Skip to main content

Full text of "On -ΣΣ- and -z-"

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 

VI.— ON -22- AND -z-. 

When we compare Homeric words containing -aa- with their 
later forms, we notice that Homeric -aa- corresponds sometimes 
to later -a--, sometimes to that -aa- which is represented in Attic, 
Boeotian and Cretan by -tt-. The examples I shall take to illus- 
trate these changes are reXe'aaai, iroaai, trpaaoeiv, || reXeaai, iroai, 

irpa<T<T(m. There are three theories which might account for this 
variation : (i) That the Homeric -aa- was in one or both of these 
cases not the historical antecedent of the later form, (ii) That 
the Homeric -era- developed differently in different circumstances, 
(iii) That the Homeric -<r<r- represented two sounds, which 
developed independently. 

Now, as to the first theory, difficult though it may be to estab- 
lish a historical connexion between the language of Homer and 
any later dialect, still, few scholars would be willing to deny such 
a connexion, for without it a large part of Greek Philology would 
be reduced to guess-work. In the present case some evidence of 
the connexion can, I think, be adduced. The change from 
Tc\e<r<rai, rroaal to nXeaai, noa-i, is phonetically simple when we bear 
in mind that we have to do, in the first case with I.E. s + s, and in 
the second with d+s. Nothing intervenes to break the series 
s + s, long ss, short s. A comparison of the usages within the 
Homeric poems themselves also throws light on this subject. 
Comparing Iliad I-VI (omitting the Catalogue) with Odyssey 
XXI-XXIV, we find : 

Iliad. Odyssey. 

instances of -o-CT- in fut. and aor. forms, 104 72 

" -aa- in dat. plur., 142 59 

" -a- in fut. and aor. forms, 71 79 

" -a- in dat. plur., 25 16 

Such figures as these cannot, I admit, be pressed to prove very 
much. Not only does Homer use many forms in -aa- without 
any philological justification, but rhythm and set phrases mate- 
rially affect the numbers; for example, IVrecn is rare throughout 
Epic compared to eneaai or meWi, while on the other hand the 

ON -22- AND -Z-. 42 1 

phrase avv xevx«™ almost ousts the form reixtaai. Nevertheless, 
the coexistence of -aa- and -a- in early Epic, and the growth of 
-o- at the expense of -aa-, particularly in verb-forms, in the later 
Epic, seem to establish the historical connexion. 

As to the relationship of Homeric vpaaaetv to the later irpaaaeiv, 
to admit the descent of the latter from the former drives us to 
hypothesis (ii) or (iii) in order to explain the variation, while to 
deny it seems impossible, since there is not the slightest evidence 
of any difference between the two. 

To establish the descent of -tt- from -o-o- is slightly more 
difficult, since Attic gives us no hint of any form previous to -tt-. 
The earliest Cretan inscriptions, however, do give evidence of 
the period before -tt-, and if we may assign the same value and 
history to Attic -tt- as to Cretan -tt-, this gives considerable 
help. The Cretan forms I shall discuss presently. 

As to the second possible theory given above, that the differ- 
ence between -aa- and -a- was due to circumstances of accent, 
position in the word, etc., it would, I think, be impossible to 
apply it consistently. There remains the third possibility, that 
-aa- in Homer represented more than one sound ; what, then, are 
the sounds that we are to assume to have been expressed by 
-cro-? In the first place, I see no other value for the -o-o- of 
TeXeWai, woaai, except dental s, either doubled or, more probably, 
lengthened. The other value must be assigned to that Epic -aa- 
which is represented in later Ionic by -aa- and in Attic, etc., by 
-tt-, that is, the Epic -aa- which arises from Ur. Gk. <ej, *i. t», 0*. 
It is hardly necessary to give examples of this well-known 
change; but I must emphasize the fact that my view differs from 
that of Brugmann in supposing a far closer connexion between 
the dental -aa- and the guttural -aa-. Brugmann supposes the 
former to have been -ss- in prehistoric Greek, and thus makes a 
form like /nAm-a difficult to explain; while the guttural -aa- 1| -tt- 
he takes to be divergent developments from some Ur. Gk. 
spirant. I think the two sets of forms can be better explained 
together. We have -aa-, which is not -ss-, arising from m., and t$; 
the obvious value to assign to -aa- is s. Both changes are illus- 
trated by the English word conscientious, but although the two 
sounds are now identical in English, the first must have been 
originally a palatal s, which we may write *s, while the latter was 
a supra-dental s ('J). The importance of this difference will 
appear later. 


Contrast with the simple and natural change from «t, r^ to s, the 
series assumed by Meyer (Gr. Gr. 3 , §282), tj — tz — is — ss. The 
first step in this series is unexampled and improbable, since the 
change in the position of the vocal organs from/ (= f) to z is no 
slight one. In the second place, why did not the « from this is 
become 5 in Attic as it does where .? follows a dental stem ? The 
only way to meet this objection is to suppose that the change 
Ti — ss was not completed till after dental 4- i had becomes; that 
is to say, there was a time when -vo- from ti had a different value 
from that of -00- from ts; and that period is attested by the 
Homeric poems. Moreover, how is -o<r- from « to be explained 
as -ss- ? The union of « and tj in -era- is to my mind the greatest 
proof of the existence of J as a stage of the development. The 
next point to discuss is the treatment of this s in later Greek. 
Attic, Boeotian and Cretan treated it in a manner markedly 
different from dental s ; they lisped it to ]>, which is now com- 
monly regarded as a phonetic approximation to the sound of -tt-. 
Ionic and the other dialects retained the symbol -00-, and possibly 
retained the sound s. Smyth (Ion. Dial., §375) hints that -00- 
was not a pure sibilant. The transliteration of -00- into Latin as 
x (e. g. in Ulixes, malaxd) seems to show that Greek -00- was 
not Latin -ss-, though Greek -<r- was Latin -s-. Inscriptional 
evidence is also forthcoming. The sign T at Halikarnassus and 
Mesembria interchanges with -00- (see Meyer 3 , 1. c, note). Now, 
if -00- was pronounced -ss-, there was no need for another sign; 
whereas, if -00- was not -ss-, an attempt at a more exact repre- 
sentation was natural ; and even if, as some say in order to 
minimise the importance of the sign, it represented a local 
pronunciation, why did the provincialism affect only the sibilant 
which the other Ionians wrote double, not that which they wrote 

What, then, was the value of Ionic -00- ? It was a sound so 
close to .j that ancient writers give us no hint of any difference, 
nor has any difference survived in Modern Greek. On the other 
hand, it was sufficiently unlike to have a different representation, 
namely the doubled sigma, since doubled dental s had been 
reduced to -0- in Ionic, and to have also a different method of 
transliteration into Latin. The sound that answers to this 
description is s, and we may conclude that Ionic has remained 
at the same stage from which Attic, etc., have advanced a step 
further. Such a conclusion, however, is not likely to go unchal- 

ON -22- AND -Z-. 423 

lenged. Dr. Blass scouts the idea of the existence of the sound 
s in Greek ; he says (Aussprache, p. 92) : " Boeckh was inclined 
to regard this" (such spellings as e'crcm/v) "as an indication of the 
sound s, and his suggestion has found many to repeat it ; it is, 
however, as unwarrantable as it is unmaintainable, and is at 
present given up. The sound J is unknown even in cultivated 
modern Greek : and if the ancients had possessed it, they would 
doubtless have made use of the proper Phoenician symbol to 
express it." Against this it may be argued that the phonetic 
correspondence of modern to ancient Greek is so slight that the 
fact that the sound J has not survived to the 19th century a. D. 
cannot at all disprove its existence in the 5th century B. C. 
Secondly, so little is known for certain about the relationship 
between the Greek sibilant signs and their Phoenician prototypes, 
that it is difficult to say what Phoenician sign would represent J 
in Greek. Without dwelling unduly on this point, I may give 
the following sketch of the question, abstracted from Taylor, 
History of the Alphabet, II, p. 95 ; Roberts, Greek Epigraphy, 
p. 8, and Hinrichs, Gr. Epigraphik, in Miiller's Handbuch. 
The Semitic alphabet possessed the following sibilant signs : 

Numerical order 

i 7 
» 15 

. Sign. 





ds, Z 


hi 18 



is, ss 

iv 21 




Corresponding to these, we have in the oldest Greek alphabets, 
the Western alphabets of Caere and Formello : 




Zeta (= 3d name above) 

ds, z 




? (Xi = 4th name above in 

? (= x in 





San (? = 1st name above) 





Sigma (? = 2d name above) 


The value of the sign I is fixed in nearly all Greek alphabets : 
the value of EB m the Western group is unknown : perhaps it 
was merely numerical ; in the Eastern group it has the form S 
and the valuer. Hence, if any alphabet were found that used 
both the remaining signs, it would be natural to assign to one of 
them the value s, and to the other the value s. Now at Halikar- 
nassus, Teos and Mesembria the two signs are found in use, and 


one of them, in the form T, which is taken to be a variant of M, 
does represent a sound which on phonetic grounds I take to be J. 
Professor Ramsay (Jour. Hell. Stud. I) derives the sign from an 
Asiatic source ; but its use at Mesembria, a Megarian colony 
founded by Chalcedon and Byzantium, makes this less probable, 
since Asiatic influence could hardly have so wide a range. Other 
alphabets possessed either M or § only and used them to repre- 
sent -<r- or -cro- indiscriminately. The confusion between the 
signs M and S is shown by the alphabets of Corinth and Meta- 
pontum, which put M after P in the place of S. 

On early monuments no difference whatever is made between 
-o~ and -fro--, but both are written § or M. It was only at a later 
period that 5§ is written, and at all times 5 appears sporadically 
instead, while I can find no instance of double M. The adoption 
of the double sign was, I think, an attempt to distinguish between 
two closely similar sounds, and the point on which the distinction 
was based was the fact that the prosodial effect of J, like that of p, 
was that of a double consonant, while s did not "make position." 
In early times, then, S represented (i) dental s, as in \v<rai; 
(ii) the same sound doubled or lengthened in Sucdo-a-m ; (iii) s in 
npavaeiv. At a later period the sound of (ii) became short in the 
dialects of Attica and Ionia, but where in the older literature it 
made position, it was written double ; then, since J likewise made 
position, that too was written -<r<x-. The reason for the prosodial 
weight of J I shall discuss later. 

So much, then, for the sign T with the value s. Turning now 
to the ancient Cretan inscriptions, given by Comparetti (Mus. 
Ital. Ill), I shall endeavour to prove the existence of the sound s 
there too. In the archaic Cretan inscriptions the sign I repre- 
sents a sibilant arising from at least four different sources : 

(i) In 100 1 and -MEN, I corresponds to Attic f, later Cretan 
S-, -88-. 

(ii) In OIOS, I arises from I.E. ti, and corresponds to Epic 
-<nx-, Ionic-Attic -a-, later Cretan -tt-. 

(iii) In ANAAIA9AI (= avabaaa^Bai), I comes from -is-, and 
corresponds as in (ii). 

(iv) In FOIIHA (= Epic oIk^o), I represents a peculiar archaic 
Cretan palatalization of k, unknown elsewhere. 

Of these four varieties of I, the last three certainly represent 
voiceless sounds ; the origin of (ii) and (iv) point to a s sound, 
while the later representation of (ii) and (iii) by -tt- points to a 

ON -22- AND -Z-. 425 

sound which was not dental -ss-. I consider that the sound of I 
in these three cases was s, and that these archaic inscriptions, 
which are assigned to the 7th century B. C, preserve evidence of 
a period, unattested by monuments in Attica or Boeotia, before s 
became p. 

The next question that must be considered is the difference 
between the guttural J and the dental J : this difference manifests 
itself in the passage of dental J to dental s under certain condi- 
tions, whereas the guttural s never becomes s. Brugmann con- 
siders that -<ro- from t{, 81 regularly passed to o- both after conso- 
nants and between vowels, and explains forms like ipea-a-a, Kpe'wav 

(Att. ipirra, Kpeirray) as analogical. But how Can olvoiirra, piXtrra 

be explained on this theory ? epeaa-a may perhaps follow irpdao-a 
and icpeioo-av follow rjo-o-cov, through parallelism of meaning, as 
Brugmann says. But I cannot see what analogy can retain the 
large class of feminines in -taaa and -aaaa. It cannot be denied 
that analogy has affected Epic verb-forms in -<r<x- and -<r- ; but in 
such cases analogy is quite as likely to work one way as the 
other ; still, vepeo-ibpai beside vepeao-apu is the only example of a 
verb with -o~ from « that I can discover, whereas ipe<raa>, ipda-a-a 
and \ia-a-opai all show -a-a--, and Attic /3XiVto> shows -tt-. Putting 
aside, then, verbal forms as ambiguous, we have two cases in 
which -o-o- from ti becomes -<x-, namely : (1) after a nasal conso- 
nant ; e. g. TiBeio-a from *ti0€vtui, *Tidevara ; the presence of a nasal 
consonant does not affect -<ro- from kj, ; e. g. ao-o-av from *dyxk<ov. 
Without the nasal consonant we have, beside Skt. dpavatz, Gk. 
oiroto-o-a, for *6jro-J : a<r<ra, with the strong vocalism of the masculine ; 
so also, beside Skt. sail, Gk. Jao-o-a, for *esntia, by the side of 
feminine participles which have the strong stem of the masculine 
as oSa-a, (ovcra, for *esontia. Brugmann (II 400) gives deicao-oa and 
perhaps rrpocppaaa-a and deppao-aa as weak feminine participles. 

(ii) The second case of the reduction of -<ro- from ti- 0i-, to -a- 
occurs in the three sets of words : 

ptcro-os, Ion.-Att. pevos, from *peSjfis, 

rdo-ffor, itoaaos, etc., Ion.-Att. Tocror, from *TOTifts, 

irp6&(Ta>, ottiVctg), Ion.-Att. 7rp6<ra> t from TrpoTfcG). 

Brugmann notices that in these cases Cretan and Boeotian show 
-tt- ; e. g. Boeot. Cret. cm-on-or, Cret. piiTov. Accordingly, if Attic 
piaos is a reduction from Homeric peao-os, the change must be 
pre-historic, since -<r<r- is not to be found in Attic, and -tt- could 


not have been reduced to -o~. The antiquity of the forms piaos, 
etc., is attested by their frequency in the Homeric poems, though 
they are not so common as pJoaos, etc. In Iliad I-VI (omitting 
the Catalogue) occur 

fiiao-os 6 times, So-o-os, etc., 17 times, npoaaa, etc., 9 times; 
peoos 2 times, 8<tos, etc., 5 times, wpoo-a, etc., 1 time. 

Total with -ao--, 32 ; with -a-, 8. In Odyssey XXI-XXIV occur 

/ita-a-os 5 times, oa-a-os, etc., 17 times, np6o-aa>, etc., 3 times ; 
piaos 2 times, Saos, etc., 10 times, npoaa, etc., 1 time. 

Total with -aa-, 25; with -a-, 13. The proportion of forms with 
-a- to those with -ao- is thus twice as great at the end of the 
Odyssey as at the beginning of the Iliad. We can thus see -aa- 
passing to -a- before our eyes : but why it should do so in these 
forms and not in x a P^< r(ra > etc -> ' s not clear. The only point of 
resemblance between the sibilant of peaaos, rSaaos and wpoaato as 
opposed to that of x a P« (r °' a > is that the former is preceded by an 
accented vowel, while the vowel before the latter is unaccented. 
But even this distinction does not appear in the case of updaaav, 
which, since it should be more properly Kpe'aaav, a form which is 
found in Ionic, might be expected to occur as *Kpeoa>v. 

Although the connexion between this position of the accent 
and the change of 's to s is not clear, the closer approximation to 
s of 's as compared with "s makes the change less surprizing. 
We may suppose that the lengthening of the vowel in nStio-a, due 
to the absorption of the nasal, obscured the sibilant sound, and 
perhaps assisted by ntieis with dental s, led to its passage to t : 
while the position of the accent in peaos may have had a similar 
effect. The two cases cannot be considered parallel, since the 
first was proethnic, whereas the second was not completed till the 
Homeric period. The difference is well illustrated by Cretan 
and Boeotian, which show -<r- in the first case, but -rr- in the 

These two dialects show such a curious likeness in their use of 
the group -rr- that they deserve special mention. They are the 
only dialect areas outside Attica that show the lisped s, and when 
we consider that there was no great connexion between them, the 
similarity of their usages is startling. 

In Crete the sound arising from ki, « is represented in four 
different manners at four different periods, the latest embracing 

ON -22- AND -Z-. 427 

the spread of the koivtj with its -<ra-. The earliest inscriptions 
show I, which, as I said above, I take to represent s. The 
Gortyn inscription shows -rr- like Attic and Boeotian ; but Cretan 
approaches more closely to Boeotian than to Attic in two points : 
First, the dental s does not become s in p.eaos, etc., and second, 
the group dental +s becomes -rr-; e.g. Cret. aor. eSarrapav, beside 
pres. barqdai, Boeot. aor. Kop.iTrdp.fvos. I do not believe that these 
are cases of assimilation of spirant to stop : is rather became 
is — s; the archaic dv8a£a6ai preserves this stage; and then s 
became J> later, like the s from Kf,, r«. The third stage of Cretan 
shows 66 for the -rr- of Gortyn and the archaic I, in 6dka68as, 
oddaKiv (Mus. Ital. Ill, p. 681), *Apicad0i (op. cit., p. 691), the dat. 
plur. with 66 from 8+s, like rr in ibarrapav of the Gortyn inscrip- 
tion and I of the archaic dv&afaBai. If 66 was merely a graphic 
variation of rr to represent p, as I think, this is conclusive against 
the theory of assimilation of s to t. This third stage shows 66 
also for <rr in I86avri (Cauer, Del. 1 42), a change most curiously 
paralleled in the Boeotian "tto of Aristophanes, and erre for co-re 
at Orchomenus. J. and T. Baunack, to explain 166am, assume 
the stages or — pi — pj> ; this may be quite correct, since there is 
no necessity to explain -66- from -or- on the same principle as 
-66- from dental +.s; for the regular appearance of -or- in the 
Gortyn inscription, e. g. in KaTiaraptv, shows that the passage of 
ox to 66 was much later than that of dental +s to rr, 66. It is to 
be noticed, however, that even in archaic Cretan, <r is assimilated 
to a following 6. 

Dr. Blass has an article in the Jahrbiicher f. Philologie for 1891, 
p. 1 seqq., on an inscription from Phaistos in Crete, containing 
the words TTPATEI and EVTA090I (?), which he assigns to the 
first century B. c, but which Halbherr, who first edited it in Mus. 
Ital. Ill, p. 559, assigns to the third century B. C. In his article 
he advances the view that 6 was the hard explosive aspirate in 
Cretan even at this late date, and that rr in the Gortyn inscrip- 
tions was a double stop. He then explains 

Gortyn **ApKarri (analogous to eSarrdpav) : later 'hpnadOt 
Gortyn npa&ba : later npar{T)u 

by assuming "eine art lautverschiebung," though he admits that 
this new "Grimm's Law in Greece" does not affect the aspirate 6. 
If, however, we may believe, on the authority of Meister, the 
Baunacks, Comparetti, and Dr. Blass himself in his Aussprache, 


that 6 was f in Cretan as early as the Gortyn period, these forms 
can be otherwise explained. At that period Cretan possessed an 
inter-dental spirant, 6, developed from the dental aspirate, and a 
supra-dental spirant, tt, developed from the spirant s. At a later 
period these two sounds were confused, and were both written 6 
or 66. Hence vopria66av (Mus. Ital. I, p. 44) = irpoooiaav || Gortyn 

larra. If the forms 7rpaT(rjei, etrirpe piTrev = Attic, k&tto- 

\oyirre66a>, quoted by Dr. Blass, belong to this period, we must 
put them alongside the form Tnjwi, Doric ASxa, Attic Zijva, and 
assume that when tt ceased to represent f, it was used instead of 
d& to express Tt. 

The history of initial «>,, tj is not so easy to trace. In the first 
place, the materials are scanty; secondly, the need of expressing 
the syllable weight of a final short vowel preceding was not felt 
to be sufficient to justify the use of a doubled initial <r, so that it 
is difficult to distinguish between s, 's, and "s when initial ; thirdly, 
dentalized gutturals cause further ambiguity ; e. g. how can it be 
determined whether Megarian <rd = riva came from *«a or from 
*n,a ? We have from k« or ti Ionic a- in aeva ; this o- frequently 
makes position in Homer, and lengthens the augment in every 
case but one. In this it appears to have the value s ; those cases 
in which a short vowel remains short before o-cij<» may be explained 
partly by assuming a poetic license, similar to but perhaps not so 
harsh as that by which Sicdfiaj/Spor appears in hexameter verse, 
partly by supposing that the poet attended occasionally rather 
to the written form of the word than to its pronunciation. As 
regards syllable weight, I equate J exactly with p, and initial p 
does not always lengthen a preceding short vowel. Brugmann 
connects doubtfully with a-eito the Attic revraopai, revTtifa, explaining 
the initial r- as a shortening of -tt- which would have appeared in 
the augmented and reduplicated forms. But if initial a in Ionic 
o-eia was J, — and we can hardly suppose that the sibilant of eoo-ev* 
was s, while that of otc vevairo was 5 — why should not initial t of 
Attic have had the same value as medial -tt- ? Another example 
from Kt or n is the Megarian ad quoted above, to which corres- 
ponds the Attic enclitic -rra, as in irocra tto, whence, by wrong 
division, arra. If -tto had ever followed a consonant, would it 
not have been written -to ? 

From ti we have Ionic criy/«poK, <rrjres (in Etym. Mag.) beside 
Att. Tripepov, rijTes. The origin of these forms from the pronominal 
stem which appears in Skt. as tya- would support the pronunci- 

ON -22- AND -Z-. 429 

ation of initial a-, t- as J and J> but for the fact that there exist 
many little-understood cases of Attic initial r corresponding to 
Ionic a--, where the origin of the two sounds seems to be ty, ; e. g. 

ripPt}, Ionic o-ipfiri, TtjXia \\ 0-17X10, Siarram || <xaa> : with these must be 

grouped the forms tv || <rv and Terrapes beside reWaper. These 
forms are quite separate from those discussed in this paper. 
Whether the -tt- of nrrapet and Sum-da represented a double stop 
or a spirant, and what was the process of its development from 
tu, I cannot at present determine. 

It will be seen that in the theory of -trtr- given above, I have 
not hesitated to ascribe more than one value to a single Greek 
sign in one and the same alphabet. The assumption that the 
Greek alphabet, like most others, did not possess sufficient conso- 
nant signs to represent accurately all the sounds of the language 
which was written in it, is one that has been frequently made by 
philologists and passed over without notice. The assumption is, 
I think, quite justifiable, for it would be a miracle if a borrowed 
alphabet could express all the sounds of the language that bor- 
rowed it. The fact that some of the Greek symbols were conven- 
tional, e. g. -tro-, -rr-, -85-, which, it must be remembered, were 
not distinguished in writing from -<r-, -t-, -8-, in early times, is to 
me less surprizing than the fact that the Greeks themselves do 
not appear to have thought the matter worthy of remark. 

This assumption, then, I make in the case off, for no one value 
has ever yet been proposed for this sign which is satisfactory in 
every case in which the sign appears. 

The following points appear to me clear with reference to the 
pronunciation of £ ; — (i) that in ifa, 'A^vafe, Boeot. 6c6£oto?, etc., it 
was zd. tfa preserves I.E. zd, while the other two forms are late 
compounds, (ii) that after the time of Alexander (see Blass, 
Ausspr., p. 91) it was dental z in Z/iipva ifiewvpi, etc. (iii) that in 
other forms it was neither zd nor z. Take, to begin with, the 
Epic-Ionic-Attic £ in <pv£a, <rx% a > Kpd£a, n-«/«rd£». In these cases £ 
is manifestly the outcome of I.E. di, gi, and many attempts have 
been made to bridge over the gulf between these sound groups 
and the sounds zd, or z. Blass (p. 125) assumes the series di — 
dz — zd, the last change being due either to metathesis or to the 
analogy of I.E. zd in tfa. Both steps are difficult, for dz is a 
long way from di, and the conversion of dz to zd is almost 
impossible. G. Meyer (Gr. Gr. 3 , §284), recognizing this, derives 
zd straight from d%, in defiance of the law " Natura saltum non 


facit." He justifies the change by appealing to the O.C.S. zd 
from di. But (i) O.C.S. is a long way from Greek ; (ii) z is not 
z ; (iii) at best this is only ignolum per ignoiius. I would explain 
O.C.S. zd by equating the z to the di and assuming the d to be 
parasitic. Hoffman (Die gr. Dialekte, II, p. 512) argues that 
original £ = dz must have become simple z\ he obtains the 
required value zd by the series di — dz — z — zd—zd. He does not 
say whether all these changes were proethnic; he does not 
parallel the adventitious d, whereas in O.C.S. & from ti does 
parallel zd from di, nor does he explain why it passed from 
spirant to stop, and that, too, after a spirant. If, then, it is only 
with the greatest of difficulty that zd can be obtained from di, it 
would seem to be quite impossible to obtain it from gi. The 
parallel between ki, ti and gi, di seems to me to be quite close. 
Just as the former pair meet at the sound J, which passes in 
certain dialects to \, so the latter pair, as I think, meet at the 
sound z, which passes in Doric to d. Now the place of articu- 
lation of the sound z is no more fixed than is that of s, and, 
furthermore, it shades off into other spirantal sounds. Modified 
in the front it becomes d; another modification produces z. At 
the back of the mouth it produces spirant 3, the spirant of the 
German tnorgen ; this in its turn can pass into spirant y or semi- 
vowel i, as it has in the English yesterday, Boeotian lav = iim> for 
iywv, and Tarentine SXios for oklyos. Every one of this series of 
connected sounds has some bearing on the history of the symbol 
f in Greek. 

I shall first give evidence for the existence of the sound z in 
Greek, and then consider the value of £ in the different dialects. 
My evidence is drawn partly from transliteration, partly from 
phonetic considerations. 

Early Latin transliterations give but little help, since the 
symbol z had become obsolete in the Latin alphabet, at an early 
period. Latin accordingly had no symbol wherewith to represent 
the Greek £, except s, which we find in Saguntum, and the 
Plautine sona, tarpessita, badisso,- comissor, etc. A curious, but 
probably quite accidental, resemblance to these last two forms is 
seen in the Tarentine vaKma<ra>, <f>pdoaa>, etc., with -00- for Ionic- 
Attic f. This gives rise to the supposition that the Tarentines 
had transformed their -f-verbs to -o-a- by analogy, like the Attic 
ApHOTTto and Thessalian ivt<j>avia<roev ; but we have one Tarentine 
verb in -&>, namely avdfa, and by derivation that should have -a<r-. 

OAT -22- AND -Z-. 43 1 

Without assuming a 'lautverschiebung' in Tarentine, I suggest 
that here the signs -f- and -<r<r- had interchanged values, £ being 
pronounced s, and -o-<r- as z. Such variations, whether due to 
analogical transference of forms or confusion of alphabetical 
symbols, certainly point to a closer resemblance between -an- and 
-£- than exists between -ss- and -zd. Early Latin can thus give 
us but little assistance, but some light is thrown on the question 
by the correspondence of Late Latin z, Greek £, to classical 
Latin consonant i and g before e z and i, and d before i. Lindsay 
(Lat. Lang., p. 49) considers that the sound in these cases wasj/, 
and that Latin 2 and Gk. £ were then pronounced z. The subse- 
quent history of the sound he gives as follows : 

Lat. consonant t) _ * }• "' 

T . . ( Low Lat. y Italian gi (= dz) 

La.^ - (written i)^ S .Italifnl 

Lat. di ) . \ y 

bpanish J 

But the disparity between the sound and the sign in Low Latin, 
together with the fact that French and Italian have now a sound 
which is closer to the original Latin than is this postulated Low 
Latin y, makes me think that the following series more closely 
represents the facts of the case : 

Lat. consonant n T T . . ~ . { T Fr ^ ch >(= *) 

T . • ( ^ Low Latin z or dz / I Italian et (= dz) 

\; g J' g (written,, or 0< S. Italian \ 

Lat. di 1 3— . \y 

bpanish i 

So that the people who wrote for Latin Julia, Greek Zot/X«a had 
not such a bad ear for sounds as to write z when they meant y: 
they pronounced the name as a modern Frenchman or Italian 
would pronounce it. Roby (Lat. Gram. I, §195) assigns the 
value dz or I to Late Latin z in these cases. 

The passage of consonant i to z through the stages i—y — 3 — z 
is illustrated by the Sanskrit hariya, transliterated from Greek 
oplfav (Wackernagel, Alt. Ind. Gramm., pp. 137, 242). This, 
however, does not support the value y for £ in ZovXem, inasmuch 
as transliteration out of a language is a vastly different thing from 
transliteration into it. For example, if Greek possessed a sound 
z, another language, having no z, might transliterate it as_y; but 
that y would not be written z in Gk., but probably *'. Greek £ 
represents, not Sanskrit y, but a palatal consonant, in 'o&w; = 
Ujjayini (Blass, Ausspr., p. 128). 


The Sanskrit language paid more attention to phonetics than 
any other that has ever existed ; is it likely that it would have 
represented zd or z by y"i The nearest equivalent to z in 
Sanskrit would be s; the nearest to z would be palatal /' or 
semivowel jj/. 

According to Meyer (Gr. Gr. 3 , §226, note) o-f is employed on 
papyri to represent Arabic and Coptic J. Granting the difficulty 
that any Aryan tongue would find in accurately reproducing a 
Semitic sibilant, still szd seems a very weak attempt to represent 
any kind of a J sound : sz is at least intelligible. 

As inscriptional evidence I may cite the archaic Cretan I, 
referred to with reference to the value J. I sought to prove that 
in three cases out of four it had that value ; in the fourth case it 
represents a voiced sound, which I take to be the voiced counter- 
part of J, namely 2. 

From Cyprus come the forms a£a0os and fa, where ( represents 
the spirantized y, that is 3. 

The Cyprian icopfc, Aeolic mpCa, £a = 81a show f as a late 
formation from 81 ; the sound here was probably Eng. j, that is 
dz, or perhaps z. 

So far I have treated only of f from I.E. di, gi; but { from I.E. 
spirant y, in £vy6v, ga, can also be explained as z. Sievers defines 
the difference between semivowel i and spirant y as due to greater 
friction. Whether that greater friction is produced by narrowing 
the air-passage or increasing the pressure of the air-current, the 
same process that produces 2 from * will, if continued, produce 3 
from 2. If we give this value, j, to I.E.j/, its representation by 
Gk. {, that is z, no longer needs to be explained by a complicated 
process such as that given by Meyer, j — dj — zd, which obscures 
the difference between I.E. i and I.E.j>. The passage from the 
spirant to the semivowel in other languages is readily paralleled 
by the English yesterday. This distinction between the palatal 
semivowel i and the palatal spirant y (=3~) is the same as that 
between the labial semivowel u and the labial spirant t>. 

Just as the voiceless s was lisped to ]> (tt) in Attica, Boeotia 
and Crete, so the voiced z was lisped to d (88) throughout the 
entire range of the Doric dialect ; e. g. Laconian povolSSei, Mega- 
rian /idSSav, both in Aristophanes ; Cretan (Gortyn) 8u«i88<», Boeo- 
tian TpujrsSSa ; initial 8 = d is seen in Laconian Adv, Cretan 
(Gortyn) 8<»i/, = Boeotian 8a>«, Sicilian Ady/cky. And just as the 
supra-dental -tt- was confused in Crete with inter-dental 6, so in 
Elean supra-dental -88- was confused with inter- dental 8 = d. 

ON -22- AND -Z-. 433 

In Aeolic we find £ = Ionic-Attic f written on inscriptions ; it is 
also attested by grammarians as the sign employed for the sound 
arising from the late union of 8j in Kap£a, and £a = 8id ; the symbols 
-<t&- also occur in such forms as /«XtV8e«/, SSeis, given by MSS and 
grammarians; the first inscriptional evidence for it is on an 
archaising monument of imperial times. 

Meyer (Gr. Gr., 1. c.) explains <rfi as due to the fact that £, 
formerly zd, had become z in the rest of Greece, and that Aeolic, 
preserving the sound zd, adopted a new sign to represent it. My 
objections to this are as follows : (i) I hold that Meyer has not 
substantiated the value zd for f in all cases, especially from g{. 
(ii) It remains to be proved that £ was simply z in the rest of 
Greece, (iii) On Meyer's own theory f in Aeolic Kapfa, etc., was 
z ; but z does not arise directly from di : the stages are di — dz — 
z; a further step, and no inconsiderable one, is necessary to 
arrive at z. (iv) Although a dialect might adopt a sign which it 
did not possess, from another dialect, it would hardly discard a 
sign which it did possess, because another dialect used it with a 
different value. According to my theory Aeolic £ was zd or z 
down to quite late times : the spelling ah was due to confusion of 
the two values ; 8(6(oros and deoo-doros were equivalent, so beside 
S«icaf« arose Sixdo-Sei, with o-S — z. What, then, became of the 
discarded symbol ( ? It may have been employed to represent 
the affricate dz in KiipCa. 

The <x8 of the Sicilian Doric of Theokritus is probably merely 
a literary form. The Doric -88- seems to have been entirely 
banished from elevated literature, its place being taken either by 
the Ionic f or the Aeolic o-S. That vo was foreign to Sicilian 
might be taken for granted, were it not for the Oscan Nm/io-Siijir, 
which occurs in a Mamertine inscription at Messana, written 
about 280 B. c. (Conway, Italic Dialects, No. 1). The alphabet 
is that form of the Ionic alphabet which came into general use in 
S. Italy. Two conjectures are open : we may suppose that the 
sound to be represented, namely Oscan intervocalic -s-, corres- 
ponded to the value of the Greek f, but that it was the fashion at 
that time in Sicily to represent this by -<r8-. Of such a fashion 
we have no other evidence except the conflicting spelling of the 
MSS of Theokritus. Secondly, if we suppose that the Oscan -s- 
corresponded to no value of -f-, that was known in S. Italy, the 
-08- would be an isolated attempt to represent Oscan -.?-. If 
there was any connexion between this Oscan -a-S- and Aeolic -a-S-, 



the sound in Oscan, on Meister's theory, would be zd, which is 
obviously impossible ; while if Aeolic -<t8- was merely a graphic 
variant for f = z, we must believe that Oscan -si- was pronounced 
-zi-, which is possible without being probable. On the whole it 
seems best to treat the two as independent. 

The question of the value of the symbol -f- in Elean is a 
curiously complicated one. In the inscriptions of the earliest 
period £ appears for I.E. and Ur. Gk. d; apparently d had been 
spirantized to d, and I explain the use of the symbol f in this 
manner : — Just as there was a period before -an- (s) became -tt- 
(f) in Attic, so there was a period before -f- (z) became -88- (<?) 
in Doric. In Elean the change of sound was not at first accom- 
panied by a change of sign. That is, f = z became f = d, where 
the origin of the sound was I.E. di, gi or spirant y. Then, when 
Ur. Gk. d became d, this too was written £. Unfortunately, the 
early inscriptions contain no sure example of a representative of 
I.E. di, gi, y. In my opinion f would be found in these cases, 
with the value d. Inscriptions after the 5th century show the 
ordinary Doric spelling -88- for £, and 8 for d; those at the end of 
the 5th century represent a transition period ; the sign is usually 
8, with rare lapses to the older f. For the whole question cp. 
Meister (II, p. 52), from whom I quote the following: "Dass 
diese beiden spirantischen Laute des eleischen Dialekts, der 
durch C Ratios = 8i}fios) und durch 8, 88 (8uyo>< = £vy6v) bezeichnete 
sich unterschieden, ist fur gewiss anzunehmen, denn w'aren sie 
zusammengefallen, so wiirde man sie nicht durch verschieden 
gewahlte Schreibung auseinander gehalten haben; worin aber 
der Unterschied bestand lasst sich nicht erkennen." To this I 
would reply that there may have been a difference of sound, 
namely, that between inter-dental d and supra-dental d, but the 
difference was never expressed. The difference of sign is chro- 
nological : it does not appear in any inscription except those two 
of the transition period, one of which shows £ once (in & = 8«) 
with 10 cases of 8 unaltered, while the other shows one f with 20 
cases of 8. 

Arkadian shows f = I.E. di, gi,y; e. g. Stxafi/Toi (Meist. II 106 ; 
Cauer 2 , 457). A difficulty arises from the appearance of £ = I.E. 
velar 3 in gpeBpov (Strabo, VIII 8. 4 (p. 389) ; Meyer, Gr. Gr. 3 , p. 
266). The stage previous to gpeBpov could not have been 8epe6pov, 
since Arkadian preserves Ur. Gr. 8 unchanged. The only other 
similar form is £{k\a> — pd\\a, given by Hesychius without a 

ON -22- AND -Z-. 435 

locality. The two forms seem to contain a peculiar product of 
velar 3 : since it is not a case of dentalization, it seems possible 
that it arose from a spirantizing of y, seen also in Cyprian, which 
y must have appeared beside the regular representatives of 3" as 
it does in y\i<papov and yetpvpa ; 8 appears in eVSeAAovrer (Collitz, 
1222. 49). Under what conditions 8 and £ appear respectively 
cannot be determined with such scanty material. The only other 
view is that Arkadian contained a mixture of dialects. 

I have now to treat of the metrical weight of the sounds 
represented by -ua-, -tt- and -£-. In the first place it must be 
remembered that any continuous consonant can, in the Epic 
dialect, make a metrically long syllable when following a short 
vowel. A stopped consonant has not this power, except in very 
rare cases. In these cases, then, the difference between a heavy 
syllable and a light syllable depended on the difference between 
continuous and stop consonant. That is, the greater amount of 
time spent, or breath used, in the production of a continuous 
consonant made the syllable containing it long as compared with 
a syllable containing a stop. In the later language, however, the 
continuous sounds /, m, n and dental s were no longer able to 
give metrical weight; and such combinations of stop and contin- 
uous sound as tr were treated in the same fashion. The change 
was, perhaps, not so much a change of pronunciation, although 
the sibilant of ihUaaaa may have been dwelt upon longer than 
that of ihUaoa, but was rather due to a change in the feeling of 
what constituted metrical weight. The other continuous sounds 
retained their power of making metrical weight even in classical 
Greek ; p, aa- (= s), £ (= z), and -tt- (= J>) and -dS- (= tt) regu- 
larly " make position," and the reason is not difficult to see. The 
production of these sounds needs a more open position of the 
vocal chords, and consequently involves a greater muscular 
exertion and consumption of breath than does that of /, m, n, s. 
The sound z or 3 has a more open position than any other sound 
not a vowel. Consequently it produced a heavy syllable in 
Greek ; and its effect in the pronunciation of English is analo- 
gous. Compare the long vowel of please with that of pleasure, 
Asia with azure, mete beside measure. 

To the Greek of the 6th century B. c. this difference of syllable 
weight afforded the most striking contrast between the sounds s 
and s. It was the point on which the different alphabetical repre- 
sentation was based. Consequently it is not surprizing to find 


that as the language decayed, and syllable weight disappeared 
before the stress accent, the difference between 5 and s likewise 
disappeared, so that no trace of it survives in Modern Greek. 
But, in my view, the sounds s and z lasted at least long enough 
to account for the spellings malaxo and ZouXcta. 

In conclusion, I may briefly summarize my position as follows : 
(a) The present views on -o-<r-, -£-, -n--, -88- are unsatisfactory 
because — 

(i) reXeWai and npaweip could not both have been pronounced 
with dental s. 

(ii) The series commonly given to show the development of «i, 
etc., in Gk. contain too many phonetical difficulties, and 

(iii) they separate n from «* and both from 8i, yi, although -(re- 
unites the first pair, and the interchange of -a-a- with -£- and the 
correspondence -rr- \\ -88- unites the voiceless with the voiced 

(0) Since a new theory is necessary, the values s, z suggest 
themselves as the representatives of Ur. Gk. «t, etc., because — 

(i) «£, t* naturally converge to J. 

7i> 8*1 3 ( = I-E. jc) naturally converge to I. 

(ii) Archaic Cretan I in F oifija shows dialectic growth of s from 
palatalized k. 

Archaic Cretan I from dental + ^ shows dialectk grQwth of 

Gortyn Cretan rr " " V j ( ^ f)fromr , 

Boeotian rr " " ) K y 

Cyprian £ in a£a66s shows dialectic growth of z from 3. 

Aeolic £ in K ap£a \ shows dialecdc growth of | or d " g f rom 8t . 

Cyprian £ in Kop£a j 

(iii) It is quite possible that the Semitic symbols should, when 
adopted, have the values assigned to them by my theory, and 
probable that in the sign T we have the fourth Semitic symbol 
with the fourth Semitic value. 

(iv) J, z and supra-dental j>, tt resemble r in their phonetic 
character as in their metrical effect. 

(v) Transliteration, where it gives any help at all, favours my 
theory, especially transliteration into and from the most scientific 
of all alphabets, the Sanskrit. 

Caius College, Cambridge. W. F. WlTTON.