lacuna here, omitting at least the instances 
cited under this rule, the first statement of the one next following, its 
paraphrase, and perhaps a part of the illustrations belonging to it. It 
is impossible to say, of course, whether a rule or two has not dropped 
out also, affecting one or more of the words which I have introduced 
into the gana of rule 54 ; but this is not at all certain, nor would the 
loss be of much consequence, considering the quality of the rules in 
this part of the section. 

67. Nor a specific appellative. 

The term saihjnd is evidently used by our treatise in the same sense 
as] by Panini (see Bohtlingk's glossary to Panini, sub verbo) and the 
Vaj. Pr. (iv. 96) : it might be tolerably rendered by our term " proper 
name." The commentator's illustrations — which, as remarked in the 
preceding note, follow immediately upon the paraphrase of rule 66, 
and are perhaps therefore defective — are apvatth&h : nyagrodhah (iv. 37. 
4 : cf. V. Pr. v. 37), kapyapah (e. g. iv. 37. 1), and vipv&mitrah (xviii. 3. 
15). He adds: bahulam iti ca vaktavyam ; 'it should have been said 
that with regard to samjna usage varies ;' and he gives, as instances of 
proper names which are separable, jamadagnyalharvana (not in AV.), 
jamat-agne (xviii. 3. 16), bharat-vajam (iv. 29. 5), para-para (vi. 65. 1), 
and v&ma-deva (xviii. 3. 16). The amendment is made with exceeding 
good reason, for the rule is absurdly comprehensive in its form of state- 
ment. It can only be said with truth that the being a saihjnA is a cir- 
cumstance which rather favors non-division, helping to excuse the pada- 
text from attempting the analysis of an obscure or anomalously formed 

548 W. D. Whitney, [iv. 68- 

3F?t II \* II 

68. Nor is the root vyadh separated. 

Compounds with this root have already been made the subject of one 
of our rules (iii. 3), and it has been there explained that the usage of 
our pada-text is to leave undivided such of them as show a protracted 
vowel before the root. The commentator cites here two of the three 
instances which the Atharvan offers, viz. hrdayavidham (viii. 6. 1 8) and 
marm&vidham (xi. 10. 26). The rule is too broadly stated, and should 
have been restricted by him, as was the preceding one : it is only when 
a protracted vowel precedes the root that the compound is left undi- 
vided; and we have, for instance, vi-vyddhin, abhi-vyddhin (both i. 19. 
1), and hrta-vyadhani (v. 14. 9). 

Vfit tfcMUkl^kM II ^11 

69. Nor the root drg, when compounded with a pronoun end- 
ing in a or i. 

The form of this rule is in one respect very unusual : such a thing as 
the fusion into a diphthong of two vowels of which the specification is 
desired is elsewhere unknown. If the reading were slightly amended, 
to sarvandmnekdr&nlena, it would answer all the purposes of a rule of 
our Pratigakhya, for the Atharva-Veda presents only a single one of 
the compounds which it appears in its present form to contemplate, 
viz. idrp (e. g. iii. 1. 2). The commentator, however, paraphrases as I 
have translated, and gives the instances t&drk, t&drfah, yddrk, yddrpah, 
idfk (iv. 27. 6), and idrpah. 

The Vaj. Pr. (v. 37) instances idrn and anyddrn among indivisible 

<fl^T5TT3% II oo ii 

70. Nor the root sah, when it ends in dt. 

Under this rule, the commentator gives us once more the same series 
of compounds of sah which we have had above, under rules ii. 82 and 
iii. 1, and which it is unnecessary to repeat here. 

Compare Vaj. Pr. v. 30. 

4|ot(elMM II u\ II 


71. Nor are indeclinables divided. 

As examples of indivisible indeclinables, the commentator offers us 
sanutah : yuyotu (vii. 92. 1), pr&tah (e. g. iii. 16. 1), uccdih (iv. 1. 3), 
uccdt (uccd, xiii. 2. 36), ntcdih (e.g. iv. 1. 3), and nic&t (nied, e. g. i. 21. 
2). The rule does anything but credit to the acnteness of the authors 
of the Pratigakhya, for no word in the text which would otherwise be 
entitled to avayraha is left unresolved oh account of its being an inde- 

iv. 74.] Atharva- Veda Prdtigdkhya. 549 

$TT5[TT f^% II o^ II 

72. Nor <%<£, when it means ' region.' 

The word d'pd, meaning ' region,' comes from the root ap, and fur- 
nishes no ground for a divisiou by the pada-text: dpd', on the other 
hand, meaning ' hope, desire,' is a later form of dpas, and comes from 
the root pans, with prefix d ; hence it is divisible. The commentator 
cites the words and. phrases d'pdbhyah (x. 5. 29), d'pdndm (i. 31. 1), 
d'pdh : dnu (vii. 9. 2), and finally, by way of counter-example, abhi-dhd- 
vdmi : d-pd'm (vi. 1 19. 3). 

The signature of the section is merely caturlha&ya dvitiyah pddah. 

BflH^M HHWfri: II ©S II 

73. Eestoration is exhibition of the natural form. 

This is simply a definition of the term samdpatti, which I have ven- 
tured, instead of transferring, to translate by 'restoration.' It means, 
as the next rule will show, the reinstating, in the pada and krama texts, 
of that form of a word which is looked upon as the original and normal 
one, to the rejection of the anomalies of Vedic orthoepy. It does not 
occur in any other of the grammatical treatises, although its corres- 
pondent sam&pddya (see below, rules 117, 124) is once found in one 
of the later chapters of the Rik Pr. (xiii. 11, 12), in a passage so ob- 
scure, without the light which the treatment of the subject in our own 
Pr&tic,akhya casts upon it, that its meaning has, very naturally, been 
misapprehended by the learned editor. 

tHNirl': 110811 

74. In the repetitions of the pada and krama texts, restora- 
tion of the natural form is made where s has been converted 
into sh, n into n, visarjaniya before k and p into s, where a vowel 
has been lengthened, t or th made lingual, an element omitted, 
or final n converted into visarjaniya. 

Most of the technical terms of this rule meet us here for the first 
time, and several of them are not employed elsewhere in our treatise. 
Cared (see iv. 123) designates the repetition, with ili interposed, made 
in the pada text of a divisible compound which is also prayrhya, or 
which ends in a vowel not subject to the ordinary rules of combination : 
for example, xalydnrte Hi sutya-anrte (i. 33. 2) ; parihdra (see iv. 1 17) 
is the like repetition made in the krama-text of a praqrhya, a divisible 
compound, a word requiring restoration to its natural form, and the last 
word before a pause. The former term is employed in a like sense by 
the Vftj. Pr. (e. g. iii. 1 9) ; the latter is peculiar to our treatise, being 
replaced in the others by pariyraha and stkilopasthita. Updcdra, ' the 

550 W. D. Whitney, [iv. 74- 

conversion of visarjaniya into a sibilant before k and p J (by ii. 62 etc.), 
corresponds to the updcdra and updcarita of the Rik Pr. (xiii. 12, iv. 
14), and dnpada is the term employed by the same treatise (iv. 27) to 
designate the conversion into visarjaniya, and consequent loss, of a final 
» before a following vowel, as taught in our rule ii. 27. Shatva, natva, 
and tutva are of obvious derivation and significance, nor is there any- 
thing calling for remark in their form, excepting the w in tutva, which 
identifies the term with a Paninean symbol (viii. 4. 41 : shtunA shtuh*). 
A quite embarrassing question now presents itself, in connection with, 
the part of the text contained in this and the following rules ; namely, 
with reference to the constitution of the pada-text which they imply. 
The actual pada-text of our manuscripts is very sparing in its use of 
cared, or repetition with iti interposed : it avails itself of that expedient 
only in the case already referred to as prescribed by iv. 123, or when a 
pragrhya is likewise avagrhya. The Rik pada employs it in one addi- 
tional case ; namely, when a word ends in a visarjaniya which is riphita, 
or liable to pass into r before a sonant, but which does not actually be- 
come r in the sanhild : it would read, for example, at ii. 32. 1 (where the 
Atharvan pada has simply antdh), antdr ity antdh. The Vajasaneyi- 
Sanhita is, according to the rules of its Pratieakhya (iv. 17-22)— with 
which, in the absence of any testimony from Weber to the contrary, we 
must suppose the usage of the known manuscripts to correspond — very 
much more liberal in its employment of the repetition; not only in the 
two cases where this is practised in the Rik pada, but also in the case of a 
simple pragrhya (thus it says dve iti dve, where Atharvan and Rik 
would give simply dve iti), in that of a word which contains a length- 
ened vowel or a lingualized consonant, and even in that of a mere di- 
visible compound, it performs cared. In short, it repeats in pada-text 
all that is repeated in krama-text, excepting (by iv. 21) su and the final 
word of a sentence. The precept of the Vaj. Pr. corresponding to this 
one of ours is to be found implied in iii. 18, 19, where direction is given 
that in the repetitions of the pada-text the remaining rules of the chap- 
ter should not be observed — they being precisely the ones which teach 
the changes which our precept specifies in detail. Now when we find 
put forth in our treatise, as its leading and principal direction for the 
restoration of the natural form in pada, a rule like the one here given, 
which classes pada repetitions and krama repetitions together, and cor- 
responds, as regards the pada, so nearly with the Vaj. Pr., we cannot 
help suspecting that it contemplates a pada-text in which, as in that of 
the Vaj. Sanhita, the repetitions of krama and pada extend over nearly 
the same classes of cases. It is actually the fact that, if we allow the 
^ada-text to be of the form in which our manuscripts give it, there are 
but about half a dozen words in the whole Atharvan text to which this 
rule and the two following, all together, have any application : while, on 
the other hand, the Pratieakhya is found to give no direction at all for 

* Shtu and shatva are also used by the little Jcrama-tre&tise belonging to the Rig- 
Veda, and called the Upalekha (Upalekha, de kramapatha libellus. Textum Sans- 
criticum recensuit, vanetatem lectionis, prolegomena, versionem Latinam, notas, 
indicem adjecit Dr. Guil. Pertsch. Berlin: 1854. 8ro), to which we shall, in the 
sequel, have frequent occasion to refer. 

iv. 75.] Atharva-Veda Prdtiqdkhya. 551 

the use of Hi alone in pada after a pragrhya, or for the innumerable 
restitutions of natural form which are made in words not repeated. I 
find myself, I must acknowledge, hardly able to avoid the conclusion 
that this part of our Pratigakhya was framed to suit a pada-text in 
which all pragrhyas, divisible words, and words requiring restoration to 
normal form, were alike repeated, or suffered cared : such seems to me 
to be the only intelligible and consistent interpretation of its rules. 
That the fourth section of the chapter contains a direction for cared 
agreeing with the nature of our extant pada-text, would find its expla- 
nation in the evident character of that section as a foreign addition to 
the main body of the work; we should have to assume that the school 
to which the treatise as a whole belonged, in its present form, framed 
its pada-text in the manner there taught, and probably suffered that rule 
to take the place of one of another character formerly contained in this 
section, and now omitted from it; while yet they did not so recast the 
section as to adapt it fully to their new method of construction of the 
pada. This may seem a violent and improbable supposition ; but it 
appears to me, after making every possible attempt to avoid it, to in- 
volve less difficulty than the interpretation of the rules of this section 
in such a manner as to make them suit the pat/a-text of the manuscripts. 
The true illustrations of our rule, then, would be of the nature of the 
following: for the conversion of * to sh, in vasosh pate (i. 1. 2), vasor iti 
vasoh; in vidmo shu (i. 2. 1), sv iti su ; in vy ashohanta (iii. 10. 12), 
asahante 'ty asahanta : for the conversion of n to », in pari nah (i. 2. 2), 
na iti nah; in prd 'naikshtt (ii. 7. I), an&ikshid ity andikshlt: for the 
conversion of visarjaniya to s before k and p, in tatas pari (i. 10. 1), 
tata iti tatah; in tokebhyas krdhi (i. 13. 2), tokebhya iti tokebhyah: for the 
lengthening of a vowel, in vidma parasya (i. 2. 1), vidme 'ti vidtna ; in 
ydvaya (i. 2. 3), yavaye Hi yavaya : for the lingualization of dental mutes, 
in bahinh te (i. 3. 1), ta iti le; in vi taxhthe (ix. 10. 19), tastha iti tasthe: 
for omission of an element, in ut thuh (vii. 52. 2), sthur iti sthuh: for the 
conversion of final n to visarjaniya and its consequent omission, in 
mahdn asi (i. 20. 4), mah&n iti mahdn. 

One other solution of our difficulties, less satisfactory, but also less 
violent, deserves to be suggested. If we could omit the words carcd- 
parikdrayoh from the rule altogether, leaving the latter to authorize a 
restoration of normal form in the pada generally, we could perhaps 
make shift to get along with such inconcinnities and omissions as would 
still remain — of which the principal would be that the treatise made no 
provision for the use of iti after a pragrhya word, and that it did not 
direct what form words should have in the numerous repetitions of the 

The commentator, offering no explanation of the rule, gives a series 
of compound words in illustration of it, which belong more properly 
under the following rules ; and to the next, accordingly, I shall take 
the liberty of relegating them. 

^^PliHrlHf STllttMjl 

75. And also, where the cause of the conversion stands in a 
former member of a compound. 
vol. vii. 70 

552 W. D. Whitney, [iv. 75- 

The commentator's paraphrase is simply p&rvapadanimiltdndm ca «Aa- 
tvddindm samapattir bhavati ; ' and restoration is made of the conver- 
sions detailed in the preceding rule when their cause stands in a former 
member of a compound.' He cites no examples, but says etdny evo 'dd- 
harandni, ' the illustrations are those already given :' namely, under the 
preceding rule. According to his exposition, then, the present rule 
would seem merely an explanatory appendage to its predecessor. But 
this is clearly inadmissible : not only onght we to have it, in that case, 
combined with the other, so as to form part of it, but, more especially, 
it would not contain the particle ca, ' and,' which positively stamps it as 
something added to the other. We cannot avoid, as it seems to me, 
understanding rule 74 of the abnormal changes of disjoined and inde- 
pendent words, and rule 75 of such as are produced by an altering influ- 
ence in the prior member of a compound. The illustrations which the 
commentator offered under the other rule, and which do, in fact, in good 
part appertain to this, are as follows : conversion of s to sh, nhheranam : 
nuecanam iti ni-seeanom (i. 3. 1 etc. : our pada, simply ni-secanam) ; con- 
version of n to n, pardyanam : pardyanam iti pard-ayanam (e. g. i. 34. 3 : 
p. pard-ayanam) ; conversion of visarjaniya to a sibilant, adhahpadam : 
adhahpadam ity adhah-padam (e. g. ii. 7. 2 : p. adhahpadam) ; prolon- 
gation of a vowel, abhivarlena : abhivartene 'ty abhi-vartena (i. 29. 1 : p. 
abhi-vartena) ; conversion of dental mute to lingual, yo visktabkndti : 
vistabhndti Hi vi-stabhndli (xiii. 1. 25 : p. vistabhndti) ; omission, $epa- 
harshanim : pepoharnhanim iti pepah- harshanim (iv. 4. 1 : p. pepah-har- 
shanlm : see above, ii. 56) ; and loss of final n, sdldvrkdn iva : sdld- 
vrkdn ive Hi sdldvrkdn-iva (ii. 27. 5 : p. sdldvrkdn-iva). The comment- 
ator does not state whether he takes his instances from the pada or from 
the krama text : according to the construction of our present pada, they 
could only come from a krama ; if the conclusion drawn above as to 
the original pada contemplated by our text is correct, they may be illus- 
trations of both. In the very rare cases in which the extant pada-text 
has occasion to repeat words showing any of the abnormal changes 
which the rule mentions, it restores the normal form : thus we have 
dustano iti dvh-tano (iv. 7. 3 : s. dushtano), pratisthe iti prati-sthe (iv. 26. 
1, 2 : s. pratishthe), dyushpatni* ity dyuhpatni (v. 9. 8 : s. dyuskpalni), 
vistabhite iti vi-xtabhite (x. 8. 2 : s. vishtabhite), and pathisadi iti jmthi- 
-sadi (xviii. 2. 12 : s. pathishadi). 

The commentator adds a couple of counter-examples — viz. parirdpi- 
nam Hi pari-rdpinam (v. 7. 2) and sutrdmanam iti su-trdmdnam (vii. 6. 
3) — to show that, when the effecting cause of an alteration of form is 
in the same member of a compound with the alteration itself, the latter 
is not reversed, and the normal form restored, by the repetition and 
resolution of the word. 

* Our pada-MSS. -write the word as I have given it, apparently infringing the 
rule; but I have no question that the th here is only an attempt to represent the 
labial spirant, or upadhmdniya, which the theory of tho Prfitic&khya requires (by 
ii. 40) ill such a place: another like case is chandnnhpakuhe iti chandaKpok*he (viii. 
9. 12 : s. chandahpak*h4—or, more properly, chandappnkxhe — by ii. B2). Before the 
iti, where no pause of avagraha intervenes between the two members of the com- 
pound, they are, of course, to be put in simple mndhi with one another: thus, dm- 
tauo. fepohnrxharfim, and, as we ought strictly to read, adhahpadam; we have also 
tadohavirdhdne iti sadah-havirdhane at xii. 1. 38. 

iv. 77.] Afharva'Veda Pr&tiq&khya. 553 

76. And where the compound is divisible. 

That is to say, restoration to the normal form is made only in such 
compounds as are by the pada-text resolved into their constituent ele- 
ments. Those words which, although they may be acknowledged com- 
pounds, are left undivided on account of special anomalies of form, 
retain also their irregularities of orthoepy. The commentator, in his 
paraphrase, represents ingydndm by ingyamdndndm avagrhyamdndndm, 
'forms which undergo division, or separation by avagraha] and adds 
again etdny evo 'ddharandni, 'the examples are those already given' — 
namely, under rule 74 (here presented under rule 75). Of counter-ex- 
amples, however, he furnishes two — viz. parishkrtd (e.g. ix. 3. 10) and 
prananli (e. g. i. 32. 1) — and then cites a verse which contains two 
more: aningyatvdt samdpattir eshu nelapadeshu tu: utpanne 'vagrahe 
cd Hra tsamdpattis tathdi 'va ca : mnrtavad apdshthavad iiy udaharet. 
The commentator's own instances belong to the class of those in which 
a cause in the former member of the compound produces an effect in 
the latter member : the words, if divided, would read pari-skrla, pra- 
-ananli: in the other two, the irregularities are mainly in the first 
members themselves, and, if s&nrta and apdshtha were resolvable, we 
should read (according to the next rule), with restoration, sunrtd-vat, 
apaslha-vat, instead of, as now, sunrtd-vat (e. g. v. 20. 6), apdshtha-vat 
(xiv. 1. 29). The rule, as these illustrations help to show, is not a mere 
additional specification to the one preceding, affecting only the cases to 
which the other applies : in that case it would have been incorporated 
with it, not made to follow it, as an independent precept ; but it concerns 
all changes occurring in the interior of divisible words, whether in the 
former or the latter member, and a part of the commentator's examples, 
rehearsed under rule 75, belong to it, and not to the latter. 

^Hlfa ^W ll oa n 

77. In which, case restoration is made, even when the word is 
farther compounded with another member. 

That is to say : a compound which, being divisible by avagraha. is 
entitled to restoration of the normal form of its constituent parts, 
retains its right even when, by farther composition, the division of its 
original members is lost. Examples are given in the commentary as 
follows: vixita-srupah (vi. 60. 1 : s. viMtassrupah), abhi-rtihpalan* : api- 
patat (vii. 64. 1 : s. abhinishpatan), visthit&h-iva (vii. 1 1 5. 4 : s. vishthit&h- 
-iva), brhaxpati-pranutfdndm (viii. 8. 19 : s. °pran»ttdnam), prshadAjya- 
-pranuttdndm (xi. 10. 19 : s. as before), and durnihita-eshhiim (xi. 9. 15 : 

* Our manuscript writes abhinisfipatan, as do also the manuscripts of the Athar- 
van padatvxl in the passage cited ; but I suppose here, as in the other similar cases 
referred to in the note to rule 75, that the »h is an attempt at representing the 
labial spirant : we have the guttural spirant, the jihvdmuliya, in like manner repre- 
sented by sh in abhi-nukkfta (x. 1. 12) and abhi-nUhkdrin (x. 1. 31). 

554 W. D. Whitney, [iv. 77- 

s. durnihitaishintm). Other instances afforded by the text are vitkan- 
dht-dmhana (ii. 4. 1), atisthd-vant (iii. 22. 6), su-praniti (e. g. v. 11. 5), 
durndma-cdtana (viii. 6. 3), anu-visiryate (viii. 10.33), abhimoda mud (xi. 
7. 26), jdgrat-duhivapnyam and svapne-duhsvapnyam (xvi. 6. 9), prlhivi- 
sat-bhyah (xviii. 4. 78), etc. Three exceptions to the rule are made 
below, in rule 1)6, and the text affords one more, as is there pointed 
out in the note. The commentator again adds a verse, but it is more 
than usually mutilated and obscure; it reads: prakrtyd manatvaih yad 
avagrhyet tathdi J va tat: vpatuhthanti prapanddiny uddharet. 

5fJ% qr^TTT f%JJ^nr£ II otj h 

78. In krama, restoration is made of a word which is taken 
together with another word than the disjoinable cause of its 
altered form. 

The commentator's paraphrase is krame parena prasamdhdne vigrhydn 
nimitt&t ; which shows us — what the necessities of the case would of 
themselves have pointed out — that the important word to be supplied 
with vigrhydt is, by inference from rule 75, nimitt&t, ' the cause of the 
altered or abnormal form.' Vigrhya denotes a word which is altogether 
independent, and therefore disjoined from others in the pada-text, a nd- 
ndpada, in distinction from avagrhya, which means ' divisible into its 
constituents {pkrvapada and utlarapada), as a compound.' In the con- 
struction of the £ni?na-text, then, where each word is in succession 
taken along with its predecessor and its successor, a word which in 
sanhitd has an abnormal form, under the influence of the former or of 
the latter, retains that form when in the same kramapada with the alter- 
ing word, but is restored to its natural form when making a kramapada 
along with any other word. The commentator cites a couple of pas- 
sages — dpo hi shthd mayobhuvah (i. 5. 1) and pari no vrndhi (vi. 37. 2) 
— but does not write them out in krama, form, so as to illustrate the 
rule : they would read dpo hi : hi shtha : sthd mayobhuvah, and pari 
nah : no vrndhi. As counter-examples, to show that restoration of the 
normal form is made in a kramapada only when the cause of euphonic 
alteration stands in a separate word, and so is left out of the krama- 
pada, we have given us two passages in their krama-form : prthivydm 
te : te nishecanaih : nishecanam bahih : nisecanam iti ni-secanam (i. 3. 1 etc.), 
and dyane te : dyana ity d-ayane : tepardyane .-pardyane durvd -.pardyana 
iti pard-ayane (vi. 106. 1). Here the sk of nishecanam and the » of 
pardyane are maintained wherever the words containing them enter 
into a kramapada, and only suffer restoration (by rule 75, above) to * 
and n in the repetition or parihdra. 

The corresponding rules in the other treatises are Rik Pr. x. 5 (r. 5), 
xi. 21 (r. 44), and Upalekha iii. 3, 4. I do not find in the Vaj. Pr. any 
special direction upon the subject. 

^for fsqr% u i^ ii 

79. A lengthened vowel is restored to its natural form before 
a pause. 

iv. 82.] Atharva- Veda Prdti'gdkhya. 555 

The illustrative passages are given by the commentator in pada form : 
as, dpah : hi : slha : mayah-bhuvah (i. 5. I : s. shlhd), and parva : axya : 
grabhitd (i. 12. 2: s. asyd). The rule, however, evidently applies not 
less to the krama than to the pada text, and is even intended chiefly for 
the former : it is our authority for shortening a protracted final when 
it comes to stand at the end of a kramapada, while it is left long when 
taken together with its successor : we read At shtha : sthd mayobhuvah, 
and parva '»yo : asyd grabhitd. 

80. In caturdlra, this is done only before the pause of separa- 

From rule 74, which prescribes restoration of the normal form of a 
lengthened vowel in both parts of a repetition, one might draw the con- 
clusion that the word here in question should be written, when repeated, 
caturalra Hi catvh-rdtrah : hence this rule, which teaches the reading 
calurdtra iti catvh-rdtrah. Our commentator cites, in &rama-form, the 
passage containing the word, caturdtrah paficardtrah : caVdrdtra iti ca- 
tuhrdtrah (xi. 7.11). 

q^Trrf%f FIFTFT II *\ II 

81. Restoration is made of alterations taking place at the end 
of a word. 

The commentator's paraphrase is paddntaviirtdndm ca shatvddindm 
samdpattir bhavati, which would seem to show that he understands the 
rule as referring to the same series of abnormal alterations which was 
detailed in rule 74. His illustrations, however, put quite another face 
upon the matter : they are pari-eti : rakshan (iv. 38. 5) and abhi-dimi : 
devdh (vi. 118. 3). Here the only changes of form which have under- 
gone restoration are the regular conversions of i into y (by iii. 39) before 
the following dissimilar vowel. We are thus guided to a different in- 
terpretation of the rule : whereas we have heretofore dealt with irregular 
or abnormal changes only, learning under what circumstances, in pada 
and in krama, they become reversed, and the original form restored, 
here we are taught that all alterations made at the end of a word, by 
the ordinary as well as the extraordinary combinations of the phrase, 
undergo restoration when the word comes to stand, in pada or in krama, 
before a pause (virdme, rule 79). It should be remarked that the final 
repetition of this rule is wanting in the manuscript, and that we cannot 
therefore be certain that we may not have lost with it other examples 
and farther exposition, which would have set the meaning of the rule, 
or the commentator's apprehension of it, in a clearer light. 

wwRmhmi ^li^ll 

82. Also of forms lingualized by the influence of a redupli- 

556 W. D. Whitney, [iv. 82- 

The Pratic&khya now goes on to inform us where restoration must 
be made of alterations which have taken place in the interior of a word, 
and not under the influence of any cause lying outside of the word 
itself. The rules in this portion of the work are in great part the reverse 
of others formerly given, when the subject under treatment was the con- 
version of pada into savhitd. Thus, the present precept is the correlative 
of ii. 91, and it is illustrated by precisely the same series of examples; 
which, however, are here given in the pada form : thus, swrkdata (i. 26. 
4), abhi : sisyade (v. 5. 9), A : susvayanti (v. 1 2. 6), sisasavah : sis&satka 
(vi. 21. 3), sisdsati (xiii. 2. 14), and susuve (xiv. 1. 43). 

83. Also of slrdis/ittyam, ndrshadena, dushtaram, traishtubkam, 
trdih&yandl, and jdspatyam. 

By rule 76, above, no compound was declared entitled to restoration 
of the natural form of its constituents, unless it was by the joorfa-text 
treated as divisible. The words here detailed constitute exceptions 
under that rule, and have their irregular alterations reversed, even 
though (partly by rule 55, and partly by 54, above) they are not avce 
grhya. Our pada, in fact, reads str&is&yam (vi. 11.3), n&rsadena (iv. 
19. 2), dustaram (vi. 4. 1), trdistubham (ix. 10. 1 : we have also other 
forms from the same theme in the same and the following verse), and 
trdihdyandt (x. 5. 22 and xii. 4. 16) ; and jdhpatyam (vii. 78. 10) is pre- 
scribed by iv. 64, although, as there remarked, our parfa-manuscript 
actually gives jdh-patyam. 

WIHM t|^TdleJIH ii t$ ii 

84. Also of a reduplication, in a form of the perfect. 

The term paroksha, ' beyond the sphere of sight, out of one's sight,' 
is also employed by Panini (iii. 2. 115 etc.), along with bHta, 'past,' 
and anadyalana, ' not on the present day,' to define the proper sphere 
of the perfect tense. We may suppose it here used alone as a name of 
the tense as being its distinctive characteristic, since the imperfect and 
aorist are also entitled regularly to one or both of the other designations. 
The commentator cites, as instances, talrpuh (xi. 7. 13 : s. and 
vavrtuh (v. 19. 13 : s. vdvrtuh) ; and, as counter-instances, to show that 
the vowel is not shortened in any other tense than a perfect, he gives 
lalaplti (vi. 111.1) and rdrajiti (vi. 71. 2). The usage of the Atharvan 
texts as concerns the reduplication was fully explained under iii. 13, the 
only rule in which the subject is treated in the earlier part of the work. 

A couple of verses follow in the commentary, they read as follows: 
dbhydsasya ca dirghatvam dirghokiteti drpyate : na tasye 'shtd sam&pat' 
tir lalajkli nidarpmam: yady abkydsasga dirghatvam nvjddindm ca 
yardakoh: savarne ca paroJcshdydm na samdpadytte kvaxit. I have not 
succeeded in amending the text so as to be able to translate the whole 

iv. 88.] Atharva-Veda Pr&tiq&khya. 557 

eUcgblMUIrfrTf ^T II t\ II 

85. Also of vdvrdhdna etc. 

The commentator's instances under this rule are varrdhdnah-iva (not 
found in AV.), nasahih (iii. 18. 5 : s. sdsahih), and varrdhdnah (e. g. i. 8. 
4). The gana might be filled up from the material collected and pre- 
sented in the notes to the first section of the third chapter, but I have 
not taken the trouble to put it together, as it is uncertain how much 
and what the authors of the treatise meant the precept to cover. 

fi^faf^UIMH<&MIH II z\ || 

86. Also of the roots Jcrp, rup, and risk, when they are anahva. 

I can find nowhere any clue to the derivation and use of anahva, and 
the range of the cases to which the rule applies is too narrow for the 
induction with any confidence of a definition from them. For the root 
krp, either the commentator furnished no instances, or the manuscript 
has omitted them : the only derivative from that root, so far as I can 
discover, which the rule can have any concern with, is c&klpat (vi. 35. 
3 : p. caklpat) ; since caklpuh and caklpe would properly fall under rule 
84. For the root rup is cited na : rurupah (iv. 7. 5, 6 : s. rurupah) ; for 
the root risk, the two passages enasah : deva : ririshah (vi. 51.3: s. ri- 
rishah) and m& : nah : ririshah (v. 3. 8 : s. ririshah). The commentator 
then asks anahvanam iti kim artham, ' why does the rule say " when 
they are anahva;"' and cites, as counter-examples, na : amimadah : na : 
arurupah (iv. 6. 3), m& : ririshah : nah (xi. 2. 29 : this is, however, no 
counter-example, but precisely analogous with the two already cited for 
the same word), and sinivali : aciklpat (vi. 11.3). So far as these in- 
stances go, anahva might be understood as designating an aorist form 
which has lost its accent; or, virtually, an aorist subjunctive. 

The text affords one other word, fu$ucah (xviii. 2. 4 : s. f&fucah), of 
the same class with those treated in this rule. Its omission must be 
understood as signifying, either that the verse containing it was not in 
the Atharvan text of the authors and commentator of our Praticakhya, 
or that their text read, with the Rig-Veda (x. 16. 1), pocah, or, finally, 
that the word escaped their notice. 

sfl«£Ul«£H ll co II 

87. Also of jiMdd 'ham. 

The commentator cites the passage in its jrarfa-form, akratvh :jihlda : 
aham (iv. 32. 5). Compare rule iii. 14, of which this is the reverse. 

Hh^JIH ll w ll 

88. Also of sdhydma. 

The commentator cites the passage in its ^ocfa-form, sahy&ma : dasam 
(iv. 32.1). Compare the previous rule, iii. 1 5. He adds a verse or two : 

558 W. D. Whitney, [ir. 88- 

sahydmejfliya saheh dlrghatvam yad drpyate : na tasye 'shtd sam.apa.ttir 
yah pabdo dtrgha eva soh: akhyate 'ntahpade hrasvo na samdpadyate 
punah. The various irregularities of form appearing in, or in connec- 
tion with, the root sah have been the subject of several previous rules : 
see ii. 82, iii. 1, iv. 70. 

r^^T«IrJ II t;^ ii 

89. Also of didayat. 

In the former rule (iii. 22), didayat was made the leading word of a 
gana composed of forms exhibiting an irregular prolongation in the sec- 
ond syllable, and we are justly surprised at not finding the statement 
here made in a corresponding manner. The commentator, in fact, cites 
precisely the same cases as before, in their pada-forva — viz. didayat (iii. 
8. 3), ushasah : vira-vatih (iii. 16. 7), and ushasdnaktd (e. g. v. 12. 6) — 
just as if the rule read here also dtd&yddtndm. 

90. Also of ndraha etc. 

Here we have the precise counterpart of rule iii. 21, above, and the 
commentator cites from the text the same three cases, viz. narakam (xii. 
4. 36), sadanam (e. g. ii. 12. 7), and asatah : indra (viii. 4. 8). 

^JMMH: shll^HMW llh» 

91. Also of the root cyu, in a form containing the causative 

Under this rule the commentator is unusually liberal of his citations : 
they are & : cyavayantu : sakhydya (iii. 3. 2), yathd : vdtah : eyavayaii 
(x. 1. 13), ang&t-angat : pra : cyavaya (x. 4. 25), cyavayan : ea : vrkah&n 
(xii. 1.51), devatdh : cyavayantu (xii. 3. 35), and puxkd : tvd : itah : cya- 
vayatu (xviii. 2. 54). These are all the cases which the text furnishes 
of causative forms from the root eyu: in every instance, the sanJiild 
prolongs the vowel of the first syllable, reading cydvayanlu etc. 

92. As also of the root yu, if the form be a verbal one. 

The commentator cites three of the numerous examples of causative 
form3 from this root, having the long vowel of their first syllable short- 
ened in pada : they are variyah : yavaya. : vadham (e. g. i. 20. 3 : s. yd- 
vayd), asmat : yavayatam (i. 20. 2 : s. ydvayalam), and varuna : yavaya 
(i. 20. 3 : p. ydvaya). He does not explain the meaning of the re- 
striction dkhydte added to the rule, nor cite any counter-example. I 
can discover no other reason for it than the occurrence of the word 
yavaydvdnah, at ix. 2. 13 : this may have been deemed by the authors 
of the treatise to contain the causative ending (kdritdnta) aya, and 

iv. 95.] Aiharva- Veda Pr&tiq&khya. 55$ 

therefore to require the rule to be so framed as to exclude it. But the 
word is divided by the ^orfa-text yava-ydvdnah, as if composed of yava 
and yavan, from yd : and this seems the best account to be given of it. 

faMlfa II ^ n 

93. Also of the roots van, yam, gratli, and gldp. 

The cases referred to in the rule are cited by the commentator, as 
follows : am&m : sam : vanayantu (vi. 9. 3 : s. v&nayantu), vi : madhyam : 
yamaya (vi. 137. 3 : s. yamaya), madhyamam : prathaya (vii. 83. 3 : s. 
prathdya), and na : im : ava : glapayanta (ix. 9. 10 : s. gldpayanta). 

The manuscript contains no final repetition of this rule, but offers, 
after the last citation, the words ingyavac ca. What to make of these 
words I do not precisely know : they may be part of a cited verse, of 
which the rest, along with the repetition of the rule, is lost; or they 
may possibly belong to an omitted rule : but I can hardly suppose the 
latter to be the case, not seeing what the meaning of the phrase should 
be, as a rule or a part of one. 

The form of our rule 93, it may be remarked, is somewhat unusual : 
we should expect at the end of it the genitive plural ending: thus, 

*[FFV II \& II 

94. Ashta is not restored to its natural form. 

The commentator gives the same citations as under the corresponding 
rule above (iii. 2) : they are, in pada-form, ashtd-padt : caluh-akshl (v. 
19.7), ashtd-pakshdm (ix. 3. 21). ashtd-parnah, ashtd-donxht.ram (these 
two are not found in AV.), ashtd-yogdih (vi. 91. 1), ashtd-cakrd : nava- 
-dvdrd (x. 2.31), and ashtd-cakram : vartale (xi. 4. 22). He also inter- 
poses, between the first and third examples, ashtd-yonih ; but this is a 
blunder, for the word is read with a short vowel in both pada and san- 
hitd (viii. 9. 21), in our Atharvan manuscripts, nor is yoni mentioned 
(iii. '2) by the Praticakbya among words before which the final vowel 
of the numeral is made long. 

%ft^: n H n 

95. Nor the root hi. 

That is to say — wherever forms of this root, having the conjngational 
suffix mi or its modifications, show in sanhitd after pro, a lingual nasal, 
this nasal remains lingual also in the pada text. The commentator's 
examples are prati-prahinmah (x. I. 5). pra : hinomi : d&ram (e. g. xii. 2. 
4), and pra : hinuta : pilrn (xviii. 4. 40). 

Kul'e 88 of the preceding chapter is to be compared. The pada usage 

as regards these forms is quite anomalous : I can only conjecture that 

it may have been adopted in order to mark the euphonic alteration as 

itself of anomalous and exceptional character : there being, so far as I 

vol. vn. 71 

560 W. D. Whitney, [iv. 95- 

have been able to find, no other cases in which a preposition lingualizes 
the nasal of a conjugational sign. 

96 : nor is restoration made in the words here 


The first three of the words detailed in this rule are exceptions under 
rule 77, above, or cases in which the normal form is not restored to a 
divisible compound, on its being farther compounded with another 
member: the pada writes them bodha-praiibodhdu (v. 30. 10 : compare 
prati-bodha, e.g. viii. 1. 13), kesara-prdbandhdydh (v. 18. 11 : pra-bandha 
is not found in the Atharvan test), and abhi-agkdyanti (v. 6. 9, vii. 70. 
3: compare agha-yaniam, x. 4. 10). The last three are exceptions 
under rule 84, above, being forms of the perfect tense with short vowel 
unrestored in the pada-text, which writes them like the sanhitd, viz. 
dddhdra (e. g. iv. 2. 1), jdgdra (e. g. v. 19. 10), and mtm&ya (v. 11. 3). 
The text affords us once mimdya (ix. 10. 21), so that the rule is deficient 
in explicitness as regards this form, and should have cited along with it a 
preceding or a following word. The other two, panishpada (v. 30. 1 6) 
and atishthipam (vii. 95. 2), might be regarded as falling under the first 
general rule (iv. 74) for restitution of original form ; or they might as 
naturally, one would think, be looked upon as special cases, falling under 
no previous rule, and therefore not needing specification here. 

Of the class of the first three cases is sam-nishadya (iv. lfi. 2), which 
equally calls for inclusion in this rule, unless the reading in our pada 
manuscript is a copyist's error, and should be amended to sam-nisadya. 

97. Nor in prapana, provided only it comes from the root pan. 

The commentator cites the only two passages in which this word is 
to be found in the Atharvan, namely yena : dhanena : pra-panam : card- 
mi (iii. 15. 5), and punam : nah : astu : pra-panah (iii. 15. 4). I cannot 
in the least understand why any such rule as this should be deemed 
called for. There is no rule, and no principle, which should require the 
restoration of the n of prapana to a dental form, nor is there any word 
in the text which exhibits an element pana whose nasal is lingualized 
by a previous constituent of a compound. So far as we can see, it is 
merely the fear lest some one should be stupid enough to mistake the n 
for an effect of the preposition pra, and so should commit the blunder 
of speaking, in pada, pra-pana, that calls out the precept. Its repeti- 
tion before the one next following is wanting in the manuscript : possi- 
bly, then (as in the case of rule 81, above), we have lost something in 
the way of exposition or illustration which would have farther enlight- 
ened us. In his paraphrase, the commentator says prapana iti paratdi- 
raka samdpattir na bhavati; but what paratdiraka is, I do not know. 

iv. 100.] Atharva- Veda Pr&tiy&khya. 561 

98. Nor in idam & shu etc., on account of their forming to- 
gether (in Jcrama) a single word. 

The commentator paraphrases padatvdt by tripadatv&t, the latter 
being apparently a technical designation for those kramapadas which, by 
rule 113, below, are composed of three words, instead of, as usual, two 
only. The rule evidently applies to the fcrama-text alone; the pada 
reading of the passages referred to does not deviate in any manner from 
the usual norm : we have idam : un iti : su, etc. But what the point 
of the rule is, as concerns the fcrama-text, I find it rather difficult to 
see. The passages cited in illustration by the commentator are nearly 
the same with those already twice given, under ii. 97 and iii. 4 : they 
are idam u shu (i. 24. 4), tad u shu (v. 1. 5), pary u shu (v. 6. 4), mahtm 
ii shu (vii. 6. 2), anya u shu (xviii. 1. 16), and stusha u shu (xviii. 1. 37). 
According as the abnormal alteration aimed at by the rule is under- 
stood to be the prolongation of the u or the lingualization of the sibi- 
lant of su, we should add to the series the farther passages vii. 85. 1 
and xviii. 3. 7, or vii. 72. 2, 73. 7, and 85. 1. I presume that we must 
adopt the former of the two interpretations : the u, in these passages, 
is nowhere to be restored to its short form in the Jcrama, since it cannot 
fulfill the condition required by rule 79, and appear before a pause. 


99. Nor in brahmanvati etc. 

The commentator cites brahman-vatim (vi. 108. 2), papyat : akshan-van 
(ix. 9. 15), firshan-vali (x. 1. 2), and vrshanyanti-iva : kanyald (v. 5. 3). 
The irregularity which renders necessary the rule is the retention of the 
lingual n as final, against the principle of rule iii. 89, above. The last 
case cited, however, does not belong with the rest, since the denomina- 
tive ending, by rule 29, above, is separable ouly after a vowel, and we 
read vrahunyanlyah (vi. 9. 1 ) and vrshanyatah (vi. 70. 1-3), without ava- 
graha: hence there is no ground for restoration. 

<{taiM<PU<iVu ^ II \00 n 

100. Nor in dirghayutva etc. 

The same passages which were cited under the corresponding rule in 
another part of the treatise (ii. 59), and no others, are here again given 
by the commentator : they are, according to the reading of the pada- 
text, dirghdyu-tv&ya (e. g. i. 22. 2), sahasracaksho iti sahasra-caksho : 
tvam (iv. 20. 5), and barhi-sadah (xviii. 1. 45, 51). 

The signature of the section is as follows : 102 : caturthasya trtiyah 
p&dah ; so that, unless rule 53 is to be divided into two, or unless the 
copyist's count is inaccurate, we have lost, somewhere in the course of 
the second and third sections, one of the rules of the text. 

562 W. D. Whitney, [iv. 101- 

The concluding section of the treatise is occupied, first, with the 
recommendation of the study of the lemma form of the text, and sec- 
ond, with the description of its mode of construction. The way in 
which it is stated and explained by the commentator is altogether dif- 
ferent from that hitherto followed. First we have presented us the 
whole of the text of the section, separated by marks of punctuation 
into the rules which compose it. Then follows the independent state- 
ment and explanation of each rule in succession ; but not, as heretofore, 
according to the set method of restatement in paraphrase, brief and 
dry illustration by examples, and final repetition : we have, instead of 
this, a free exposition, drawn out at considerable length and with some 
unction, much more in the style of the known comments upon the 
other treatises of the class. This not only authorizes, but compels us to 
conclude that the remaining part of the commentary is by another hand 
than that which furnished the preceding. And the difference in style 
of the text itself no less justifies us in believing that the section was not 
an original part of our treatise, but is a later appendage to it. Whether 
or not it takes the place of another similar body of rules in the original 
Pratic&khya, and was substituted for them as being a fuller and more 
satisfactory exhibition of the subject, it would not become us to attempt 
to say too conficlentl}' : the near agreement of the preceding chapters 
in respect to extent (each containing not much more than a hundred 
rules) would favor the supposition that it had been tacked on as new 
matter to the treatise, carrying with it a new division of the preceding 
rules of the chapter into three instead of four sections : the subject of 
krama may have been formerly disposed of in a few brief rules forming 
part of the last section : but our Praticakhya has in too many of its 
previous rules made allusion to or implication of the Aroma-text (some- 
times even naming it and contemplating it. alone), to allow our assuming 
with plausibility that the construction of that text was not from the 
beginning one of the subjects with which the treatise dealt. 

The krama. is not treated by the Taitt. Pr. ; it is disposed of by the 
Vaj. Pr. in the closing rules of the final section of its fourth chapter 
(iv. 179-194), not ocenpying a whole section. In the Rik Pr. it takes 
up two chapters, the tenth and eleventh, each of which is by itself a 
complete krama treatise ; the former giving (in fourteen verses) a concise 
exhibition of the subject, the Utter (in thirty-seven verses) setting it 
forth with much greater fullness of detail. It is also the exclusive 
theme of the Upalekha, of unknown date and authorship, to which 
reference has already been made (see note to rule 74, above). The 
corresponding rules of all these authorities will be cited or referred to 
in connection with those of our own text. 

101. Study of the Veda is duty. 

In the comment we read three times, instead of the full form of the 
rule, vedd dharmah (or dharmam) simply ; but doubtless by a copyist's 
omission. The commentator adds to the rule the more detailed state- 

iv. 104.] Alharva-Veda Prdtiqdhhya. 563 

tnent karmafeshabhutdt : veda\dhyayanam\ dharmam ahur yajntkah — 
because sacrifices are performed by means of the Veda, and sacrifice is 
obligatory. He makes reference to a couple of verses or sayings which 
inculcate the necessity or advantage of sacrifice, thus : svart/alrdmo 
aghayat&m ity anena mantrena paldudauastakhyam [fataudandkhgam?] 
karma krlva srargam s&dhuyed iti ydjnikdmndnam. He then antici- 
pates and explains rule 104, below, concluding na vind reddir yojnas 
tdyate, ' sacrifice is not performed without the Vedas ;' and lie winds up 
his exposition with vedddhyayanam dharma ity ddits&tram, ' this is the 
first rule.' 

102. On the part of one desiring a condition of light after 

Prelya, literally 'having gone forward, having departed, deceased,' is 
by the commentator, with many words, explained to mean 'having 
quitted this world and gone on to another.' Upon jyotishtvam he dis- 
courses as follows : jyotirbkdvo jyotishtvam : uklam hi : ye vd iha yajnair 
drdhnuvans tesham etdni jyotinshi yany amuni nakuhatranfti: jyotir 
diptibh&vam ity arthantaram. The next rule he introduces by the ques- 
tion, " is it merely the reading that is duty f the answer is, no : how is 
it then ?" 


103. In the manner as handed down by those who understand 
the sacrifice. 

A ydjnika is defined as ' one who studies or understands the sacrifice ' 
(yajnam adhlte yajnam, vidur vd). Not merely the study of the Veda, 
but its study according to the traditional methods of those versed in 
sacred things, is declared meritorious. Amndna is defined by pathana, 
' reading,' and the commentator continues: "and how do the ydjnikas 
read ? ' with the verse svaryakdmo aghayatdm [already referred to, un- 
der the first rule of this section] one must secure paradise :' but it is 
objected 'that is an act of sacrifice, not a study of the Veda: hence 
merit is acquired by sacrifice, not by the study of the Veda :' this, how- 
ever, is not so" — as the rule which is next to be given is intended to 

thrirlirlH qsp%"^T: ll \o8 II 

104. There is no performance of the sacrifice without the 

Hence, as the Vedas are an indispensable aid to the performance of 
that in which duty consists, the declaration that their study is a duty is 
one to which no objection can be made (iti niravadyam vedadhyayanam 
dharma iti). 

664 W. D. Whitney, [iv. 105- 

^ 3H?Tl*|{ CTlrtfWrTT: II W» 

105. In the sacrifice, again, the worlds are established. 

I add the whole argument by which the commentator proves this 
pious proposition : dydvr viyad avanis trayo lokd yajne pratishthitah : 
kalham: nirvdpddisanskrtaih havir anffuahtaparvamdtrena pakayd 'vat- 
tarn antar nidhanasvdhakarana agn&u hutam jyotirdh&mabhdvena pari- 
natam jyotirbh&vena dydulokam, dh&mabhdvend 'ntariksham punar vrxhti- 
bh&vena parinalarh prlhivim y&li : evarh yajfie lokdh pratishthilah. The 
properly prepared sacrifice, duly offered in the fire, becomes light and 
smoke : the light goes to the sky, the smoke to the atmosphere, and, 
becoming rain, returns again to the earth : thus it reaches all the three 
worlds — and, if one chooses to look upon it in that light, establishes 
and supports them all. 

106. As are the five races in the worlds. 

The five races, the commentator says, are men : they are established 
in the worlds, the worlds in the sacrifice, the sacrifice in the Veda, and 
the Vedas are dharmapesliabhUtdh : dkarmdddivaiamatiddnpat karmani 
peshaldm ca gachanti: which last sentence is corrupt and obscure. From 
this, he goes on to point out the necessity of the study of the paeZa-text, 
and, as a help to it, of the Praticakhya, anticipating the rule which is 
next to follow: "the connection and distinction 1 of the appellation 
(abkidh&na) and the subject of appellation (abhidheya.) is not, without 
the study of the pada-tzxt; the recognition of the terminations is not 
assured, without the study of the pada-text : hence, in order to an un- 
derstanding of the mantra, its pada-text (pad&ni) mnst be studied ; and, 
by one who studies the pada. the Pratic&khya must necessarily be 
studied, in order to the resolution of doubts (avapyam saihpayacheddya 
pr&tipdkhyam adhyeyam) : and the uses of the study of the pada are 
farther set forth in the following rale." 

107. The study of the pada-text is for the sake of gaining 
knowledge of the endings, the beginnings, and the proper forms 
of words, and of their accent and meaning. 

The commentator explains and illustrates this rule at considerable 
length, and by means of examples which are for the most part taken 
from our present Atharvan text. First-, as he says, we are told that a 
rinsing of the mouth with water (udakdcamanam) is prescribed to be ac- 
companied by the pddas of the verse pain no devi (pain no devy&h pa- 
ddih : the verse is found at i. 6. 1 ) ; and here, without study of the pada, 
one fails to know that the first pdda ends with e (abhishtaye : in sanhitd, 

1 I — MS. iambadkdu antanam ca. 

iv. 107.] Atkarva-Veda Pr&licdkhya. 565 

it is abhishtaya). The next following example is intended to illustrate 
the difficulty, without the joarfo-text, of finding upon occasion the right 
beginning of a word : it reads rtubhyas tvd yaja ity drtavebhyas tvd 
yaja ity atrd "kdrddi na tu jndyate. The reference here is to iii. 10. 
10, rtubhyas tvd "rtavebhyah . . . . yaje ; eight separate recipients of 
offering are enumerated in the verse, and it seems intended that, in 
liturgical use, tvd yaje, which the verse gives once for all, should be 
appended to each separately ; this is intimated, though obscurely, by a 
prefixed direction : ashtakdydm : rtubhyas tve Hi 1 vigraham ashtdu. 
Again, " without study of the pada, the Vedic forms of words {ydidikdh 
fabddh) are not known : as for instance, afvavatim strdisiiyam, 2 etc. ; 
in sanhitd they have different forms, viz. apvdvaiim (xviii. 2. 31), strdi- 
sMyam 2 (vi. 11.3): therefore the pada-text must be studied (adhyeydni 
paddni)." The next point made is the necessity of pada, study to the 
understanding and right application of the rules respecting accentua- 
tion : " the brahmayajna etc. (? brahmayajndnddi) are directed to be 
made with the employment of the three accents (trdisvaryena) : here 
one who does not study the pada is unpractised (apravinaK) as regards 
the words : here, in the passages brahmdudanam pacati (xi. 1.1) etc., one 
is to speak not with the accents, but with monotone, at the pitch of 
acute (? tatra brahmdudanam pacati 'tyevamddishu 'ddttaprutyd ekafru- 
tyd td na svarena adhiyila) : now beware lest there appear here the 
fault of a mantra deprived of [its proper] accent. In the Atharvan rites, 
excepting the ydga, in the tyings on of an amulet, etc., in the performance 
of the sacrifice (? yajnavehe), the employment of the mantras is taught to 
be made with the use of the three accents." Finally, the assertion that 
study of the pada is necessary in order to the comprehension of the 
meaning of the text is supported with much fullness of illustration : as 
instances are cited vi hara (v. 20. 9 : this, however, may be no citation, 
but part of the exposition), alasdld 'si (vi. 1 6. 4 : we could wish that 
the pada actually taught us more about this obscure verse), yavdn ne 
'd ad an (vi. 50. 1), un iti (passim), sam v asnd 'ha dsyam (vi. 56. 3), 
tad v asya retah (ix. 4. 4) ; and farther, with special reference to the 
element of accent, ye asmd/kam tanvam (ii. 31.5), and svddv' admV Hi 
(v. 18. 1) : and the conclusion is "here, and in other instances, one who 
does not study the pada would spoil the sanhitd ; hence, for these rea- 
sons, the pada must be studied." I add the whole text, which in places 
is corrupt and obscure, and of which the value is too small to make an 
elaborate attempt at restoration necessary or advisable : mantrdrthaf ca 
padddhyayanad vind najndyate: vdkyam hi padapo vibhaktam anuvya- 
nakli : tac ca padddhydyt sandhim ca pade chedam tu faknuydd vibhak- 
tam .' vi hara : alasdld 'si : yavdn ne 'd addn : un iti : sam v dsnd 'ha 
dsyam: tad v asya retah: ityevamddishu sanhitdydm ca bhavati:rya- 
jati Hy atra sanhitikah sa pra kurydt: tathd uddtlasvaritodayena vighd- 
tam ajdnan : ye asmdkam tanvam : anyatrd 'pi nihanyeta : svddv admi 
'ti : atra ca svaritam kurydt: tathd uddttdntasya purvapadasyd 'nuddttd- 
ddv utlarapade tat (asya 'ntasthdpattdu svaritam akshamnarihe janasyar- 
the 'ty anyatrd 'pi tat kurydt: evam ddy anyatrd 'py apadddhydyi san- 
hit&m vindpayet : tasmdd ebhih kdrandir avapyddhyeydni : Mm ca : 

1 MS. rtumabhyastyeti. 3 MS., both times, trais&yam. 

566 W. D. Whitney, [iv. 108- 

108. The study of the krama has for its object the fixation of 
both sanhitd and pada. 

The trne reading of this rule is a matter of some doubt. Prefixed 
to the commentator's exposition, the manuscript gives simply sanhiid- 
ddrdhydrtltam, and the commencement of the comment implies or re- 
quires no more than that : it might seem, then, that we had here only 
an addition to the last rule, " the study of pada is in order to the fixa- 
tion of sanhitd." But this would be a lame conclusion to the argument 
of this part of the section, which must be intended finally to bring out 
the importance of the Aroma-text. And as the exposition closes with 
pretty clearly assuming as the full form of the rule under treatment 
that which is presented above, and s the prefixed text of the whole 
section so far favors the latter as to read sanhitdpadaddrdhydrtham, I 
think there can be little question that it is to be received as here given. 

The pada, the commentator tells us, must be studied for the sake of 
the establishment of the sanhitd. He defines ddrdhya by drdhasya 
bhdvah, and cites the rule of Panini which teaches its formation (var- 
nadrdh&dibhyah shyan ca, 1 Pan. v. 1. 123).' He goes on : " when here, 
in the sanhitd, a doubt arises, the student of the pada will give a solu- 
tion of the doubt. Then what need of a study of the krama? On 
this point, it says : 'the study of the krama has for its object the fixa- 
tion of both sanhitd and pada.'" What follows is corrupt, and only in 
part intelligible : sanhitd ca svasamsihd ca bhavati : yaitdmany ehtpada- 
dvipaddc ca pragrhy&vagrhyasamdeh&panodanam. Finally, he intro- 
duces the next rule by saying idam c& 'param kdranam kramddhyaya- 
nasya, ' and here follows another reason for the study of kruma; 

109. And the origination of accent is not seen in pada or in 

That is to say — as we are doubtless to understand it — in the pada we 
have before us only the accent of the uncompounded elements ; in the 
sanhitd, only that of the combined phrase : how the one grows out of 
the other is shown by the krama, which gives everything in both its 
separate and combined state. The commentator defines upajana by 
utpatli, and declares it unperccived in pada (padukdle), while it actually 
takes place in krama (kramakdle). As an illustration, he takes xvddv y 
admi' Hi (v. 18. 7): here, in pada, we have an oxytone and an unac- 
cented syllable, which form a circumflex, while in the sanhitd the cir- 
cumflex farther suffers depression (nighdta, the vikampita of our rule 
iii. 65, above), and the circumflex itself only appears in krama (in svddv 
admi, where the cause of depression of the svarita is not present) : 
hence, he concludes, the krama ought to be studied. He adds : " now 
comes the description : of what sort, it is asked, is this krama :" the 
following rules of the section will answer. 

1 MS. var^adfdhdditvdt : dhyaii : 

iv. 112.] Atharva-Veda Prdtigdkhya. 567 

110. Two words form a /fcra»ia*word. 

The commentator is very brief upon this rule : he says " the study 
of krama being now assured (prasiddha), two combined words form a 
single Arama-word ; their combination will be taught hereafter [in rule 
122], where it says 'according to the rules' (yath&fdstram)" 

The corresponding rules of the other treatises are Vaj. Pr. iv. 180, 
Kik Pr. x. 1 (r. 2) and xi. 1 (r. 1), and Up. i. 14. With the exception 
of the latter, they are more comprehensive than ours, including some- 
thing of what here is made the subject of following rules. The precept 
of the Vaj. Pr. covers our rules 110-113. 

111. With the final of this is made farther combination of the 
following word. 

The term antena is explained by avas&nena, ' close, end :' we might 
have rather expected the reading anlyena, ' with the last word of each 
krama-word as already defined. To parasya is supplied padasya, in 
the comment, as in the translation. The commentator takes the trouble 
to tell us that to the end of this following word is then to be farther 
appended its successor, and so on, so that one constructs the krama by 
thus successively combining the words of the text by twos. Were this 
rule not given, he says, the former one might be erroneously understood 
as prescribing that we should form our krama-words by taking first the 
first and second words of a verse, then the third and fourth, then the 
fifth and sixth, and so on ; while this shows us that we are to take the 
first and second, then the second and third, then the third and fourth, 
and so on. We may take, as an illustration, the last line of the first 
hymn of the Atharvan (i. 1.4 c, d), in constructing the krama-text of 
which only this simple and fundamental rule would come into action : 
it would read tarn prutena : prutena gamemahi : gamemahi ma : ma. pru- 
tena : frulena vi : vi radhishi : radhishi Hi radhishi (by iv. 117). 

The Vaj. Pr. and Rik Pr. combine this rule with the preceding : the 
TJpalekha (i. 15) states it separately, and in a distincter manner than our 
treatise : layor uttareno Htaram, padarh samdadhyat. 

HMilri tf^TIT HU^II 

112. A last word is not combined with its successor. 

By antagatam, literally ' a word gone to, or standing at, the end,' is 
meant, in verse, the closing word of a half-stanza, or one preceding a 
pause : in a prose passage, it doubtless indicates a word preceding one 
of the pauses of interpunction by which a numbered passage, or verse, 
is divided into parts. A pause, which interrupts the ordinary combina- 
tions of sandhi, interrupts those of krama also : there is no need that 
the Aroma-text should exhibit the euphonic connection of words which 
vol. vii. 72 

568 W. D. Whitney, [iv. 112- 

in sanhitA do not euphonically influence one another. The rule, as the 
commentator tells us, is intended to restrict the too great extension 
(atiprasaktam) of the one which precedes it. That the final word, thus 
left uncombined, suffers parihara, or repetition, is taught in rule 117. 
The corresponding rules in the other treatises are Vaj. Pr. iv. 180, 
Rik Pr. x. 8 (r. 9), 11 (r. 18), and xi. 21 (r. 44), and Up. i. 16. 

=ftfnr ?4<M4rtH^ifH imsu 

113. Three words form a &rama-word, if the middle one of 
them is a pure vowel. 

The term aprkta we have met with before (i. 72, 79), as used to 
designate a word composed of a single vowel or diphthong, uncon- 
nected with any consonant : the commentator, after exposition of its 
meaning, paraphrases it by avyanjanamiprafuddhakevalasvarah, 'a pure 
and entire vowel, unmixed with consonants.' He cites, as an instance, 
dhiyA : A : ihi (ii. 5. 4) : here the Jcrama reading is not dhiyA " : e 'hi, 
but dhiye " 'hi: to which, by rule 115, would follow again e 'hi, and 
then, by the present and other rules, ihy A nah : A nah : na ill nah. 
The only aprkta words which the text contains are the preposition A, 
the particle u (p. un ili: see i. 72, 73), and their combination o (p. o iti: 
see i. 79). It is doubtless to point out and call attention to this mode 
of treatment of the A in the krama-text, that our Atharvan pada manu- 
scripts quite frequently write a figure 3 after the word which follows 
it: thus, in the instance cited, the manuscript gives dhiyA : A : ihi : 3, 
at i. 1. 2, punah : A : ihi : 3, etc.* 

All the Arawia-systems have this feature : compare Vaj. Pr. iv. 1 80, 
181 (which calls such a lemma-word, composed of three members, a tri- 
krama), Rik Pr. x. 2 (r. 3), xi. 2 (r. 3), and Up. i. 17. The two latter 
authorities, however, except the compound o, and would have it treated 
like any ordinary word. The Vaj. Pr. is obliged to note (iv. 1 83), as 
farther instances of trikramas, mo shu nah and abhi shu nah, where, if 
the kratna were performed in the nsual way, the sanhitA reading of 
lingual n in nah would not be capable of exhibition; and like reasons 
compel it (iv. 184) to establish, in a few cases, krama-viords of four con- 
stituents, as iirdhva it shu nah. The Rik systems, also, are not a little 
complicated by the necessity of attending to such special cases occurring 
in their text, and which once cause a Arama-word to contain even five 
members. The fact that such complicated cases of sandhi do not happen 
to be met with in the Atharva-Veda saves our treatise the like trouble. 

^ri^^RSr^tf^ff^H Hi: yylsHH n \\8 u 

114. The grounds of this are the fusion of vowels into a sin- 
gle sound, the combination of vowels, prolongation, and lin- 

* The addition of the figure is usual, but not invariable, in the first books of the 
tert ; later, it is only made here and there. The figure is never inserted after u. 

iv. 115.] Atkarva-Veda Prdti$dkhya. 569 

The longer of the two £rama-treatises incorporated into the text of 
the Rik Pratigakhya is the only other authority which gives any reason 
why aprkla words should not be independently reckoned in construct- 
ing the krama-text. It says (R. Pr. xi. 2, r. 3) that the omission takes 
place " for fear of nasalization ;" that is, lest the particle should, if suf- 
fered to stand at the end of a krama- word, receive a nasal pronunciation. 
The entire, disagreement of the two explanations offered is noteworthy, 
and may be taken as an indication that neither is authoritative, and as 
a permission to us to find a better one, if we are able. It seems to me 
more likely that the weakness of the vowel-words a and w, unsupported 
by consonants, and their liability to disappear in or become obscured 
by the final of the word which precedes them, as if they were mere 
modifications of its termination, was the cause of their exceptional treat- 
ment. A similar suggestion has already been made (see under i. 73) as 
to the way in which the pada-text deals with u. The commentator's 
exposition of the rule is elaborate, but deficient in point. To illustrate 
the item ekddefa, ' vowel-fnsion, substitution of one vowel-sound for two 
or more others,' he takes dhiyd : a : ihi (ii. 5. 4) : here, he says, by the 
operation of the rule samdndksharasya (iii. 42), dhiyd and & become 
dhiyd : the d of the latter, combined with the i of ihi, becomes e, and 
the result is the one word dhiyehi: ergo, this is the way the combination 
must be made, otherwise there would be no krama-v/OTd (tasmdd ity 
anena samdhdnena bkavitavyam: anyathd kramapadam eva na sydt). 
The conclusion appears to me an evident non-seguitur, a mere restate- 
ment of the original proposition. For the second item, he selects the 
example ihi : d : nah (ii. 5. 4) : here, if we compound thy a and d nah, 
a vowel-combination (svarasandhih) is made of the i, by the rule svare 
ndmino 'ntahslhd (iii. 39). How this vowel-combination furnishes a 
ground for the krama-word ihy d nah, he does not attempt to point out. 
For the third and fourth items, the chosen instance is idam : un iti : su 
(i. 24. 4). This, too, is to be regarded as (in krama) forming a single 
word. The following text is corrupt, and I subjoin it, instead of at- 
tempting a restoration and translation : tasyd ca na samdhih : tathd hi : 
idamushv&dishv asanhitikam dlrghatvam : cared 'sya tripadasya madhya- 
bhavdd ishyate: idamfoshv ity evambhtitasydi 'va rupdkhyatdy&m ava- 
syam tripadam vanfakramena na bhavatitavyam : atrdi 'va 'padatvam : 
tad api tripadamadhydvayavam. The prolongation of the u in this and 
similar cases would indeed seem to furnish a reason for the construction 
of the kfama-v/ord out of three members, since the long vowel could 
not properly appear if the particle were made the final of one such 
word and the beginning of another; but I am unable to see how the 
lingualization of the sibilant should have any effect in the same direc- 
tion, since there would be no difficulty in reading u shu as a lemma-word, 
if the u were treated in the ordinary manner. 

115. A and o are made to begin a word again. 

That is to say, after d and o have been included, as middle members, 
in a triple &rama-word, they are again taken as initials of the word next 

670 W. D. Whitney, [iv. 115- 

following. The commentator's examples are gop&yatd " , sm&kam : & 
'smdkam (xii. 3. 55 etc.), dhiye " 'Ai : e 'hi (ii. 6. 4), and havir o shu : o 
shu : o ity o (vii. 72. 2). 

The same usage, as concerns d, is taught also by the other treatises : 
compare Vaj.Pr. iv. 182, EikPr. x. 8 (r. 11) and xi. 18 (r. 34), Up. iv. 13. 

116. ?7is merely to be repeated. 

The mode of repetition of the particle is taught in the next rule but 
one. This rule is, as the commentator explains it, intended to forbid 
the combination of u (like d and o) with the next following word to 
form a new krama-word (anyayoganivrttyarthah). As an example, he 
gives us sa u suryah : un ity un iti (xiii. 4. 5). 


117. ^Repetition with iti interposed, or parihdra, is to be made 
of pragrhyas, of words admitting separation by avagraha, of 
those requiring restoration to the natural form, and of those 
standing before a pause. 

The commentator simply expounds this rule, without bringing up any 
instances to illustrate it. The kinds of words specified are to be re- 
peated, or spoken twice, in the krama-text (kramakdle), the name of 
the double utterance being parihdra : and this parihdra is to be made 
with interposition of iti : having performed one of the two utterances, 
one is to say iti, and then repeat the word. 

The mode of repetition is, as has already been noticed (under iv. 74), 
called in the Rik Pr. by the related name parigraha (e. g. R. Pr. iii. 
14). The Vaj. Pr. (iv. 187) styles it sthitopasthita, which title is also 
known to and defined by the Rik Pr. (x. 9 and xi. 15). The Up. (iv. 1 2) 
knows only parigraha. The forms to be repeated are, according to the 
doctrine of the Vaj. Pr. (iv. 187-193), a divisible word (avagrhya), one 
in the interior of which appears a prolongation or a lingualization, a 
pragrhya, a riphita of which the r does not appear in sanhitd. and a 
word preceding a pause (avasdna). The first and the last three of 
these elasses are, indeed, treated in the same manner by all the other 
authorities (compare R. Pr. x. 6-8, r. 7-9, and xi. 13-14, r. 25 ; Up. iv. 
4-1 1) ; but, as regards the words which in sanhitd undergo an ab- 
normal alteration of form, there is a less perfect agreement among 
them. The Rik Pr. and Up. specify as requiring repetition in krama 
/besides sundry special and anomalous cases), words having their initial 
vowel prolonged, and those in the interior of which there is a change 
not brought about by external influences — that is to say, due to eu- 
phonic causes within the word itself. Whether the Vaj. Pr. includes 
among the repeatable words those having a prolonged initial, or whether 
any eases of this kind occur in the text to which it belongs, I do not 

iv. 120.] Atharva-Veda Prdtigdkhya. 571 

know. Our own Jer am a-sy stem, it will be noticed, while in one respect 
more chary of the repetition than the others, in that it repeats no 
riphita words, in another respect is vastly more liberal of its use, apply- 
ing it in the case of every word which requires restoration from an ab- 
normal to a normal form, according to the rules given in the preceding 
section of this chapter. There is no limitation made, either by the text 
or by the commentary, of the term samdpddya ; so far as I can see, 
every word in the text which undergoes in sanhild any of the changes 
detailed in rule 74, above, must suffer parihdra. The Atharvan krama is 
thus made a more complete and elaborate index of the euphonic irregu- 
larities occurring in its text than is that of either of the other Vedas. 

By way of introduction to the following rule, our commentator says, 
at the close of his exposition, that the words mentioned in this precept 
have their repetition made with a single iti : we are next to be told that 
in the repetition of the particle u two are required. 

118. The particle u requires two iti's. 

That is to say, when w is repeated, each occurrence of the word is 
followed by iti, and we have un ity un Hi. None of the other treatises 
supports this reading : all would prescribe simply un ity un. 

%l3«llft*{l4pl SFTtsFR II ^ II 

119. The reason of this is its nasalization and protraction. 

The commentator explains as follows : " the nasalization of this par- 
ticle m when followed by iti is taught by the rule ukdraaye Hdv aprktct' 
sya [i. 72] ; if, then, it should not be distinguished by (ddriyeia) a 
second iti, it would be deprived of its nasal quality — as also of its pro- 
traction [since this also, by i. 73, is prescribed only before iti]. There- 
fore, considering its prescribed nasality and protraction, u must always 
be repeated with a double iti." 

MHTOHclcl ll \^o ll 

o o *\ x 

120. A protracted vowel is, in repetition, to be treated as if 

The commentator's exposition is : plutap cd 'plutavac ca pariharta- 
vyah : aplutena tulyatdm prdvahitavyah : parihdrakdle :purushah : d : ba- 
bh&vana : atra d ity akdrah plutah : sa aplutavatd parihartavyah : d ba- 
bMv&nz iti babhuve Hi vaktavyam ; ' a protracted vowel is to be repeated 
as if it were unprotracted ; i. e., it is to be reduced, in parihdra, to 
equivalence with an unprotracted vowel : thus, in the passage purushah : 
d : babhitvdn.3 (x. 2. 28), the d is a protracted a: it must be repeated 
along with [or, in the form of] an unprotracted a; we must read d ba- 
bhuvdns iti babhuve 'ti.' The reading of the manuscript is unfortu- 
nately corrupt at the end, where the required *rama-form is to be given : 
the scope and intent of the rule will be examined under the one next 

572 W.D.Whitney, [ir.120- 

following, which also concerns only the passage here cited by the 

121. And a nasal vowel, in its first occurrence, is to be made 

I again add the -whole comment : yah purvam anun&siko drshtah sa 
pariharakale fuddlmm krtvA pariharlavyah : etad evo 'dAharanam : atr&i 
'va purusha A babhuvAn3 ity aoanAne : iti : avasAne A iti purvam anuna- 
siko drshtah fuddhah parihartavyah : babhuve Hi bahhuv&n; 'the vowel 
first seen as nasal is, in parihAra, to be repeated pure [i. e. free from na- 
sality] : the instance is the one already given : here, the vowel first appear- 
ing as nasal before the pause — by the rule purusha A babhuvAn ity ava- 
s&ne (i. 70) — is to be repeated pure : thus, babhuve Hi babhuvdn.' The 
most obvious and natm - al understanding of this would be that the nasali- 
zation is only to be retained in the first utterance of the word, and that 
in parihara, by this and the preceding rules, both protraction and na- 
sality should be lost altogether; so that the krama would read A babhii- 
vanz: babhuve Hi babhuva. But the rules in the first chapter to which 
the commentator refers expressly require the nasality to be retained 
before a pause, and forbid the protraction only before iti, so that they 
would appear to teach babhuve , ti bahhuvAiiz ; which, as we see, is the 
actual reading of the commentator under this rule, while, under the pre- 
ceding, the reading is too corrupt for us to understand what he intends 
to give us. The best manner, as it appears to me, of reconciling these 
apparent discrepancies is to take purvah in the present rule as belong- 
ing with the predicate instead of the subject, and as indicating the 
former of the two occurrences of the repeated word in parihAra, thus 
making the translation such as it is given above ; and farther, assuming 
the same thing to be implied also in rule 120, the repetition, or parihA- 
ra, there referred to, being, in a restricted sense, the occurrence of the 
word before iti: the Upalekha employs parigraha, in part, in the same 
sense.* It may be, however, that we ought to confess a discordance 
between the teachings of our treatise here and in the first chapter, and 
to understand the krama reading here prescribed to be babhuve Hi ba- 
bhuva — or, if purvah be interpreted in the manner proposed, babhuve Hi 
babhuvan. A like case occurring in the Rig- Veda (x. 146. 1) is, accord- 
ing to the Upnlekha (vii. 9, 10 : the Eik Pr. seems to take no notice of 
it), to be treated in the manner laid down in our first chapter : vinda- 
Un3, for vindati, is in krama to be read vindati Hi vindaiin3. What 
is the doctrine of the Vaj. Pr. in a similar instance has been mentioned 
in the note to i. 97. 

122. The successive combination of words into krama-yrords 
is to be made according to the general rules of combination. 

* See Pertsch's preliminary note to chapter v. 

iv. 123.] Atharva-Veda Prdtigdkhya. 573 

The commentator expounds this rule in a clear and pertinent manner. 
He says: "it has been said in a former rule [iv. Ill], 'with the final of 
this is made farther combination of the following word :' there, how- 
ever, the method of combination is not taught (samdhAnavidhanath no 
, ktam) : wherefore the present rule is here added. The term yatkA- 
f&stram means ' according to the several rules (yad yac ck&stram) :' 
whatever mode of combination of separate words is taught in the pada- 
$ Astra (yad yat padafAstre padAnAm samdhAnalakshanam uklam), that 
has force also here in the formation of each single lemma-word. This 
is expressly stated, in order to guard against the danger of understand- 
ing a krama-vtord to be composed of disconnected vocables (? krama- 
padalvAd anyafabdapankayo 'cyate) : this must not be the case." 

The Vaj. Pr. has no precept corresponding to this, evidently regard- 
ing it as clearly enough implied in the general direction that two words 
"are to be combined" (sam dudhAti: iv. 180) to form a &rama-word. 
It is, however, distinctly laid down by the other treatises (R. Pr. x. 5, 
r. 6, and xi. 21, r. 44 ; Up. iii. 3-5). 

yjj^lolil^-cHiuf SFrajrqfFl^R^: ll ^ II 

123. The ^acfa-repetition of a divisible pragrhya is to be made 
in the manner of that of krama, with separation by avagraha in 
the latter recurrence of the word. 

The commentator begins with explaining pragrhyAvagrhya to be a 
determinative and not a copulative compound (yanminn avagrhyatvam. 
[pragrhyatvam, ca] ekaxminn era yugapad bhavati), and goes on as fol- 
lows : " such a word, in its /rarfa-repetition (carcAy&m), is to be treated 
as in krama : carcA means twofold utterance (dvirvacanam) : that takes 
place in the pada-text (padakAle) just as in the krama-text (kramakAle) ; 
that is, one repeats (pariharel) the form of the krama-text. In krama, 
both pragrhyas and divisibles suffer repetition ; in pada (padethu), on 
the other hand, only a divisible pragrhya is repeated. In such a repe- 
tition, how is separation by avagraha made ? The rule says, ' with sepa- 
ration by avagraha in the latter recurrence of the word :' that is, the 
latter or second recurrence of the word is to receive avagraha, and not 
the first : for example, vir&pe iti vi-rtlpe (x. 7. 6, 42)." 

The usage of both our Atharvan parfo-text and that of the other Vedas 
in regard to carcA, or repetition, has been fully set forth in the note to 
iv. 74, above, as also the doubt which may reasonably be entertained 
whether the usage here taught, and followed in the extant manuscripts, 
is that which the rules of the preceding section contemplate. 

The prescription in this rule, and in this alone, of the employment 
of avagraha in separating the constituents of a compound word when 
it appears for the second time in the repetition, after iti, seems necessa- 
rily to imply that, in the repetitions of krama, separation by avagraha is 
not to bo made at all, either before or after iti, but that we are to read, 
for instance (i. 1. 1), ye trishaptdh : trishaptAh pariyanti : trisapta iti 
trisaptAh : pariyanti vipvA : pariyanti Hi pariyanti, etc. To regard the 
specification uttarasminn avagrahah of our rule as in such manner re- 

574 W. D. Whitney, [iv. 123- 

trospective as to reflect its prescriptive force, through kramavat, back 
into the rule for &ra»»a-repetition, or parigraha — understanding the 
meaning to be, ' in ^Kwfti-repetitions, the second recurrence of the re- 
peated word is to suffer division by avagraha : as is to be the case also in 
Arania-repetitions' — would be, I should think, much too violent. Never- 
theless, the Eik Pr. (x. 10, r. 16, and xi. 16, r. 31) and TJpalekha (iv. 3) 
distinctly teach that the avagraha is to be used in the repetition of com- 
pound words, after iti. I cannot find that the Vaj. Pr. prescribes the 
separation either in pada or in krama, although it is regularly made 
by the commentator on that treatise in the examples which he cites, and 
Weber passes it over without remark. 

tHWWMI^M «[^HIo|^H^iiV$ii 

124. Words requiring restoration, if occurring before a pause, 
are to be spoken in their sanhitd form. 

The three last rules of the section and of the treatise concern the 
treatment of such words as, while they stand at the end of a half-verse, 
or in any other situation before a pause, also exhibit in sanhitd some 
abnormal peculiarity of orthoepy which, by the rules of the preceding 
section, requires restoration to the natural form. A word of this class, 
instead of being combined with its predecessor into a &rama-word, and 
then repeated, is, before its repetition, to be spoken once more in sanhitd 
form. The commentator takes the example sd vrkshan abhi sishyade 
(v. 5. 9 : p. sisyade, by iv. 82), and, without writing it out in full in the 
krama form, says that we must utter sishyade again, and then repeat it, 
sisyada iti sisyade. He adds: "so also may be brought forward as in- 
stances (uddhdrydh) pranitaye (vi. 23. 2 : p, pra-nitaye) and all other 
like words, having the cause of alteration within their own limits." 
This last restriction, as we shall see, he insists upon more distinctly 
under the next following rule. 

This special point is left untouched in all the other fcrama-treatises. 

125. Of this, furthermore, the name is dsih&pita, 

Weber (p. 283) regards punah here as a part of the title, which he 
understands to be punardsthdpita. This would not be in itself unlikely, 
but it is not favored by the commentator, who, both here and under 
the following rule, treats dsthdpita alone as the term designating the 
word to which the rule relates. He omits punah altogether, in his para- 
phrase of the present rule, as being superfluous : tasya sanhitdvad vaca- 
nasya : dsthdpitam ity evam samjnd bhavati. He then goes on to state 
more at large the restriction hinted at under the last rule : namely, that 
a word which is samdp&dya, or liable to restoration, as being altered at 
its commencement by the influence of the preceding word, is not to be 
treated in the manner prescribed by these rules : " that is to say, where 
cause and effect (nimittandimittike) are in one word ; for instance, 
sishyade : here the conversion into sha produced by the reduplication 

iv. 126.] Atkarua-Veda Prdtigdkhya. 57o 

holds over, owing to the fact that the reduplication and the altered 
sibilant are in the same word. But this is not the case in the passage 
strip Am anu shicyate (vi. 11. 2), [or, in pada form] tat : striydm : anu : 
sicyate: here no repetition in the sanhita form takes place [or, the repe- 
tition does not take place in the sanhita form : iti na sanhitdvad bha- 
vati\ ; and why ? because cause and effect are declared to stand in two 
separate words : anu is a proposition ; it produces conversion into sh 
by the rule upasargdt etc. (ii. 90), and that conversion is heard (prutah) 
in a separate word ; here, then, let the process not be performed (? tend- 
tramidibhul) : for this reason is this explanation made." It might still 
seem doubtful, after all this lengthy exposition, whether such a word as 
sicyate was regarded by the commentator as not to be separately spoken 
at all, or as to be separately spoken, only not in sanhitd form, as follows : 
striydm anu : anu shicyate : sicyate : sicyata iti sicyate ; but the latter in- 
terpretation seems to me the more probable. 

H ^J^: gf^pfoiu^u 

126. That is a Icrama-w or d of a single member, and is also to 
be repeated. 

The commentary upon this rule is not so clear as the rule itself seems 
to be, without comment or explanation : it reads as follows, with only 
the most obvious emendations : parihartavyap ca sa dsthitasamjnapa- 
drpah \dsthdpitasamjndikapadahf\: yo 'sau sanhitddvirvacanena nirdip- 
yate : nimittandimittikayor bhinnapadasthatvdt : sa parihdryap ca bha- 
vati. If the intent of this is to limit the application of the rule to 
those words whose cause of alteration is situated in a preceding inde- 
pendent word, it can hardly be accepted. What follows is still more 
corrupt and less intelligible : sanhitdvad viddci bahulam iti yap chandast 
Hi: varndlopdgamahrasvadtrghapluta dtmaneshd parasmdi vibhdshd api 

Finally, to close up the commentary, two verses are given us, but so 
much mutilated that hardly more than their general sense (and even 
that only in part) is recognizable : nataMbudhyd nacapdstra drshtyd 
yathdmndnam anyathd ndi 'va kurydt: dmndtam parishannasya pdstram 
drshto vidhir vyatyayah purvapdstre : dmndtavyam andmndtam prapd- 
the 'smin rkvdcatpadam : chandaso 'parimeyatvdt parishannasya laksha- 
nam : parishannasya lakshanam iti. 

The signature of the chapter and of the work has been already given 
in the introductory note, but may be repeated here : iti gdunakiye catur- 
ddhydyike caturthah pddah : caturddhydyibhdshyam, samdptam : prir 
astu : lekhakapdthakayoh pubham bhavatu : prtcandikdydi namah : pri- 
rdmah: samvat 1714 varshe jydishthapuddha 9 dine samdptalikhitam 
pustakam. I may also be permitted to add the propitiatory heading 
of the manuscript, which was, by an oversight, omitted to be given 
in its proper place : it reads orh namah sarasvatydi namah : om namo 
brahmaveddya : athdngirasah. The last word is, as I cannot doubt, a 
copyist's error for atharvdngirasah. 
vol. vit. 73 

576 W. D. Whitney, [add. note 1. 

Additional Notes. 

1. Analysis of the Work, and Comparison with the other Pralif&khyas. 

So far as concerns the agreement or disagreement of the other Prati- 
§akhyas with that of the Atharva-Veda in respect to the doctrines 
taught in the latter, the comparison has already been made in detail in 
the notes to the text. I have thought, however, that it would be ad- 
visable to append here a systematic view of the contents of our treatise, 
and a brief statement of the correspondences of the rest,, in order to 
exhibit more clearly the sphere of the former, and to show how far 
those of the latter coincide with it, and how far they cover more or 
less ground than it occupies. 

I. Introductory and explanatory. 

Introductory, object of the treatise, i. 1, 2 ; definition of terms, i. 3, 42,43,48, 49, 
92, iv. 73, 125 ; interpretation and application of rules, i. 95, iii. 38. 

II. Phonetics and Euphony : production, classification, properties, and com- 
bination of SOUNDS. 

1. Simple sounds : formation and classification of consonants, i. 10-13, 18-31 ; do. 
of vowels, i. 27, 32-41, 71 ; quantity of vowels and consonants, i. 59-62; accents, 
i. 14-17. 

2. Sounds in combination, and resulting modifications : possible finals, i. 4-9, ii. 
3 ; final vowels not liable to euphonic change (pragrhya), i. 73-81, iii. 38 ; syllable, 
i. 93; division of syllables, i. 55-58; quantity of syllables, i. 51-64: — kinds of in- 
dependent circumflex accent, iii. 55-61, 65 ; kinds of enclitic do., iii. 62-64; evoca- 
tion and modification of accents in words and sentences, iii. 67-74: — conjunction of 
consonants, i. 49, 50, 94, 98, ii. 20 ; abhinidhana, i. 43-47 ; yama, i. 99, 104 ; nasikya, 
i. 100, 104; svarabhakti, i. 101, 102, 104; sphotana, i. 103, 104, ii. 38; karshana, ii. 
39 ; euphonic duplication of consonants {varnakrama), iii. 26-32. 

III. Construction of combined text, or sanhitA. 

Prolongation of initial, final, and medial vowels, iii. 1-25 : — combination of final 
and initial vowels and diphthongs, iii. 39-54, ii. 21-24, i. 97 ; resulting accent, iii. 
65, 56, 68, 65, 66 ; resulting nasality, i. 69 ; final vowels not liable to combination, iii. 
38-36: — combinations of final and initial consonants: final non-nasal mutes, ii. 2-8, 
13, 14, 38, 39 ; final nasals, ii. 9-12, 26, 26, 28, 30-37, iii. 37, i. 67 ; final semivowels, 
ii. 19, 21-24, iii. 20 ; final visarjaniya, ii. 21, 24, 40,43-52, 54-59, 62-80; initial con- 
sonants, ii. 7, 15-18 : — final nasal before a vowel, ii. 27, 29, i. 68, iii. 27 ; final visar- 
janiya before a vowel, ii. 41, 42, 44-53, 66 :— lingualization of t and th, ii. 15, 16 ; 
do. of n, iii. 75-95 ; do. of s, ii. 81-107 : — insertion of s, iii. 96. 

IV. Construction of krama-text. 

Importance of krama-text, iv. 108, 109; its construction, iv. 110-122, 124-126, 
74-100, i. 70, 97. 

V. Construction of disjoined, or pada-text. 

Importance of pada-text, iv. 107 ; combination or separation of verb and preposi- 
tion, iv. 1-7 ; do. of suffixes, iv. 13-37, 46-48, 53, 65, 66 ; do. of compounds, iv. 8-12, 
21, 22, 27, 28, 38-46, 49-64, 67-72 ; restoration of the original form of words, iv. 
74-77, 79, 81-97, 99, 100; repetition in pada, iv. 123 ; special cases, i. 72, 73, 81, 
82; enclitic accent in pada, iii. 68, 69, 72, 78. 

VI. Sundries, Special cases, etc. 

Study of Veda recommended, iv. 101-109: — special irregularities of formation, 
i. 63-66, ii. 26, 60, 61, iii. 7, 43 ; special case of accent, i. 96 ; list of protracted 
vowels, i. 105 ; quantity of nasalized vowels in interior of word, i. 88-91. 

add. note 1.] Atharva- Veda Prdtigdkkya. oil 

In regard to the matters embraced in the first of the general divisions 
here laid down, it may be said that our treatise is much more curt and 
concise, and more ready to pass without notice what may be assumed 
as already known, than either of the others. Definitions of terms are 
far from numerous, and the whole department of paribhdshd, or expla- 
nation of modes of phraseology, of extent, bearing, and application of 
the rules, and the like, which in all the other Prati$akhyas occupies 
considerable space, is here almost wholly wanting. This is in part 
owing to the simpler and less artificial mode of arrangement adopted 
in our work. 

In the division of Phonetics and Euphony, the discordances among 
the different authorities affect chiefly matters of detail, and are not of a 
character to call for notice here. The other three treatises include or 
imply a list and enumeration of the sounds of the spoken alphabet, 
which is wanting in our own. All, also, after treating the subject of 
the formation of articulate sounds in the outset in a manner nearly 
according with that here followed, return to it in their later and less 
genuine chapters, and discuss it anew with a straining after greater 
theoretic profundity. The niceties of consonantal combination, as abhi- 
nidk&na etc., make nearly the same figure in all the four : for minor 
differences, see the notes on the text. The Kik Pr., to its specification 
of possible finals (which is wanting only in the Taitt. Pr.), adds that 
also of possible initials, and of compatible or conjoinable consonants 
(xii. 1-4). 

Under the head of the conversion of pada-text into sanhita, the au- 
thorities differ only on minor points, or by the treatment of special cases 
appertaining to the text with which each has to deal. 

Thus far, the subjects treated are those which no Pratigakhya can 
pass over in silence : those which remain are not essential to the com- 
pleteness of a work of this class, and are accordingly found altogether 
wanting in one or more of the treatises. Thus, the IVth general divis- 
ion, the construction of the krama-text, is not touched upon in the 
Taitt. Pr., and the parts of our own and of the Rik Pr. which concern 
it are open to the suspicion of being later accretions to the text. The 
Vth division, the construction of the pada-text, receives still more scanty 
attention, being entirely passed over in the Taitt. Pr., and represented 
in the Rik Pr. only by a few scattering rules relating to special cases, 
analogous with those found in the earlier chapters of the Ath. Pr. ; only 
the Vaj. Pr. joining the latter in treating it at large, although in a less 
complete and elaborate manner. 

A more detailed comparison will be necessary under the last head, 
that of miscellaneous and extra-limital additions to the body of the 
work, of matters more or less akin with its proper substance, and aux- 
iliary to its object, yet omissible without detriment to its completeness 
as a Pratigakhya. As concerns the study of the Veda, the first subject 
mentioned in our analysis, the Rik Pr. offers a very interesting chapter 
(xv. 1-16) on the mode of instruction followed in the schools of Vedic 
study ; the Vaj. Pr. recommends Vedic study (viii. 35-42), and tells (i. 
20-26, viii. 32-34) who should pursue it, and under what circumstances. 
All the other treatises give explanations of single irregular and exeep- 

578 W. D. Whitney, [add. note 1. 

tional words and forms, of which the list will be more or less extended 
according as we include in it all those words which the pada-text does 
not analyze, or only such as are of especially anomalous character. The 
Rik Pr. alone among them (i. 6) catalogues the few protracted vowels 
occurring in its text. Like our treatise, it also teaches (in a late chap- 
ter, xiii. 7-10) when interior nasalized vowels are long : the Taitt. Pr. 
(xvi. 1-31) goes farther, and laboriously catalogues all the nasalized 
vowels contained in its text, short or long, which are not the result of 
euphonic combination. Of this same class of appendices, which are 
designed to call attention to points in the text where especial liability 
to error is held to exist, are the following : the Rik Pr. (ii. 5) gives a 
list of instances of hiatus within a word ; the Vaj. Pr. attempts (in part, 
in a bungling manner, and with very sorry success) to point out words 
which do not end with visarjaniya (iv. 26-32), which contain one y or 
two (iv. 149-159), or single, double, or triple groups of consonants (vi. 
25-30) ; the Taitt. Pr. adds to its list of nasalized vowels only (xiii. 
8-14) that of words showing an original lingual n. The Vaj. Pr. (iii. 1, 
viii. 50, 51) defines a word, and both it (i. 27, viii. 52-57) and the Rik 
Pr. (xii. 5, 8, 9) distinguish and define the parts of speech, while all 
the three give a list of the prepositions (R. Pr. xii. 6, 7 ; V. Pr. vi. 24 ; 
T. Pr. i. 15) : and the Vaj. Pr., after its fashion, carries the matter into 
the domain of the absurd, by laying down (viii. 58-63) the divinities 
and the families of rskh to whom the several classes of words, and even 
(viii. 47) of letters,' belong. The Rik Pr. (x. 12, xi. 12) and Vaj. Pr. 
(iii. 148, iv. 77, 165-178, 194) treat of the samaya or samkrama, the 
omission of verses or phrases which have already once occurred in the 
text. Such omissions are abundantly made in all the manuscripts of 
the Atharva-Veda, sanhild as well as pada, but the Praticakhya takes 
no notice of them. The Rik Pr. has a chapter (xiv. 1-30) on errors of 
pronunciation, from which, by careful comparative study, important in- 
formation on phonetic points may be drawn. It also, in its three closing 
chapters (xvi-xviii), treats with much fullness the subject of metre, 
which no one of the others even hints at. The Taitt. Pr. devotes a 
single brief chapter (xviii. 1-7) to the quantity and accent of the aus- 
picious exclamation om. The Vaj. Pr. has an interesting, although 
rather misplaced, series of rules (vi. 1-23) respecting the accentuation 
of verbs and vocatives in the sentence, and also makes a foolish and 
fragmentary attempt (ii. 1-45, 55-64) to define the accent of words in 
general. With its rules respecting the ritual employment of different 
toneg and accents (i. 127-132) is to be compared what the Taitt. Pr. 
(xxii. 12, xxiii. 12-20) says of tone and pitch. The Vaj. Pr., finally, 
remarks briefly and imperfectly (iii. 17, 137, 138) on the omission in 
the Vedic dialect of certain terminations. 

It is thus seen that the Atharva-Veda Praticakhya does not greatly 
differ in its range of subjects from the other treatises of its class ; being 
somewhat less restricted than the Taitt. Pr., and somewhat less com- 
prehensive than the remaining two, the Vaj. and Rik Pratigakhyas. Its 
style of treatment is marked by sundry peculiarities, of which the most 
striking and important is the extensive use which it makes of ganas in 
the construction of its rules. It is this which has enabled it, while in- 

add. note 1.] Atharva-Veda Prdticdkhya. 579 

eluding so much, to be at the same time so much the briefest of the 
four works. This approximates it, also, to the character of the general 
Sanskrit grammar, as finally and principally represented to us by Panini. 
The close connection between the two is farther shown by many other 
circumstances which have been pointed out in the notes upon the text 
— by the contemplation in numerous rules, both general and special, of 
phenomena of the general language rather than those of the Atharvan 
vocabulary alone, by a more liberal introduction of grammatical cate- 
gories than any other of the Praticakhyas makes (the Taitt. Pr. is its 
antithesis in this respect), and by the exhibition (not the use, as signifi- 
cant) of some of the indicatory letters employed by Panini himself. It 
is very sparing of its references to the opinions of other authorities, 
Qaunaka and Qakatayana being the only grammarians whom it cites by 
name. The latter of the two appears, both from the text and the com- 
mentary, to have stood in an especially near relation to the authors of 
our treatise. Qaunaka, although his opinion is rejected in the only rule 
where his name appears, is yet mentioned in a way which may be re- 
garded as implying his special importance as an authority; it being 
thought necessary to teach expressly that his dictum upon the point 
referred to is not binding. There is nothing, at any rate, in the mode 
of the reference, which should militate against the claim apparently 
implied in the name of the work, that it represents in the main the 
doctrines of a Qaunaka, and belongs to a school which derives its name 
from him. 

Whether the peculiarities pointed out are of a nature to determine 
the chronological relation of our treatise to the other Praticakhyas is a 
difficult and doubtful question. The discussions of this point hitherto 
made appear to me nearly barren of any positive results. They are 
all more or less based upon the assumption that the appearance in a 
Pratiijakhya of a later phase of grammatical treatment or of grammati- 
cal phraseology is an unequivocal evidence of later composition. That 
this is so is not readily to be conceded. Since Praticakhyas are no 
complete grammatical treatises, but only the phonetical text-books, and 
the manuals of rules for conversion of pada into sanhita, belonging to 
special schools, and since they imply a vastly more complete grammati- 
cal science than they actually present, it is not to be denied that any 
one of them might include more or less of the form and the material 
of that science, as its compilers chose, or as the traditional usage of 
their school required. Thus, for instance, there would be no implausi- 
bility in supposing that the Taittiriya Pratigakhya, though so much 
more limited in its grammatical horizon than those of the Vajasaneyi 
and Atharvan, was actually composed at a later date than either of 
them, and deliberately adopted the method of treating its material ac- 
cording to the letter rather than the meaning, as being better suited to 
the character of a Pratigakhya, which concerns itself only with pho- 
netic form, and not with sense. If such a supposition admits of being 
proved false, it can only be so by a more searching and wary investiga- 
tion and comparison than has yet been made, or than is possible before 
the full publication and elucidation of all the treatises. It is very doubt- 
ful whether any one of the Praticakhyas has escaped extensive modifl- 

580 W. D. Whitney, [add. note 1- 

cation, by alteration, insertion, and addition, since its first substantial 
construction. The fact that in the Eik Pr. all that is essential to make 
out such a treatise is contained in the first half, or chapters i-ix, is 
strongly suggestive of the accretion of the later chapters, and the char- 
acter of more than one of them lends powerful support to such a sug- 
gestion. That the Vaj. Pr. has suffered interpolation and increment is 
the opinion of its editor, and we cannot help surmising that its weakest 
and most impertinent portions, especially those in which the expression 
seems intended to conceal rather than convey the meaning they cover, 
are the work of a very late hand. None of the other treatises is dis- 
figured by such features. In point of dignified style, and apparent 
mastery of the material with which it deals, the first rank belongs un- 
questionably to the Rik Praticakhya ; with what inaccuracies and de- 
ficiencies it may have to be reproached, its editors have not ascertained 
for us : a careful testing of the rules by the text whose phenomena they 
were meant to present in full will have to be made for all the treatises 
before their comparison can be rendered complete. The results of such 
a testing as regards the Atharva-Veda are to be found scattered every- 
where through the notes upon the text, and do not need to be summed 
up here. I fulfil, however, a promise previously given (note to i. 1), by 
making summary reference below to the rules in which our treatise over- 
steps the limits of the subject which it is treating, or of the sphere of 
such a work as it professes to be, or in which it contemplates euphonic 
combinations and words not to be found in its text. 

Treatment of matters purely concerning the ^orfa-text, in the por- 
tions properly devoted to sanhitA, is made in rules i. 72, 73, 81, 82, iii. 
64, 68, 69, 72, 73 : a like thing is done for the Ammo-text at i. 70, 97, 
iv. 74 etc., 98. Combinations not Atharvan are had in view at i. 47, ii. 
9, 11, 12, 14, 15, 26, iii. 27, 46 ; words not Atharvan, at i. 77, 86, 87, ii. 
25, 51, iii. 2, 92, iv. 28, 62, 69. The bounds set in the first rale of the 
work are transgressed, by the inclusion of matters of word-formation 
and derivation, and the explication of forms which have no other quali- 
ties in pada than in sanhitA, at i. 63-66, ii. 18, 33, 34, 52, 59, 60, 61, 
82, 87, 89, iii. 1, 2, 3, 7, 8-11, 43, 49, 57. 59, 60, 61, 75, 78, 87, 90. 
Among these last cases, a few are palpably and grossly out of place ; 
but the greater part may be explained and excused by supposing that 
the pada-text implied by the Praticakhya is an ideal one, which our 
established and recorded pada-texta nearly approach, but do not alto- 
gether coincide with. 

a. Belatim of the Pr&tig&khya to the existing Text of the Atharva-Veda. 

In attempting to determine the relation of our treatise to the only 
known text of the Atharva-Veda, by the help of the citations which 
the former contains, it is necessary, of course, to consider as one the 
text and its commentator, since the ^ana-method pursued by the treatise 
relieves it from quoting more than a small part of the words and pas- 
sages to which it was intended to apply. And even with the help of 
the commentator, since he fills out the ganas but in part, we are able to 
find references to no more than a portion of the phenomena of the text 
to which the view of the makers of the Praticakhya was directed. This 

add. note 2.] Atharva- Veda Prdtigdkhya. 581 

state of things deprives our investigation of much of the definiteness 
and certainty which it ought to possess, and which would attend a simi- 
lar examination of any other of the Pratigakhyas by means of its Veda ; 
yet it is desirable even here to make the comparison, which will be 
found not barren of valuable results. The index of passages in the 
Atharvan text cited by the Praticakhya and its commentary, to be 
given later, furnishes in full detail the principal body of the material of 
investigation ; and from that we derive at once the important informa- 
tion that to the apprehension of the Praticakhya the Atharva- Veda com- 
prehended only the first eighteen books of the present collection. The 
two single apparent references to passages in book xix, the one made by 
the commentator (under ii. 6V), the other by an authority whom he 
cites (under iv. 49), are of no account as against this conclusion : the 
absence from the rules of the treatise of any notice of the numerous 
irregularities of the two- closing books, and the want of other citations 
in the commentary than the two equivocal ones referred to, are perfectly 
convincing. This testimony of the Pratigakhya, moreover, agrees en- 
tirely with that which we derive from a consideration of the character 
of those books and the condition of their text : no pada-text of book 
xix and of those portions of book xx which are not taken bodily from 
the Eig-Veda is known to be in existence, and it is not at all likely that 
there ever was one ; the text could hardly, in that case, have become so 
corrupt. The citations run through all the other books of the Athar- 
van ; they are more numerous, as was to be expected, in the earlier 
books, and in parts of the text they are but thinly scattered ; yet no 
extended portion of the first eighteen books can with plausibility be 
supposed not to have lain before the commentator for excerption. As 
regards single passages, there is room for more question : although our 
lack of the complete ganas greatly interferes with a full discussion of 
this point, we are able to discover phenomena in the existing text of 
which the Pratigakhya, even as at present constructed, plainly fails to 
take notice. Some such cases of omission the commentator himself 
has perceived, and calls attention to,* but those which have escaped 
his notice also are much more numerous.f Many, probably the greater 
part, of these are to be set down to the account of the authors of the 
treatise, as results of their carelessness or want of accuracy : but that 
all of them can be thus disposed of does not appear to me likely ; it 
seems a more probable supposition that in our authors' Atharvan single 
passages and single readings were wanting which are met with in the 
present text. The question, however, hardly admits of a positive solu- 
tion : it would aid us not a little in coming to a conclusion upon it, did 
we know precisely what is the completeness and accuracy of the other 
treatises, as tested by their respective texts. 

Differences of reading offered by the manuscripts as compared with 
the Pratigakhya form another main branch of the evidence bearing upon 
the question under consideration. That which I have collected, how- 

* See under ii. 63, 66, 101, iii. 60, iv. 16, 18, 67. 

f All, so far as my own search for them has been successful, have been set down 
in the notes on the text, above: see under ii. 63, 72, 93, 96, 97, 101 102 iii 5 12 
13, 26, 33, 45, 61, 80, iv. 13, 16, 39, 50, 57, 63, 68, 86, 96. ' ' ' ' 

582 W. D. Whitney, [add. note 2. 

ever, is not of decisive character, and hardly furnishes so much ground 
for suspicion of a discordance between the present text and that of the 
authors and commentator of our treatise as was derived from the cita- 
tions. That the manuscripts neglect the refinements of Vedic orthoepy, 
such as the yama, nasikya, svarabhakli, and sphotana, and the duplica- 
tions of the varnakrama, is a matter of course. Other theoretical niceties 
of a similar character, as the aspiration of a final mute before a sibilant (ii. 
6), and insertions between a mute and sibilant (ii. 8, 9), wc are equally 
prepared to see neglected in the written text, and we should not think 
of founding upon their absence the suspicion that the manuscripts rep- 
resented the Veda of another school. Some peculiarities of euphonic 
combination — the insertion of t between n and s (ii. 9), the assimilation 
of n toj (ii. 11), the omission of a mute between a nasal and another 
mute (ii. 20), the conversion of to to nasal I before I (ii. 35), the retention 
of visarjaniya before a sibilant and following mute (ii. 40) — have been 
noted as followed or disregarded by the copyists of our codices with 
utter irregularity and absence of rule. Their treatment of a or a before 
r (iii. 46) is a more distinctive trait, and may possibly rest upon a dif- 
ference of scholastic theory. Their writing of dhdk or dh for ddh (i. 
94) is of no significance, being more or less common in all Vedic MSS., 
while opposed to all sound phonetic theory and doctrine. Nor do I re- 
gard as of importance the great discordance of the manuscript treatment 
of the visarjaniya with that which the Pratic&khya teaches (ii. 40) — viz. 
the neglect of the guttural and labial spirants, and the retention of 
visarjaniya, instead of its assimilation, before a sibilant : all the written 
Vedic texts, so far as I know (with, at least, but rare and unimportant 
exceptions), follow in these respects the usage of the later language, 
and not the requirements of the Vedic phonetic grammars. A few 
single cases have been pointed out in the notes, where all or nearly all 
the manuscripts give readings of words differing from those which the 
rules of the treatise require : but most of these* are of a kindred class 
with those last noted, or concern the conversion or nonconversion of 
visarjaniya into a sibilant, and are therefore of doubtful value : upon 
such points our Atharvan manuscripts, closely connected as they are 
with one another in origin, not unfrequently disagree. These being set 
aside, only two or three indubitable cases of violation of the Praticja- 
khya rules in the existing galcha of the Atharvan remain,! and these 
admit of ready and plausible explanation as errors of copyists. 

We come now to consider the remaining department of the evidence, 
or that afforded by the references and citations in the text and com- 
mentary which furnish words and phrases not to be found in the extant 
Atharvan text. Such references and citations are very numerous, oc- 
curring in or under nearly a fifth of the rules which the treatise con- 
tains.;]; Much the greater part, however, of the considerable body of 

* See under ii. 62, 78-76, 80, 86, 98, 107, iv. 75, 77. 

t See under iii. 76, 79, iv. 64. 

\ See the notes to i. 4, 14-16, 20, 25, 28, 44, 47, 49, 52, 56, 58, 65, 68, 77, 78, 86, 
87,89-91, 98; ii. 2,5,6,9,12,14,16, 19,20,25,26,28,31,40,41,47,50-52,63,64, 
74, 82, 83, 86, 90, 102 ; iii. 1, 2. 5, 8, 1 1, 20, 27, 80, 32, 87, 89, 40, 42, 44-46, 48, 50, 51, 
53, 54, 56, 58, 64, 75, 77, 78, 90, 92 ; iv. 18, 26, 28, 56, 61, 62, 67, 69, 85. 

add. note 2.] Atharva- Veda Prdtigdkhya. 583 

non-Atharvan material thus presented us is recognizable at first sight as 
of no force to show any discordance between the Atharva- Veda of the 
Praticakhya and that of the existing manuscripts. It is, as has been 
often pointed out in the notes, a peculiarity of the authors of our treatise 
to give their rules a wider scope than the vocabulary of the Atharvan 
requires, in many instances contemplating and providing for combina- 
tions of sounds which are found nowhere in the whole body of the 
Vedic scriptures, and for which, accordingly, the commentator is obliged 
to fabricate illustrations. Moreover, even where the Atharvan furnishes 
numerous or innumerable examples of the application of a rule, the 
commentator sometimes prefers to draw upon his own fancy, instead of 
citing its text (notable instances of this are to be found especially under 
iii. 42-50). This being the case, it is evidently impossible to draw any 
distinct and certain line of division between what may be cited from an 
Atharvan text not agreeing with the one which we possess, and what is 
derived from other sources. But there are a certain number of sentences, 
among those given by the commentator, which have more or less clearly 
the aspect of genuine citations from a Vedic text ; and although some 
among them might be regarded as instances of carelessness on his part, 
he quoting by memory from another source than his own Veda, we 
cannot plausibly extend this explanation to them all : it must remain 
probable that in part, at least, they were contained in some hitherto 
unknown fdfchd of the Atharva-Veda. The sentences referred to are 
as follows : prd , mA ca roha (i. 14-16), pund raktarh v&sah (i. 28, ii. 19, 
iii. 20), rtunr rtubhih (i. 68), ami afafre (i. 78), tad abMtam (ii. 2), 
dhdtar dehi savitar dehi punar dehi (ii. 47), samaho vartate (ii. 50), yad 
aho r&pdni drpyante (ii. 51), yadd 'ho rathamtaram sdma giyate (ii. 51), 
bhuvo vipveshu savaneshu yajrliyah (ii. 52 : found in Rig-Veda, x. 50. 4), 
dvish krnute r&pdni (ii. 63), dydush pilar nyann adhardn (ii. 74), vi srpo 
virappin (ii. 102), tatarsha puroddgam (iii. 32), vdrshyodakena yajeta 
(iii. 32), sahasrarcam ide atra (iii. 54), svargena lokena (iii. 78), mahi 
tvam (iv. 26), jamadagnydtharvana (iv. 67), and vavrdhdnah-iva (iv. 
85). In two instances these citations are directly referred to in a rule 
of the text (ii. 51) : in all the other cases where the treatise itself men- 
tions or implies words not found in the Atharvan,* it seems to me un- 
necessary to see anything but the tendency of the rule-makers to give 
their rules a wider bearing than the nature of the case required. 

The identity or near correspondence of many of the fabricated illus- 
trations furnished by the commentator with those given by the scho- 
liasts to Panini has been remarked in many instances,f and is a very 
noteworthy circumstance, as adding a new proof to those already else- 
where given of the more intimate relation of the grammatical system 
of our treatise than of that of any other of the Pratigakhyas with the 
general Sanskrit grammar; and also, as indicating the antiquity and 
the persistence in use of at least a part of the examples selected to illus- 
trate the Paninean rules. 

* They are i. 77, 86, 87, ii. 25, iii. 2, 92, iv. 28, 62. 

f See under i. 68, ii. 14, 25, 40, 52, 63, 83, 90, iii. 27, 30, 32, 39, 40, 44, 45, 48, 50, 
61, 63, 77, 90, iv. 28 ; analogies might also be pointed out in Panini for the exam- 
ples under i. 49, 98, ii. 6, 9, 26, iii. 42, and a more thorough and careful search than 
I have made would doubtless bring to light additional correspondences. 
vol. vii. 74 

584 W . D. Whitney, [add. note 3. 

3. 2%e Consonantal Combinations of the Atharva- Veda, and their Pho- 
netic Form according to the Rules of the Prdtifdkhya. 

In the course of the notes upon certain portions of the text of the 
Pra.tic3.khya, I found it highly desirable, or almost necessary, to ascer- 
tain how many consonantal combinations of certain classes were to be 
found in the Atharva- Veda, and with what frequency they occurred. I 
was hence led to draw out a complete list of all the combinations of 
consonants which the text contains. Later, in examining and compar- 
ing with one another the nicer points in the phonetic theory of the 
treatise, particularly those which the written alphabet does not attempt 
to represent, I thought it worth while to make a practical application 
of all the phonetic rules to the collection of combinations already drawn 
up, writing each one out iu the form which the rules would require it 
to assume. The result is the following scheme, which has seemed to 
me of sufficient interest to be worth appending to the present work. 

To make out a complete list of the consonantal groups of our text is a 
work only of time and patience : to determine in every case what is the 
Praticakhya's doctrine as to its true phonetic form is less easy, since it 
involves the application of rules which sometimes appear to trench upon 
each other's spheres, and of which the reconciliation cannot always be 
satisfactorily effected. I have not, however, been willing to assent to 
the opinion which Weber (p. 247) expresses, that any of the modifica- 
tions prescribed are absolutely inconsistent with, and exclude, one an- 
other. It is not easy to see how, in that case, the phonetical treatises 
should present them side by side without any apparent misgivings, and 
without notifying us that the application of certain ones exempts us 
from the necessity of making certain others. At any rate, I shall here 
follow as accurately as I can all the directions which our Praticakhya 
gives, expressing now and then such doubts as may suggest themselves 
respecting the mutual limitations of the rules : if the resulting combina- 
tions sometimes look strange, intricate beyond measure, and unutterable, 
the fault will lie with our Hindu authorities. 

One circumstance deserves to be specially noted here : namely, that 
the loss of a rule or rules from the midst of the passage of the Prati- 
cakhya treating of duplication (see under iii. 28) doubtless loads our list 
with a few more doubled consonants than it should properly bear. I do 
not venture, however, to fill out the lacuna by conjecture : a suggestion 
or two will be made farther on as to what the lost rules may in part 
have contained. 

I. Groups not liable to phonetic modification. _ These are, «, of two 
consonants : cy, chy,jy, ts, pv, bv, bhv, mv, yy, 11, 11, xk, xkh, fc, pch, $y, 
ff, sht, shth, shn, shsh, st, sth, sn, ss, <pp, <pph ; @, of three consonants : 
Urn, isy, t'sv, xkl, zkr, xksh, fey, ffm, ffy, ffr, $cl, ffv, shty, shtr, shtv, 
shthy, shny, shnv, sty, str, stv, sthy, ssk, sst, ssth, ssn, ssp, ssm, ssy, ssr, ssv, 
<j>pr, <ppl ; •/, of four consonants : xkshv, shtry, ssir. 

This class, it will be noticed, is composed, of combinations containing 
two consonants of the same organ (sasthdna), where, by iii. 30, no du- 
plication is made — the constitution of the group being, at the same 
time, not such as to necessitate abhinidhdna, or any of the euphonic 
insertions taught at i. 99-104. To the groups containing, in the 

add. note 3.] Atharva- Veda Prdti^dkhya. 585 

printed text, a final visarjaniya as their first member, I have given the 
form required by the Praticakhya at ii. 40, representing (after Muller's 
example) the jihvdm&Uya spirant by z, and the upadhmdntya by <p, as 
I have also done in one or two isolated cases in the notes on the text. 

II. Groups suffering abhinidhdna only (by i. 44). These are as fol- 
lows : a, of two consonants : kk, gg, ggh, nk, nkh, ng, ngh, nh, cc, cch,jj, 
nc, nch, nj, ddh, nth, nd, ndh, tt, tth, dd, ddh, nt, nth, nd, ndh, nn,pp, 66, 
mp, mph, nib, mbh, mm ; §, of three consonants : kksh, hkr, nkl, nkv, nksh, 
nkhy, ngy, ngr, ngl, nghr, cchr, cckl, cchv, jjy, jjv, richy, nchr, nchl, nchv, 
njy, ddhy, ndy, tty, ttr, ttv, ddy, ddr, ddv, ddhy, ddhr, ddhv, nty, ntr, ntv, 
nts, ndy, ndr, ndv, ndhy, ndhr, ndhv, nny, mpr, mpl, mps, mbr, mbhr, 
mml; y, of four consonants: nkshn, nkshv, ntry, ntst, ntsth, ntsp, ntsv, 

All these groups are of such a character that, by the rule already re- 
ferred to (iii. 30), they are not subject to duplication of their initial 
consonant ; which latter, however, by i. 44, must, as being followed by 
another mute, undergo the weakening process of abhinidhdna. Eepre- 
senting, as I propose to do, this weaker utterance by a type smaller and 
set a little lower, the pronunciation will be kk, kksh, nkshn, etc. 

III. Groups suffering duplication only: 1st, with duplication of first 
consonant ; «, groups of two consonants : yv (as yyv ), Ik, Ig, Ip, Iph, lb, 
Im, ly, Iv, vn, vy, vr, pn, fp, pm, pr, pi, pv, shk, shp, shm, shy, shv, sk, sp, 
sph, sm, sy, sr, sv ; (?, «of three consonants : Igv (as llgv), prv, pvy, shkr, 
shpr, shmy, sphy, smy, svy ; — 2nd, with duplication of second consonant; 
«, groups of two consonants : hy (as hyy), hr, hi, hv ; @, of three con- 
sonants : hvy (as hvvy). 

IV. Groups suffering duplication and abhinidhdna. 1st, with abhini- 
dhdna of one consonant ; <*, groups of two consonants : ky (as kky), kr, 
kl, kv, ksh, khy, gy, gr, gl, gv, ghr, ghv, ny, nv,jr,jv, ty, dy, dr, dv, dhy, 
dhr, dhv, ny, nv, ty, tr, tv, thy, thv, dy, dr, dv, dhy, dhr, dhv, ny, nr, nv, 
py, pr, pi, pp, ps, br, bl, bhy, bhr, my, mr, ml, lp, Ih ; (t, groups of 
three consonants : kshn (as kkshn), kshm, kshy, kshv, ghry, nvr, try, trv, 
dvy, dvr, dhry, nvy, nvr,psy; y, of four consonants : kshmy (as kkshmy) ; 
— 2nd, with abhinidhdna of two consonants; «, groups of two conso- 
nants : kc (as kkc), kt, kth, kp, gj, gd, gdh, gb, gbh, nj, nt, nd, ndh, nn, np, 
nm, nh, tt, tp, db, dbh, nn, nm, tp, tph, dg, db, dbh, np, nb, nbh, nm, nh, pt, 
bj, bd, bdh, mn, mn; $, of three consonants : kty (as kkty), ktr, ktv, 
kthy, kpr, gdhy, gdhv, gbhy, nty, ntv, ndhy, npr, ttv, tpr, dbr, dbhy, nny, 
tpr, tpl, dbr, dbhy, dbhv, npr, nps, nbr, nbhr, nmy, pty, ptv, bdhv, mny. 

V. Groups suffering one or both the above modifications, with interpo- 
sition of yama. 1st, with yama and double abhinidhdna ; o, groups of 
two consonants : jn (asjjn), tn, thn, dn, dhn,pm; (?, of three consonants: 
cny (as ™ny),jny, tny, dhny, sthn (as sthtlin) ; — 2nd, with yama of sec- 
ond consonant, and triple abhinidhdna: nghn (as nghghn), jjn, njn ; — 
3rd, with duplication of first consonant, yama, and triple abhinidhdna; 
a, groups of two consonants : kn (as kkkn), kn, km, khn, gn, gm, ghn,jm, 
tm, tm, dm, dhm, pn, bhn, bhn; ft, groups of three consonants: kny (as 
kkkny), gny, ghny, dmy, pny ; — 4th, with duplication of first consonant, 
yama of second, and quadruple abhinidhdna : kthn (as kkthfkn). 

The Vaj. Pr. (iv. Ill) has a rule expressly prohibiting duplication of 

586 W. D. Whitney, [add. note 3. 

the first consonant in a case like the last, where a consonant suffering 
yama follows it. Such a precept is very possibly one of those lost in 
the lacuna exhibited by our MS. after iii. 28 ; and I should hope that 
the Praticakhya might have extended the exemption from duplication 
also to any consonant of which yama is made, thus including our 3rd 
division, and allowing us to say simply hkn, etc. There were enough of 
duplication, one would think. 

VI. Groups suffering one or more of the above modifications, with 
interposition of ndsikya ; viz. duplication of second consonant, ndsikya, 
and abhinidhdna; a, groups of two consonants : hn (as A"»n), hn, hm; 
(I, groups of three consonants : hay (as h"nny), hny. 

VII. Groups suffering the interposition of svarabhakti only : 1 st, of 
the longer svarabhakti: rf (as r°f), rsh; — 2nd, of the shorter; a, of 
three consonants : rjy (as r'jy), rls, rsht, rshn ; §, of four consonants : 
rtsy (as r'tsy), rshny. 

I have assumed that, in the last class of cases, the rule forbidding the 
duplication of a consonant before another of the same class prevailed 
over that prescribing the duplication after the r. In the absence, how* 
ever, of any direction as to this point in either the text or its comment- 
ary, I am by no means confident that my understanding is the true one. 

VIII. Groups suffering svarabhakti, together with one or more of the 
preceding modifications. 1st, with svarabhakti (shorter) and abhini- 
dhdna : rtt (as r'tt) ; — 2nd, with svarabhakti and duplication ; a, with 
longer svarabhakti: rh (as r°hh); b, with shorter; a, groups of two 
consonants : ry (as r'yy), rl, rv ; §, groups of three consonants : rvy (as 
r'vvy), rvr, rfv, rshm, rshy, rshv, rhy, rhr, rhv ; — 3rd, with svarabhakti 
(shorter), duplication, and abhinidhdna ; a, groups of two consonants : 
rk (as r'kk), rg, rgh, re, rch, ry, rn, rt, rth, rd, rdh, rn, rp, rb, rbh, rm ; 
(?, groups of three consonants : rksh (as r'kksh), rgy, rgr, rny, rty, rtr, 
rtv, rdy, rdr, rdv, rdhy, rdhr, rdhv, rny, rpy, rbr, rbhy, rbhr, rmy ; y, 
groups of four consonants : rkshy (as r'kkshy), rtvy ; — 4th, with svara- 
bhakti (shorter), yama, and double abhinidhdna : rtn (as r'tm), rdhn ;— 
5th, with svarabhakti, duplication, yama, and triple abhinidhdna: rtm 
(as r'ttim). 

Two or three of these combinations are liable to the doubt expressed 
after the last class. It may also excite a question whether r and A are 
to be duplicated when following one another : I have, however, in clas- 
sifying such groups, interpreted the first part of rule iii. 31 as meaning 
4 r and h are not liable to duplication when either is the first consonant 
of a group :' this would leave each free to be doubled when preceded 
by the other. 

IX. Groups liable to sphotana, together with some of the preceding 
modifications: viz. to duplication, double abhinidhdna, and sphotana; 
a, groups of two consonants : t-k (as tfk), t-k, t-kh, d-g, d-gh, p-k, p-ch, b-g, 
bj, wk, n-kh, wg, n-gh ; ft groups of three consonants : tkr (as tt'kr), 
fkv, t-ksh, d-gr, n-kr, n-ksh, n-gr. 

With regard to the question whether the groups commencing with 
n are to be ranked in this class, see the note to ii. 38. If denied spho- 
tana, they would fall under IV. 2. a. 

It would seem most natural to regard a mute as relieved by sphotana 

add. note 3.] Atharva- Veda Prdtigdkhya. 587 

of its modification by abhinidh&na ; and if any of these subtle euphonic 
changes mutually exclude one another, the two in question roust cer- 
tainly do so. It is not absolutely impossible that the Hindu gramma- 
rians may have regarded the name sphotana itself as implying suspension 
of abhinidh&na, and so have saved themselves the trouble of specifically 
teaching the fact of such suspension : yet I cannot think this likely, 
and so have combined the modifications as they are given above. The 
admission of a nullifying effect in the sphotana would remove merely 
the abhinidh&na of one consonant ; changing, for instance, dtFg to dd s g. 

It remains only to take notice of a peculiar class of cases, where a 
final surd mute comes to stand before an initial sibilant, and their com- 
bination undergoes (according to the rules ii. 6, 8, 9) certain euphonic 
modifications, which are not usually represented in the manuscripts or 
in the edited text. 

X. Combinations of a final surd mute with an initial sibilant. 1st, 
with aspiration of the surd: t-s (as ths), tsy, tsr, t'sv, Vstr ; — 2nd, with 
aspiration, duplication, and abhinidh&na: irs (as ikhs), t-p, p-s, fcsth, 
p-sv ; — 3rd, with insertion and abhinidh&na: t-s (as tts), t-sv, n-s (as nks), 

Akin with a part of this last class is the combination of n and s, with 
insertion of t ; I have not brought it in here, because the inserted letter 
is regularly given in the printed text. It might be made a question 
whether all these inserted mutes are not liable to be converted into 
aspirates by the action of rule ii. 6, so that we ought to speak tths, 
nkhs, nths, etc. But, considering the want of explicitness of the treatise 
upon this point, and the uncertainty whether the inserted mute is prop- 
erly to be regarded as appended, in the character of a final, to the 
former word, I have thought myself justified in adopting for the com- 
binations in question the simpler mode of utterance. 

In order to facilitate the determination, with regard to any given 
group, of its occurrence in the Atharvan text and of the phonetic form 
to which the rules of the Praticakhya reduce it, I add an alphabetical 
list of all the groups, each followed by its theoretic mode of pronuncia- 
tion, and by a reference to a passage of the text where it occurs. As 
in the preceding examples, a sound which has suffered abhinidh&na is 
marked by a smaller letter ; a yama, or nasal counterpart, has a straight 
line above it ; the n&sikya is expressed by a superior n ; the two svara- 
bhaktis, shorter and longer, by a heavy dot and a little circle respect- 
ively ; the sphotana, by a superior s. Where a group has the phonetic 
form given it only when composed of final and initial letters taken to- 
gether, the division between final and initial is made by an interposed 

I. Groups of two consonants : 

kk (kk: iv. 19. 6), kc (kkc: i. 6. 3), kn (kkkn: viii. 10. 18), kt (kkt: i. 
23. 1), kth (kklh: i. 10. 3), kn (kkkn: i. 23. 1), kp (kkp: i. 14. 1), km 
(kkkm : i. 25. 1), ky (kky : ii. 33. 2), kr (kkr : i. 12. 1), kl (kkl : ii. 33. 3), 
kv (kkv : i. 27. 1), ksh (kksh : i. 2. 3), k-s (km : iii. 1.4); khn (kkhkhn : 
v.31.8), khy (kkhy: iii. 3. 2) ; gg (gg: ii. 36. 7), ggh (ggh: iii. 19.6), 
91 (ggi ■ iv- 15- 2), gd ( gg d : iv. 3. 1), gdh (ggdh :_i. 10. 2), gn ( gff }n : i. 6. 2), 
gb (ggb: iii. 27. 6), gbh (ggbh : iii. 6. 6), gm ( ggg m : ii. 2. 3), gy (ggy ; vii. 

588 W. D. Whitney, [add. note 3. 

44. 1), gr ( g gr : i. 10. 1), gl (ggl : iv. 4. 7), gv ( g gv : iv. 19. 5) ; ghn ( gg hljM : 
i. 28. 4), ghr ($hr : iv. 3. 1), ghv (gghv : xix. 55. 5) ; nk (Ak : i. 12. 2), 
nkh (Akh : iv. 10. 1), ng (Ag : i. 12. 2), ngh (Agh : iv. 11. 10), nn (An : v. 
22.2), nj (Anj: v. 4. 8), nt (AAt: xvii. 17), nd (And: xiii. 2. 20), ndh 
(nndh : i. 25. 1), nn (Ann : ix. 1. 3), np (AAp : xii. 2. 1), nm (AAm : v. 1. 
1), ny (Any : v. 26. 12), nv (Anv : iii. 4. 1), ns (Aks : iv. 11. 8), Ah (AAh : 
iv. 19.7). 

cc (cc : iv. 1. 3), ech (ech : iii. 12. 3), ey (ey : i. 3. 6) ; chy (ehy : ix. 5. 4) ; 
jj (jj: i. 11. 4), ./ft" (£n : i. 7. 6), jm (jijm : iii. 30. 6), jy (jy : i. 1. 3), jr 
(jjr : i. 7. 7), jv (jjv : iii. 24. 2) ; nc (nc : i. 4. 1), rich (nch : iv. 8. 4), nj 
(nj: i.2.4). 

fie (tfk: iv. 18. 3), tt (ttt: i. 11. 1), /p (up: v. 14. 8), tm (turn: vih. 
6. 15), jy (tty: xix. 44. 6), t-p (tthp: ix. 5.' 21), ts (tts: xi. 1. 2); ddh 
(ddh : v'ii. 97. 7), db (ddb : vi. 96. 2), dbh (ddbh : xii. 2. 48), dy (ddy : ii. 
2.1), dr(ddr: xi.7. 11), dv (ddv: iii. 11. 5); dhy (ddhy : i. 18.4), dhr 
(ddhr: vii. 95. 3), dA» (ddhv : viii. 8. 11) ; n«A (»<A : x. 9. 15), nd ( n d : 
ii.*7.3), ndh (ndh:' xix. 29.6), nn(nnn: v.20.11), nm(nnm: i.13.3), 
ny Uny : i. 9. 2), nv (nnv : i. 13. 4). 

fie '(tt'k : i. 24. 2), l-'kh (tt*kh : iv. 1 1. 10), tt (tt : i. 9. 1), tth (ttk : i. 8. 
4), tn (tm : i. 9. 2), tp (up : i. 12. 2), tph (uph : vi. 124. 2), tm (turn : i. 
18. 3), ty (tty : i. 7. 4), tr (ttr : i. 1. 1), tv (ttv : i. 7. 6), ts (U : i. 15. 3), 
fs (ths : i. 13. 3) ; thn (ththn : ii. 30. 1), thy (tthy : iii. 4. 7), thv (tthv : xiii. 
2. 22) ; dg(ddg: i. 27. 3), d-g (dd*g : i. 2. 3), d-gh (ddfgh : v. 21. 8), dd 
(dd : i. 28. 1), ddh (ddh : i. 7. 7), dn (ddn : vii. 45. 2), db (ddb : ii. 29. 1), 
dbh (ddbh : i. 20. 1), dm (dddm: i. 2. 1), dy (ddy : i. 1^1), dr (ddr : i. 3. 
8), dv (ddv : i. 2. 2) ; dkn (dhdhn : i. 14. 1), dhm (ddhdhm : i. 22. 1), dhy 
(ddhy : i. 8. 2), dhr (ddhr : iii. 12. 2), dhv (ddhv : i. 4. 1) ; wk ( nn 'k : i. 
14.2), n-kh (nn"kh: v.19.3), n-g (nn"g: ii. 1.2), n-gh ( nn s gh: iii.10.ll>, 
nt (nt : i. 15. 2), nth ( n th : i. 27. 1), nd (nd : i. 7. 1), ndh (ndh: i. 4.3), 
nn ( n n : i. 7. 5), np (nnp : i. 25. 1), nb («»6 : v. 26. 6), nbh (nnbh : ii. 9. 4l 
»m (nnm : i. 10. 4), ny ( n ny : i. 2. 1), nr ( n nr : ii. 35. 2), nv ( n nv : i. 1. 1), 
wsA (nteA : viii. 9. 17), n-» («<*: viii. 5. 16), nh (nnh : i. 12. 2). 

pk (pp s k: viii. 9. 20), p-ch (pp'ch : vi. 48. 3), pt ( pp t: i. 1. 1), pn 
(pppn: ii. 11. 1), #p (pp: iii.H.5),pm (ppm: iii. 31. 1), py (ppy : i. 
14. 3), pr ( p pr : i. 7. 2), pi (ppl : iii. 6. 7), pv (pv : iii. 2. 5), pp (ppp : 
vi. 2. 2), ps (pps : ii. 2. 3), ps ( P phs : i. 6. 2) ; b-g (bb*g : xviii. 2. 6), # 

(J*; : viii. 4. 1), ft?' (44 s ; = xix - 21 - x )' bd ( bbd '• "• 24 - 6 )' bdh ( bbdh : v - 18- 
4), 66 (66 : xix. 21. 1), br (bbr : i. 6. 2), bl (bbl : xi. 9. 19), bv (bv : xiii. 1. 
15) ; bhn (bblMn : iii. 8. 6), bhn (bbhbhn : x. 3. 3), bhy (bbhy : i. 4. 3), bhr 
(bbhr : i.'l. 1), b'hv (bhv : v. 2. 7) ; mn (mmn : ii. 36. 2), mn ( mm n : u. 25. 
2), »ip (nip: iii. 26. 5), mph ( m ph: x. 6. 6), nib ( m b: i. 4. 1), mbh (mbh: i. 
6. 2), mm ( m m: iii. 30. 1), my ( m my: i. 11. 2), mr ( m mr: ii. 24. 3), ml 
( m ml: viii. 6. 2), mv (mv: vi. 56. 3). 

yy (yy.l 1.3), yv (yyvA.M.4,); rk (r'kk:xa.S.i),rg (r' g g: vi. 69. 
2), rgh (r- g gh : i. 22. 2), re (r'ec : i. 2. 3), reh (r'cr-h : xx. 34. 12), rj (r'jj : i. 
2. 1), r» (/••«» : i. 10. 4), rt (r'tt: i. 28. 2), rth (r'tth: i. 7. 6), r<* (»••«** : l. 
22. 2), rdh (r-ddh: i. 9. 3), rn (r' n n: i. 1. 3), rp (r'pp: i. 2. 1), r6 (r>bb: i. 
1. 1), r6& (r'bbh: i. 11. 2), n» (r' m w: i. 4. 1), ry (r'yy: i. 3. 5), rl (r>ll : 
i. 18. 1), rv (r'vv : i. 2. 2), rp (r°p: iv. 4. 7), rsh (r"sh : i. 5. 4), rh.(r°hh: 
i. 8. 4) ; «: (Ilk: v. 17. 4), Z^ (llg : ii. 36. 1), lp (lip : iii. 8. 1), Iph (Uph : 

add. note 3.] Atharva- Veda Prdtigdkhya. 589 

x.2.1), lb (lib: iv. 2. 8), Im (llm: ii. 30. 3), ly (lly: vi. 11. 3), 11 (11: ii. 
32. 5), 11 (11: iv. 38. 5), Iv (llv: ii. 31. 1), If (lip: vi. 30. 2), Ih (llh: v. 22. 
5) ; vn (vvn: vi. 2. 3), vy (vvy : i. 3. 1), vr (vvr : ii. 1. 1). 

Xk (xk: i. 4. 3), xkh (xkh: i. 6. 4) ; fc (fc: i. 5. 4), fch (pch: ix. 3. 14), 
$n (fpn: i. 11.4), $p (gfp: iv. 5. 6), pm (ffm: i. 2. 2),fy (py: i. 7. 5), p- 
(pp: i. 1. 2), pi (£pZ: i. 31. 3), pv (p$v: i. 1. 1), ff (ff. i. 5. 2); shk 
(shshk: i. 11. 3), sht (sht: i. 3. 1), shth (shth: i. 2. 4), sAn (sAn: i. 13. 4), 

shp (shshp : i. 1. 2), shm (shshm : i. 12. 3), shy (shshy : i. 14. 1), shv (shshv : 
i. 2. 1), shsh (sksh: xi. 5. 2) ; sk (ssk: i. 13. 2), st (st: i. 1. 2), sth (sth: i. 
16. 1), sn (sn: v. 5. 8), sp (ssp: i. 1. 1), sph (ssph: iv. 7. 3), sm (ssm: i. 1. 

4), sy (ssy: i. 2. 2), sr (ssr: i. 2. 4), sv (ssv: i. 2. 3), ss (ss: i. 4. 2) ; <pp 
((fp: i. 1. 1), <pph (<pph: iii. 15. 4). 

hn (h" n n: ii. 34. 5), hn (h" n n: ii. 33. 3), hm (h n m m: i. 8. 4), hy (hyy: 
i. 8. 4), h'r'(hrr: iv. 15. 14), AZ (AW: xviii. 3. 60), hv (hvv: i. 4. 3). 

II. Groups of three consonants : 

kksh (kksh: xix. 24. 2), My (kkty: i. 12. 3), ktr (kktr: iii. 30. 6), ktv 
(kktv: iv. 31.4), kthy (kkthy: vi. 9. 1), kthn (kkthihn: xx. 136. 4), kny 
(kkkny: i. 23. 3), kpr (kkpr: ix. 1. 10), kshn (kkshn: ii. 33. 3), k»hm 
(kkshm: vii. 89. 1), kshy (kkshy: i. 8.3), kshv (kkshv: ii. 2.1), krsth 
(kkhsth: xx. 134. 3); gdhy (ggdhy. iv. 33. 1), gdhv (ggdhv: v. 18. 10^, 
9 n V (ffggny- ix - 1- 12 ). 9 hh V [gg^V- *• 6. 12) ; ghny (gghghny: iii. 30. 1), 
ghry (gghry: xii. 2. 4) ; nkr (nkr: v. 20. 9), nkl (nkl: vi. 138. 1), nkv 
(nkv: x. 7. 1), nksh (hksh: i. 5. 4), nkhy (nkky. iv. 16. 5), Agy (hgy: iv. 
37. 2), ngr (ngr: xiv. 1. 38), ngl (hgl: iv. 38. 3), nghn (ngkghn: vi. 76.4^, 
nghr (nghr: xiii. 1.46), ■hty (nnty: xiii. 3. 5), ntv (nntv: xviii. 2. 60), 
■hdhy (nndhy: viii. 3. 6), npr (nnpr: x. 8. 19), nvr (nnvr: xv. 18. 5). 

cchr (cchr: iii. 12. 2), cchl (echl: x. 2. 1), cchv (cchv: xviii. 3. 18), cny 
(ceny: xii. 4. 30), jjn (jjjn: ii. 12. 1),jjy (jjy: vi. 61. l\jjv (pv: iii. 11. 8), 
jny (jjfly. xiv. 1. 44) ; nchy ( n chy: xi. 9. 17), nchr (achr: iv. 8. 3), hchl 
(whl: xviii. 1. 33), nchv (achv. xi. 10. 23), njn (a£n: iii. 14. 4), njy (njy: 
i. 9. 1). 

ttv (tttv: iii. 4. 1), tpr (ttpr: xi. 4. 12), tsv (ttsv: viii. 9. 9) ; ddhy 
(ddhy. xi. 1. 31), dbr (ddbr: xi. 8. 30), dbhy (ddbky. iii. 3. 4) ; ndy (ndy: 
vi. 138. 2), nny (many : Vi. 77. 2). 

t-kr (u*kr: ii. 12. 6), t-kv (tt s kv: ix. 9. 17), t-ksh (tfksh : ii. 8. 2), Uy 
(tty: v. 6. 10), ttr (ttr: vi. 107. 1), ttv (ttv: i. 23. 4), tny (any. ii. 12. 1), 
tpr (ttpr: ii. 34. 2), tpl (ttpl: v. 5. 5), try (ttry: ii. 8. 2), trv (ttrv: vi. 40. 
2), tsm (tsm : vii. 52. 2), tsy (tsy : xi. 3. 37), <-sy (<A»y : iii. 13. 4), tsr 
(thsr: xii. 4. 34), tsv (tsv: ii. 5. 4), t-sv (thsv : iii. 3. 1) ; d-gr (difffr: ii. 10. 
6), ddy (ddy: iv. 19. 6), ddr (ddr: ii. 10. 1), ddv (ddv: vi. 107. 2), ddhy 
(ddhy. xii. 2. 3), rfrfAr (drfAr: vi. 87. 3), ddhv (ddhv: iv. 14. 2), dbr (ddbr: 
i. 32. 1), dbhy (ddbhy: ii. 33. 5), dbhv (ddbhv: xviii. 4. 26), rfwy (dddmy. 
v. 23. 13), cfoy (ddvy. iv. 19. 6), dw (drfw: vii. 90. 1) ; dhny (dhdhny: iv. 
1. 1), dhry (ddhry: iv. 29. 4) ; wkr (nn'kr: ii. 31. 2), wksh ( nn ?ksh: x. 4. 
8), n-gr (nn'gr: viii. 7. 11), nty (nty. i. 3. 7), ntr ( n tr: i. 3. 6), ntv (ntv: 
i. 4. 2), nts ( n ts: viii. 6. 8), ndy (ndy: ii. 14. 2), ndr (ndr: viii. 8. 9), redi» 
{nrf»: ii. 11. 3), ndhy ( n dhy. vi. 25. 3), ndhr ( n dhr: iv. 16. 7), ndhv 
(ndhv: v. 13. 7), nny ( n ny. v. 5. 5), npr (nnpr: i. 10. 2), nps ( nn ps: x. 3. 
14), nbr (nnbr: i. 29. 1), nbhr ( nn bhr: v. 22. 12), nmy (nnmy: xi. 10. 13), 
nvy (nnvy. iv. 3. 1), nvr (nnvr: xv. 3. 1). 

590 W. D. Whitney, [add. note 3. 

pty (ppty: i. 28. 4), ptv (pptv: iv. 5. 6), pny (pppny: iv. 9. 6), psy 
(ppsy: x. 9. 7), psv (pphsv: i. 4. 4) ; bdhv (bbdhv: i. 8. 2) ; mny (mmny: 
iv. 7. 5), mpr ( m pr: i. 7. 4), mpl ( m pl: ii. 33. 4), mps ( m ps: x. 5. 43), mbr 
(mbr: i. 8. 4), mbkr ( m bhr: vi. 120. 2), mml {mini: vi. 66. 3). 

rksh (r'kksh: xviii. 2. 31), rgy {r'ggy: ix. 2. 14), ryr (r'^yr: iii. 2. 5), 
W ( r yV : v "- 22. 1) rny (r'yjiy: ii. 25. 1), r« (»••*<: ii. 7. 5), rira (»••«»: i. 
1. 3), rtm (r'tttm: iii. 8. 6), rty (vtty: iii. 31. 2), rtr (r'ttr: x. 1. 30), rtv 
(r'ttv: i. 4. 3), rts (r'ts: v. 7. 1), rrfy (r-ddy : vii. 6. 1), rrfr (r'ddr: vii. 28. 
1), rrftf (r-ddv: v. 12. 5), rrfAra (r'dhdhn: iv. 39. 1), rrfAy (r'ddhy: vi. 94. 
3), rrfAr (r'ddhr: vi. 88. 1), rdAw (r'ddhv: iii. 26. 6), rny (r' n ny: ii. 14. 3), 
rpy {r'ppy' ix. 2. 22), r&r (r'bbr: iii. 20. 5), r6Ay (r'bbhy: i. 12. 4), r&Ar 
(r'bbhr: v. 1. 1), rmy (r'ntmy: iv. 5. 5), rvy (r'vvy: iii 17. 3), rw (r'vvr: 
i. 16. 1), r^« (r'fpv: ii. 33. 3), rsAi (r'sht: iv. 18. 7), rsAn (r'shn: x. 2. 1), 
rsAw (r'shshm: iii. 4. 2), rsAy (r'shsky: vi. 18. 1), rsAi> (r'shshv : xviii. 3. 
2), rky (r'hky: iii. 1. 2), rhr (r'hkr: xii. 5. 29), rkv (r'hhv: vii. 56. 3) ; 
Igv (llffv: xii. 3. 32). 

#H (*££: ii. 2. 5), xkr (xkr: ii. 32. 1), xksh (xksh: ii. 8. 5) ; fey (fey: 
x. 1,13), prv (ffrv: xiv. 2. 26), pvy (ppvy: viii. 3. 15), ppm (f$m; v. 31. 
8), SSV (PW- v - 5 - 8). ff r {99 r '- I"- !'• 2 ), ^ OtfJ: v. 20. 7), ffv (ffv: 
viii. 5. 11); shkr (shshkr: ii. 34. 1), shty (shty: i. 12. 1), shtr (shtr: viii. 
2.27), shtv (shty: i. 22. 3), shthy (shthy: i. 9. 3), shny (shny: i. 3. 1), 
shnv (shnv: iii. 19. 5), shpr (shshpr: iv. 10. 4), shmy (shshmy: ii. 32. 3) ; 
sty (sty : ii. 32. 3), str (str: i. 8. 1), stv (stv: i. 10. 3), sthn (sththn: iv. 12. 
1), sthy (sthy: iv. 12. 3), sphy (ssphy: xi. 3. 9), smy (ssmy: iv. 32. 6), svy 
(ssvy: v. 28. 10), ssk (ssk: ix. 7. 3), sst (sst: i. 8. 3), ssth (sslh: i. 31. 2), 
ssn (ssn: vi. 115.3), ssp (ssp: iv. 16.4), ssm (ssm: v. 22. 10), ssy (ssy: 
ii. 10. 7), ssr (ssr: ii. 3. 3), ssv (ssv: i. 19.3); fpr (<ppr: i. 7. 5), <fpl (q>pl: 
ix. 7. 12). 

Any (h"nny: x. 8. 18), Aray (h" n ny: vi. 110. 3), Avy (hvvy: iv. 17. 2). 

III. Groups of four consonants : 

kshmy (kkshmy: i. 18. 1), nkshn (nkshn: v. 20. 1), nkshv (nkshv: iii. 
12. 6) ; t-str (thstr: vi. 11. 1), ntry (ntry: ii. 31. 4), ntst (ntst: ii. 35. 2), 
ntsth (ntsth: v. 2. 4), ntsp (ntsp: viii. 3. 7), ntsv (ntsv: iv. 5. 1), ndry 
(ndry: x. 2. 9); rkshy (r'kkshy: vii. 85. 1), rtvy (r'ttvy. viii. 10. 22), rtsy 
(r'tsy: x. 1. 21), rshny (r'shny : vi. 24. 2) ; xkshv (xkshv : viii. 3. 7), shtry 
(shtry: iv. 1. 2), sstr (sstr: vii. 95. 3). 

There is no group of five consonants in the Atharva-Veda : if, how- 
ever, the order of two words in i. 8. 1 had been reversed, we should 
have had a group of six, viz. ntstry (pum&nt stry akah). The fact de- 
serves to be remarked here, although familiar to all students of the 
Vedas, that by no means all the groups of four and three consonants, 
or even of two, were, in all the cases of their occurrence, actually such 
groups to the makers of the hymns : in a majority of the passages 
where a y or v follows two or three other consonants, and very fre- 
quently where they follow a single consonant only, they are, as the 
metre shows, to be read as i or u, or i or ■&, constituting separate sylla- 
bles. Those combinations which seem most difficult of enunciation are 
thus often relieved of a part or the whole of their harshness. Earely 
(as at v. 28. 10), an apparent group of three consonants is to be resolved 
into two separate syllables. 

add. note 4.] Atharva- Veda Prdtigdkhya. 691 

4. Longer Metrical Passages cited by the Commentator. 

In the notes on the text, I have passed over two extracts of consider- 
able extent, made by the commentator from unnamed sources, and I 
now offer here the text of them. The first is found under rule i. 10; it 
runs as follows : 

apara aha: caturtho hakdrene Hi: 

I. 1. pancdi 'va prathamdn spar fan dhur eke mantskinah, 
teshdm gunopasamcaydd dnyabhdvyam pravartate. 

2. jihvamuliyapashasa upadhmdniyapancamdh, 

etdir gundih samanvitd dvitiyd iti tan viduh. 

3. ta eva saha ghoshena trtiyd iti tan viduh, 
ushmand r.a dvitlyena caturthd iti tdn viduh. 

4. prathamdh saha ghoshena yadd syur anundsikdh, 

tdn ahuh pancamdn spar fans lathd varnagundh smrtah. 

5. na tu hi vyanjanasamdhir asamyogo bhavet punah, 
samyogaf ca prasajyeta krarno vdcyah punar bhavet. 

6. dvitvaprdptif caturtheshu hakdro hy alra kdra, 
dvittyeshu tu tan na 'sti sasthdne tannivdrandt. 

1. pippalyddishu yad dvitvam svardc chabdavidhih krtah, 
jnapakanf ca dvitiydndm dvitvaprdpter iti sthitih. 

8. gunamdtrd nu tatrdi 'shdm apuriiam vyanjanam kvacit, 
apurne vyaftjane kramah samyogaf ca kuto bhavet. 

9. prthaksattvdni pafydmas tulyalingdni kdni cit, 
na teshdm lingasdmdnydd ekatvam pratij&yate. 

10. sattvaprthaktvdd dvdilingyam yad eteshu nibodha tat, 
tathdi 'va paHcavargena gunamdtrena tulyald. 
The other passage is the commentator's introduction to the fourth 
chapter : 

samdsdvagrahavigrahdn pade yaiho 'vdca chandasi fdkatdyanah; ta- 
thd vakshydmi catushtayam padam ndmdkhydtopasarganipdtanam. 
II. 1. dkhydtam yat kriydvdci ndma sattvdkhyam ucyate, 
nipdtdf cddayah sarva 1 upasarg&s tu prddayah. 

2. ndma ndmnd , nuddttena samastam prakrtisvaram, 

na yushmadasmadvacandni na cd "mantritam ishyate, 

3. ndmd 'nuddttam prakrtisvaro gatir anucco vd ndma cet sydd 

uddtlam, kriydyoge galih purvah,' 
samdso ydvanto 'nuccdh samarthdns tdn samasyate. 

4. yatrd 'neko 'nuddlto 'sti paraf ca prakrtisvarah, 
dkhydtam ndma vd yat sydt sarvam eva samasyate. 

5. sopasargam tu yan nicdih purvam vd yadi vd param,* 
uddttena samasyante tathdi 'va supratishthitam. 

6. uddttas tu nipdto yah so 'nuddttah kvacid bhavet, 
samasyante tathd vidhim itihdso nidarfanam. 

7. naghdrishdm susahe 'ly evamddiny uddharet, 
sahe Hy anend 'nuddttam param ndma samasyate. 

8. anuddttena co 'ddttam svabhdvo yatra co 'cyate, 
sahasuktavdkah sdntardefdh gatakrato nidarganam. 

1 cddayo satva. * gatuhurvah. * purvavdyadiv&padah. 

vol. vii. 75 

592 W. D. Whitney, [add. note 4. 

9. anuddtto gatir 1 madhye purvapardu prakrtisvardu, 
p&rvena vigrahas tatra purushe 'dhi samdhitdh. 

10. uddttdnugatir yatrd 'nuddltam param padam, 
purvena vigrahas tatra sam mbhfatyd nidarpanam. 

11. yatro 'bhe prakrtisvare pttrvam yac ca param ca yat, 
varjayitvd "dyudattani sarvam eva samasyate. 

12. na "kkydtdni samasyante na ca "khydtam ca nama ca, 
ndma ndmno 'pasargdis tu sambandhartham* samasyate. 

13. na yushmadasmadddepd anuddttdt paddt pare, 
namopasargagatibhih 3 samasyante kadd cana. 

14. mam anu pra te pra vam ilyevamddiny uddharet, 
etadargconudattani idam asya tathdi va ca, 

15. namopasargagatibhih 3 samasyante kada cana, 

brhann esham ya en&m vanim dyanti pary endn pary asye Hi 

16. anuddtto gatih sarvdih samastah svaritddibhih, 
samsravyena 1 durarmanya dcdrye Hi nidarcanam. 

lY. pra-pard-ni-sam-d-dur-nir-avd-'dhi-pari-vini ca, 
aty-ubhyapf-su-'d-apd ya upd'-'nu-pratir vinpatih. 

18. ekdkshard uddttd ddyuddttds tathd 'pare, 
abhi Hy anta upasargdh kriydyoge gatis tathd. 

19. ddyuddttd dapdi Heshdm uccd ekdkshard nava, 
vinpater upasargdndm antoddttas tv 'abht 'ty ayam. 

20. achd- 'ram- astam- hasta- IdngAla 7 - tirah-puralf-punar- jtamah'- 

-kshiti-vdti , "-phalt-hin-srug ll -vashat-prddur l ''-uld-kakajd- 
-svdhd-svadhd-prat-svaralale Hy upasargavrtlini yalhdstdta- 
Of this passage, the preliminary sentence (or verse) has been already 
translated, in the introductory note to the fourth chapter. The first 
verse gives a definition of the four parts of speech — or rather, a defini- 
tion only of the verb and noun, the prepositions and particles being 
referred to as the two series commencing respectively with pra and ca 
(compare Pan. i. 4. 57, 58, and the lists of prepositions in the other 
Prati§akhyas). Thenceforward, from the second to the sixteenth verse, 
inclusive, the extract is occupied with giving directions for the combina- 
tion or non-combination, in the pada-text, of the various parts of speech, 
as they stand in juxtaposition with one another in connected discourse. 
The text of the Veda is assumed as existing in an utterly disjoined state, 
each independent element being known in its phonetic form (including 
its accent), and as such and such a part of speech ; and the attempt is 
made to, define the cases in which the elements form compounds with 
one another. The problem, however, is evidently much too extensive 
and difficult to be so briefly solved — if, indeed, any solution of it is 
possible without taking into account also the inflectional forms of the 
nouns and verbs — and the system of rules laid down is only fragment- 
ary : but I have not taken the trouble to test them by the text, so as to 

1 anuddttonugavir. s samvadhdrtham. s namop. 

* aaihsdvyani. 5 abhipi. * apdydnnityupd. 

' hastoldng&lam. " punak ' navah. 

M vdkshiti. " hinsralc. a gudur. 

add. note 4.] Atharva- Veda Prdliqdkhya. 593 

determine where their deficiencies lie. The second verse informs us 
that a noun which has its natural accent is compounded with another 
which is unaccented, except in the case of an enclitic pronoun of the 
first and second persons, or a vocative : thus we must combine tri : sap- 
td'h (i. 1. 1), patd : vrshnyam (i. 3. 1), into tri-saptd'h, patd-vrshnyam. 
The next verse and the one following treat of the combination of nouns 
and verbs with prepositions and other words employed as prepositions : 
these receive here and in what follows the name gati, which Panini also 
(i. 4. 60 etc.) uses. Verse five has for its subject such compounds as 
are instanced by s&pratishthitam (xii. 1. 63 : p. sd-pratisthitam). The 
sixth verse applies to compounds into which a particle enters as a con- 
stituent, and cites ilihdsdh (xv. 6.4: p. itiha-dsdh) as an example ; and 
the first line of the next verse adds the farther instances naghdri- 
shd'm (viii. 2. 6 ; 7. 6 : p. nagka-rishd'm) and s&saha (vi. 64. 3 : p. s&saha). 
Hence to the end of verse 8, compounds with saha (and its substitute 
sa) are defined, the chosen illustrations being sahdsitktavakah (vii. 97. 6 : 
p. sahas&ktavdkah) and sd'ntardepdh (ix. 5. 37 : p. sd-antardepdh) : pa- 
takrato is also added, but apparently only by a blunder; or rather, the 
reading is probably false and corrupt, as the metre helps to show. 
Verse 9 prescribes the mode of combination in such a case as purushi 
'dhi samd'hitdh (x. 7. 15), where an unaccented preposition stands be- 
tween two others, both of which have their proper accent. The next 
verse takes for its example sdm stibhiityd (iii. 14. 1), where two accented 
prepositions precede an unaccented noun, and the former of them is to 
be made independent. Verse 11 has no example, and, although easily 
enough translated, its meaning is to me obscure. We are then told 
what combinations are possible : verbs are not compounded with verbs 
or with nouns ; but nouns with nouns and with prepositions. A verse 
and a half follows, denying the enclitic forms of the first and second 
personal pronouns the capability of entering into compounds, and citing 
as instances md'm dnuprd te (iii. 18. 6) and prd vdm (e. g. vii. 73. 5) : 
and three additional lines extend the same exception to the enclitic de- 
monstrative pronouns, and give the examples brhdnn eshdm (iv. 16. 1), 
yd endm vanlm dydnti (xii. 4. 11), pdry enan (ix. 2. 5), and pdry asya 
(xv. 12. 7). The sixteenth verse declares an unaccented preposition 
capable of composition with a following word, however accented, but 
illustrates only their composition with a circumflexed word, as samsrd- 
■vyena (e. g. i. 15. 1), durarmanyah (xvi. 2. 1), and dcdryah (xi. 5. 3 etc.). 
The subject now changes, and verses 17-19 give us a list of the twenty 
prepositions and a definition of their accentuation. Finally, the last 
verse (or prose passage) attempts to give a list of those words which 
are treated as if they were prepositions, although properly belonging to 
other classes. This list is a somewhat strange one, with regard both to 
what it includes and what it omits. A part of the forms which it con- 
tains are in frequent use, and familiarly known as bearing marked 
analogies with the prepositions proper. Such are achd, tirah. purah, 
puna/i, kin, prddith, and prat. Others, as aram and astam, are more 
remotely connected with the same class. Vashat, svdhd, and vdtt are 
in the Atharvan compounded only with kdra and krta, and hardly in 
such a manner as should require their inclusion in the list. Svadhd 

594 W. D. Whitney, [add. note 4- 

and sruk form no other compounds than svadhdkdra and srukkdra (ix. 
6. 22) ; phali forms phalikarana (xi. 3. 6) ; kakajd forms kakajakrta (xi. 
10. 25). Ndmas enters into namaskdra, namaskrta, and namaskrtya 
(vii. 102. 1), which last affords actually good ground for special treat- 
ment, as does hastagrhya (e.g. v. 14.4), on account of which hasta is 
ranked with the others. For lahgula, kshitt, and uld, I can find noth- 
ing at all in the Atharvan : there is room in the case of the two last, 
and especially of the third, to suspect corrupted readings. What may 
be hidden in svaralald, I have not been able to discover, nor how the 
last word in the extract, which apparently has to do with the accent of 
the words treated of, is to be amended into intelligibility. There are 
two words which we especially miss in this list of upasargavrtttni, and 
can hardly believe to have been originally absent from it: they are 
antah and Avih: I cannot, however, find by emendation any place for 
them in the text as it stands. 

5. Corrections and Emendations. 

I add here a rectification of certain errors in the body of the wort, 
which have attracted my attention as I have been engaged in preparing 
the indexes ; as also, corrections of such errors of the press as I have 
hitherto discovered. 

At p. 355, rule i. 24, for fsr^TJT read fi*|m 

At p. 371, rule i. 47, translation, for palatal read lingual. 

At p. 378, 1. 16, for kdrpanyam read kdrpanyam. 

At p. 409, in commenting upon ii. 15, 16, I failed to notice that the 
implication in the former rule is of catavargdbhydm, ' after palatal and 
lingual mutes,' and mutes only : hence this rule has nothing to do with 
any cases practically arising in the conversion of pada into sanhita, and 
rule 16 covers all the lingualizations of t and th after sh, whether in the 
same or a different word. 

At p. 419, rule ii. 29, at the end, for °<CMl<^ read "d^MW. 

At p. 436, comm. to ii. 62, 1. 4, the hyphen representing the avagraha 
has dropped out from vifvatahpdnih. 

At p. 436, comm. to ii, 62, 1. 28 : another like instance of repetition 
is found at xx. 132. 12, where all the MSS. read punah punah. 

At p. 437, rule ji. 64, translation : for t« read does. 

At p. 442, rule ii. 76, at the end, read °q^ (broken letter). 

At p. 448, under rule ii. 92, 1 was so heedless as to refer to vy dsthan 
(xiii. 1. 5) as an anomalous form from sthd, forgetting for the moment 
that it was, in fact, the irregular aorist of the root as, although I had 
formerly interpreted and indexed it as such. The ^arfa-reading is v{ : 
dsthan, as given by our manuscript ; but it seems to require amendment 
to vl : as that, having rohilah as subject. 

At p. 450, rule ii. 97, comm., 1. 3 ; for stusha u shu read stusha u shu. 

At p. 456, comm. to iii. 5 : a compound analogous with oshadhi-ja is 
prthivisad (pjrthivishadbhyah : xviii. 4. 78), which has also a claim to 
the attention of the Praticakhya in this part. Its sh was prescribed 
by ii. 100, 

add. note 5.] Atharva- Veda Prdlicdkhya. 595 

At p. 514, at the top, introductory note to chapter iv, I have ex- 
pressed myself in a manner which misrepresents and does injustice to 
the Vaj. Pr. Although that treatise does not make the restorations of 
normal form in pada the subject of detailed treatment, yet its rules iii. 
18, 19 (as pointed out in the note to iv. 74, above) virtually cover the 
ground, with more or less completeness. 

At p. 539, under rule iv. 53, I have omitted the reference belonging 
to rks&mabhy&m : it is xiv. 1. 11. 

At p. 544, rule iv. 59 would have been more accurately translated "nor 
is division made before any member having an inserted s as its initial," 
etc. Whether, however, there is any propriety in regarding the inserted 
sibilant of iuvishtama, and of the other words cited, as the initial of 
the second member of the compound, is very questionable. 

At p. 583, 1. 32, for vavrdh&nah-iva read vavrdhanak-iva. 

The indexes call for but a few words of explanation. 

Into the first I have admitted all distinct references to single passages 
of the Atharva- Veda, made by either the text or the commentary of the 
Praticakhya, distinguishing those of the latter always by an appended 
c. Words or brief phrases found at two or more places in the Athar- 
van (and so referred in the notes to the first of them, with a prefixed 
e. g.) are made no account of unless they are of peculiar and distinctive 
character ; and, when noticed, they are marked here also by an added 
e. g., or, if found only in a series of passages occurring in the same connec- 
tion (as in different verses of the same hymn), by an added etc. I have 
also included in the index all passages to which important reference 
has been made in the notes on the text, as for the purpose of amending 
a reading, giving account of an emendation made in the edited text, 
stating the manuscript authority favoring or opposing a given reading, 
or the like : these are distinguished by an n affixed to the Praticakhya 

The second, or Sanskrit index, comprises, in the first place, the whole 
vocabulary of the treatise itself, both its grammatical phraseology and 
its citations from the Atharvan, the latter being denoted by a prefixed 
a. I have added as much of the vocabulary of the commentary as 
seemed to me worth the trouble, adding always a c to the reference. 
The pseudo-citations of the commentator, or the illustrations which he 
fabricates or derives from other sources than the Atharva- Veda, are also 
([excepting the phrases given in the latter part of add. note 2) included 
in the index, and marked with a prefixed q. This same indicatory 
letter is set before the few words quoted in the text of the treatise which 
are not found in the Atharvan. 

In the third index it has not seemed to me worth while to make de- 
tailed references to the doctrines of the other treatises referred to in 
the notes; they may always be found stated in connection with the 
treatment of the related subject by our own Praticakhya. 


W. D. Whitney, 


1. Index of Atharvan Passages, 

Cited or referred to in the Text, Commentary, and Notes. 

An e. g. added to the Atharvan reference marks it as denoting one of two or more 
passages where the word or phrase cited is found; an etc., as denoting the first 
verse of a hymn or passage to which the reference applies. A c added to the 
Praticakhya reference marks it as belonging to the commentary; an n, to the 
editor's note. The passages of the commentary given in add. note 4 are indicated 
by 4. II. 16 etc. 

AV. i. 






1 etc. 





















ii. 98c, iv. lc. 

i. 79c. 

i. 82, iv. 3c. 

iv. 9c. 

iii. 16. 

ii. 32c, 90c. 

iii. 16c. 

i. 5c. 

ii. 84c, iv. 75c, 

ii. 32c. 
ii. 90c. 

iii. 18c, iv. 29c. 
ii. 101c, iv. 78c, 

iv. 107c. 
iii 70c. 
iv. lc. 
ii. 27. 
iv. lc. 
ii. 66. 
ii. 5c. 
ii. 106c. 
U. 30. 
ii. 40c. 

i. 101c, iv. 79c. 
ii. 60. 

a 40c. 
a 30c. 

ii. 40c, 62c. 

iv. 3c, 4.IL16. 

iL 30c. 

iv. lc. 

iii. 13c. 

ii. 98c. 

ii. 27c. 

a 48c. 

ii. 40c, iv. 56c. 

iv. 92c. 

iv. 92c. 

ii. 27c. 

i. 28c, iL 19c, 97c, 


98, 114c. 
iii. 52. 
iv. 21c. 

ii. 91c, iv. 82c. 
ii. 38c, iii. 13c. 
iv. 3c. 

AV. 1. 

AV. ii. 









ii. 96c. 


ii. 101c. 


iv. 72c. 


iv. 57c. 


iii. 12c. 


ii. lie, iv. 49c 



ii. 5c, iv. 16c. 


g. iii. 34c. 


a 84c. 















15. 1 etc 



a 2c. 

ii. 73c. 

ii. 94c. 

iv. 5, 14c. 

a 70c, iv. 34c. 

a 70c. 

UL 9c. 

i. 103c, ii. 38c, iv. 

ii. 40c, 40n. 
a 48c, iii. 70c. 
ii. 82c, ia lc, 13c, 

iv. 70c. 
i. 94n, iii. 38c, 

66c, iv. 113c, 

114c, 115c. 
iii. 18c, iv..29c. 
ii. 14c. 
ii. 94c. 
a 27c. 
iv. 55c. 
iv. 16c. 
ii. 46c. 
ia 80n. 

iii. 62c, 67c, 74c. 
ia 62c, 67c, 74c. 
iii. 56c. 
ii. 15c. 
i. 28c, a 19c, iii. 

ffi. 16c. 
ii. 44c. 
iv. 37c. 
i. 45c, 68c, iv. 41c, 

75c ' 
iv. 26n. 
ii. 84c. 

AV. ii. 








AV. ia 





















ii. 55c. 

i. 3c etc. 

i. 45c, 68c. 

iv. 107c. 

i. 47c, iii. 27c. 

iv. 18c. 

iv. 33c 

iv. 31c 

i. 66c, iv. 31c. 

ii. 85. 

ii. 67. 

iv. 30. 

iL 31c, 85n. 

iv. 35c. 

ii. 57c 

a 33c. 

ii. 57c. 

ii. 10c, 17c, iii. 

18c, iv. 29c. 
ia 27c. 
iv. 91c. 
iv. 56. 
iv. 20c. 
ia 76c 
i. 27c, 53c, 83c, a 

iii. 91c. 
iii. 80c 
a 98c. 

iii. 16n (to 1. 1. 6). 
ia 22, iv. 89. 
i. 78c, ii. 40c. 
i. 99c 
iv. 29c. 
ii. 104c 
iv. 33c. 
ii. 63c. 
ii. 72. 

ii. 84n, iv. 107c. 
ii. 92. 
ii. 87c. 
iii. 64c. 
iii. 17c. 
ii. 106c. 
iii. 59c, 65c 
ii. 2c, 45c. 
ii. 44c. 
ii. 44c. 

Atharva- Veda Prdtigdkhya. 


AV. iii. 

AV. iv. 

AV. v. 


ii. 32c, 4. II. 10. 


ii. 40n. 


ii. 91c, 103c, iv. 


iii. 52c. 


ii. 31c, iii. 84c, iv. 

82c, 124c, 125c. 


iv. 97c 



iv. 98c. 


iv. 97c. 


i. 66. 


. iv. 96. 


ii. 48c. 


iv. 83. 


iv. lc, 75c 


iii. 17c, 22c, ir. 


i. 47c. 


i. 75c, it 22c, iii. 



ii. 39c. 

33c, 56c. 


ii. 55c. 


ii. 59c, iv. 100c. 


it 53c. 


iii 6c. 


. ii. 63c. 


iv. 30. 


iv. 85c. 


ii. 80c. 


iv. lc. 


4. II. 14. 


iv. 58c. 


ii. lie. 


it 7c. 


ii. 85c 


iv. 96. 


i. 75c, ii. 22c, iii. 


i. 82c. 


iii. 90c. 

33c,ir. 49c. 


i. 88c, iv. 36c. 


L 68c, ii. 29c. 


ii. 103. 


iii. 91c 


iv. 34c 


iii. 54c. 


ii. 23. 


iv. 14c. 




i. 81c, iii. 87c. 


iii. 71c 


iv. 4c. 


i. 74c. 

6 etc 

. ii. 91c, iv. 82c. 


iii. 64c 


i. 81c, iii. 40c. 


iv. 65. 


iv. 48c 


iii. 21n. 


i. 81c. 


ii. 84. 


ii. 68c. 


i. 68c, ii. 42c, iii. 


ii. 74n. 


iv. 69c. 


26. 1 etc 

. iii. 65c 


iv. 18c. 


iii. 72c. 


iii. 3c. 


iv. 67c 


iii. 3c 


iv. 33c. 


iv. 16c. 


iii. 81. 


iv. 45c. 


iv. 57c. 


ii. 104c 


ii. 35c. 


iv. 11. 


i. 27c, 53c, 83c, ii. 


i. 63c, it 2c, iv. 


iii. 80n. 






ii. 39c. 

3 etc 

. ii. 35c 


i. 81c. 

17. 1 (?) 

iii. 53c, 55c 


iv. 50n. 


iii. 15, iv. 88. 


ii. 61c. 


iv. 56c. 


iii. 56c. 


ii. 2c. 


ii. 32c 


ii. 47c, iii. 14, iv. 


iv. 107c, 109c. 



iv. 96. 

AV. iv. 


i. 68c, 81c, ii. 29c. 


ii. 55c. 


ii. 44c, 103c. 


i. 69c. 


iv. 22c. 


ii. 5c, iv. 71c. 
iv. 26c 


iv. 12c, 30. 


ii. 40c. 



ii. 105c. 


iv. 4c. 


ii. 56, iii. 6c, iv. 


iii. 46c. 


iii. 2c, iv. 94c. 


5 etc 

. iv. 38c 


it 61c 




ii. 80n. 


iii. 13c, iv. 84c. 

iii. 46n. 


ii. 9c, iii. 18c. 


iv. 18c. 


iv. 16c. 


ii. 31c. 


ii. 69c. 


ii. 93c. 


i. 28c, ii. 19c, iii. 


ii. 7c 


iv. 86c. 



iii. 89c, iv. 107c 


ii. 95. 


iii. 59n, iv. 67c. 


ii. 62. 


ii. 84c. 

38. 5 e. g 

. ii. 35c, iv. 81c. 


ii. 23, 82c, iii. lc, 


ii. 62c, 85c 

39. 1 etc 

. iii. 49c. 

iv. 70c 


iv. 86c. 


ii. 6c, 43c 


ii. 42c 



2 etc 

ii. 75. 


ii. 83c 


iii. 94c. 


ii. 20c. 


i. 34c. 

9. 9 e. g. ii. 67c. 


ii. 62c. 


ii. 85c. 


ii. 67c 


ii. 40c. 


iv. 5c. 

AV. v. 


i. 46c. 


ii. 62n. 


ii. 5c. 


ii. lie. 


iv. 12c. 


i. 76c, 77n. 


i. 103c, ii. 38c, iv. 


iii. 9c. 


i. 76c, ii. 97c, iii. 



iv. 48c. 

4c, 13c, iv. 98c. 


iii. 93c. 


ii. 54c. 


ii. 49. 


ii. 104c. 


i. 85c. 


iv. 54n. 


iv. 18c. 


ii. 22c 


iv. 86c 


iii. 34n. 


iv. 53c. 


ii. 70c 


iv. 45c. 


it 87c. 


ii. 102c. 


ii. 47c. 


iv. 31c 


iii. 38c, 66c. 


i. 71c, ii. 13c. 


iv. 84c 


iv. 14c 


iv. 15c. 


iii. 38c, 66c. 


iii. 27c 


ii. 77. 


i. 96, 96n, 105. 


ii. 63c. 


i. 66c. 


ii. 5c, iii. 34n, 


iv. 99c. 


iii. 7. 

4. II. 15. 


ii. 65c. 


iv. 96. 


W. D. Whitney, 

AV. v. 

AV. vi. 


iv. 96. 


AV. vi. 



i. 80c. 



i. 74n. 



i. 37e, ii. 

85n, iv. 





ii. 74c, 74n, 80n. 



iv. 38c. 



iv. 93c. 



i. 27c, 53c, 83c, ii. 



103. 2, 3 


i. 91c, iv. 




i. 39c, ii. 

88, iv. 






ii. 94c. 



iii. 60c. 


15. le 

g. iv. 16c. 



iv. 56c. 






iv. 107c. 


17. 1 etc. ii. 103c. 



ii. 91c, iv 




ii. 78. 



iv. 124c. 



iv. 46c. 



i. 74c. 



iii. 95c. 



i. 47c. 



iv. 56c. 



i. 46c. 



iii. 34n. 

121. 1 e 


ii. 48c. 



ii. 40c. 



iii. 96, iv. 




iv. 4c. 



iv. 32c. 



iii. 24c, iv. 18c. 



i. 68c, ii. 




ii. 36c. 



iv. 78c. 



iv. 20c. 



iv. 16c. 


ii. 65c. 

AV. vii. 


ii. 28 53c 
iv. 55c. 






iv. 107c. 



iv. 86c. 



ii. 26c. 


iv. 56n. 



ii. 37c, iv 




ii. 21c, 24c, iii. 


35c, iv. 




i. 80c. 



iii. 24c. 



ii. 72n. 



iv. 39c, 4. II. 7. 



iv. 67c. 



ii. 15c. 



iii. 52c. 



ii. 15c, 21c, 24c, 


iii. 35c, 38c, 66c. 



iv. 18c 



ii. 72c. 


71. le 

g. ii. 85c. 



iv. 84c. 



ii. 40c. 



iv. 55c. 



iii. 78c. 


AV. vii. 

ii. 5c. 


iv. 32c. 

ii. 65c. 


ii. 18c. 

iii. 80n. 


ii. 46c 

ii. 44c. 


iv. 30. 

iv. 19c. 


iv. 9c. 

ii. 40c, 66c. 


iii. 80n. 

iii. 2, iv 



iii. 78c. 

i. 17c. 


iv. 77c. 

iii. lie, 

iv. 61c. 


ii. 48c. 

iii. 56c. 


iii. 18n. 

i. 88c, iv. 36c. 


i. 79c, iv. 115c. 

it 46c. 


iii. 69c. 

i. 78c. 


iii. 71c. 

iii. 69c. 


ii. 2c. 

ii. 30e, 

v. 78c. 


iv. 64, 83. 

ii. 65n, 



i. 69c. 

iii. 61n. 


ii. 48c, iv. 38c. 

iv. 99c. 


iii. 58c. 

iii. 65n. 


ii. 86. 

iii. 59c, 



iv. 93c. 

iii. 86c. 


ii. 97c, iii. 4c 

iii. 69c, 

iv. 84c. 


ii. 48c, iv. 71c. 

iv. 50c. 


ii. 107. 

ii. 101c. 


ii. 65c. 

iv. 13c. 


iv. 96. 

ii. 2c. 


i. 88c, iv. 36c 

iii. 14c. 


iii. 34n. 

iii. lie, 

iv. 61c. 


iv. 12c, 4. II. 8. 

iii. 85c. 


i. 94n. 

iv. 81c. 


ii. 99c, 105c, iii 

iv. 72c. 


g. ii. 86, 86n. 


iii. 93c 

i. 34c. 


i. 85c. 

iii. 58c. 


i. 99c 

iv. 14e. 


iv. 28c. 

ii. 51c, 

v. 52c. 


ii. 56. 

i. 85c. 


iv. 77c 

iii. 43c. 



i. 71c. 

iv. 93c. 

AV. viii. 

iii. 59c. 


iii. 82c 

ii. 65. 

2.6e.g. iii. 6, 4.II.7. 


ii. 76. 


iii. 76c, 91c 

i. 74c. 


iv. 34c 

ii. 15c. 


ii 32c 

iv. 26c. 


iii. 80c 

ii. 97c, iii. 4c, iv. 


i. 28c, ii. 19c, iii. 



iv. 75c. 


ii. 32c 

ii. 41c. 



ii. 26c. 


ii. 102c 

iv. 72c. 


iii. 21c, iv. 90c. 

iii. 89c. 


iv. 34c. 

ii. 31c. 


iv. 8. 

ii 44c. 


Hi 63c 

iii. 61. 


iii. 10c 

i. 65c. 


ii. 9n. 

ii 71c. 


iii. 84c 

iii. 56c, 



iii. 84c 

iii. 94c. 


iv. 25c. 



iv. 18c 

iii. 56c. 


i. 52c 

iii. 76. 


ii. 35c. 

ii. 58. 


iv. 22c. 

iii. 5c. 


ii. 94. 

i. 94n. 


iv. 56c. 

iv. 60c. 


ii 7c, iii. 52e. 

ii. 40c. 


Iii. 3c, iv. 68c 

Aiharva- Veda Prdticdhhya. 


AV. viii. 

AV. ix. 

AV. xi. 


ii. 52. 


i. 103c, ii 

38c, iv. 


iii. 94c. 


iv. 18c. 

26c, 55c, 83. 


i. 94n. 


iv. 18c, 55c. 


iii. 60c. 


i. 94n. 


ii. 74n. 


ii 93c 


iv. 42c 


iii. 82c. 

21(?) i. 52c. 


iv. 58c 


iii. 76c, iv. 77c. 


ii. 4c 


iv. 2c, 7c 


iv. 16c. 


iv. 24c 


i. 87n. 


ii. 95. 


iv. 15c 


iv. 27c. 


i. 50c. 

12 etc 

. iii. 46c, 65c. 


ii. 2c. 

AV. x. 


iii. 65c 


i. 77n, ii. 8c. 


iv. 99c. 


iv. 49c 


ii. 38c 


iv. 95c. 


iv. 86c 


ii. 62. 


iii. 65n. 


iii. 51c. 


ii. 96c. 


iv. 91c. 


i. 105, 105n. 


i. 39c. 


i. 38c. 


iv. 6. 


ii. 39c. 


iii. 35n. 


i. 105. 


i. 23c, ii. 9c 


ii. 40c. 


iii. 55n. 


ii. 4c, 40c. 


i. 94n. 


iv. 57c 


iv. 94c. 


iii. 18c 


iii. 2c iv. 94c 


iii. 46c, 64c 


ii. 18c 


iii. 70c. 


iv. 21c. 


i. 75c, iii. 



ii. 8c. 


iv. 55c. 



iii. 69c. 
iii. 72c 

3 etc 

i. 17c, iii. 57c, 4. 
II. 16. 

AV. ix. 


iv. 39c. 


ii. 51c, iv. 52c. 


ii. 68c. 


iii. 19c 


i. 84c, ii. 34c, 89c 


i. 87c. 




ii. 13c 


iii. 80c, 4. II. 15. 

iv. 6, 120c, 121c. 


ii. 62. 


iv. 92n. 


iii. 2c, iv. 



iv. 51c, 80. 


ii. 84c, iv. 33c. 


iii. 80c. 


ii. 50c. 


iii. 89c. 


ii. 20c. 


iv. 84c. 

10 e. 8 

. iv. 58c. 


i. 66c. 




i. 4c, 61c. 


iii. 60c. 


iv. 42c. 


iii. 2, iv. 94c. 
ii. 45c 


iv. 49c. 


iv. 15c. 



iii. 59n. 


i. 65c 


iv. 107c. 


iv. 30. 


iii. 10c. 


iii. 46c. 


ii. 34c. 


iii. 59c, 65c. 


iv. 62c. 


iv. 91c 


ii. 62, iii 73c, 76c, 


ii. 70. 


ii. 43c 

iv. 77c. 


iii. 94c. 

22 e 

g. iv. 83. 


iii. 10c 


i. 87n, ii. 47c, iv. 


iv. 72c 


ii. 42c. 


36 e 

g. ii. 92c 


iii. 76c, iv. 77c 


iv. 60c. 


ii. 43c. 


iii. 55n. 


iii. 19c 


ii. 87c 


iv. 68c. 


iv. 34c. 

5 etc. iv. 40c. 


iv. 44c. 


iv. 10c. 

AV. xii. 


iv. 12c, 4. II. 8. 


ii. 43c. 


i. 85c. 


i. 84c, ii. 34c, 89c 


ii. 65c. 


iv. 4, 48c 


ii 105c, iii. 75c. 


g. iv. 123c 


ii. 20n. 


iii. 79c 


4. II. 9. 


iv. 59c. 


iii. 79c. 


iv. 5c 


iii. 78c. 


iv. 42c 


iii. 82c. 


iv. 91c. 




iii. 92c. 


ii. 82c, iii lc, iv. 


i. 52c, 62c, 105, 


iii. 49c 




iii. 93c 


ii. 47c, 4. II. 5. 


ii. 96c 


ii. 41c 


iv. 43c. 


i. 58c, 100c, ii. 


i. 4c 


iv. 56c. 

41c, iii. 31c. 


iii. 74c 


ii. 100c. 


iv. 7c 


ii. 54c. 


. iv. 95c 


iv. 2c, 7c. 


i. 66c 


iii. 88c 


ii. 42c. 


ii. 100c. 


i. 71c. 


iii. 47c 


iii. 55n. 


ii. 102c. 


iii. 47c. 


i. 20c. 


i. 94n. 


iii. 47c. 


iv. 13c. 


iii. 73c. 


iv. 5c. 


iv. 27c 


iv. 93c. 

AV. xi. 


ii. 76c 


iv. 99c. 


iii. 51c, iv. 107c. 


iv. 40c. 


iv. 19c 


ii. 8c. 


ii. 20c. 


ii. 102c. 


i. 45c. 


ii. 32c 





W. D. Whitney, 

AV. xii 

AV. xiv. 

AV. xviit 


a 68. 


ii. 18c, iv. 62c. 


i. 74c, a 45c- 


iv. 26c. 


iv. 35c. 


ii. 17c. 


ii. 18c. 


i 50c. 


ii. 97c, iii 4c, iv. 


iii. 34. 


iv. 53c. 



iv. 91c. 


iv. 24c. 


i. 74c. 


iv. 2c. 


ii. 65c. 

45 etc. ii. 59c, iv. 100c. 


ii. 62c. 


iv. 18c, 48c, 76c. 


iv. 34c. 


iii. 92c. 


ii. 92. 


i. 88c, iv. 36c. 


i 28c, ii. 19c, 20c 


ii. 64. 


ii. 100c. 

iii. 20c. 


u. 91c, iv. 82c. 


g. iii 83c. 


iv. 40c. 


ii. 36c, iii. 60c. 


ii. 51c. 


iv. 91c. 




ii. 76. 

55 etc 

. iv. 115c. 


ii. 70. 


iv. 86n. 


iii. 55n. 


ii. 31c, iii. 95c. 


i. 103c, ii. 38c, 


iii. 65n. 


ii. 54c. 


ii. 100c. 


4. II. 15. 


ii. 31c. 


iii. 25c. 


ii. 52c. 


ii. 65c. 


ii. 31c. 


i. 86c, 86n. 


iii. 60c. 


iii. 17c, 46n, iv. 


ii. 15n. 


ii. 20c. 



iv. 37c. 


i. 76e. 


ii. 52c. 


iii. 55n. 


iii. 60c, 


iv. 63c. 


iii. 21, iv. 90. 


iv. 50c. 


ii. 7c. 


iii. 59n. 


ii. 31c. 


iv. 39c. 


i 97, 105. 


iv. 54n. 


iv. 91c. 


iii. 55n. 


ii. 14c. 


iii. 79c. 


ii. 4c. 


iii. 18c. 


i 80c. 


ii. 102c. 


iv. 67c. . 


iii. 65n. 

AV. xv. 


iii. 9c, iv. 67c. 


iv. 62c. 

2. 1 etc 

. i. 90c, ii. 25c. 


iii. 25c, iv. 15c. 


ii. lie. 


iii. 53c, 55c. 


i 82c. 


iii. 92c. 


4. II. 6. 


i. 94n. 


i. 105. 


iii. 53c, 55c. 


i. 80c, 84c, ii 34c, 


iii. 40c. 

89e. 96n. 
iii. 12n. 

AV. xiii 


iv. 62c. 



ii. 9n, 46c. 


4. II. 15. 


iii. 10c. 


iii. 65n. 


iii. 8c. 


iv. 17c. 


i. 50c. 

14 3 

iii. 55n. 


i 94c, iii. 9c, 9n. 


ii. 107c. 


iii. 45n. 


ii. 47c. 


iv. 75c. 




ii. 7c. 


i. 16c. 


iii. 53c, 55c. 


i. 94n, iii 85c. 


ii. 84c. 


ii. 63c. 


iii. 55n. 

AV. xvi. 


. i. 74n. 


iv. 5c. 


ii. 102c. 


ii. 35c. 


ii. 91c, iv. 29c, 


4. II. 16. 


iii. 55n. 



iii. 85c. 


iii. 38c, 66c. 


iii. 27c. 


iii. 55n. 


i 38c, iv. 95c. 


ii. 62c. 


ii. 71c. 


iii. 79n, iv. 2c, 7c. 


iv. 7lc. 


ii. 50c. 


i 37c. 


ii. 69c. 


iii. 9c, 9n. 


ii. 106c, iii. 87c. 

AV. xvii. 


i 94n. 


iv. 57c. 


u. 50c. 


iii. 34n. 


ii. 9n. 


i 65c, ii. 10c, 17c. 


iii. 17n. 

AV. xviii. 


iii. 86c. 


iii. 65n. 


i. 79c. 


ii. 99c. 


iv. 13c. 


ii. 7c. 


ii. 100c. 




iii. 34c, 86c. 

36c. iv. 116c. 


iii. 8c. 

AV. xix. 


iv. 13c. 


ii.97e, iii. 4c. iv. 


ii. 22n. 


iii. 65c. 



iv. 49c. 


i. 74c. 


iii. 73c. 


ii. 22n. 


iii. 36c. 


iii. 25. 


ii. 67e. 


ii. 71. 


ii. 40c. 


i 77n. 


iii. 30c. 


iii. 65n. 


iv. 50c. 

Aiharva- Veda Pr&tig&khya. 


S. Sanskrit Index. 

A prefixed a marks a word or phrase a9 a citation from the Atharva-Veda ; a 
prefixed q, as a quotation in the text or commentary not taken from the Atharvan. 
Other signs are used as in the preceding Index. 

a, iv. 69. 
a. alcar, ii. 44. 

akdra, i. 36, 101, ii. 53. 
92, iii. 13, 53, 55, 77. 
q. akravan, iii. 53c, 55c. 
q. akshatarcah, iii. 32c. 

akshara, i. 14, 93. ci.ekdk-, 
a. agni, ii. 96. 
q. agni, iii. 40c 
q. agnicit, ii. 14c 
q. agrepitva, iv. 26c 

agkosha, i. 12, 13c, ii. 3, 4, 

anga, iv. 16c. 
a. ahguli, i. 66. 

ahgushthaparvamdtra, iv. 
a. angebhyak pari, ii. 67. 

aneati (r. afic), iv. 63. 

r. anj : anuvyanakti, iv. 

amimdtrd, iii. 65. 

aiah, i. 49. 

atigdyana, iv, 16c. 
a. atishthipam, iv. 96. 

alra, ii. 39, 81. 
q. alra, i 56c, 78c, ii. 2c, 20c, 
iii. 39c. 

adrskta, iv. 109. 
a. adka, iii. 25. 

adharakantha, i. 1 9. 

adhardushtha, i. 25. 
a. arfAe. ii. 104. 

adhika, i. 105c. 

adhikdra, iii. 96c. 

adhisparcarn, i. 9, ii. 24. 

adhyayana, iv. 101, 107, 

o»a<« (r. an), iv. 39. 

anantara, iii. 74. 

anartha, iv, 3. 

anahna, iv. 86. 

andrsha, i. 81. 

auihga, iv. 12. 

anihgya, iv. 49c, 76c. 

anucca, 4. II. 3. 

anuttama, i. 99, ii. 2, 20. 

anudarpana, i. 8c. 

anuddtta, i. 15, 96, iii. 57, 
58, 67, 71,74, iv. 2. 

anundsika, i. 8c, 10c (4ja. 
1.4), 11, 27, 53, 67, 69,!q. 
83, ii. 35, iv. 119, 121. la. 

<j«uprarf4«a, i. 12. 
a»MS!idra, i. 26c 
anushman, i. 94. 
aneka, iv. 2. 
anekdkshara, iv, 1 5. 
ante, i. 8, 75, 80,88, ii. 47, 

99, iii. 1,19,53,57,77, 

78, 90, iv. 14, 55, 66, 69, 

70, 91, 107, 111, 124. 

cf. paddnta. 
antahpade, i. 83, ii. 33, iii. 

59^ iv. 88c 
antahahd, i. 30, ii. 32, iii. 

39J58, iv. 107c 
antagata, iv, 112, 117. 
antar, ii. 48,62. 
antoddtta, iv. 26, 30c, 38c, 

4. II. 19. 
antya, i. 63, 92. 
anya, i. 49, 52, 102, iv. 77. 
anyayukta, iv. 3. 
anyayoga, iv. 116c 
apa, ii. 95. 
apanodana, iv. 108c 
apt, i. 35, ii. 16,63,89,92. 

93, iii. 59, 68,72,79, iv. 

apurna, i. 1 0c (4. 1. 8). 
apurva, iii. 57. 
aprkta, i. 72,79, iv. 113. 
aprajd, aprajdta, iv. 56c. 
apraydvan, iv. 56. 
apravina, iv. 107c. 
apluta, i. 97, iv. 120. 
npsu, ii. 100. 
abibhar, ii. 44. 
a&Ai, ii. 104. 
abhidhdna, abhidkeya, iv. 

abkinidhdna, i. 43,44,49c 
abhinipdta, i. 43c 
abhinishtdna, i. 42, 105c. 
abhinihita, iii. 55 intr. n., 

55, 65. 
a&Az vi tanu, iv. 3. 
afiA.?* vipacydmi, iv. 4. 
a6Ae sydma prtanyatah, ii. 

abhyaghdyanti, iv. 96. 
abhydsa. ii. 91, 93, iii. 13, 

iv. 82, 84. 
ami, i. 78. 
amaydvi, iv. 18c 
a»i»aA, ii. 52. 

ay, iii. 40. 

ar, iii. 46. 
q. arkah, i. 58c, iii. 31c. 
q. arcd, i. 68c, iii. 31c. 
r. arch : updrckati, prdr- 

ehati, iii. 48c. 
artha, i.74,iv.l07,108. cf. 

anartha, matvartha. 
r. ardh : drdhnuvan, iv. 

r. ardh: updrdhnoti,prdr- 

dhnoti, iii. 48c. 
ardha, i. 101. cf. mdtrdr- 

ardharca, iii. 74c. 
r. arsh : updrshdti, prdr- 

shall, iii. 48c 
av, iii. 40. 
ava, iv. 30. 
avakrshta, i. 43c 
avagrhya, iv. 42, 44, 49c, 

55c, 108c, 117, 123. 
avagraha, i. 45, iii. 64, 69, 

78, 74c, 85, iv. intr. n., 

iv. 7, 44-46c, 55c, 76c, 

avati (v. av), iv. lie. 
avatla, iv. 105c 
avattam, iii. lie, iv. 61c 
avani, iv. 105c. 
avarna, iii. 44, iv. 56. 
avacd, i. 97, 105. 
avasanna, i. 43c 
avasdna, i. 70, iv. 111c. 
avasdnika, avasita, i. 8c, 

avyaya. ii. 48, iv. 71. 
ashta, iii. 2, iv. 94. 
ashtama, i. 102. 
asfitddanxhtra, iii. 2, iv. 

q. ashtdparna, iii. 2, iv. 94c. 
r. as : samam/ate, iv. 1, 26c, 

27c, 46c, 4. II. 3 -7, Il- 
ls, 15 ; samastah, 4. II, 

asti (r. as), ii. 101. 
asmad, 4. II. 2, 13. 
a. asme, i. 77. 
a. asydh, ii. 70. 

asuara, iii. 74c 
a. aAij*i, altar, ii. 50, iii. 77. 
a. ahdr, ii. 46. 
a. ahordtre, iv. 62. 


W. D. Whitney, 

d, iii. 95. 

dkdra, i. 35, 79, 96, ii. 22, 
27, 55, iii. 38, iv. 14, 1 15. 
dkshipta, i. 16. 
dkhyd, ii. 38c. iv. 39. 
dkhydta, i. l,iv. 1,88c, 92 

dgama, iii. 78, iv. 8c, 59, 

dcamana, iv. 107c. 
dt, ddanta, iii. 1, iv. 70. 
q. ddarpah, iii. 32c. 

ddi, i. 66, ii. 83, iii. 18,28 
48, iv. 18, 29, 50, 58, 59, 
61,63, 107,115:— i. 65, 
66, 82, 85, ii. 25, 27, 29, 
103, iii. 4, 12, 16,21-3, 
34,47,62,76, 92, iv. 3- 
dditah, i. 17. 
ddepa, i. 63, 77c, ii. 84, 4. 

II. 13. cf. ekddepa. 
ddya, iii. 23. 
ddyuddtta, iv. 38c, 4.II.1 1, 

18, 19. 
dntarya, i 95. 
dnpada, iv. 74. 
Anyaiareya, iii. 74c. 
dpatti, i. 68, iii. 58. c£ earn- 
a. dpdka, ii. 94. 
a. d babhuvdzn, i. 70, 97 
105, iv. 6. 
dmantrita, i. 81, ii. 47, iv. 

26c, 4.11.2. 
dmndna, iv. 101c, 103c, 

dmredita, ii. 62c, iv. 40. 
dy, iii. 40. 
a. dyuh, ii. 75. 

dr, iii. 48. 
a. drtni iva, i. 82. 
drsha. cf. andrsha. 
dv, iii. 40. 
a. dear, ii. 44. 
a. dvih, ii. 63. 
a. dfd, iv. 72. 
a. dfi:h, i. 105, iv. 6. 
a. deist, i. 105. 
q. date, iii. 27a 

dxthdpita, i. 48, iv. 126. 
dsya, i. 27 c. 
dA«A, ii. 39. 

r. i. cf. pretya, etc., etc, 
ifoira, i. 96, iii. 66, iv. 69. 
r. ing : ingyate, iv. 26c ; 

ihgita, iv. 49c ; ingya- 

mdna, iv. 76c. 
inga. cf. aninga. 

ingya, iv. 44c, 76, 93c. cf. 

ichd, iii.18.iv. 29. 
a. iddydh, ii. 72. 

iii, i. 2, 8, 66, 70, 77, 96, 
97, 101, 105, ii. 23, 28, 
29,39,45,100, iv. 6, 45, 
a. iti,\. 72, 81,82,97. 

ittva, i. 77c. 
a. i<fam u shu, iii. 4, iv. 98. 
a. iddzm, i. 105. 
a. iyttem, i. 105. 
a. iva, i. 82, iv. 41. 
iuarna, i. 21c, iii 44. 
ilia, i. 2. 

i&orw, i. 74, iii. 57c. 
iya*, i. 89. 
ishatsprshta, i. 30. 

witdra, i. 63, 72, ii. 53, 60, 

iii. 36, iv. 116, 118. 
ucca, 4. II. 19. cf. anucca. 
uecdih, i. 14. 
un (part. «), iii. 4. 
uttama, i. 6, 11, 99, ii. 5, 

20. cf. anultama. 
uitara, iv. 123. 
■uttarapada, ii. 94c, iii. 23 

iv. 60, 107c. 
uttararupa. iii. 74c. 
, ud. ii. 18, iv. 62. 
udaya, iii. 27c, 65. 
uddtta, i. 14, 17,77, iii. 58, 
65, 66, 67, 70, 74, iv. 1 
cf. anud-, antod-,ddyud-. 
■oddttapruii, iii. 71, iv. 

uddharana, i. 48c, ii. 81c, 
iv. 75c, 76c, 121c. 
q. uddhatah, iv. 62c. 
q. ud dhara, iii. 40c. 
upajana, iv. 109. 
upajdta, iv. 10, 12c, 46c, 

upadhd, i. 92, ii. 27, 29, 

42,53,55, 81, iii. 27. 
upadhmdniya, i. 10c, 25c. 
upapada, ii. 28c. 
a. upabaddha, ii. 27. 
a. iiparshanti, iii. 47. 

upasarga, i. 1 , ii. 90, iii. 1 1 , 

48,79, iv. 1,35c, 37, 4. 


upasargavrtti, iv. lc, 2c, 

upasrshta, iv. 36. 
updcdra, iv. 74. 
ubha, ii. 35. 
a. ubhaya, iv. 21. 

MtJanta, iii. 45. 
q. ushtrah, iii. 30c. 

ukdra, i. 74, iii. 60. 
q. ittaye, iii. 40c. 
a. ud/iah, ii. 52. 

Ashman, i. 10c, 31,46, 68, 
101, ii.25,26, 32,33. cf. 
and«A*, ao«A-. 

rfcdra, ii. 99, iii. 85. 
q. rnoti, iii. 27c. 
a. rto, iii. 24. 

rti, i. 101c. 
a. rftinr ut srjate vafi, ii. 29. 

rvarna, i. 20c, 37, 71, iii. 

Vcdra, i. 4. 
Ivarna, i. 20c, 89. 

e&a, i. 40, ii. 47, iii. 41. 
ekapada, iv. 108c, 126. 
ekamdtra, i. 59. 
ekap-uti, iv. 107c. 
ekdkshara, iv. 16, 55, 56, 

4.11.18,19. cf. o«ei-. 
tkddefa, i. 69, iii 66, iv. 

ekdra, i.34,76, 97, iii. 44, 

60, 63, 65. 
eke, i. 33, 101. 
a. eta, iv. 48. 

etdvattva, ii. 28c, 44c 
a. ena, iii. 80. 
a. end ehdh, iii. 84. 

eva, i. 103, ii. 33, iv. 36, 

evam, i. 2, 66, ii. 29. 
a. esha, ii. 57. 

ojfcdra, i. 41, iii. 50. 

okdra, i. 34, 80, iii. 45, 61, 
63,65, iv. 115. 
a. onyoh, iii. 61. 
a. odanfch, i. 105. 
a. onhadhi, iii. 5, 6. 

oshtha. cf. adhardushtha. 

oshthya, i. 25. 

dukara, i. 41, iii. 51. 

ka (k), ii. 9, 62, 87. 

ka (suff.), iv. 25. 

kantha. cf. adharakantha. 

kanthya, i. 1 9. 

r. ton : idmayomdna, iv. 

a. kar, ii. 44. 

karana, i. 18, 18c, 26c, 29, 

60. cf. bhUtakarana. 

Aiharva- Veda Prdticdkhya. 


a. karat, ii. 65. 
a. karam. ii. 65. 

fcaroti (r. iar), iv. 27c, 68 
a. karna, ii. 65. 
q. kartr, i. 4c. 

karman, iv. 101c, 106c. 
karmandman, iv, 29. 
karmapravacaniya, iv. 3. 
karmacexha, iv. 101c 
kanhana, ii. 39. 
q. karshali, iii. 32c. 

kavarga, i. 20c, ii. 40c, 87c 
Jtovt, iii. 65. 
q. kdmartah, iii. 46c. 
a. kdmya, iv. 40. 

kdritdnta, iv. 9 1 . 
q. kdrpanya, i. 65c. 

id/a, ii. 39. cf. kramak; 
padak-, parihdrak: 
a. i«r«, ii. 65. 
a. &nw<M, ii. 65. 

fcr/, iv. 26c, 27c, 35c, 45c. 
a. kr/i, ii. 65. 
a. £r<i>a, iv. 27. 
a. £jo7m, ii. 65. 
a. krpd, i. 65. 

krpi(r.klp), i. 6t, iv. 86. 
feiWa, iii. 36, 38, iv. 113c. 
kecaoeshta, iii. 43c. 
a. kemraprdbandhdydh, iv 

q. koshanam, iii. 75c. 
q. kratvd, ii. 64c. 

r. fcram : anukramishyd 

mah,ii. 81c; on«iran<o, 

ii. 81c; upakramyate, i, 


krama, i. 10c (4.1.5, 8), iv. 

kramakdla, iv. 109c, 117c. 

kramaja, i. 58. 
kramapada, iv. 110, 122c. 
kramavat, iv. 123. 
kramddhyayana, iv. 108. 
kriydvdcin, 4. II. 1. 
kriydyoga, 4 II. 3, 18. 
q. kroslitar, iii. 30c. 
kvacit, iii. 54. 
r. 4*Aip : dkuhipta, i. 16. 
kshubhna (r. ksliabh), iii 

ksltdipra, iii. 55 intr. n., 
58,65, iv. lie. 

q. khatvendra, khatvodakam, 
khatvdirakd, k/mtvdi'i- 
kdyanah, iii. 44c, 45 c, 

a. khanvakhdzi, khdimd- 
khdti, i. 96, 105. 

q. khanati, i. 20c. 

oa», i.47c, ii. 9c. 

jranti (r. jam), i. 86. 
gavdm, ii. 70. 
gavishti, ii. 23. 
gaveshana, ii. 23. 
? «no, i. 1,10c (4. 1. 1,2, 4 

8, 10). 
(7«ru, i. 43c, 52. 
. goduh, i. 3c, etc., etc 
. gosani, ii. 103. 
r. jrroA : avagrhyate etc., 

iv. 35c, 38c, 50c, 56c, 

76c, 77c. cf. avagrhya, 

grahana, i. 68c, ii. 28c. 
^tapi (r.gld), iv. 93. 

0Ao*Aa, i. 10c (4.1. 3,4). 

cf. aghoxha. 
ghoshavant, i. 13, ii. 2, 11, 

43, 54. 

ha, i. 47, ii. 9, iii. 27. 

ea, i. 2.5, 9, 31,32,45, 48, 
89-91, 104; ii. 3,4,14 
100, 103, 106; iii. 18, 
73,82,87,94,95; iv. 5, 
9,11, 18,22,26,28,32, 
61, 63,75, 82, 85,100, 
i. cakra, iii. 2. 

catur, i. 1. 

caturtha, i. lc, 10,10c (4. 1 
3.6), 101, 102, ii 7. 
x. caturdtra, iv. 80. 

r. car : uccdrya, i. 29c. 

eared, iii. 96c, iv. 74, 114c 

eavarga, i. 7, 21c, ii. 14, 26 
39, iii. 94. 

cavargiya, ii. 11, 15c 
q. cdyakak, iii. 40c. 
q. cinoti, ii. 26c. 

cet, ii. 38, iv. 39. 

cydvayati (r. cyu), iv. 91. 

chakdra, ii. 17. 
.. cfiandas, ii. 62. 
chandas, iv. intr. n., lie. 
35c, 126c 

q. chddayati, ii. 26c 
chdndasa, iv. 26c, 36c 

r.jan : upajdyeta, iv. 35c. 
cf. upajana, upajdta. 

japana, i. 43c 

jara, iv. 53c. 

jurat, iv. 53. 

jahdti (r. hd). ii. 46. 
n.jdgdra. iv. 96. 

jdta, iii. 65 intr. n. cf. pa- 
q.jdtiya, iv. 28. 

jdtya. iii. 55 intr. n., 57, 65. 
a. jdspatya, iv. 64, 83. 

jihvd. cf. madhyajihva. 

jihvdgra, i. 22, 24. 

jihvdmula, i. 20c. 

jihvdmuHya, i- 10c, 20, 20c. 
&.jiva, ii. 76. 

jivantim oshadhim, iii. 6. 
a.jihidd 'ham, iii. 14, iv. 87. 

r.jnd : pratijdnate, iii. 55 
intr. n. cf. pratijnd etc. 

jridna, iv. 107. 

jyotishtva, iv. 102. 

nakdra, ii. 10. 

ta, ii. 9. 

takdra, ii. 8. 

tavarga, i. 22c, ii. 14, 26, 

39, iii. 94. 
tavargiya, ii. 12. 
q. h"io(e, ii. 14c, 26c, 40c 
tutva, iv. 74. 

rfa, i. 29c 
q. dayamdnam, ii. 31c 
q. diyate, ii. 12c, 14c. 

dha, i. 29c 

«a, i. 47, ii. 9. iii. 27, 75. 
nakdra, ii. 1 2. 
natva, iv. 74. 

<, Hi. 11. 
<o, ii. 9. 
takdra, ii. 8, 13, 61, 83, iv. 

tatah, i. 35. 
a. tatas pari, ii. 66. 
q. tatra, ii. 26c. 

tad, i. 103, ii. 7, 15, 39, iii. 
31,65, iv. 36, 111, 125. 
a. tad, ii. 85, iv. 48. 

taddhita, ii. 83, iv. 13, 26c, 

27c, 4 5c, 55c. 
tanmdnin, iv. 29. 
tama, iv. 16, 46. 


W. D. Whitney, 

tara, iv. 16. 

a. diddya, iii. 22, iv. 89. 

a. nara, iii. 9. 

q. tarati, ii. 26c, iii. 37c. 

dirgha, i. 38, 61, 73, 84 

navati (r. mi), iii. 82. 

tavarga, i. 24c, ii. 26, iii. 94 

iii. 1,42, iv. 33, 50, 74 

q. r. nag : nirnashtah, pari- 

tavargiya, ii. 15, 17. 


nasktah,pranas/itah, iii. 

tdtil, iv. 20. 

a. dirghdyntva, ii. 59, iv. 100. 

90c. ' 

a. taid, i. 105. 

a. d*«A. ii. 60, 63. 

nafi (r. nag), iii. 90. 

a. tan agre, ii. 85. 

q. durahnah, iii. 77c. 

ndda, i. 13,43. 

tdlavya, i. 21. 

a. durndntan, iii. 84. 

ndndpada, ii. 16, iii. 79, 

tikshna, iii. 55 intr. n. 

q. dumaxlitah, iii. 90c. 

iv. 27. 

tu, ii.39. 

a. duxhtara, iv. 83. 

ndman, i. 1, iv. 126, 4- II. 

tulyatd,i. 10c (4.1.10). 

q. dushpitam, ii. 63c. 

1-4, 7, 12, 13, 16. cf. 

iulyalmga, i. 1 0c (4. 1. 9). 

q. dudabhah, ii. 60c. 

karman-, sarvan-. 

tulyavrtti, iii. 55 intr. n. 

q. dudhyah, ii. 60c. 

ndmin, ii. 29, 42, 81, 87, 

a. tuvixlitima, iii. 96, iv. 69. 

q. dundcati, ii. 60c. 

iii. 11,39. 

q. tmhtah, i. 98c. 

drdlta, iii. 55 intr. n. 

a. ndraka, iii. 21, i v. 90. 

Mh/o, i. 8, 1 Oc (4 1. 3), ii. 

drpa (r. cfarp), iv. 69. 

a. ndnhada, iv. 83. 

2, iii. 19, iv. lie. 

q. drxhad, i. 3c, etc., etc. 

ndsikd, i. 26. cf. mukhan-. 

tdirovyanjana, iii. 55 intr. 

devatd, iv. 49. 

ndsikya, i. 26, 26c, 100. 

n., 56c, 62. 

dgubhi, iv. 21. 

nigama, i. 77 c. 

a. tain tvam, ii. 84. 

a. dyduh, ii. 74. 

a. nih, ii. 63. 

q. toxhanam, iii. 75c. 

q. r. </rw : ddravati, pradra- 

niqhdta, iii. 65, iv. 107c(?), 

a. tyam. iii. 25. 

vati. i. 56c. 


t'rd, iv. 14. 

a. drughana. iii. 76. 

nitya, i. 68c, iii. 56c. 

tri, iv. 113. 

dronika, i. 23. 

nidargava, i. 29c, 98c, etc. 

a. tri, ii. 98. 

doandna, iv. 49, 60c. 

nidhana, iv. 105c. 

a. trih, ii. 64. 

a. rfuar, ii. 45. 

nipdta, i. i, 79, 80c, 4. II. 

tridhdtva, ii. 65c. 

Are, iv. 110, 118. 


tripada, iv. 98c, 114c. 

dnih, iii. 26. 

nimitta, iii. 76c, iv. 76, 

trimdtra, i. 62. 

doiiiya, i. 10, 10c (4.1.2, 

78c, 125c, 126c. 

q. trixhtubh, i. 3c, etc., etc 

3, 6,7), ii. 6, iii. 22. 

q. nirahnah, iii. 77c. 

a. tredfid, iv. 66. 

(fo^wi, i. 10c (4.1.6,7). 

nirdega, iii. 96c. 

a. trdixhtubha, iv. 83. 

deipada, iv. 108c. 

nirvdpa, iv. 105c. 

trdixoarya, iv. 107c. 

doimdtra, i. 61. 

a. nivatas prndli, ii. 78. 

a. trdihdyana, iv. 83. 

dvirukta, iv. 44. 

nivdrana, i. 1 Oc (4. 1. 6). 

«»a, iv.26. 

dinrbhdia, iii. 27c. 

nivrtti, iv. 1 1 6c. 

a. <we, i. 77. 

doirvacana, iv. 117, 123c, 

nicdih, i. 15, 4. II. 5. 

q. tmru, ii. 26c. 


q. nt«d, iii. 1 lc, iv. 61c. 

dvivacandnta, i. 75, ii. 47. 

q. nild, i. 4c. 

<Ad, iv. 15. 

dvdilingya, i. 10c (4.1.10). 

q. nexhtram, iii. 30c. 

dvyakshara, iv. 64c 

ndimittika, iv. 125c, 126c. 

q. danshtra, iii. 2. 

q. danddgram. iii. 42c. 

dliarma, iv. 101. 

jt>o, ii. 62. 

dnddli (r.dd), iv. 61. 

dharma$efiha, iv. 106c. 

a. paksha, iii. 2. 

dadhdli (r. <2Ad), iv. 63. 

r. rfAd : vidhiyate, iv. 27c, 

q.panti, ii. 20c. 

q. dadhi, i. 4c, etc., etc. 


pancajana, iv. 106. 

q. dadhindrah, iii. 42c. 

</Ad, iv. 13. 

pancapadi, i. 88, iii. 5, 59. 

dantamula, i. 28. 

cMata, ii. 90, iii. 48, 79, iv. 

pancama, i. 10c (4.1.4), 

dantya, i. 24. 


ii. 67. 

dargana, iv. 27, 78. 

a. dMA, iii. 25. 

q. patujdtiya, iv. 28c. 

a. <faf», i. 63. 

r. jtwrf/t : pathet, i. 8c. 

da, iii. 1 1. 

no (»), i. 47, ii. 9, 89, iii. 

panati (t.pan), iv. 97. 

a. dddhdra, iv. 96. 

27, 37, 75. 

q. panditajdtiya, iv. 28c. 

q. <£m, i. 87. 

na,\. 7, 8,41, 65, ii. 22,30, 

a. patdti, ii. 77. 

ddnim, iv. 23. 

36, 39, 51, 58, 102, iii. 

a. pathi, ii. 1 00. 

ddrdhya, iv. 108. 

70, 86, iv. 33, 47, 94, 

r.pad: samdpadyate etc., 

a. dMpa, i. 63, ii. 60. 


iv. 84c, 88c; utpanna. 

a. dwah, ii. 68. 

nakdra, i. 67, ii. 10, 26, 34. 

iv. 76c ; pratipddita, iv. 

a. divi,'il 100,101. 

naksltatra, iv. 102c. 


<&f, iv. 72. 

q. nadim, iii. 37c. 

a. joadb, ii. 72, iii. 2, 10. 

r. dtp: nirdipyate, iv. 126c. 

napunsaka, i. 84, ii. 52. 

pada, i. 8c, iii. 95, iv. intr. 

q. diddn&ati, i. 87c. 

q. nayati, ii. 5c, iii. 37c, 75c. 

n., 106c, 107c, 108, 109, 

Atharva- Veda PratvpMiya. 


110,113,123c. cf. an- 
tahp-, dnp-, ultarap-, 
ekap-, kramap-, trip-, 
dvip-, ndndp-y purvap-, 
padakdla, iv. 109c, 123c. 
padajdta, i. 1. 
padatva, iv. 1 6c, 26c, 27c, 

30c, 35c, 37c, 98. 
padafak, iv. 107c 
padafdxtra, iv. 1 22c. 
padddi, iii. 53, 55n, iv. 35c. 
padddhyayana, iv. 107. 
padddhydyin, iv. 107c. 
paddnta, i. 45, 54, ii. 2, 3, 
21, iii. 26, 55, 89, iv. 81. 
paddntya, i. 3. 
padya, i. 1, 3, 4, 57. 
a. panishpada, iv. 96. 
para, i. 50, 55, 71,82, 101 
ii. 13, 25,26, 31,40, 60, 
66, 70, 76, 106, iii. 25 
31,41, 57, 70, 78, iv. 
paratah, i. 99c, ii. 53c, iii. 
55. ' 
a. parama, ii. 94. 

pararupa, iii. 52. 
a. pard, iii. 80. 
a. pari, ii. 105, iii. 88, iv. 58. 
a. paridhish patdti, ii. 77. 
paripdtha, i. 105c. 
parihdra, iii. 96c, iv. 74, 

parihdrakdla, i v. 1 20c, 

pari/idrya, iv. 116, 126. 
q.parittih, iii. lie, iv. 61c. 

parokshd, iv. 84. 
q.parna, iii. 2. 
q. paryahnah, iii. 77c. 

parvan, iv. 42c, 53, 77. 
q. pavanam, iii. 40c. 

pavarga, i. 25c. 
a.pdnsu, i. 85. 

jtxfcfa, i. 93c, iv. 107c. 
a. pddam angulim, i. 66. 

pddatrtta, iii. 55 intr. n.,63. 
a. pitar, ii. 72. 
q. pitrgyah, iii. 42c. 
q. pitrartham, iii. 39c, 58c. 
q.pibati, i. 25c. 
pibati (r. pd), iv. 26c. 
pidana, i. 29c. 
pidita, i. 43, 43c. 
q. pumydna, ii. 25c. 
pwmlinga, iv. 46c. 
q. puncca, ii. 25. 
q. punpeora, punskdma, pun- 
sputra, punks/mra, ii. 

a. punar, ii. 48. 

punar, iv. 105, 115, 126. 
a. punar naydmasi, iii. 81. 
q.pumddua, ii. 26c. 
/ram, ii. 25. 
pumans, i. 91. 
a. purah, i. 63. 
■i.purusha d babhAvdsn, i. 

q. purushah, i. 20c, 25c. 

purana, iv. 38. 

a. purydna, iii. 83. 

/>ur»a,'i. 38, 56, 67,69,92, 

94, 104, ii. 7,15, 38, iii. 

36, 38, 41, 53, 57, iv. 4, 


purvapada, ii. 94c, iii. 76, 

iv. 75, 107c. 
purvarupa, i. 50, iii. 74c. 
q. purvdhnah, i. 68c, 100c. 
iii. 31c, 77c. 
prkta, i. 10c. cf. aprkta. 
a. prndti, ii. 78. 
p-iAai, i. 10c (4.1.9,10), 
iv. 104. 
a. prthivi, ii. 68, 100. 
a. prshtha, ii. 69. 
a., pra, ii. 76, iii. 80. 
q. prakdrena, ii. 64c. 
prakrti, ii. 31c, iii. 33, 54, 

iv. 50c, 77c. 
prakrtidarfana, iv. 73. 
prakrtisvara, 4. II. 2-4, 9. 

pragrhya, i. 73, iii. 33, iv. 

108c, 117,123. 
pratijnd, i. 101c. 
pratijhdna, i. 8, 9c. 
prativeshtita, i. 22. 
pratishiddha, iv. 66. 
pratiihedha, iv. 16c, 18c, 

pratishthita, iv. 105. 
q.prattam, iii. lie, iv. 61c. 

prat yak, i. 28c. 
q. pratyan, ii. 9c, iii. 27c. 
a. prat.yancds.rn, i. 105. 
pratyaya, i. 50c, ii. 87, iii. 

3,'iv. 13c, 16c. 
prathama, i. 6, 8, 10c (4.1. 
1,4), iii. 2 1,38. 
a. pratlmma, ii. 75. 
a. prapana, iv. 97. 
prapdtha, iv. 126c. 
prabhrti, iv. 85. 
prayatna, i. 27c, 29c. 
prayojana, iv. 114, 119. 
praflishta, i. 39c. 
prasamdhdna, i v. 78c, 111 

q. prasahanam, ii. 82c, iii. 1c, 
iv. 70c. 

prasdrana, iv. 37c. 

prastirna, i. 24. 

prdkflishta, iii. 56c. 

prdna, iv. 39. 
a. prdnati, prdnanti, iv. 57. 
a. prdtar, ii. 48. 
q. prdtar, iii. 75c. 

prdtijna, i. 1, 2c. 

prdtipadika, iii. 78. 

prdtiedkhya, iv. 106c. 

prdpta, i. 2, 77c, iv. 36c. 

prdpti, i. 10c (4.1.6,7). 
a. prdctoh, i. 105. 

prdflishta, iii. 55 intr. n., 
56, 65. 
q.prdsAa*, ii. 82c, iii. lc, iv. 


q.prdhnah, i. 58c, 100c, iii. 
31c, 77c. 

pretya, iv. 101. 

prepsu, iv. 29. 
q. plakshah, i. 98c. 

^Mto, i." 38, 62, 106, iv. 6, 
120. cf. apluta. 

pluti, i. 105c. 

a. babhuva, iv. 6. cf. <i £a- 

a. barhih, ii. 100. 

6a/a, iii. 55 intr. n. 

bahula, iii. 8, 13, 17, iv. 
35c, 67c, 126c. 

bahuvacana, i. 78, 84. 
q. brhadbhih, i. 44c. 
a. bodJiapratibodhdu, iv. 96. 
a. brahma, ii. 66. 
a. brahmanvant, iv. 99. 

brahmayajna, iv. l()7c. 
q. brahmdupagavah, iii. 51c. 

q. bhargah, i. 58c, iii. 31c 
q. bhavdn, ii. 12c, 26c. 
a. bhdnu, iii. 87. 

A, iv. 31. 
a. bhuvah, ii. 52. 

r. 6Am : bhavati, ii. 39. 

bhuta, ii. 82. 

bhutakarana, iii. 49. 
a. bhuydih, i. 105. 

bhyah, iv. 3 1 . 

bhydrn, iv. 31. 
q. bhrdshtram, iii. 30c. 

ma, iii. 37. 

makdra, i. 67, ii. 25, 31. 
mandala, iii. 55 intr. n. 
moh, iii. 17, iv. 17. 
matvartha, iv. 8c, 47. 
q. madhu, i. 4c, etc., etc. 


W. D. Whitney, 

q. madhushnam, iii. 42c. 

madhya, iv. 113, 117. 

madhyajihva, i. 21. 

madhyama, iv. 42. 
a. manaa pdpa, ii. 79. 

manishin, i. 10c (4.1.1), 
iv. 35c. 
a. manuahyat, iv. 65. 

mantra, iv. 101c, 107c. 

maya, iv. 24. 
q. mahdn, ii. 1 2c, 26c. 
a. mdnita, i. 85. 
a. mdtra, iv. 22. 
q. mdtrartham, iii. S9c, 58c. 

miifrii, i. 38. cf. anum-, 
eknm , trim-, dvim-. 

mdtrdrdha,i. 17,50. 

mdn (r. man), i. 87. 

mdlendrah, mdlodakam, 
mdldirakd, mdlditikd 
yanah, iii. 44c, 45c, 50c. 
a. mitra, iii. 9. 

milhah, iv. 113c. 

mindti (r. mi), iii. 86. 

mipra, iv. 1 1 3c. 
a. mimdya, iv. 96. 

mnkha, i. 1 8. 

mukhandxika, i. 27. 

muahi (r. m«»A), ii. 76. 

7mMAa», i. 22c. 

murdhanya, i. 22, 28c, 63 
ii. 60. 

mula. cf. dantam-, hanum-. 

mrdu etc., iii. 66 intr. n. 

mrdujdtiya, iv. 28c. 

mrduprayatna, i. 29c 
q. me, i. 77. 
q. mekhaldvi, iv. 18c 
q. medhdvi, iv. 18c 

ya, i. 68, ii. 21 , iii. 35, 57, 
iv. 29. 

yakdra, ii. 41, iii. 18. 
q. yajushtaram, ii. 83c. 
q. yajussdt, ii. 83c. 

yajna, iv. 105. 

yajnatati, iv. 104. 

yathd, i. 99, iv. 103, 122. 

yad, i. 71, iv. 50. 
a. yorf, iv. 48. 

yatna, i. 13c, 14, 26c, 99. 

yami (r. yam), iv. 93. 
a. yas patih, ii. 70. 

yd, iv. 30. 

ydjnika, iv. 101c, 103. 
q. ydti, i. 49c. 
a. ydtumdvat, iv. 8. 

ydvayati (r. y»), iv. 92. 

yuite, iii. 89. cf. anya- 

yushmad, ii. 84, iv. 26c, 4- 

q. yushme, i. 77. 
a. yoga, iii. 2. 
a. yondv adhy dirayanta, iv. 


ra, i. 68, iii. 20. 
q. rathamtara, ii. 51. 
q. rathe, i. 49c. 

rd/a<i (r. rdj), ii. 36. 
a. rdtri, ii. 5 1 , iii. 8. 
a. rdyas posha, ii. 80. 

rt*At (r. ri&h), iv. 86. 

rwiro, iv. 35c. 

rupi (v. rup), iv. 86. 

rudhi, iv. 1 6c. 

r&pa. cf. parar-, purear-. 
a. riy>a, ii. 51. 

repAa, i. 28, 37, 58, 64, 66, 
71, 101, ii. 19,29,42, 
87, 106, iii. 31,75. 

raw, ii. 47. 

la, i. 39c, iii. 93. 

lalcdra, i. 5, 39, 46, 64, 66, 

ii. 13, 35. 
lakshana, i. 94c, iv. 12c, 

122c, 126c. 
laghu, i. 51. 
q. lavanam, iii. 40c 

lakshana, iii. 55 intr. □. 
q. Idyakah, iii. 40c. 
linga,i. 10c (4.1.9). 
x.lup: lupyate, iv. 16c, 

60c, 64c, 66c 
q. lidham, i. lc. 
legavrtti, ii. 24. 
foia,'iv. 105,106. 
lopa, i. 67, ii. 18, 32, 66, 

89, iii. 20, 36, 91, iv. 

21c, 74. 

va, i. 26c, ii. 21, iii. 35, 67. 
vakdra, it 22, 28, 37, iv. 

vacana, i. 101c, iv. 124, 4. 
II. 2. cf. dvirv-, dviv-, 
vatu, iv. 48. 
r. vad : vadanti, iii. 65 ; 
apa vadati, ii. 63c, 101c, 
iii. 60c. 
vani (r. van), iv. 93. 
a. vandane 'va vrksham, ii. 

q. vaputhtaram, vapusidt, ii. 

83c. ' 
q. vayati, i. 49c 
q. vaydh, ii. 28c. 
vari, iii. 24. 

varga, ii. 38. cf. eat-, tav- t 

vargdntya, i. 13c. 
vargottama, i. 26c 
vargiya. cf. cav-, tav-,tav-, 
varja, ii. 67, 68, 70, 84, iii. 

25, iv. 56, 59. 
varji'-a, iii. 95. 
varna, i. 25-28c, 37c, 40, 

92. cf. a»-, iv-, uv-, rv-, 

Iv-, sat)-. 
varta, iii. 1 2. 
q. vartah, i. 58c, iii. 31c. 
vaau, i. 88, iv. 35. 
vasu, iii. 9, iv. 30, 45. 
vamdhdtarah, iv. 45. 
vaatutah, iv. 35c. 
t><«, i. 10 - 2, iii. 57, iv. 27. 
vdtah, ii. 28c 
Vdtaya, ii. 6c. 
txtea, iii. 24. 
q. vdyu, i. 4c, iii. 40c 
a. tidr, ii. 45. 

vdtrdhdna, iv. 85. 
»«, iv. 39. 
vikampita, iii. 65. 
vikalpa, iv. 27c. 
tJtirta, iv. 81. 
vihrtshita, iv. 12c. 
vigrhya, iv. 78. 
vigraha, iv. intr. n., iv. 8, 

27c, 107c, 4.11.9, 10. 
vighdta, i. 104, iv. 107c. 
mo* (r. vid), i. 90. 
t)i(ima, iii. 16. 
vidhdna, i. 41c, iv. 122c 
vidhdrana, i. 43, 43c 
vidhi, i. 41. 
vinata, iv. 82. 
vindma, iv. 34, 114. 
viparita, ii. 38c. 
viparyaya, ii. 38. 
viprakarsha, ii. 39. 
vibhakta, iv. 107c 
vibhakti, i. 77c, ii. 51, iii. 

78, iv. 30c 
vibhdshd, i. 2, iv. 126c 
viyat, iv. 105c 
t)tr<i/, i. 3c, etc., etc 
rtrdma, ii. 38, iv. 79. 
ci t)ar, ii. 44. 
vivrta, i. 31,34. 
mvrtti, iii. 63, 74c. 
tstpcsAa, i. 18, iv. 12c. 
vifpati, vippatni, iv. 60. 
trapua, iii. 9, iv. 23. 
viaarjaniya, i. 5, 42, ii. 26, 

26, 40, iii. 29. 
q. vittd, iii. lie, iv. 61c 

vipsd, iv. 19. 
q. vrksha, i. 98c, ii. 28c, 40c 

Atharva- Veda PrdtigdkJiya. 


a. vfkshdn vandni, ii. 28. 
v'rtti, i. 8, 8c, 9c, 40, 41c, 

95. cf. tulyavrtti, Ufa- 

vrddha, iv. 55. 
zrddhimant, iv. 55c. 
a. vrdha, iii. 24. 
veda, iv. 104. 
vedddhyayana, iv. 101. 
vdiydkarana, i. lc. 
vyanjaka, i. 103c. 
vyanjana, i. 43, 55, 60, 98, 

102, ii. 57, iii. 26, 62, 

iv. 25, 50. 
vyatyaya, iv. 1 3c, 1 26c. 
r. vyadh : pratividhya, iv. 

vyadhi (r. vyadh), iii. 3. 

iv. 6S. 
vyavadhdna, i. 99c, 100c. 
vyavaxthita, iv. 27c. 
vyavdya, ii. 92, 93, iii. 93 
vyavdyin, ii. 88c. 
vyaveta, i. 98, 101c, iii. 

vydkarana, i. 2c. 
vydsa, iii. 68, 72. 

fa, ii. 6, 9, iii. 32, 93. 
a. pakalyeshin, iii. 52. 

poAd, iv. 105c. 

fakdra, ii. 10, 13, 17. 
a. gaci, ii. 71. 

patdudana, iv. 101c. 

fabda, iv. 8c, lie, 16c, 27c 
etc., 107. 
q. (ami, i. 78c. 
a. para, iii. 16. 

yds, iv. 19. 

Qdkatdyana, ii. 6c, 24, iv. 
intr. n., 30c. 

Qdkalya, iv. 49c. 

fdkhd, i. 2c. 

Qdnkhamitri, i. 93c, ii. 6c, 
iii. 74c. 
q. fdn (r. pd), i. 87. 

pttrfro, iv. 11 0c, 1 22, 126c. 
q. fifdnxati, i. 87c. 

fuddha, iv. 113c, 121. 
a. pun, ii. 61, iii. 10. 
q. fete, i 49c, ii. 6c, 9c. 
a. fepaharxhanim, ii. 56. 

fnbhana, iv. 46c. 
q. gobhanajdtiya, iv. 28c. 

Qdimaka, i. 8, 9c. 

frathi (r. (rath), iv. 93. 

pr«<). cf. eiop-, uddttag-. 
a. creyah, ii. 62. 
q. cvayanam, iii. 40c. 

p»dso, i. 12,43. 


sAa, ii. 6, 9, iii. 32, 90 

shakdra, i. 22c, 23, ii. 16, 
81, iii. 75. 
a. thai, i. 63. 

q. shande, i. 49c, ii. 6c, 9c. 
q. shannavati, ii. 16c. 

shatva, iv. 74. 

shaxhthi, ii. 71 , iv. 45c 
a. shodagin, iv. 51. 

sa, ii. 6, 9, iii. 32, 93. 
sah, iii. 55, iv. 126. 
a. sah, ii. 57. 

samyukla, i. 49, 50c. 
xamyoga, i. 1 0c (4. 1. 5, 8), 

5 1.56, 94c, 98, 102c, 104, 

iii. 28, 57. 
samxrta, i. 36. 
sampaya, iv. 12c, 1 06c. 
xamxprshta, i. 37, 37c, 40, 

sanhitd, ii. 1 , iii. 96c, iv. 

107c, 108, 109. 
sanhitdvat, iv. 1 24. 
sanhitika, iv. 107c, 114c. 
sakdra, ii. 8, IS, 62, 87, iv. 

sankhyd, i. 99, iv. 27c. 
saeati (r. *ac), ii. 68. 
satijnd, i. 13c, 48c, iv. 16c, 

67, 125c, 126c. 
)at(va, 4. 1.9, 10, II. 1. 
r.sad: sanna, i. 43, 43c; 

avasanna, i. 43c; pari- 

akanna, iv. 126c. 
sadi (r. tad), ii. 99. 
a. sadyah, ii. 62. 

son, i.86. 
a. sanntar, ii. 48. 
sandeha, iv. 51, 103c. 
sandhdna, iv. 114c, 122c. 

cf. praxandhdna. 
sandhi, i. 10c (4. 1. 5), 49c, 

iii. 65 intr. n., 74c, iv. 

107c, 114. 
xandhija, iii. 56c. 
sandhya, i. 1, ii. 37, 38c, 

iii. 37. 
xandhyakshara, i. 25c, 40, 

iii. 40. 
sanna, i. 43, 43c. 
saimipdta, ii. 38c. 
xaptami, i. 74. 
a. sow/, ii. 36, iv. 39, 58. 
a. samatita, iv. 88. 
samartha, 4. II. 3. 
xamdna, i. 14, 27c, 50. 
samdnapada, i. 99, ii. 15, 

iii. 68, 72, 75. 
samdndkshara, i. 105c, iii. 


samdpatti, iv. 73, 74, 75c, 

samdpddya, iv. 117, 124. 
samdmndta, iv. 103. 
samdsa, ii. 62, 62c, 63, iv. 
intr. n., 9, 27, 35c, 43, 
44c, 45c. 
q. samidbhih, i. 44c. 
a. samudra, iv. 54. 

sam dirayan ta, ii. 80. 
q. sarpishtaram, tarpisxdt, ii. 
83c. ' 
sarea, iv. 69. 
xarvatra, iii. 60. 
sarvandman, ii. 44, iv. 69. 
savarna, i. 27c, iii. 42, iv. 

56c, 84c. 
savid/ia, iii. 64. 
a. savya, ii. 95. 

tasthdna, i. 1 0c, 1 0c (4. II. 
6), ii. 13, 15, 31,40, iii. 
a. aas padishta, ii. 58. 

saxrara, i. 10lc. 
a. sahaxra, sahasraxdtama, 
iv. 45. 
*aAi (r. soA), ii. 82, 92, iii. 

l,iv. 70,88c. 
sat, ii. 82. 
a. sddha, iii. 7. 
a. sdirosdAo, iii. 23. 
a. xddhu, iv. 30. 

sdmdnya, i. 2, 10c (4.1.9), 
q. sdye, i. 49c, ii. 6c, 9c. 
a. sdhydma, iii. 15, i v. 88. 

siei (r. sic), ii. 92. 
q. 8t«cort, ii. 90c. 

xiddhatva, iv. 6. 
a. ximanta, iii. 43. 
q. xividnta, iii. 43c. 

sm, iv. 32. 
q. suijan, iii. 27c. 

sun. ii. 97. 
a. mtprdvyd, iv. 1 1. 

subhishaktama, iv. 46. 
a. sumna, iv. 30. 

»«ira, iii. 96c, iv. 101c. 
srji (r. sor/), ii. 102. 
s»7ri (r. xarp), ii. 102. 
a. trxhtd h, i. 1 05. 
a. soma, ii. 96. 
q. somasut, ii. 14c. 
soxhmav, i. 10, 94. 
skandi (r. xkand), ii. 104. 
stainbh (r.), ii. 1 8. 
stambhi (r. xtambh), iv. 62. 
4tm<W (r. star), ii. 105. 
a. */r/o, ii. 86. 
a. stoma, ii. 96. 
a. strdiehuya, ii. 88, iv. 83. 


W. D. Whitney, 

sthd (r.), ii. 18, 92, 93, iv. 

tthdna, i. 18c, 22c, 26-28c, 

sthiti, i. 10c. 
sparfa, i. 6, 29, 44, 99, ii. 

20, 26, 81, iii. 89. cf. 

tprfi (r. sparg), ii. 102. 
sprshta, i. 29, 30, 33. 
sphdrji (r. sph&rj), ii. 102 
sphotana, i. 103, ii. 38. 
smarati (r. smar), ii. 102. 
a. roa, ii. 86. 

wrapt (r. map), ii. 86. 
a. svapna, iv. 30. 
r. «xir : svaryate etc., i. 

93c, iii. 56c, 67, iv. lie. 
a. svar, ii. 48. 
q. svar, iii. 75c. 

svara, i. 4, 13, 32, 55, 93, 

98, 101, 104, ii. 2,21 

27, 41, iii. 27, 28, 32, 36, 

89, 91, iv. 6, 29, 56, 107 

svarahhakti, i. 101, 101c. 
svaravant, iii. 74c. 
svarati (r. svar), ii. 102. 
svarita, i. 16, 17, iii. 57 

svargakdmo aghdyatdm. 

iv. 101c, 103c. 
. svarshdh, ii. 49. 
svdra, iii. 55 iotr. n. 
svdrtha, i. 1 05c. 
svdhdkarana, iv. 1 06c. 

ha, i. 13e. 

AaAara, i. 10c (4. 1. 6). 47 

58, 100, ii. 7, iii. 31, iv. 

hani (r. han), i. 86. 
r. han : ni hanyeta, iv. 

107c; vi hanyate, i. 

hanwnitla, i. 20, 28c 

Aantfi (r. han), iii. 91, iv. 

r. har : uddharet etc., iv. 

76c, 77c, 124c, 4. II. 7, 

14; pariharet, etc., iv. 

120c, 121c, 123c, 126c 

cf. uddharana, parihd- 

ra, parzhdrya. 
harati (r. har), iv. 62. 
havih, ii. 63. 
hi, ii'. 101. 

hinoti (r. hi), iii. 88, iv. 95. 
hina, i. 43, 43c. 
hotffyah, iii. 42c. 
r. hnu : apa hnute, vi 

hnute, i. 58c, 100c, iii. 

r. hmal : apa hmalayati, 

vi hmalayati, i. 58c, 

100c, iii. 31c. 
hrasva, i. 51, 59, 83, iii. 

27, 48, iv. 85. 

Atharva- Veda Pr&tiq&fchya. 


3. General Index. 

The references are as in the preceding Indexes. 

a,d: are throat-sounds, i. 19n; utterance 
of d, i. 35 ; obscure utterance of a, i. 36 ; 
final v retained after a, ii. 22 ; d be 
tween two vowels, how combined, iii. 
38 ; combinations of a and d final with 
initial vowels, iii. 42-52 ; initial a ab- 
sorbed by final e or o, iii. 63, 54 ; full 
exposition of the Atharvan usage in 
this respect, iii. 54n ; resulting circum- 
flex accent, iii. 55. 
Abhinidhana : defined, i. 43 ; -when ap 
plied, i.44-47; also called dsthdpita,\A8 
Abhinihita circumflex, iii. 55 ; its com 

parative tone, iii. 55 intr. n. 
Accents: general definition, i. 14-17 ; ac- 
cents resulting from euphonic combina 
tion and construction of sentence, iii. 
55-74 ; comprehensive exposition of 
accentual theory, iii. 65n ; modes of des- 
ignating accent in the manuscripts, iii 
65n ; special case of acceut, i. 96 : — see 
also Acute, Circumflex, Grave. 
Acute accent defined, i. 14; acute tone of 
grave syllables following a circumflex, 
„ iii. 71. 

Agnivecya, quoted by Taitt. Pr., ii. 40n. 
di : palatal diphthong, i. 21n; its pro- 
nunciation, i, 40n, 41; its combination 
with following vowel, iii. 40, ii. 21. 
Alterant vowels, ii. 29n. 
Anudatta : — see Grave. 
Anusvara, not a constituent of the spoken 
alphabet acknowledged by the Prati- 
cakhya, i. lln. 
Anvataieya, quoted in commentary, iii, 

Aspirate mutes, i. 10, lOn; become non- 
aspirate before aspirates, i. 94. 
Atharva- Veda, existing text of: its rela- 
tion to the Praticakhya, add. n. 2 ; man- 
uscripts of, intr. n. ; their mode of des- 
ignating accent, iii. 65n. 
Atharva- Veda Pratic&khya : its distinctive 
name, manuscript material, character 
of its commentary, etc., intr. n. ; its 
scope, as definedby itself, i. 1, 2 ; school 
to which it belongs, i. 2n ; its contents 
analyzed and compared with those of 
the other Praticakhyas, add. n. 1 
lation of the Atharvan text which it 
contemplates to the existing Atharva 
Veda, add. n. 2. 
&u: labial diphthong, i. 25n; its pronun 
ciation, i. 40n, 41 ; combination with 
following vowel, iii. 40, ii. 22. 
Augment, combination of, with initial r 

and r, iii. 49 ; its interposition does not 
always prevent lingualization of the 
initial s of a root, ii. 92. 
Avagraha, pause dividing the parts of a 
compound: its length, iii. 74n; rules 
for its use in padatext, iv. 7-72 ; do. 
in pa<&»repetitions, iv. 123; whether 
to be used in iramo-repetitions, iv. 

6, a labial mute, i. 25n. 

bit, a labial mute, i. 25n. 

Bharadvaja, quoted by Taitt. Pr., ii. 7n. 

c, a palatal mute, i. 2 In. 
p : a palatal spirant, i. 21n, 31n ; its pho- 
netic value, i. 21n ; its combination, 
when initial, with preceding final t and 
n, ii. 10, 13, 17. 
Caityayana, -quoted by Taitt. Pr., ii. 7n. 
Qakala, Cakalya, quoted by Rik Pr., Vaj. 
Pr., and Panini, i. 8 In, ii. 17n,40n, iii. 26n. 
akalya, quoted in commentary, iv. 49n. 
akatayana : quoted in text, ii. 24 ; in 
commentary, ii. 6n, iv. intr. n., iv. 30n ; 
by Rik Pr., Vaj. Pr., and Panini, i. 8n, 
40n, ii. 24n, 32n, 40n, iii. 30n. 
Cankhamitii, quoted in commentary, i.98n, 

ii. 6n, iii. 74n. 
Case-endings, when separable from theme 

in padatext, iv. 31-34. 
Caunaka: quoted in text, i. 8; in com- 
mentary, i. 9n ; by Vaj. Pr., ii. 6n ; 
his relation to the Praticakhya, intr. n., 
add. n. 1. 
Cerebral mutes : — see Lingual. 
ck: a palatal mute, i. 21n; conversion of 

initial f into, after final dentals, ii. 17. 
Circumflex accent defined, i. 16, 14-16n, 
17; kinds of independent circumflex, 
iii. 55-61 ; vikatnpita modification be- 
fore acute or circumflex, iv. 65 ; kinds 
of enclitic circumflex, iii. 62-64; oc- 
currence of enclitic circumflex, iii. 67- 
70 ; comparative tone of different kinds 
of circumflex, iii. 55 intr. n. 
Commentary on the Praticakhya, charac- 
ter of, intr. n. ; character of its last sec- 
tion, iv. 101 intr. n. 
Compounds and secondary derivatives: 
when divisible by avaaraha in pada- 
text, iv. 8-72; treatment of double, 
triple, etc. compounds, iv. 10-12, 42- 
46 ; compounds not divisible, iv. 47-72 ; 
list of Atharvan compounds left undi- 
vided, iv. 64n. 


W. D. Whitney, 

Conjunction of consonants, i. 49, 98 ; as 
eimilation of former to latter constitu- 
ent, i. 50 ; how divided between sylla- 
bles, i. 66, 58 ; conjunction not dis- 
solved by euphonic insertions, i. 104 ; 
duplication in consonantal groups, iii. 
28, 30-32 ; complete list of Atharvan 
combinations of consonants, and exhi 
bition of their theoretical phonetic 
form, add. n. 3. 

Consonants : classification and description 
of, i. 10-13, 19-31 ; belong to what syl- 
lables, i. 55-58 ; quantity of, i. 60 ; 
groups or combinations of :— see Con 
junction of consonants. 

Contact of organs, degree of, in forming 
the different classes of sounds, i. 29-36. 

Conversion of sounds made to their next 
of kin, i. 95. 

d, a dental mute, i. 24n. 

d: a lingual mute, i. 22n ; softened be- 
tween vowels, i. 29n. 

Dalbhya. quoted by Vaj. Pr., li. 9n. 

Dentals (t, th, d, dh, n, I, s) : how formed 
i. 24 ; assimilated after palatal and 
lingual mutes in the same word, ii. 1 5 
do. after xh in all cases, ii. 16 ; change 
a following f to ch, ii. 17 ; » inserted 
after final n before surd dentals, ii. 26. 
30 ; anomalous change of dental to lin- 
gual in certain words, i. 63, ii. 60; tes 
toration of dental in pada-tvxt, iv. 74 
etc. See also the different letters. 

Derivatives, secondary, when divisible in 
padalext : — see Compounds. 

Desideratives, certain forms of, exhibit 
long nasal vowels, i. 86, 87. 

dh, a dental mute, i. 24n. 

dh : a lingual mute, i. 22n ; softened be 
tween vowels, i. 29n ; found written 
double in Atharvan MSS., i. 94n. 

Diphthongs (e, di, o, du) : character of, 
40, 41 ; euphonic combinations of, when 
final, iii. 40, ii. 21, 22, iii. 35; do. when 
initial, with a and d, iii. 50-62. See 
also the different letters. 

Duplication of final consonants, iii. 26, 27, 
29 ; of constituents of a group, iii. 28, 
80-32 ; product of duplication after r 
and h belongs to what syllable, i. 68 ; 
duplication not observed in Atharvan 
manuscripts, iii. 32n. 

«.* a palatal diphthong, i. 21n; pronun 
ciation of, i. 34, 40n ; a prayrhya as 
dual termination, i. 76 ; do. as termina- 
tion of certain pronominal forms, i. 77 ; 
its combination with following initial 
vowel, iii. 40, ii. 21 ; absorbs sometimes 
AQ initial a, iii. 53-54 ; full exposition 

of Atharvan usage in this respect, iii. 
54n ; resulting circumflex, iii. 56. 
Elision of sounds : — see Omissions. 

Finals, possible, i. 3-9 ; final mutes, i. 6- 
9, ii 3 ; whether surd or sonant, i. 8n ; 
suffer abhinidhana, i. 45 ; belong to pre- 
ceding vowel, i. 57 ; changes of in »a»- 
hita, general rules, ii. 2-6 ; final conso- 
nant duplicated, iii. 26. 

Final vowels, possible, i.4; when exempt 
from euphonic combination, iii. 33-36 ; 
prolongation of final vowels, iii. 4, 16, 
19, 20, 26. 

Final syllables heavy, i. 64. 

Forms of declension : their final vowels 
when prolonged, iii. 1 9n ; when divisi- 
ble in padaiext, iv. 31-34. 

g, a guttural mute, i. 20n. 

Ganas, part of the Praticfikhya's gram- 
matical system, intr. n., i. 65n. 

Gargya, quoted by Itik Pr., i. 8n. 

Gautama, quoted by Taitt. Pr., ii. 7n. 

gh, a guttural mute, i. 20n. 

Grammarians quoted by name in the text, 
i. 8, ii. 24. add. n. 1 ; in the commenta- 
ry, i. 9n, 93n, ii. 6u, iii. 74n, iv. intr. n., 
iv. 30n, 49n ; in the other Piaticakhyas : 
— see their names in this Index. 

Grave accent defined, i. 1 5 ; a grave syl- 
lable receives enclitic circumflex when, 
iii. 67-70; when pronounced at pitch 
of acute, iii. 71-74. 

Groups of consonants: — see Conjunction 
of consonants. 

Gutturals (k, kh, g, gh, n, x, r, f , I), how 
formed, i. 20. See also the different 

h: a throat-sound, i. 19n; a spirant, i. 
3 In; its phonetic character, and value 
as a sonant, i. 1 3n ; nasal mutes suffer 
abhinidhdna before, i. 47 ; ndrikya in- 
serted after, i. 100; combination of ini- 
tial h with preceding final mute, ii. 7 ; 
not doubled, but causes duplication, as 
first of a group, iii. 31. 

Heavy syllables, i. 62-54. 

i, i: palatal vowels, i. 21n ; prdgUthta cir- 
cumflex produceii by fusion of two j's, 
iii. 56 ; t is pragxbya as ending of loca- 
tive case, i. 74 ; do. of dual, i. 76 ; do. 
of ami, i. 78. 

Insertions in groups of consonants, i. 99- 
102, ii. 38 ; do. in making combinations 
of consonants, ii. 8, 9, 17. 25-30. 

Instrumental case, final vowel sometimes 
prolonged, iii. 19. 

Irregular and anomalous combinations and 

Atharva-Veda Prdtigdkhya. 


substitutions, i. 63-66, ii. 60, 61, Hi. 43 ;] 
do. accent, i. 96 ; do. form, iii. 7. 

iti : used in pada-text after a pragrhya, i. 
72, 74n ; in repetitions of krama and 
pada texts, iv. 117, 118, 119, 123; its 
combination with iva, i. 82; do. with a 
protracted vowel, i. 97. 

iva : treated in pada-text as forming com- 
pound with preceding word, iv. 41 ; its 
combination with iti after a pragrhya, 
i. 82 ; its irregular combination with 
preceding final syllable in Atharvan, 
ii. 56n. 

j: a palatal mute, i. 21n; converts pre- 
ceding or following n to », ii. 11, 15. 

Jatya circumflex, iii. 67 ; its comparative 
tone, iii. 55 intr. n. 

jh: a palatal mute, i. 21 n; not found in 
Atharvan, i. lOn. 

Jihvamultya(x) : a guttural spirant, i. 20n, 
3 1 n ; visarjaniya converted into it be- 
fore surd gutturals, ii. 40; this rule 
not observed in MSS. and edited text, 
ii. 40n ; a f> w times written with *A in 
Atharvan MSS., iv. 77u. 

k: a guttural mute, i. 20n; inserted after 
n before a sibilant, ii. 9 ; visarjaniya 
converted to * or sk before, when ini- 
tial, ii. 62-81; converts « of suffix to 
sh, ii. 87. 

KAe yapa, quoted by Vaj. Pr.. ii. S2n. 

Kandainayana. quoted by TAitt. Pr., ii.40n. 

Karshiina, result of combination of final 
lingual and initial palatal, ii. 89. 

Ktulialiputra, quoted by Taitt. Pr., ii. 7n. 

Kaiindinya, sthavira-Kaiindinya, quoted 
by Taitt. Pr„ ii. 7n. 

kh, a guttural mute, i. 20n. 

Kratnalext: recommendation of study of. 
iv. 108-109; mode of construction of. 
iv. 110-126; restorations of normal 
form in, iv. 74 etc. ; special points re- 
lating to, i. 70, 97. 

Krama-word, how composed, iv. 110, 113, 

Ksbaipra circumflex, iii. 58-61 ; its oc- 
currence in declension, iii. 59-61; its 
comparative tone, iii. 55 intr. n. 

I : a dental semivowel, i. 24n, 30n ; a pos- 
sible final, i. 5 ; enters into I, i. 39 ; suf- 
ers abhinidhdna before spirants, i. 46; 
exchanges with r in certain words, i, 
64-66; assimilates preceding t, ii. 13; 
changes preceding m and n to nasal I, 
ii. 35. 

I: a guttural vowel, i. 20n; contains /, 
i. 89. 

Labials (p, ph, i, bh, m, v, <p, «, u, o, du,), 

how formed, i. 25. See also the differ- 
ent letters. 

Light syllables, i. 61. 

Unguals (t, th, d, dh,n, sh): how formed, 
i. 22, 23 ; anomalously substituted for 
dentals in certain words, i, 63, ii. 60; 
lin^ualize preceding n and t, ii. 12, 14; 
lingualize following dental, ii. 15; sh 
inserted after filial n before, ii. 26 ; 
combination of final lingual and initial 
palatal, ii. 39; restoration of dental for 
lingual in pada and krama texts, iv. 74 
etc. See also the different letters. 

Locative case : i and « are pragrhya as 
ending of, i. 74. 

Long vowels, i 6 1 :— and see Prolongation 
of vowels in sanhitd. 

m: a labial nasal mute, i. 11, 25n; nasal- 
ization of a vowel after the loss or con- 
version of, i. 67. 68 ; m converted into 
visarjaniya, ii. 25 ; assimilated to a fol- 
lowing mute, ii. 31 ; lost before semi- 
vowels and spirants, ii. 32, 33 ; before 
I, becomes nasal I, ii. 35 ; this change 
disregarded by the MSS. and edited 
text, it. 35n ; when retained unchanged 
before r and v, ii. 36, 37 ; when not 
liable to farther alteration, iii. 37. 

Macakiya, quoted by Taitt. Pr., ii. 2 In. 

Mandukeya, quoted by Rik Pr., iii. 56n. 

Manuscripts of Atharva-Veda:— see Athar 

MimSnsakas, quoted by Taitt. Pr., ii. 7n. 

Mora, measure of quantity, i. 59n. 

Mutes : produced by, and named from, 
complete contact of organs, i. 29; par- 
ticular mode of formation and designa- 
tion of the different series and their 
constituents, i. 6n, 10-13. 20-22, 24,25; 
which of them are allowed as finals, i. 
6-9. ii. 3 ; suffer abhinidhdna when fol- 
lowed by another mute or when final, i. 
44, 4 5 ; t ake sphotana or karshana when 
combined in inverted order of series, 
ii. 38, 39. See also the different series 
and letters. 

n : a dental nasal mute, i. 1 1 , 24n ; suf- 
fers abhinidhdna before h. i. 47 ; nasal- 
ization of a vowel after the loss or con- 
version of, i. 67, 68 ; t inserted after it 
before a sibilant, ii. 9 ; converted to n 
before p and j, ii. 10, 11 ; following p 
becomes ch, ii. 17; combination of n 
with Unguals, ii. 12, 15, 16; converted 
when final, to visarjaniya (i. e. has a sib 
ilant inserted after it), ii. 26-28, 30 
converted to r, ii. 29 ; these combina- 
tions historical, not phonetic, ii. 26n 
lost before spirants, ii. 34; converted 


W. D. Whitney, 

before I to nasal I, ii. 35 ; this conver- 
sion not always made in the MSS. and 
edited text, ii 35n ; when duplicated be- 
tween vowels, iii. 27 ; when not liable 
to further euphonic change, iii. 37; n 
changed to n, iii. 75-95; phonetic the- 
ory of the change, iii. 94n , restoration 
of » in pada and krama tixts, iv. 74 etc. 

n: a guttural nasal mute, i. 11, 20n; in 
what words found as final, i. 6n ; suffers 
abhinidhdna before h, i. 47 ; k inserted 
after it before a sibilant, ii. 9 ; when 
duplicated between vowels, iii. 27. 

ft: a palatal nasal route, i, 11, 2 In; n 
changed to, before e and sonant pala- 
tal, ii. 10, 11 ; do. after a palatal, ii. 15. 

n : a lingual nasal mute, i. 11. 22n ; when 
a final, i. 6n ; suffers abhinidhdna before 
h, i. 47 ; { inserted after it before a sib 
ilatit, ii. 9 ; n converted into, before and 
after a lingual mute, ii. 12, 15,16; when 
duplicated between vowels, iii. 27 ; other 
conversions of n to n, iii. 75-95 ; recon 
verted into n in pada and krama texts 
iv. 74 etc. ; n retained as final in pada 
text, iv. 99. ' 

Nasal mutes (re, n, re, n, m) : how formed, 
i. 11, 27; yama and ndsikya inserted 
before, i. 99, 100 ; duplication of, be 
tween vowels, iii. 27 ; final surd mutes 
become nasals before them, ii. 6 ; after 
a nasal, a non-nasal dropped before a 
non-nasal, ii. 20. 

Nasal semivowel (/), i. 27, ii. 35. 

Nasal vowels, i. 27 ; make a heavy sylla- 
ble, i. 63 ; result from loss or conver 
sion of m and n, and from combination 
with a nasalized vowel, i. 67-69 ; spe 
cial case of nasal protracted vowel, i 
70; its treatment in krama, iv. 121 
the particle u nasal in pada-text, i. 72 ; 
character of r and r when nasal, i. 71 ; 
nasal vowels in interior of words u«ual 
ly short, i. 83 ; when long, i. 84-91 
mode of transliterating them in this 
work. ii. 85n. 

Nasalization of a vowel, when made, i. 

Nasikya, a nose -sound, i. 2 6n ; when in- 
serted, i. 100. 

Negative compounds, treatment of in 
pada-text, iv. 56. 

Nose-sounds (ndsikya, yamas), i. 26. 

Noun, i. 1, ii. 44n, add. n. 4. II. 1 ; com- 
position with other parts of speech, add. 
n. .11.2-15. 

o: a labial diphthong, i. 25n; pronuncia- 
tion of, i. 34, 40n ; is pragrhya as par- 
ticle, i. 79 ; do. as termination com- 
pounded with particle w, i. 80 ; do. of 

vocative in pada-text,, i. 81; its com- 
bination with following initial vowel, 
iii. 40, ii. 21 ; ah converted to o,ii.53, 54; 
absorbs initials, iii. 53, 54 ; full exhibi- 
tion of Atharvan usage in this respect, 
iii. 54n ; resulting circumflex, iii. 55. 

Omissions: of initial s, ii. 18; of r before 
r, ii. 19; of a non nasal mute after a 
nasal before a non-nasal, ii. 20 ; of final 
y and v, alter a vowel, ii. 21-24; of m 
and n, before semivowels and spirants, 
ii. 32-34; of final visarjaniya, ii. 55- 
59; restoration of omitted sounds in 
pada and krama texts, iv. 74 etc. 

Organs employed in producing articulate 
sounds, i. 18-28; their distinction in 
each case as passive and active organ, 
i. 1 8n ; degree of their coutact in the 
different classes, i. 29-36. 

p : a labial mute, i. 25n ; visarjaniya con- 
verted to s or sh before, when initial, 
ii. 62-81. 

Pada-text: its importance, iv. 107; con- 
struction of, iv. 1-100; compounding of 
verb and proposition in, iv. 1-6 ; do. of 
these and other parts of speech, add. 
n. 4. II. 2-16; use or omission of ava- 
graha in compounds, iv. 7-72 ; differ- 
ences in this respect between different 
padatexts, iv. 12n, 13n, 26n, 39n, 54n, 
56n, 58n ; restoration of natural form9 
of words, iv. 74-100; repetition of di- 
visible pragrhyas, iv 1 23 ; whether, in 
this respect, the pada-text of the treat- 
ise and that of the MSS. agree, iv.74n ; 
enclitic accent in joado-text, iii. 64, 68, 
69, 72, 73 ; treatment of particle u in, 
i. 72, 73 ; do. of pragrhyas, i. 74n ; vo- 
cative in o is pragrhya in, i. 81 ; combi- 
nation of iti and iva in, i. 82. 

Padavrtta enclitic circumflex, iii. 63-64; 
its comparative tone, iii. 65 intr. n. 

Palatals (c, eh, j,jh, n, y, c, i. i, e, di) : how 
formed, i. 21 ; probable phonetic value, 
i. 21n, ii. 17n; not found as finals, i, 7; 
wand t before palatals, ii. 10, 11, 13, 
14 ; dental after palatal becomes pa- 
latal, ii. 15 ; c inserted after n before 
surd palatal, ii. 26; final lingual before 
palatal causes karshana, ii. 39. 

Paninean indicatory letters and symbols 
used in Praticakhya, iv. 16c; Paninean 
examples and illustrations given in com- 
mentary, add. n. iS. 

Panini : his rules cited in commentary, i. 
In. 2n, iv. 108n ; his doctrines compared 
with those of the Praticakhya, passim. 

Particle, i. 1, add. n. 4. II. 1 ; composition 
with other parts of speech, add. n. 4. II. 

Aiharva- Veda Prdtigdkhya. 


Parts of speech, i. In, add. n. 4.TI.1; 
their various combinations, forming com- 
pound words, add. n. 4. II. 2-16. 

Pauses in recitation of Veda, their length, 
iii. 74n. 

Paushkarasadi, quoted in vdrttika to Pac- 
ini and by Taitt. Pr., ii. 6n, 7n, 17n. 

Penultimate letter of a word styled upa- 
dhd, i. 92. 

ph, a labial mute, i. 25n. 

PlAkshlyana, quoted by Taitt. Pr., ii. 40n. 

Plakshi, quoted by Taitt. Pr., ii. 7n, 40n. 

Pluta:— see Protracted. 

Pracaya cr pracita accent, iii. 71n. 

Praclishta circumflex, iii. 66; its compar- 
ative tone, iii. 55 intr. n. 

Pragrhya : import of the term, i. 73n ; 
what finals are pragrhya, i. 73-81 ; 
how treated in pada-text, i. 74n, iv. 1 23, 
74n ; do. in krama-iext, iv. 117; they 
are exempt from euphonic combination 
in sanhitd, iii. 33. 

Praticakhya : — see Atharva-Veda, Rik, 
Taittii iya, and Vajasaney i Praticakhyas. 

Pratihata enclitic circumflex of Taitt. Pr., 
iii. 62n. 

Prepositions, i. 1 ; list of, add. n. 4. II, 
17-19; other words construed like, 
add. n. 4. II. 20, iv. In, 2n ; preposi- 
tions lingualize initial sibilant of root, 
ii. 90; exceptions, ii. 102-107 ; their 
combination with initial r or f of root, 
iii. 47, 48; they lingualize n of root, 
iii. 79 ; exceptions, iii. 79n ; when sep 
arated from or compounded with verbs, 
iv. 1-7, 36-39 ; their independent use 
and construction, iv. 3n. 

Prolongation of vowels in sanhitd : of 
final of first member of a compound, 
iii. 1-3, 9-12, 12n, 24; of final of a 
theme in declension, iii. 6, 6, 8 ; do. be- 
fore suffixes, iii. 17, 18; of final of a 
word, iii. 16, 19, 25 ; full and system- 
atic exposition of Atbarvan usage in 
this respect, iii. 1 6n ; prolongation of 
first syllable of a word, iii. 15, 21 ; of 
reduplication, iii. 13, 14 ; of particle u, 
iii. 4 ; other cases, iii. 7, 22, 23 ; prolon- 
gation very rare except of d, iii. 1 6n ; 
restoration in pada and krama texts 
of a lengthened vowel, iv. 74 etc. 

Pronoun, name for, ii. 44n. 

Protracted (pluta) vowel : has three mo- 
ras, i. 62 ; list of protracted vowels in 
Atharva-Veda, i. 105; kinds, accent- 
and designation in MSS., i. 105n ; pro- 
traction when omitted in pada and 
krama texts, i. 97, 105n, iv. 120; pro- 
traction of vikampita circumflex sylla- 
bles, iii. 65n. 

Quantity of syllables,!. 51-54; of vowels, 
i. 59-62; of nasalized vowels, i. 83-91. 

a semivowel, i. 30n ; how formed, i. 
28 ; different views as to the classifica- 
tion of, i. 28n ; enters into r and r. i. 
37, 38 ; exchanges with I in certain 
words, i. 64-66; is followed by svara- 
bhakti before any other consonant, i. 
101, 102 ; is lost before r, and the pre- 
ceding vowel lengthened, ii. 19, iii. 20; 
inserted after final n before a vowel, ii. 
29 ; m when retained before, ii. 86 ; 
visarjan.ii/a converted into, ii. 42, 43 ; 
do. in certain words after a and d, ii. 
44-52 ; not doubled as first in a group, 
but causes duplication, iii. 31 ; converts 
s of suffix to sh, ii. 87 ; converts suc- 
ceeding » to re, iii. 75 etc. 

r, r : guttural vowels, i. 20n ; contain r, i. 
37, 38; phonetic value of, i. 37n; how 
nasalized, i. 71; mode of combination 
with preceding final a and d, iii. 46-49 ; 
discordant usage in this respect of Pra- 
ticakhya, MSS., and edited text ex- 
plained, iii. 46n ; convert succeeding n 
to m, iii. 76 etc. 

Reduplication : initial « of root converted 
to sh after, ii. 91 ; do. notwithstanding 
the interposition of, ii. 93 ; prolongation 
of vowel of, iii. 1 3 ; restored to its nor- 
mal quantity in pada and krama texts, 
iv. 82, 84-87, 89, 96. 

Repeated words, treatment of in pada- 
text, ii. 62n, iv. 40, 44. 

Repetitions in pada and krama texts of 
words having certain peculiarities : 
when made in krama, i v. 1 1 7 ; do. in 
pada, iv. 123 ; different usage of the 
different padatexts in this respect, iv. 
74n; restoration of normal form in 
case of repetition, iv. 74 etc. 

Restoration of normal forms of words in 
pada and krama texts, iv. 73 etc. 

Rik Praticakhya : editions of, intr. n. ; 
mode of citing it here followed, intr. 
n. ; general comparison with the pres- 
ent work, add. n. 1 ; its doctrines cited 
or referred to, passim. 

s: a dental spirant, i. 24n, Sin; t inserted 
after t before, ii. 8 ; initial s of certain 
roots lost after ud, ii. 18; visarjaniya 
converted into, before k and p, ii. 62-80; 
s converted into sh before k and p, ii. 
81; do. in other cases, whether final, 
medial, or initial, ii. 82-101; even 
when an augment or reduplication in- 
tervenes after the cause of conversion, 
ii. 92, 93; exceptions, ii. 102-107; ir- 
regular case of insertion of, iii. 96 ; its 


W. D. Whitney, 

insertion prevents division in jtxrda-text, 
iv. 68, 69 ; < restored from sh in pada 
and krama texts, iv, 74 etc. See also 
Sibilants and Spirants. 

Sanhita-text : its construction the subject 
of chapters ii. and iii., ii. 1. 

Samkrtya, quoted by Taitt^Pr., ii. 2 In. 

Semivowels (y, r, I, v) : how formed, i. 30 ; 
meaning of the name, i. 30n; m omitted 
before, ii. 32; alterant vowels before 
dissimilar vowels converted into, iii. 39 ; 
resulting kxhdipra circumflex, iii. 68-61. 
See also the different letters. 

sh: a lingual spirant, i. 22n, 31n; special 
mode of its formation, i. 23 ; phonetic 
value, i. 23n ; dental mute becomes 
lingual after, ii. 16; * converted into, 
before k and p, ii. 81 ; in other cases, 
ii. 82-101 ; exceptions, ii. 1 02- 107 ; con- 
verts succeeding » to n, iii. 75 etc; re- 
converted into * in pada and krama 
texts, iv. 74 etc. 

Short vowels, i. 59. 

Sibilants (f, .«/*, s): aspiration of final 
mutes before, ii. 6 ; great discordance 
of authorities upon this point, ii. 8n ; 
after n, n, and n, k, t, and t inserted 
before, ii. 9 ; sibilants inserted after 
final n, ii. 2ft ; historical ground of this 
insertion, ii. 26n; preceding mtsarjaniya 
assimilated to, ii. 40 ; this rule not fol- 
lowed in M8S. and edited text, ii. 40n ; 
not duplicated after r and h, if fol- 
lowed by vowels, iii. 32. See also Spi- 
rants, and the different letters. 

Similar or homogeneous sounds, i. 27n. 

Sonant letters: defined, i. 13; list of, i. 

Sphotana: defined, i. 103; when it takes 
place, ii. 38. 

Spirants (h, x. f, sh, *, ?>, h): how formed, 
i. 31 ; /suffers abhinidhdna before, i. 
46 ; longer svarabhakti inserted before, 
after r, i. 101 ; loss of in and n before, 
ii. 32-34; final visarjaniya before ini- 
tial surd converted into, ii. 40 ; partial 
disobedience of this rule by MSS. and 
edited text of Atharvan, ii. 4(ln. See 
also Sibilants, and the different letters. 

Strong cases, name for, i. 88. 

Suffixes, peculiar names of certain, i. 88, 
iii. 17, iv. 16n, 20, 21,48; final vowel 
of theme lengthened before, iii. 17, 18; 
when separable or not separable from 
theme by anagraha, iv. 13-72. 

Surd consonants defined, i. 12. 

Svarabhakti : what, and when inserted, i. 
101, 102. 

Svarita : — see Circumflex. 

Syllables, i. 93; quantity of, i. 51-54 
mode of division of, i. 55-58, 104. 

t : a dental mute, i. 24n ; inserted after t 
before s, ii. 8 ; do. after n before sibi- 
lants, ii. 9 ; mode of combination with 
f and I, ii. 13, 17; the combination of 
t and f how made in the MSS. and ed- 
ited text, ii. 17n; do. with preceding 
or following palatals and Unguals, ii. 
14, 15. 

t: a lingual mute, i. 22n ; when a final, 
i. 6n ; t inserted before * after, ii. 8 ; 
inserted after n before sibilants, ii. 9. 

Tairovirama enclitic circumflex of Vaj. Pr., 
iii. 62n. 

Tairovyafijana enclitic circumflex, iii. 62 ; 
its comparative tone, iii. 55 intr. n. 

Taittiriya Pratic&khya : mode of citing it 
here followed, intr. n. ; general com- 
parison with the present work, add. n. 
1; its doctrines cited or referred to, 

Tathabliavya enclitic circumflex of Vaj. 
Pr., iii. 69n. 

th, a dental mute, i. 24n. 

th, a lingual mute, i. 22n. 

Throat-sounds (a, d, h, h), i. 19. 

u, & : labial vowels, i. 25n ; u a pragrhya 
as locative ending, i. 74 ; do. as dual 
ending, i. 75. 

u (particle); technical designation of, iv. 
16n ; how treated in pada text, i. 72, 
73 ; finals compounded with are pra- 
grhya, i. 79, 80; what words are so 
compounded in Atharva-Veda, i. 80n; 
when prolonged in sanhitd, iii. 4 ; when 
not liable to euphonic combination in 
sanhitd. Hi. 36 ; treatment of in krama- 
text, iv. 113, 114, 116, 118, 119. 

TJdatta : — see Acute. 

Ukhya, quoted by Taitt. Pr., ii. 21n. 

Upadhmaniya (<p) : labial spirant, i. 25n, 
31n ; visarjaniya converted into it be- 
fore surd labials, ii. 40n ; this rule not 
observed in MSS. and edited text, ii, 
40n; a few times written with sh in 
MSS., iv. 75n, 77n. 

Upalekha: edition of, iv. 74u ; its doc- 
trines cited, iv. 78 etc., etc. 

v : labial semivowel, i. 25n : SOn ; its pro- 
nunciation, i. 25n : dropped when final 
after any vowel but d, ii. 21, 22 ; ex- 
ceptions, ii. 23; or it has attenuated ut- 
terance, ii. 24 ; dn once made nasal d 
before it, ii. 28 ; m when retained un- 
changed before it, ii. 37. 

Vadabhikara, quoted by Taitt. Pr., ii. 6n. 

Vajasaneyi Pratic&khya : edition of. intr. 
n. ; general comparison with the pres- 
ent work, add. n. 1 ; its doctrines cited 
or referred to, passim. 

Atharva- Veda Prdticdhhya. 


Valmiki, quoted by Taitt. Pr., ii. 17n, 40n. 

Varnakrama: — see Duplication of conso- 

Vatsapra, quoted by Taitt. Pr., ii. 21 n, 

Vatsya, quoted in commentary, ii. 6n. 

Veda: its study recommended, iv. 101- 

Vedamitra, quoted by Rik Pr, i. 29n. 

Verb, i. 1, add. n. 4. II. 1 ; its composition 
with modifying prepositions, iv. 1-7 

Vikampita modification of independent 
circumflex, iii. 65; mode of designating, 
iii. 65n ; occasional protraction of vi- 
kampita syllable in MSS., iii. 65n. 

Visarjaniya (A) : a throat-sound, i. 19n; a 
spirant, i. 31n; a possible final, i. 5; 
called abhinishtdna, i. 42 ; conversion 
of m to, ii. 25; do. of n to, ii. 26-28, 
80 ; historical origin of this combina 
tion, ii. 26n ; assimilation of to follow 
ing surd, ii. 40 ; disagreement of the 
grammarians upon this point, ii. 40n ; 
discordance with this rule of the prac- 
tice of MSS. and edited text, ii. 40n ; 
dropped in edition before sibilant fol- 
lowed by surd mute, ii. 40n ; becomes 
y before a vowel, ii. 41 ; and the y is 
dropped, ii. 21 ; but becomes r after an 
alterant vowel, ii. 42, 43 ; and, in cer- 
tain words, alter a and a, ii, 44-52 ; ah 

converted to o, ii. 53, 64 ; it is droppe d 
after d, ii. 55 ; do. in sah and eshah, ii. 
57, 58; do. in special cases, ii. 66, 69 ; 
anomalous combinations of, ii. 60, 61 ; 
converted to s or sh before k and p, ii . 
62-81; not duplicated, iii. 29 ; restored 
from s or sh, and restored to n, in kra- 
ma and pada texts, iv. 74 etc. 

Vocatives in o, pragrhya only in pada- 
text, i. 81. 

Vowels (a, a, i, i, «, u, r, r, I, e, di, o, du) : 
belong to various classes of sounds, i. 
19n-21n, 25n; degree of contact of or- 
gans in production of, i. 32-36 ; nasal 
vowels, i. 27n ; quantity of vowels, i. 
59-62 ; combinations of vowels, iii. 39- 
54; resulting accent, iii. 55-61, 65, 66. 
See also the different letters, also Nasal 
vowels, and Diphthongs. 

Vriddhi derivatives from compounds, how 
treated in jt>aia-text, iv. 55. 

y: palatal semivowel, i. 21n, 30n; drop- 
ped when final after a vowel, ii. 21 ; 
or has attenuated utterance, ii. 24; 
visarjaniya before a vowel converted 
into, ii. 41. 

Yamas: in part sonant, i. ISn; are nose- 
sounds, i. 26n ; when inserted, i. 99 ; 
their phonetic value, i. 99n ; how writ- 
ten in the commentary, i. 99n.