THE PRONUNCIATION OF ATTIC GREEK ~ IN THE SIXTH AND FIFTH CENTURIES B.C.

Summary

There is no consensus of opinion on the phonetic value of ~ in Classical Attic. Scholars incline to show preference for either dz or zd, gene~a,lly the latter. This paper offers arguments for the less popular view.

The determination of the phc, net~c alue of ~ in 6th- and 5th-century Attic still offers scope for debate because of its essenti~ diffic~ty x). The habit of the more cautious investigator is to present the rival values, dz and zd, and either to leave the question open or else to admit a phonetic dualism *). Those who must have a definite answer, on the other hand, are apt to show preference for one of the values, generally zd. a) This interpretation has the sanction of 'aatiquity', in effect that of the late Greek grammarians Dionysius Thrax (2nd century B.C.) and Dionysius Halicarnassensis (lst century B.C.)4), and was supported by Erasmus In the 16th century 5) and favoured by the authority of Brugrnann towards the end of the 19th ~'). More recently it has been upheld by E. H. Sturtevant 7), M. Grammont s),

l) F. Blass, Uber die Aussprache des Gviechischen ~ (Berlin, 1888). t) B. F. C. Atklnson, The Greek Language ~ (London, 1931); A. C. J uret,

Phondtique grecque (Strasbourg, 1938). s) M. Lejeune, Traitg de phondtique grecque (Paris, 1947). 4) Dionysius Thrax, Avs grammatica (Tkxv~q y~a~av~×~). Ed. G. Uhlig

(Leipzig, 1883); Dionysius Halicarnassensis, De compositione verborum (Ilcpl o ~ n c t o g 6vott~x~o~) in H. Usener et L. Radermacher, Dionysii Hali- carnassei opuscula (Leipzig, 1899).

5) Desiderius E~asmus, De recta latini graecique sermonis pronuntiatione, 1525. Erasmus's 'etacist' pronunciation ot Gx-eek was opposed to J. Reuchlin's (1455~ 1522) ei~:acist' variety, which treated ~ as 2:.

6) K. Brugrrann, Grundriss der vevgleichenden Grammatik der indoger- manischen Sprachen I ~ (Strassburg, 1897) and Kurze vergleichende Grammatik der indo-germanischen Sprachen (Strassburg, 1904); also K. Brugmann und A. Thumb, Griechisikz Grammatik 4 (Munich, 1913).

~) The Pronunciation of Greek and Latin (Philadelphia, I940). 8} Pkongtique du grec anciet, (Lyon-Paris, 1948).

64

and others 9). In direct contrast to this interpretat,~,on of ~ as a con- sonszltal group, we have G. Curtius's view ~o) that the character represents the affzicate dz, and this view is stiU r e~a t ed b y m ~ t writ~ers of school grammars in Central and Western Europe n) either unchanged or with slight modifications like that suggested by L. Rous- sel, who would have us believe that the end of the affricate represented by ~ is 'tr~s faiblement teint~e d'un chuintement un peu pareil/k la fin du g italien devant e, i', which gives d3 z~). All investigators are agreed however that from the 4th century B.C. onwards ~ had its modem w~lue z as).

Our subject is strictly limited in compass: we are concerned to establish no more than the pronunciation of the character ~ in Attic Greek up to the time when it definitely became a voiced ~-~o~ o~ha..., (z). This means that our material w~a be t~ken primarily from the records of a Greek dialect in the 6th and 5th centuries B.C. Yet investigation of only this material, however scrupulous, would ob- viously not be enough. We shall also have to consider the evidence of pre-Classical Attic and of the ancient Greek dialects, and to this we must add the evidence of the Italic dialects and of Old Church Slavonic, as well as the values of ~ and z in present-day Greek and Romance respectively. All our material therefore, except the scienti- fically studied testimony of modern speech, which alone has phonetic relevance and cogency, belongs not only t 9 a written language but to a language used at a remote period of time, and its interpretation therefore is inevitably a matter of more or less plausible conjecture.

9) A. CarnGy, Manuel de linguistique grecque (Paris, 1924); U. Mengin, Ecriture et prononciation du grec ancien (Paris, 1948).

Io) G. Curti_,s und E. Windisch, Grundzi~ge der griechischen Etymologie I - - I I (Leipzig, 18791.

11) W. H. D. Rouse, A First Gree,~ Course (London, 1906) and The Sounds o/ Ancient Greek (London, 1935); L. Roussel, La, prononciation de r attique classique (Paris, 1921); E. D:rerup, Die Schulaussprache des Gviechischen (Paderborn, 1930).

I2) Op. cir. in fn 11.

~s) Cf. G. N. Hatzidakis, EinIeitung in die neugviechiscke Grammatik (Leip- zig, 1892) and Zrwo~o~ {o-top~ ~ k.:~,~,~tx~g y),~nmtg (Athens, 1915).

65

II

The character ~ now occupies sixth place in the sequence of the Greek alphabet, but as a numeral sign (~') it had the value 7, having been preceded by the obsolete digamma, (~), which immediately followed ~ +~6v and stood for 6. It was one of the twenty-three ×~8~tx, or q~otvt×/~t~, Tt~&tzi~x~ ai), originally sixteen, which the Greeks, ac- cording to Classical tradition ~5), had taken from the Phoenicians in the 9th century B.C. and possibly even earlier xr). The earliest Attic inscriptions, which are slightly earlier than the 8th century, have I for ~, and this 'primitive' form of the character occurs also in the Ionic alphabet, which was introduced in Athens under the archon z~uc~iue, in "^~ • ,vo ~.~. This character corresponds exactly in shape to one variety of Phoenician zain 17) (cf. Heb. T), which also appears as / and I. The form oI ~ , x with short horizontal strokes is found in the earliest Semitic inscription, viz. that on the tomb of King Ahir~m of Byblos (c. 1500--1000 B.C.), in 8th- and 7th-century Attic (750--600 B.C.), Rhodian (650 B.C.), and Cretan (650 600 B.C.)~a), and in Etruscan xg), Oscan 20), and Messapian ~1). The other Greek form of ~, viz. the uncial Z, occurs in 9th-century North Phoenician (Zen- jirli Inscription) and 8th-century Old Aramaic, which displaced this dialect az). The dating of the characters which we have adduced here suggests that the forms / and I preceded Z, and this inference is

a4) Herodotus, Historiae V, 58--59. Edd. H. R. Dietsch et H. Kallenberg (Leipzig, 1887).

15) Plinius Secundus, Naturalis historia V, xiii, 67. Edd. L. Janus et C. Mayhoff (Leipzig, 1906--09) ; Tacitus, ,4nnales, xi, 14. t~ez. H. Furneaux (Oxford, 1894).

~) D. Diringer, The Alphabet 2 (London, 1949). ,7) This name occurs in Doric as ~&v (Herodotus I, 139), which corresponds

to Ionic and Attic ~ y ~ . ,s) K. Meisterhans und E. Schwyzer, Grammatik der attischen Inschri/ten 3

(Berlin, 1900); W. Larfeld, Handbuch der griecMschen Epigraphik I (Leipzig, 1917) ; H. Pernot, D'Hom~re d nos ~ours (Paris, 192 I).

,9) D. Diringer, op. tit. in fn 16. ~0) C. D. Buck, A Grammar o[ Oscan and Umbrian (Boston, 1904). zi) A. Kirchoff, Studien zur Geschichte des griechischen Alphabets s (Berlin,

1877). ~2) Z. S. Harris, A Grammar o/ the Phoenician Language (New Haven,

~936). 5

66

supported by the meaning of the Semitic name of ~. Zayin (I) in Hebrew and zain (Estrangela \) in SyTiac both mean 'weapon', presumably some sort of missile. This meaning has been traced back to Egyptian z~n/zcon ~). The hieroglyph representing 'arrow' (~) is naturally horizontal, like that of 'bolt' (m), whereas the corre- sponding Semitic character is, as naturally, uptight, and it is this, as Classical antiquity and modem scholarship agree, that served as model to Greek ~. The Greek name of the character however is very curious. In the Septuagint, Hebrew zayin becomes ~ (Psalm CXIX, 49), which is also the Ethiopian name (z~ 'weapon'), and it would appear that this name, as ~Tq (z~; cf. Classical Arabic zd), was adapted to the name of the following character ~v~(~), a Greek ren- dering of ]~t (c[. Hebrew n)u). Such an explanation seems to be

well founded, for names of adjacent zettez~, ~ ~,, ,~,.,j,~,-~,, ,u.,,,,~,~, tend sometimes to approximate to one another phonetically (cf. Russian vosem' 'octo', which is an adaptation to sere" ' septem'; also Russian oktjabr' 'October', whose form was determined by those sentiabr' 'September' and nojabr' 'November'). In Greek, of course. the name ~1"~, like that of the other letters of the alphabet, has no Greek meaning, but it can be readily interpreted in terms of Semitic. Here we have, on the one hand, a parallel to the truncated and obscure names of the Latin characters, only one of which (cf. English zed), a later bon'ewing from Greek, is easy to associate with its original, and, on the other hand, to Cyrillic, which can mostly interpret its letter-names as Slavonic words (e.g. s--dz~lo 'very' and 3~zemlja 'earth', both of which are associated with Greek ~).

III

The characters of the Greek alphabet, represent both single sounds and sound-groups (cf. n, ~ with ~, ~), and the sound-groups are either 'aspirated' plosives (e.g. ~, ,¢~, ~.) or combinations of a plosive with a sibilant fricative (e.g. ,J~, ~.). The former constitute the complete triad ph, th, kh and might have been indicated graphically by the symbol of a plosive followed by that of the glottal fricative h, viz.

~8) G. R. Driver, Semitic Writing #om Pictograph to Alphabet (London, 1948).

2~) G. R. Driver, op. cit. in fn 23.

67

~', , ' , x' (cf. I'.~., K.~. in Thera and Melos :inscriptions), or, less satis- factorily, by starting with the 'rough breathing' (e.g. ~, '~, 'x; cf. 'p), but the consistent triple correlation of /orris, lenis, and aspirata (the last strictly a 'prosodic' modification of the corresponding/orris) demanded separate characters for each class, and the availability of these in later times was put to good account in the representation of sounds unknown to the older language (e.g. ~, ~, X). The prosodic nature of aspiration in Greek has a parallel in the more elaborate, 'pentadic' system of Sanskrit plosives, which includes the aspirated lenes, bh, dh, d.h, jh, gh. But Sanskrit has no close parallels to offer to Greek

and ~ (cf. Latin x), which analyse into it's and ks respectively 25). Here we find sibilation conceived as a prosody, but the series of such sounds with a sibilant 'release' is incomplete, not extending to ts, which figvres in Byzantine and Modem Greek as v~. In point of fact these consonantal groups--- and they could have been adequately written ~ and ×a are not true affricates, lacking as they do the homorganic association of plosive and fricative found in ts.

It would appear then that the pholonogical principle of 'one sound one letter' was not rigorously applied in the G~eek alphabet or

rather that its application was vitiated bv inaccuracies in the phonetic anMysis of contemporary speech. \V~, may contrast the majority of the characters ~ all the vowel symbois and most of the consonant symbols, which uphold the principle, with a small minority, which illustrate an attempt to treat not only a;piration but sibilation prosodically. To this minority of characters b, longs ~, but this charac- ter, as normally interpreted (viz. zd or dz), falls outside the foregoing categories, for zd is merely a loose consonaWal group which requires no separate symbol to represent: it, and dz :i:.~ a pure affricate which does (cf. Old Church Slavonic and Modem Macedonian s).

IV

The orthographic approach to our problem is, as we have seen, inti- mately associated with the phonetic. This may be either static or historical. Examination of the sound-systems of the most diverse

~5) The compound charac ter t rans l i te ra ted ks is the nearest app roach to ~. Cf. Sanskri t aks.ah ' axle ' with Greek ~o~ , L i t i n axis. The Greek ~ and + were originally wri t ten ×a and na (Attic xa and q~" respectively).

68

languages discovers the presence of affricates, which are invariably regarded as single consonants. The 'typical' affricate ts is found in Indo-European, Uralian, Caucasian, and Japanese; common to English and Russian; dz is found in Italian, Macedonian, and Georgian; d3 figures in English, Italian, Ukrainian, and Turkic; and less common types of affricates are not far to seek (e.g. German ib/¢, Swiss German kx). We shall find however that they are not always indicated by separate characters (cf. English ch with Czech ~, English i with Ukrai- nian dS,, Slovak dz with Macedonian s). In no language, to our know-

• = ' 1 S ' ledge, are the metamese of the denti-alveolar affricates, viz. st, zd, r , 3d, represented by a single charac*~er. The reason for this seems obvious enough: these consonantal groupings and others like them, in which the second member is a plosive, do not strike the ear as consonantal units any more than do juxtapositions of sibilant and sonant (e.g. sn, st, sl) or of sibilant and another type of fricative consonant(e.g, s[, 3v). A priori then Greek ~, as a single character, could have stood for either a single, phonetically 'atomic' consonant or an affricate.

V

Before studying the evidence of Attic Greek and the other ancient Greek dialects, let us examine the evidence of Semitic, from which, as we know, the Greek alphabet derives. Comparative Semitic lin- guistics 2,3) has established an hypothetical protoglossa (Common Semitic) to summarise the points of resemblance between the recorded Semitic languages. This protoglossa, like Classical Arabic, appears to have distinguished ~ and z, which coalesced in z in Canaanitic. Accordingly we are authorised to assume on theoretical grounds that Phoenician zain (I and Z) was pronounced z at the time when the Phoenician alphabet was taken over by the Greeks. It has been suggested however that it may have a 'double sound, like Greek ~' 27), because the Phoenician demonstrative adjective ~ is almost alway:~ written with prothetic ~¢ (e.g. ".~¢ n a ~ 'this monument'). This suggestiov. ultimately rests on the traditional interpretation of ~ and possibly

~*) C. Brockelmann, Grundriss der vergleichenden Grammatik der semitischer, Sprachen I (Berlin, 1908); G. Bergstrfisser, Ein[Ohrung in die semitische Sprachu4ssenschafl (Munich, 1928).

t~) Z. S. Harris, op. cit. in fn 22.

69

also on the knowledge that z is ts/tf in Hit t i te ~). The conception of ~ r as a 'double sound' appears to receive some support from its use along with a, ~r, and ~ in Greek transcriptions of Phoenician sade (~) phonetically emphatic s (of. Arabic d. dd), in the Punic period, which begins in the 5th century B.C. and ends in 146 B.C. (e.g. Zcopo~/ Tupo~ for "l~ in Appian, Punica I). By the middle of this period however Attic ~ had become z, and we find this Hellenistic pronunciation reproduced in the Greek loans of the Talmud (e.g. lt~:~t z~md < ~:o~$6g 'broth', ~ zfn~ < ~ v ~ 'girdle'; ~ z~g < ~uy6v 'yoke'; r ~ t 6rez < 6pu~0t 'twing') ~*).

It wiU be clear by now that the conjectured value of Phoenician zain as the simple sibilant z does not help us to interpret Attic: ~ as

t r l , ~ 1 1 1 - ~ l , , ~,.~...~..~,4' " 4 . t . ^ ~d .1 . n a , ,~ , , ,~ ~ u , , u m . , , = . , t . century ~" D. but the fact that ~ and z coalesced in z in Phoenician suggests the possibility that zain may have been a composite sound at least dialectaUy. The choice of the character by the Greeks to represent a 'double sound' in their own language must have been guided by its possession of at least a z- element, which was lacking in Greek except as a positional variant of the s-phoneme. But whether the z-element came first or second in the 'double sound' is not easy to establish Z. Hm~s 's 30) obserwat:ion that the Cypriote Phoenician demonst~*:~tive adjective T is always preceded by I~ inclines u, to ~ssume that it came first, which would make ~ the symbol of the affricate dz.

VI

The Indo-European t,rotoglossa, defined in the 19th century chiefly by the Neograrmnarian School, is, like its Semitic counter- part, a complicated set cf concordances based on the comparison of forms recording its later evolution at different historical periods. The inevitable disparity i1~ time diminishes to some extent the validity

~8) E. H. S t u r t e v a n t and E m m a A. Hahn, A Comparative Grammar o/ the Hittite Language (New Haven, 1951). On p. 25 we read: ' . . . i n H i t t i t e words, z has the va lue of t or d ptus ~'. We are also reminded t h a t : ' T h e r e q s no cer ta in proof w h a t the phonet ic cha rac te r of the s ibi lant was, bu t the f r e q u e n t double wri t ing indicates t~a t it was a voiceless sound. '

~0) S. Krauss , Grieehische und lateinische Lehnw6rter im Talmud, Mid~asch und Targum I - - I I (Berlin, 1898~99) .

30) Op. c.it. in fn 22.

71

7(T) as parallels to < (cf. Boeotian &LO ’ live’ with Attic c&w, &.+I

Cretan 4xLva ‘girdle’ ‘I judge’ with Attic

i7q&a& ‘to count’ with

The phonetic interpretation of c as zd, which appears f:o emerge cx3ncIusively from a consideration of the foregoing examples, is, in

testimony of two Greek grammarians - Dionysius Thrax, who

flourished c. I ensis, who came cients” of such investigators testimony in detail %). We

may summa&e this here by saying that both gramrrarians, describe the symbols [, (J and < as representing in Classical times, i.e. at least three hundered years before Dionysius Thrax, not simp le, but ‘double’, or composite, sounds occupying the space of two s:mbol:ised con- sonants in the syllable to which they belong. Both interpret t: as zd, i.e. as a compound whose second constituent is the plos:Cve, in contrast to 4 and +# whose second onstituent is t ilant , viz. XD arld m. 36)

VII

(c) ci(t$), Sansktit iivdb “living’, Avestzm ’ fiIA T&h Am Gnthir &us Old C~UIY& Shvcmic

f v&u ***u&a y&v) u-r---+ ~--v-j --- ------I -VT- v--_-4

Armenian Ksam, Welsh by@. Further proof

a sedectisn of them shall be given here, viz. (a) Qq ‘leaven’, ‘, Lath ifk, Old Church Slavonic jzlclca Lithuati cuy6v ‘yoke”, Hixtite yzckan, Sanskrit y Old Church SLwonic ago (6 *&p),

onetics, tradition itself appears, at le;lst once, &. Our author this time i

in Modern Persian: lass’, Persian

73

- labial, dental, and velar, Archinus says of the dental sounds that they axe made ‘with the blade of the tongue near the teeth’ (~i;z X~U& Y& yhcjqr; xcqh ‘&q &%v+raq) and uses 6 as an illustration, adding significantly : ‘and for this reason c is produced in that place’ (xab Bt& mike rb < xa& ~akqv yew&r$a~ dp &av). This suggests that a d-element was present in c, and if < were merely zd, which is an ordinary compound requiring no special notation, there would have been no need to use it as :kn illustration, for the d-element is identical with 8.

The subsequent development of <, whether it is taken to be xd or dz, begins m the 4th century .C. and aft,er, w en we come across misspellings suc!r a5 iivaf3 IL& t?rs++jqxasv, ~p..&mmLV, zp<-

pcw&& Z&, < , and Zp’;Qva. Here we observe a confusion and i rchange of the character d and c, both of which, apparently, require, to be interpreted as z, his has been the normal pronwnciation of < in literary Greek since the inauguration of the XOW~ :

it is ,the pronunciation of this character in ehenistic, Byzantine, an odem Gree e statement that Greek drZ (zd) becomes z from the

4th century onwards must naturally be limited by reference to dialectal dev ments and present pr over the Greek-speaking area. Roman grammarians lstian era like Terentius

aurus (Znd century) and th century) state that c was still pronounced & (ts) in contemporary Greek as spoken by the unlettered. J. ‘Psichari 8’) quotes forms such as nomindzo, nomilzzo, and tioanBnzo as being in use on the island of Chios in the 19th century;

t a) mentions sraf6cu (with dz) as occuring in Carpathos, and ‘@) FecOFdS ~&%9~, (i.e. ‘dgavokos) for ~K$~o~oL; and

s in the the now expatriate Cappadocian dialect of age. These, of course are all probably later develo

merits, but they are ge possibility of the existe Greek.

*‘) See F. Blass (Op. cit.), who quotes audouin (Bdetin de co~rtu~xm-

dams h&biqwe IV, pa 366). 88) O@. cit. in In 2. 8@) Modma Greek I’N Asi Misaov (Cambridge, 19 16).

74

VIII

The value of [ in Italic and Romance is not Unifom, a it embodies va,iations, often contiderable, of both time and place.. The oldest

. re,co,-ds are pre-Ch&ian and come from oxan and Lkbrian m), both of which drew their alphabets from a Grtxk ductus q’)f Chalcidian type (i.e. with 1 for 2) through the alien Etruscani In bc*th languages the sPbol I(Z) appears to have been pronounced ho we have, for instance, Oscan zicokmz and $24~ conespondirlg to Latin diem and act respectively, and Umbrian artemenzarzc corresponding to Latin infermenstyitinz. This value of < may be due to a local, or dialecta,l, Greek interpretation of the symbol ol)a In the &can Tabula Bantina of the 2nd century B.C., which is written in the Latin alphabet, 5 (for I) however is used with its Oscan value 43). This appears to have been ideniicai t&h the Hel.lenJGcl, ;c+;fi Greek va!ue of cP viz _zj e_x~qt in

certain unusual spellings of Imperial times, where x represents &, for instance oze (&-&?), aeta (diaeta), zebza (dii%zcs), and Z&s&

(Diorrysius). Before z *was adopted into Latin from the Greek however, the sound it stood for was represented by ss in 0 d Latin (e.g. fi2dw2

( $&, bad&ire x ‘PO(8i&xv, atlicissa’re < &TTWL&IY), which implies a long

sound as well as unvoicing. In the modern Romance languages the character z has different

phonetic values, which however are constant for ’ each Ianpage :

thus 2 is 2 in French, ts/dz in Italian (e.g. one ‘barley’; mezzo ‘ha middle’), and $ (e.g. danza ‘dance’) in Castilian Spanish. Its French value recurs in English and Dutch; its Italian value its in German and Icelandic. In the Slavonic Latin alphabets, in. Hungarian, and in Rumanian z is z. The Italian pronunciation of the character as liz may reflect its phonetic value in Appennine Greek, which because of its comparative isolation did not undergo the ‘deaffrication’ oi ?I

in Balkan Greek. A parallel to this is lxesented by the ‘colonial’ Cretan, some of whose early inscripti ns have c for the consonantal groups ~(0) + j (e.g. G&s ‘such’ <

~&~xw). Paul K oqoc;, cf. Attic Baoc; @?‘?‘~og

re SC t h mer 43) suggests that c here is the affricate ts, as in ~!;ca;l and Urnbrian,

75

IX

The Shmnic contri ution to our problem involves the phonetic

value of a character and the apparently parallel developmen% of a consonantal group in Old Chur Slavonic (Old Bulgarian). Cyrillic S, which occurs to-day only in acedonian where it has the value of dz, has a Greek and a Latin red, but not phonetic counterpart. Its modem aced&an value appears to have been shared by l&h- century Old Bulgarian, whose two alphabets, Glagofitic and Cyrillic, carefully distinguish t.he affricate represented by this character from z, represented in Giagolitic y a Mike! character and in Cyrillic by either a hooked z or an uncialised minuscule 3 &). Obviously 5 must have ha its modern pronunciation (z) in the 9th century, -when Const e (St Cyril) a disciple adopted the Greek uncials to represent e sounds of th acedonian variety of Slavonic spoken on the periphery of his native essalonica 4:s). In view of this, another character was needed. to represent the a,ffricate, once symbolised by <, and the inventors of the earliest Slavonic alphabets had to go outside the contemporary Greek ductus for an appropriate symbol, unless Cyrillic s is a stylised reversal of it may well be such a reversal is su ported by the existence ernative character for

dzin Cyrillic. This character has a short s rough the correspond-

ing character for z, wlhich is obviously from the Greek uncial

2. Here we have made a very helpful discover-v: the same basic character, derived from Greek *<, seems to

_

epresent both dz and z, and the difference between their phonetic va es is adequately shown

by what is essentially an ‘incorporated’ diacritic mark. The relation of dz to ,z is made visually clear, and the former is regarded as a phonetic unit requiring a separate symbol, as Classical Greek required 5 for what must also have been a phonetic unit and not merely a juxta-

as suggested by the spelling ~6. velopment of a consonantal group in Classical Qprian is more apparent than real. The inter-

44) J. Vajs, RukovtY hldi~kske' pnkogm~ie (Psague, 1932) ; Je. F. Karskij, Slav janska ja kidlovska ja pale .kgraf ijsz. (Leningrad, 1928).

4s) See my articles ’ The Old EIulgarian Language-Type’ (Archivum Lisguis- tiwm I, 2, Glasgow, 1949) and ‘!;+ources of Old Church Slavonic (Tlze Slavcmic

and East Ezcrcqfwan Rmiew XXWII, 7 1, London, 1950).

prdation of <, emanaW$ fmm *d + j, as zd seems to receive

finnation from the origin of $& w&h m be traced back the other Slavonic ckVd0

COnsOnantal group; Russian has 4 Serbian has 8, Polish and Slc~

1 1-q dz, Czech has 2 and Slovene has j (e.g. Old

margin’, R&an W&Z, Ser’bian da, Polish mkda, Sllovak, me&g, Czech me&e, Slovene me@, all of them going back to an I.-E. 1 #

and we are led to assume that fd represents of the constituents of df or a simplification of

either a transposition #dZ (cf. Old B

X

wie have considered a representative sekction of the evide advanced by comparative study for the phonetk ~~bpiX?ta~bn of < and may now proceed to an examination of the relevant lexical material in the light of modern phonetic knojMledge. Study of the &t&ution of ahis character in the c1aSsiCa.l ttic voc listed in standard dictionaries 47) shows that it occurs ody i

in both words and syllables, i.e. it follows a pause or a vowel (e.g. &$VU~C ‘I yoke’, c?jAos ‘zeal’, ?&&I ‘yoke’, c rian’, 6Coq ‘branch’, ~+KJ ‘I think’, &pxdi&~

save’), The character hefore ut and obu; it is commonest before q and o and, unlike 6 d J1, is never final; medially, i.e. at

of a syllable, it tends to follow t and a more oft&n than not. For purposes of syllabification is treated as a unit and is a&

tittefi to begin a syllable (e.g. tpp~~-~i-&

this function it res most profinently at the be- he other consonantal gr&ps d

ut reference to

iagram: ._._

Palatalisatio

rication

a horizontal association, whit the dental and alveolar are similar mav be seen in t

the direction of

d” (cf. Lakian sog’is ‘judge’ “’

ussia ise

aa) See my review of K. Kepes The Slavo~eic and East

(< d + i) tends als~ to change into the pdatal PlmiVe (e.g. Czech ted’ (now’). ant the normal tendency is to produce slants. The hush-sibilant figares in such affricates in twcj guises - either as d3’

(cf. Polish &‘, Serbian a) or as d3 \ ‘Serbian and Macedonian &). The

former may pass into 1, l as it has in Slovene (cf. Serbian m&j with

Slovene meja), the latter into ‘3 as LJ French (cf. French joie with q )

Old French joi and English joy). The possibility of a change of 3 into 2 is seen in Baltic (cf. Lithuanian z’i&i ‘to know’ with Latvian

z&it), and the same language-group illustrates the intimate association of g, and dz (cf. Lithuanian &iz ’ ‘to drink’ with Latvian &&) without

d3 to connect them. Similarly Persian z is directly preceded by a PaHavi j (d3) (cf. Pahlayi roj *day’ with Persian roz). The development of d5 into &z 1s * however as ‘natural’ as that of 3 into z and may be

seen by comparing standard Italian d3, as in gid ‘formerly, already’, giovane ‘youth(ful)’ with Venetian dx in &a, &ovine. Thus a the lines of development in our diagram are accounted for, and in the light of them we can now establish the value of Classicd Greek c as a phonetic entity,

XI

Assuming the manifest ori@d unity of the phonetic ‘compound’

represented by r and its subsequent SimpbfiCatiOn, we CUlnot COIKei~??

it as zd, if our scheme of historically attested phonetic developments is v&d, for such a ‘compound’ would be easily analysable into its components z and d, or, in other words, into distinct phonemes, which even the phonetically untraine ear would be inclined to keep apart as distinct acoustic impressions. These components are specified by the grammarians of antiquity as constituting c, but they are placed in the order z + d, in contrast to the components of 6 and JI, whose sibilant element (s) comes last. This interpretation of c, as we already know, is confirmed by numerous etymologies, from the opa pfionetic structure m2

ue Bcos, whose

be elucidated by the comparative method, to the transparent ‘Af+at& with its transformation of the ending -83~~ and such misspellings as A&Jw~. Various dialectal forms offer SQPPoe here: in Aeolic and Asian Ionic -a& often ccrrcspclnds to %tic -r- fe.& x$& for Z&G;, ypov&$w for cppo\r~ic~). There are also instances of C as the equivalent of zd in transcripts of foreign names

hG ‘f2~~%$ CM Persian Awamazdi):, of the substitution af

79

6 for 1: in intervocalic position, which is normally 4g) interpreted as representing syncope of z parallel to that of TV between consonants (cf. Ae&-Ionic @So < F&?& with PGLw ‘pedii’ < *p&w), and of the 10s~ of v alike before G and < (cf. CW&MU ‘I draw together’ with ~U?&JW.U ‘I gird up’). It is clear from this catalogue of evidence that the items in it are not of equal value: some of them belong to Classical Attic, others to divergent dialects; some are established ‘by the application of the comparative method, others from the native resources of the language. The possibility of different treatments in cognate dialects must be necessarily conceded, and this will eliminate. the dialectal evidence which is not strictly c gt3nt to our argument here. Spelling errors too, unless they are suf ciently frequent, may be discounted, and too much emphasis need not be laid on parallels involving CF and x (6). Although s and z are phonologically correlative, they are acoustically distinct, with at least voice as the differential, and can therefore function independent y. In saying this we do not underestimate the validity of the evide e for the interpretation of < as zd, but muster and phonetic assessment of the available data would seem to show that c symbolised dz, to which zd may well have been assimilated in Classical Attic. \Ve are inclined to concur with M. Lejeune 50), that < represents a unifor unciation, but, unlike him, we believe that zd must have become 11”~ and not the reverse. The character < must have been allocated to a phonetic and phono- logical unit, and this could have been only dz, which is one outcome of the transformation of both *dj and *gj (cf. Zel;~ with Sanskrit &uw~, Cypriote CE ‘earth’ with Attic ~7) as well as one member of the ‘heteroglottic’ ratio-voiced affricate/j (cf. Q$-q with Sanskrit y$$ It was also natural for dz to discard its plosive element in course of time. This may be illustrated, for instance, by the form acx,uicx (2nd century B.C.), which reflects Doric ?&J& and &tic CyipSa ‘loss’ (cf.

however ~aphc for ~q.da in esychius). Initially, o for the sound z may also be seen in the Greek name (CE$~~) of the Phrygian goddess Zemele (cf. Avestan Za;t,z, Lithuanian 2%unynrz ‘Earth Goddess’) 51). -_

48) E. H. Sturtevant, The Pronz.wiaGg?z oj Greek and Latin (Philadelphia, 1940).

50) 09. cit. in fn 3. 51) P. Kretschmer, ‘Semele und Pinnvsos’ (Aus der A now&z, Berlin, 1890) ;

F. Altheim, Terra6 Mater (Giessen, 193 lj ; L. H. Gray, The F~wdxt~o~u of tk Iranian Rsiigiora-s (Bombay, 1929).

The explanation of the tnditiond equation c = zd, which goes back to Dionysius Thrax, must L qa sought chiefly in the para,& Attic

and Aeolic spellings of identic lexunes (e.g. Z&65: C8&). If ()sgi&

c WS.S pronounced dx, as we have tried to prove in this paper, & ia Aeolic must be a development from it by metathesis later than the 9th cent-uqy- .v*RJ. \I&&. ‘12 r LJ Old Bugari~~~~ Jd) u). The reverse _proce~ appears

to have taken place in Attic, where dz WAS paramount and copse- quently attracted to itself the original juxtapositions of 2 and d icf, Attic V,w ‘I sit clown’ with Doric E&I < *CW&GJ).

-- ---

51) See fn 46.

sa) ff the intqxetarion of Homeric a~ in, say, pbac~1q ‘middle’ and Attic (3 in @GO< * 1s mepted as representing an earlier ts (< ~~)1 this pPlo*etic ~~~~~~~~e~t of ‘ers a complete parallel to that of c as the affricate & viz. ‘@ ) LIZ > 2. In other words, because of the regressive assimilatio’n of ~~~~~~a~~ts in (2 reek, dz becomes Z, as LS becomes S, whereas & can give only d.