R.S.P. BEEKES PRE-GREEK.

A LANGUAGE RECONSTRUCTED
(second version, November 2003)

To the memory of Edzard Johan Furne e

CONTENTS
A. INTRODUCTION B. PHONOLOGY 1. The phonemic system of Pre-Greek 2. Characteristic sound(group)s 1. au; 2. b; 3. bd; 4. gd; 5. gn; 6. dn; 7. kt; 8. kc; 9 mn; 10. ou; 11. pf; 12. rd; 13. rkn; 14. rn (rd, nd); 15. s; 16. sb; 17. sg; 18. sk, st; 19. stl; 20. tq; 21. fq; 22. cm, cn; 23. w; 24. geminates 3. Prothetic vowel 4. s mobile 5. Variations: consonants 5.1. Voiceless / voiced / aspirated stop 5.2. Prenasalization 5.3. Nasalization 5.4. Labial, m and u! a. p, b, f/m b. p, b, f/(#) c. m/(#) 5.5. Stops / s(s), t or stop + s, t 5.6. Velar / labial / dental: labio-velars 5.7. Dentals / liquids (a. r/l; b. d/l; c. t,q/l; d. n/l; e. d/r; f. n/r) 5.8. Simple / geminate 5.9. s- / zero, (before consonant and) before vowel 5.10. K-, T- / zero 5.11. n-, l- / zero 5.12. Metathesis, and aspiration shift 5.13. Secondary phonetic developme nts 5.14. Other variations 6. Variations: vowels 6.1. Single vowels, timbre

6.2. Long / short 6.3. Single / diphthong 6.4. Rising diphthongs? 6.5. Secondary vowels (or syncope) C. MORPHOLOGY 1. Reduplication 2. Suffixes Introduction The material: -ab-, -ag-, -agg-, -ad-, -aq-, -ai-/e(i)-, -ai(#)-o, -aib-, -aiq-, -ain-, -, -air-o-, -ak-, -al(l)-o-, -amb-, -amn-, -am-o-, -an-, a n-, -and-, andr-, -anq/t-, -ann-, -ax-, -ap-o-, -ar, -ar-, -as-a/o-;ass-, -at-, -aur-a/o-, -ac-, -gda, -gr-, -ed-, -ez-a, eir-o-, -ell-a/o-, -el-a/o-, -em-o-, -emn-o-, -enn-a, -er-a, -et-o-, -eur-, -eut-, -hb-, -hq-, -hk/c-, -hl-, -hn, -hr, -hr-, -hs(s)-, -ht-, -htt-, -hy-, -q-o-, -qr-, -ib-, -igg/k/c-, -i d-, -idna, -i q-, -ik-, -i l-, -ill-, -imn-, -in-, -i n-, -ind-, -inq-, -is-, -it-, -ic-, -kn-, -m-o-, -n-, -x-, -op-, -or-, -oss-a, -ott-a, -oul-o, -our-, -ous(s)-a, -pn-, -pt-, -r-, -rg-, -rd-, -rn-, -sk-, -s-o, -ss-, -st-, -stron, -t-o-, -tt-, -ub-, -ugg-, -ud-, -udna, -uq-, -ui-a, -uk-, -u k-, -ul-, -um-, -umb-, -umn-, -u n-, -und-, -unq/t-, -unn-, -ux-, -up-, -ur-, -u r-, -us-, -ut-, -uc-, -fq-, -f-o-, -wk-, -wl-, -wm-, -wn-, -wp-, -wr-, -wss-, -wt-. 3. Word end 3.1. in vowel: 1a. -a; 3.2. in -r: 3.4. in -n 3.5. in -a j D. THE UNITY OF PRE-GREEK [I am indebted to professors Melchert and Ruijgh for their comments.] 3.3. in -x, -y 3a. -x 1b. -i, -ij; 3b. -y 1c. -u, -uj 2c. -or; 1d. -euj 1e. -w, -wj 2d. -wr 2a. -ar; 2b. -ir;

A. INTRODUCTION In 1956, Kuiper wrote a study on Greek substratum words which opened a new chapter in the study of the field. Furne!e then wrote a dissertation on the subject (1972), in which he presented twenty years of research and which is up to now the handbook for the subject. As Kuiper was also my promotor, I knew the book from the beginning (see my review in Lingua 36, 1975). The short overview which follows here is based on Furne !e's material and on my own research during thirty years. Furne!e's book met fierce criticism and was largely neglected. [An exception is R.A. Brown's Pre-Greek Speech on Crete, 1985.] That was a major mistake. Pre-Greek words often show variations which are not found in inherited words. It is obvious to study these variations. That is what Furne!e did. It appeared, as Kuiper had shown, that these variations show certain patterns, so that they can be used to recognize Pre -Greek elements. Two points of criticism on Furne !e are possible. One is that he considered almost all variations as expressive, which is certainly wrong. It is evident that the variants are due to the adaptation of words of a foreign language to Greek. We shall see below that in this perspective many variations can be understood. The second objection to Furne !e is that he suggested etymological connections between Greek words, as variants of a Pre -Greek word. When several possible variations are used, many combinations become possible, and here Furne !e has gone too far in a number of cases. We can hardly reproach the author in thes e cases, however, as he was exploring new ground: it is only to be expected that in such a situation one goes sometimes too far. The method, however, was correct; what we have to do, as always in the case of etymology, is to see which suggestions are proba ble and which are not. Of course, in many cases we cannot reach certainty, but this is no objection. On the whole, Furne !e's material is indeed Pre-Greek with a very few exceptions, and so we have a large corpus of material. His index gives 4400 words; as many derivatives and variants are given (and a few Indo -European words that were discussed in the text), I guess that it contains some 1000 Pre-Greek etyma. Also, Furne !e often adduces new material, (even now) not mentioned in the etymological dictionaries (mostly glosses from Hesychius). I have in general given only a few names, and no material from outside Greece and Asi a Minor. The comparison with Basque or Caucasian languages has not been considered as this is not my competence; I think it possible that there are such connections, but that must be left to others. My suggestions for reconstructions are not essential. One may ignore them and just consider the variations themselves. Often these variations are explained as incidental phenomena (assimilations, influence of other words, etc.), and such explanations may be correct in some cases. But if we know which variations frequently occur, we are warned to consider Pre -Greek origin if we find them. The existing etymological dictionaries often seem to `avoid' the conclusion that a word is a substratum element. It is remarkable that Chantraine was quite aware of the question in his Formation, but has very often withdrawn his - in my view correct - evaluation in his dictionary. It seems as if substratum elements were not welcome. A question that is a real problem, is that a word is often called a loan from an Anatolian langua ge, while it may be just as well a word from the substratum in Greece. It is generally accepted, on the basis of the place names, that the same language was once spoken in Greece and in (western) Asia Minor. [A point for further study is to establish how far to the east related names are found. It is my impression that such forms are found in the south as far as Cilicia.] But it is mostly impossible to distinguish between substratum words and - (mostly) later - loans from Asia Minor. A word may have been taken over through commerce etc., as happens between two neighbouring countries, or in the time when Greeks settled in Asia Minor, which happened probably as early as in the 14th century. I think that, from a methodological point, it is better to consider su ch words as Pre-Greek, and only to take them as - normal - loan words when there is reason to do so, but it is clear that here we may often make mistakes. A good example is tolu/ph `clew, ball of wool ready for spinning'. The word is clearly related with Luw., Hitt. talupa/i- `lump, clod'. The Greek word is typical for Pre -Greek words: CaC-up- (with a = o before u); there is no IE etymology (Melchert, Orpheus 8 (1998) 47-51 does not convince). So it is Pre-Greek / Anatolian. [I shall abbreviate

this, when necessary PrGrAn.] Also, `clew...' is not a word that you bring home from overseas; it is an everyday word, which the Greeks took up at home. I completely agree with Furne !e's interpretation (35 n. 33) that the word was brought to Greece by the settlers from Anatolia who brought their language, which, from another perspective, we call Pre-Greek to Greece. So it is a loan fron an Anatolian language, but from the one that was al so spoken in Greece before the Indo-European speaking Greeks arrived there. The essential point is that it should be recognized that substratum words are a frequent phenomenon. One may regret this, for example because Indo -European words can be much better explained, but this is irrelevant; it is simply a fact that has to be accepted. My approach is rather that it is fascinating that in this way we can learn something about the old languages of Anatolia, and of the role of Anatolia in early history. And, of course, it is part of the oldest history of Greece. As to `Pelasgian' and related theories which assume an Indo -European substratum in Greece, these theories have failed, and I no longer mention them (in my etymological dictionary). The theory has bee n extensively discussed by Furne !e (37-68). `Pelasgian' has done much harm, and it is time to definitely reject it. The latest attempt was Heubeck's `Minoisch-Mykenische' (1961 Graegraeca), where the material was reduced to some ten words; the theory has been tacitly abandoned, I think. (Another matter is the problem of su~j / uj and the z-, for which Ruijgh assumes an Indo-European para-stratum. Here the evidence seems so clear that an explanation of this kind may have to be assumed.)

B. PHONOLOGY 1. The phonemic system of Pre-Greek The consonants. The fact that voiced, voiceless and aspirated stops interchange shows that voice and aspiration were not distinctive in Pre-Greek. On the other hand, the Mycenaean signs for rjo, rja and tja show that palatalization probably was. This is confirmed by the sign pte which will go back on pye. (In our material cf. qa/pta. I wonder whether krosso/fqon points to py > pt but then realized with aspiration.) Further, the signs two, twe, dwo, dwe, nwa, swa, swi point to labialization as a distinctive feature, i.e.: t"o, t"e, d"o, d"e, n"a, s"a, s"i. Note that palatal and labial forms are found with resonants and stops. The existence of labio -velars is confirmed by qasireu basileu/j, etc. (See further Beekes, Glotta 73 (1995/6) 12f.) This results in a system: p p y p" t ty t" k k y k" c c y c" s s y s" r r y r" l ly l" m m y m" n n y n" Of course, it is possible that one or more of the supposed phonemes did not occur (e.g. my: a palatalized m is a rare sound).

dak"n-, in the first form giving a labial (this time aspirated), f, in the other being rendered by -uc- with anticipation of the labial feature, while the velar remains a velar. Then we can understand au'ch/n - Lesb. amfhn from *ank"e#n;

We can now use this insight in explaining forms. Thus, da/fnh - daucn(a)- can now be explained as

giving au'chn (with loss of the nasal, a development known from Armenian). Perhaps also possible is * ak"e#n > au'ch/n with prenasalized * ank"e#n > amfhn. Such interpretations may in individual cases be wrong, but that is no reason not to try it. On the other hand strange variations become underst andable if we start from a limited set of assumptions.

the latter form is directly understandable (with f from the labio-velar); the first went through * an"ke#n (or *a"nke#n)

we know that * ly gave ll in Greek. Thus in Achilleus. [Chantraine's treament of this name is charac teristically averting; and in the Addenda (CEG 4) this is not corrected.] Note also that Achileus with one l occurs only in Homer. This may show that earlier the variation was greater; as is only to be expected. In a similar fashion any may have given ain, and ary air, and also eir (with colouring of the vowel; see on the suffix). This is confirmed by the m are much less frequent or rare. The idea is nicely confirmed if Hitt. Appaliunas is the forebear of Apollo (Beekes, fact that double l is very frequent (Fur. 387: "Dutzende von apellativische Beispiele"), whereas double r, n and

The palatalized phonemes may explain other developments. Thus I wonder whether ll might continue ly;

JANER 2, 2003). Also we have -air-, -ain- but no *-ail-. Thus asy may have given ais, cf. ka/baisoj, which has a

v.l. ka/basoj. In rendering such a foreign word, the palatalizati on may one time have been represented, another time have been neglected. This is a phenomenon we often find - it was the main cause of the variations in Pre-Greek forms. The interpretation is further confirmed by the parallel development of labialized cons onants. Thus I suppose

kalar"-op-. Another form which shows the remarkable interchange a/au is a'rasca/dej - au'rosca/j. Here one might assume ar"ask-at-. (Note that the labial element would at the same time explain the o as a variant of a.) I think this gives the solution for the hopeless etymon alox, aulax, wlax, eu'la/k-a . I assume al"-ak-. It gives aulakthrough anticipation, alok- through colouring; thus the first two forms, which are best attested, are directly clear. Further au/eu/w interchange frequently. Also o'lok- is unproblematic. [Homeric acc. wlk-a is hopeless; it is the only form which has no vowel between l and k, and therefore may be due to some accident of the tradition; does it stand for * o'lak'(a), *w'lak'(a)? Or is it w'lk- < *wlak- with syncope?]

that ar" resulted in aur (s. the suff.). In this way, we may understand kalau~roy (beside kolo/r-ob-on) as

I do not know whether in suffixes of the structure VC a diphthong is allowed. Cf. the forms in -ai#oj. Otherwise one might think of -ay"-, or even -awy-, but such sounds are rather rare in the languages of the world. Cf. Lak-ed-ai!m-wn, if it should be thus analysed (for -ed- cf. Te/nedoj, Le/bedoj). An instance of -ai- due to a palatal consonant may be e'xai/fnhj - e'xapi/nhj - afnw (a brilliant combination by Furne !e 158 etc.) which may contain -apy- (the palatalization was ignored in the last form). (Comparable to the development in the second word is knw/y - kinw/peton, from kyn-? with u < i before ou in kunou~pej?) There is evidence for an affricate, somewhat like [ts]; I shall note it with /c/. See on the variations B 5.5 (esp. on the forms of Asklepios). We may suspect a series c, cy, c". There may have been a sound like the Hittite h$. It may have been present in a cluster h$s, which was represented by x. See B 5.5. But I do not see what it may have become in another position. The language probably had a y and a w. Initial ya- will often have lost its y-, but it may sometimes be represented by i'a-, iamboj, 'Ia/swn (?, long a) [but -uia may have been -uy-a (a Pre-Greek y may have had a different development from inherited y.) In the same way, disappeared in some cases, giving gai~a : ga~ (see on the suffix -ai-/-e(i)-). Iason ( Argoj) from i#waso-, with long i]. The ending

-aia < *-ay-a? with a variant -eia (note that there is no -oia). Cf. Phnelo/peia, Persefo/neia. Perhaps the y Initial *w may often have been lost ( anax). But wa- may also have been rendered by oa-, as in 'Oaxo/j, Iloj). We find ua-

Cret. #axo/j. Thus also 'Oi leu/j (which has been considered as identical with the root of

(which became automatically u`a-) in u`a/kinqoj, Cret. #a/kinqoj. [Furne!e 377 assumes a prothetic u- in the latter

word, which seems improbable to me.] Thus perhaps also ua/eloj. The diffrences will be due to the time when the loan was borrowed, and whether the Greek dialect stil had a u!; many possibilities arise in this way. - Still another treatment we see in the word for `truffle', for which we find ouiton, oidnon (-tn-), udnon (-tn-), or iton. I think that these are all renderings of * wit-. [Again Furne !e 184 assumes a prothetic vowel, #it- / o'#it-, which does not seem to be the right solution. He further assumes a variation * wit- / wut-, which seems also impobable to me, though

the variation i/u is attested.] Rather u- is a form of oi-, with the -o- changed under influence of the -i-; cf. Lejeune, the third century B.C.; variation oi/u is found more often in Pre-Greek words. The case nicely shows that the variations in Pre-Greek words are tentative renderings of the sounds of a foreign language, and therefore have to be taken seriously. bra/kalon: r`o/palon H. probably shows a development *wrak- > brak- [as Fur. 147 remarks on soro/a: paliou/rou eidoj H. (`sorb-apple') continues * sorw- (Lat. sorbus, Fr. `sorbier'; Fur. 230). kalau~roy "Die landla#ufige Etymologie <connecting> r`e/pw ... ist wohl ohne weiteres aufzugeben."]

Phone t. 174 with n. 2; note that Greek did not allow -ui- before consonants; of course oi became u in Boeotian in

with o`-, h`-, w`-). Several are doubtful; best is ai`masia/ (ai`moi/). One might conclude that the language had no h. This would agree with the fact that aspiration is not a distinctive feature in the stops. - The result is remarkable for hrwj, Ellhnej and Hfaistoj (but note that Myc. apaitio does not have ha-). Of course aspiration can have been deledw/nh. But Prof. Ruijgh points out to me that Mycenaean had place names ( haratua) and personal names (hakumijo) with initial h-; it occurs also in inlaut ( pihala, korihadana ); cf. further emaa2 (/Hermaha#s/ `Hermes') . The vowels. I think that this language had only three vowels, a, i, u. The Greek words have very often e and o, but this would not be surprising: the three vowels have a wide phonetic range, and the phoneme / a/, e.g., may often have sounded as [ e] or [o]. What makes me think so is in the first place the fact that the suffix -system has a, i, u, but not e, added secondarily in Greek. Cf. the variatio n in afqa / afqa and e'ledw/nh / e`ledw/nh which is a variant of

It seems as if there was no initial aspiration. Furne !e has a few words with a`-, e`- (one or two with i`-; none

o. E.g. we have ag, ig, ug; agg, igg, ugg; and aq, iq, uq; anq, inq, unq but no forms with eg(g), og(g) etc. (I saw
`Rh/skonqoj; and olonqoj, but as a variant of olunqoj, and mhlolo/nqh with a variant mhl(ol)a/nqh.) This cannot be due to chance. The idea is confirmed by the fact that variations a/e and a/o are (both) very frequent (while e/o hardly occurs). It agrees with the fact that the prothetic vowel is only a- (B 3.). Also the palatalized and labialized consonants will have influenced the vowels. - Furne!e (340) has a rule a > o before o, w, u (kalubo/j / kolubo/j); this can now be understood as the o-like realization of /a/ before high rounded vowels in the following syllable. Further I doubt that there was phonemic length. Greek quite often has only a form with a long vowel, but this can be due to the fact that in Greek length is distinctive: a vowel must be lo ng or short, interchange is (normally)

not tolerated. Vacillation is found, as in qrina/kh - qri~nax (see B 6.2). Cf. obrimoj - bri mo/j, bri!mh (though this `chaff'; it is therefore probable that they contain the same suffix -ur-; but in the first word the u is short, in the second long. This idea is confirmed by the fact that it explains the foi~nix-rule of Greek accentuation. T $% The nominatives in -ix, -ux are properispomenon (if possible) though the suffix has a long vow el elsewhere, e.g. foi/nikoj. If there was no distinctive length, the phonetic length was determined by other factors, e.g. short before two consonants, long(er) before a single consonant (as we find also in the Germanic substratum words). It agrees with the fact that the suffixes -ik, -uk- were of Pre-Greek origin. (We don't have it with -a x, perhaps because i and u are shorter than a; also we find hesitation between long and short vowel in Pre -Greek words more often with i and could be explained differently). A quite different argument is the following. acuron and pi/turon both mean

u than with other vowels; see 6.2.)

the diphthong was contracted early, whereby the u/u! disappeared. For -euj (from *-huj) one would also posit * au. That /a/ was realized / heard as [e, e ] is no problem. [The length may come from prevocalic * au. The e-colour may partly come from a preceding palatalized consonant (' Acilleu/j, 'Odusseu/j?)]. (Are we allowed to compare a > Ion.-Att. h?) For the feminine in -Myc. -eja I agree with Ruijgh, Etudes $ 212, that it is an independent Pre-Greek suffix, cf. Kuqe/reia, Phnelo/peia. Another point is that suffixes often have a long vowel ( -in-, -is-, -a s(s)-): they are often found in the pre-final syllable) and standing before one conson ant; were they stressed? I counted long a's in Furme!e's index, which gave only 13 instances; anyhow whis asks for an explanation. Note that an h often represents a (ga qulli/j / ghq-), and as our knowledge of the relevant dialects is rather limited, we often simply do not know whether h is an old a or e. If we did not have si/da roj we would not know that it is an old a. Few people know that Lh~mnoj represents La~mnoj. Note Krh/th. Note the suffixes -hl-, -ht-. Well known words with h = e : sph/laion Diphthongs. I assume two diphthongs, ai and au. If there were no e and o, other diphthongs are not to be expected. The diphthong eu is rare (Fur. 353 A5; I found some 12 instances in all of his material); it interchanges with au. Furne!e (339 A2) calls ei "(in mehreren Fa #llen) nur eine Nebenform von ai". oi is also rather rare; we find ou perhaps more often, but interchanging with other vowels. See on the variants of the vowels (B 6.1). Accentuation. I noted hesitation in: abramij / -mi/j, ai'gw/lioj / -io/j, acuroj / -o/j, acwr / a'cw/r, ko/rudoj / korudo/j, koru/daloj / korudallo/j, me/dimnoj / medimno/j, si/kuoj / sikuo/j, uriscoj / u`risso/j. Note also near-identical forms like lukayo/j / lu/koyoj. This may not mean that the language had n o clear stress; it may just be that the Greeks who took a word over were not always sure about the stress. But the phenomenon may be important heuristically: inherited words only seldom show such variation. A difficulty may seem the names in -w(j); they may continue * au (cf. Tlw/j, Hitt. talawa-) I suggest that

2. Characteristic sound(group)s In Pre-Greek words we find some sounds or clusters that are rare in PIE words. The words may be checked in Furne!e. (In brackets I give variants.) 1. au

Of course, au occurs in PIE words, but only when it derives from * h2eu (mostly in initial position) or eh2u (which is rare). Exx. blau~dej, brau/kaj, grau/kalaj, ka/naustron, kasau/ra, trauxa/na; Labraundo/j. 2. b As is well known, * b was rare in PIE. In Pre-Gr. words it seems to occur relatively more often. Exx. ablaroi, ablaj, a'bur/bhloj, arbaloj, a'rbo/n, a'rbu/lh, a'ta/rbaktoj, ba/rbiloj, busso/j, qo/ruboj, ki/baloj, It is often found word initially. - We know that b may also go back to a Pre-Greek `labio-velar' (labialized velar): basileu/j < Myc. qasireu.

3. bd The cluster is possible in PIE words, but it is rare; see on b. Exx. abdellon, abdhra, abdhj, ibdhj, au'to-ka/bdaloj, ki/bdhloj, kuba/bda, si/bda, ti/bdainon; Kombdilipia. [Note that mo/lubdoj rather continues -wd-; Beekes, MSS 59, 1999, 7-14 .]

4. gd

Cf. Fur. 318 n. 5. There is nothing against PIE gd, but it is infrequent. Of course, the group is reminiscent of bd.

Exx. agduj, a'mugda/lh, gdoupe/w (cf. ktupe/w), igdh, kri/gdanon, lu/gdh, ma/gdwloj. 5. gn Exx. i'gnu/j (i'knu/j). - On cn, fn see on the suffixes. 6. dn The sequence is rare in IE words, where it can only arise at a morpheme boundary. Exx. a'kidno/j, a'lapadno/j, a'ra/cidna, lepadno/j (la-), si/pudnoj. 'Aria/dnh. 7. kt The group is regular in PIE, but in Pre -Gr. it is found with variants; see B5.5. Exx. a'bi/ukton, ba/ktai, di/ktu. 8. kc The group can hardly be of IE origin, but it is not frequent. I noted: ba/kcar, la/kca, sa/kcar, sukca/j; Ba/kcoj, Bri/akcoj, Bu/kcij. -kc is the geminate of c. Cf. on pf, tq. 9. mn The group is quite possible in PIE words, but it is frequent in Pre -Greek. Exx. a'mfi-ke/lemnon, #edimnoj, iamnoj, basumnia/thj, krhmno/j (Beekes), krimno/j, la/mna, lwrumno/n, me/rimna, prhmna/j, r`o/damnoj, si/gumnon, si/drimnon; 'Atu/mnioj, Kartemni/dej. 10. ou The diphthong is of course perfectly IE (but not in suffixes), but it is found several times in Pre -Greek. Exx. sendou/kh, ski/ouroj, strouqo/j, ta/gcouroj, tou~foj, fa/ndouroj, fasi/ouloj (o/w), fou~skoj, clou/nhj. 11. pf The group can hardly be of PIE origin, but it is rare in Pre -Greek words too. Like kc it is the geminate of f. Exx. a'rciza/pfhj (?); Sapfw/ (Yapfw/). 12. rd On a morpheme boundary the group is possible in PIE. Exx. a'ge/rda, kapard-eu~sai, kardama/lh. 13. rkn A rare group. Perhaps there is no reason to speak of a group.

Exx. abarkna, berkni/j. 14. rn (rd, nd) Exx. ki/sirnij (-nd-), a'ce/rda (-na), skape/rda. See on the suffixes. 15. An s occurs both word-initially and between vowels, where is has disappeared in inherited words. (Of course, in a few cases Greek got a new s.) Init.: sa/rutta, sagu/rion, sa/naptin, sa/ndalon, sa/ndux, sarrufqei~n, sekou/a, se/linon, sibu/nh, si/gumnon. Intervoc.: a'gasulli/j, agcousa (egc-), aiqous(s)a, ai`masia/, aisakoj, ai'sa/lwn, ai'simna/w (-umn-), a'i/sonej, ai'suhth/r, aisuloj, ai'su/f(i)oj, aleison, dro/soj, ka/dousa, kasalba/j. After resonant: alsoj, ba/lsamon, gelso/n, ge/nsimoj, ma/rsi/uppoj. 16. sb Exx. asboloj, qi/sbh, Asbetoj. -sb- may continue Pre-Gr. -sg"-: Myc. tiqajo may be /thisg"aios/ Qisbai~oj. 17. sg The group is hardly known from inherited words ( sbe/nnumi is problematic).

Again this group is hardly known from IE words. It may sometimes continue -cg-, as in a'musge/la , 'Asgela/taj (see 5.5). Exx. a'lisge/w, u`sgi/nh, fa/sganon, a'musge/la , a'sga/ndhj, pisgi/j; 'Asgela/taj. 18. sk, st These groups are well known from IE, but mostly in word initial po sition. See on B 5.5. Exx. be/skeroi, bu/stax, ku/steroi, lastagei~. 19. stl Though the cluster contains nothing that could not be IE, it occurs more often in substr. words (?). Exx. astliggej, stleggi/j. 20. tq The group can hardly be of PIE origin. I n Pre-Greek, it is a variant of tt and ss (see 5.5). It is sometimes clearly the geminate of q: 'Atqi/j : 'Aqh/nh. Exx. i'tqe/la, ko/tquboj. Petqaloi/. 21. fq The cluster is possible in inherited words. Exx. na/skafqon. 22. cm, cn Rather rare in IE; Fur. 110 assumes that the nasal caused the aspiration. Exx. daucmo/j, daucna-, saucmo/n.

23. w The w is of course perfectly IE, but occurrence of *o was very restricted. Otherwise, it must represent vowel + laryngeal ( eh3, oH) which are only possible in certain circumstances. Exx. a'makrw~tij, amwton, anqrwpoj, a'nwni/j, a'pofw/lioj, a'rrwde/w, a'skalw/paj, a'skamwni/a, a'skw/lia, #askw/ndaj, a'smwlei~n, ballwth/, kasswri/j, kinw/peton, laimw/rh, metw/pion, lwrumno/n, us(s)wpoj, fa/gwroj, fasi/wloj 24. geminates See also B 5.8 on single / gemin. Indo-European had no geminates. Of course, there arose geminates in Greek, but they are not very frequent. I doubt whether Pre-Greek had geminates, but there occur several in Pre -Gr. words. (I compare e.g. Brixhe, Dialecte grec de Pamphylie, 1976, 95, that there were no geminates in this language.) As Pre-Greek had palatalized phonemes, I wonder whether l' was (often) represented by ll in Greek. And then perhaps also n' as nn, r' as rr. For ss and tt see B 5.5. Unclear dd, kk, pp, and mm (palatal m' is a rare sound). I note some exx. ka/ttoj, ma/ttaboj, me/ttej, mi/ttoj; Pittako/j. We must also recall the instances of kc, pf, tq (above). Liquids ll: a'lla/bhj, a'llopi/hj, a'mi/llakan, ballwth/, ba/tella, bde/lla, bi/llin, pa/tella, - mm: klemmu/j, nn: a'ga/nna, ble/nnoj, gi/nnoj, laca/nna; Di/ktunna, - rr: a'rra/baka, bi/rrh, bi/rrox, karro/n Sibilants ss: aiqous(s)a, afrissa, gi/ssa. 3. Prothetic vowel The definition is `initial vowel that is present or absent in (nearly) identical forms'; for we cannot say whether the vowel disappeared or was added under certain circumstances; still another possibility is that it represents a kind of laryngeal sound, that was sometimes heard as a vowel and and sometimes not. The vowel is in most cases an a'-. The numbers by Fur. 368ff. are as follows: a + 90, o 10, e 5, i 3, u 0, h 6, ai 2. (For u I have excluded u`a/kinqoj and usswpoj, for ai I have excluded ailinoj, which is a cry.) Note that - in general - a varies with o, e, ai. Indeed we have cases where proth. o varies with a; also for e (ei'kl- : ai'kl-, e`yi/a : a'yi/a). The h interchanges three times with a. ai interchanges once with a, once with o. Though not all other cases can be removed, it is clear that the phenomenon regards, originally, a only - which agrees with my assumption that the language had no e and o. Exx. a'gasulli/j : ghqulli/j, a'kiri/j : ki/rrij, a'kornoi/ : ko/rnoy, a'cradamu/la : cramadoi~lai, a'nari/thj : nhri/thj, a'ska/laboj : (s)kalabw/thj, a'cu/nwy : ku/nwy. 4. s mobile A large number of words shows an initial s- before consonant, which is absent in nearly identical forms. It occurs before stop or m (so not before r, l, n); the stop is mostly voiceless, sometimes aspirated. Fur. 390f. Exx. ge/lenoj / sce/linoj, (s)kida/fh; ki/keroj : ski/gkoj; (s)kordu/lh (u ); ba/taloj : sp-; pe/leqoj : sp-; fatta/ghj : sp-; qrigko\j (trigco/j) : stri/gco/j; topei~on : stuppei~on; (s)mh/rinqoj; (s)mu/raina. A prothetic vowel may appear before an s- mobile (Fur. 390 n. 8): a'ska/laboj : skalabw/thj : kalab/a/j; a'sfa/ragoj : sfa/ragoj : fa/ragx; a'ska/lafoj : ka/lafoj. (Uncertain is: skuri/ttw : a'gkuri/ttei.)

Stops dd: addai, addix, - kk: a'kkaloj, be/lekkoj, la/kkoj(?) - pp: agrippoj, lou/ppij - tt: bi/ttakoj, la/tta,

5. Variations: consonants 5.1. Voiceless/voiced/aspirated stop This is the most frequent phenomenon. Furne !e devotes a hundred pages to it (101-200). Examples: bu/stak- / bu/stag, ge/rsumon / ke/rsimon, ktu/poj / gdoupe/w, a'ca/rnaj / a'ka/rnax, isklai / i'scala/w, ki/clh / ki/gkloj, kre/muj / cre/muj, ble/n(n)a / ple/nna, broko/j / pro/kon, kalau~rop- / kollo/robon, kupa/thj / kuba/lhj, agrippoj / agrifoj, ku/proj / ku/feron, ompnh / o'mfu/nein, baska~j / faska/j, kinabeu/esqai / kinafeu/ein, skolu/bra / skolu/fra, a'tra/faxuj / a'dra/faxuj, kaia/taj / kaia/daj, konto/h / kondo/j, bru/qakej / brutiggoi/, qro/na / tro/na, qrugona/w / trugona/w, ge/lgiq- / ge/lgid-, ko/ruqoj / ko/rudoj, spu/raqoj / sfura/d-. 5.2. Prenasalization Before a stop a nasal may appear. This `prenasalization' is one of the most typical phenomena of Pre-Greek, and it is found very frequently. Exx. Fur. 267-291. a'cu/nwy / a'gcu/nwy, gi/glumoj / gi/gglumoj, gou~roj / aggouroj, ka/cruj / ka/gcruj, ku/cramoj (i/) / kigkra/maj, usplhx / usplhgx, fa/ragx -ggoj / (a')sfa/ragoj, kolu/bdaina / kolu/mbaina, korufh/ / ko/rumboj, sala/mbh / sala/bh, tra/fhx / tra/mpij, a'tra/faxuj / a'ndra/faxuj, ki/dafoj / kinda/fh. Often the nasal is followed by the voiced variant of the stop (Fur. 271 n. 16), but an aspirate is also often found (cf. -inqoj). [Prof. Melchert notes that in western Anatolian voiceless stop became voic ed after nasal, which could be due to substratum influence.] This was the reason that a sound law aspirate > voiced after nasal was assumed for (inherited words in) Greek (Schwyzer 333), which is incorrect. A prothetic vowel (above B 3.) and reduplicatio n (section C 1.) create the possibility for prenasaliazation; examples are given at the places indicated. 5.3. Nasalization Nasalization is called the process whereby a stop is replaced by the nasal of its series, i.e. a dental by n, a labial by m. In the case of the velars, there is the problem that Greek had no sign for a velar nasal; the Greek spelling in this case could not distinguish between nasalization and prenasalization. The case of the labials ( p/m etc.) is discussed in the next section. The phenomenon is less frequent than prenasalization. Examples for the dentals are (Kuiper 216; Furne !e does not discuss the phenomenon): ki/dafoj (kinda/fh) / ki/nadoj, kidafeu/ein / kinafeu/ein, kla/doj / klw/n (klw~nax), ka/qoj / kani/aj, flhdw~nta / flh/nafoj. 5.4. Labials, m and u There are three interchanges: labial stop / m, labial stop / # and m/#. Labial stop / m. The variation is a case of nasalization discussed above (5.3). Fur. 203 -227. The stop is mostly b. Exx. a'rbu/lh / armula (n. pl.), ba/rbitoj / ba/rmitoj, burro/j / mu/rsoj, kubernh/thj / kumernh/thj, ku/mindij / ku/bindij, luka/baj / luka/maj, ma/skh / ba/ska, mu/stax / bu/stax, sko/lumoj / sko/luboj, sti~mij / sti~bij, fa/rmakon / fo/rbanta; a'ph/nh / a'mana/n (acc. sg.), spa/roj / smari/j, sfa/ragoj / sma/ragoj. Labial stop / #. Fur. 228-242. Exx. te/qhpa, qa/poj / qau~ma, ko/baloj / kaualo/j, kassaba/j / kasau/ra, kra/mboj / krau~roj, le/paj / la~aj(?), si/bda / si/dh. The variation m/#. Fur. 242-247. A difficulty here is, of course, that Greek mostly did not preserve a #, so

that we often just find zero and the # must be reconstructed, which gives uncertainties. - In this case we must perhaps reckon with a development u! > b (as in mo/lubdoj, Myc. moriwodo ). Exx. basumni-a/thj / basun-i/aj, kri/mnon / kri/non, me/dimnoj / #edimnoj, si/gumnoj / si/gu noj (si/gunnoj). The evidence comprises 8 or 9 words in -mnoj. It is found six times word initially: mh~lon / hlon, monquleu/w / o'nquleu/w; note me/roy / a'e/roy (eiroy), where the latter forms could continue * a'-#eroy, *e'-#eroy. Note further ku/amoj / ku/mhca < *ku#-am-, *kum-hk-. A few other instances are less clear. 5.5 Stops varying with s(s) or stop + s, t This kind of variation is quite complicated. I distinguished no less than 11 (or 16) different types. They may be presented as follows (C = consonant): a. labials b. velars 1. C/Ct 2. C/Cs 3. C/sC 4. Ct/Cs 5. Ct/sC 6. Cs/sC (y/sp) 7. Cs/ss 8. sC/ss 8c. C/ss dentals 9. t/ss 10. t/st 11. ss/st t/ss t/st ss/st p/pt p/y (p/sp) pt/y k/sk kt/x kt/sk (x/sk) x/ss sk/ss k/ss k/kt

Interpretation. We are not concerned with stops simply extended with an s or t, as this would (mostly) not have led to variants; also extension with an s or a t, in further identical forms, would be linguistically quite strange. Also the fact that mostly exactly an s or a t is involved is a fact that must be explained. The most complicated instance is 5., showing kt/sk. As one might expect, the most complicated phenomenon gives the most information and can be best solved. One expects in this case a cluster with k, i.e. a consonant before or after the k; one of the two resulting clusters will then have had metathesis. As Greek had metathesis tk > kt (and no metathesis of sk or x), we may assume that this worked here too. Thus we reconstruct probably unknown to Greek, which resulted either in s or t. This is of course a dental affricate, i.e. a sound like [ ts]. For convenience I shall note it as c (without a diacritic). May be it was a c%, but as long as we have not shown that there was a distinction between c and c%, we can disregard this. E.g. a'musge/la / a'mugda/lh will have been *amucgala, represented in the first instance as * amusgala or *amudgala, the latter giving * amugdala. Thus Asklepios was called 'A(i)sklapio/j or 'A(i)glapio/j. I think that he name was * Acklap-, giving * A(i)sklap- or *A(i)dglap-; in the latter form metathesis did not operate because * Agdlap- was not tolerated in Greek. Of course, often we will find only one variant. The strange feature or phoneme may also be dismissed altogether, as in dikei~n beside di/skoj and di/ktuon. Now one might suppose that all variants in this group are due to an affricate, but this seems less probable to for an earlier stage an interchange sk/tk. And this interchange can be easily explained by assuming a consonant,

me, as consonant clusters are rather rare, and notably as there are no suffixes beginning with a consonant (except n, r etc.). Of course, we may not be able to identify in each case what happened exactly. I will shortly review the 11 (16) types (I call the labials 1a etc., the velars 1b etc.). 1a. pt may represent a single phoneme py, as we saw in B 1. Exx. (Fur. 315ff): gnup- / gnupt- (gnupet-), kolu/mbaina / kolu/bdaina, ki/baloj / ki/bdhj, lu/ph / lu/pta; without variants note kro/ssofqon, sarrufqei~n. Note that mo/luboj / mo/lubdoj continues *moliwdos. 1b. kt is most probably explained like 5b, discussed above (so 1b is a part of 5b). Exx. (Fur. 319ff): arakij / a'ra/kthn, moge/w / mocqe/w, peleka/n / spe/lektoj, a'kaki/a / ka/ktoj. 2a. y may result from * pc. (It is remarkable that there is no 2b. k/x, as x is unproblematic in Greek.)

Exx. (Fur. 324ff): bi/ttakoj / yi/ttakoj (si/ttakoj), de/fw / de/yw(?), kanbi/on / kanyh/j, pw~roj / ywri/thj. 3a, b. ( p/sp, k/sk) Both can come from * cp, ck. Exx. (Fur. 292 n. 2): qi/sbh / qi/bij. (Fur. 295ff.): be/koj / be/skeroi, icla / isklai, ma/kella / ma/skh (ba/skh), mu~koj / mu/skoj, fa/kelon / fa/skwloj. 4a, b. pt/y and kt/x can continue * pc, kc and belong together with 2a, b.

Exx. (Fur. 263 A3): difqe/ra / diya/ra, (Fur. 318, 324): calubdiko/j / Ca/luyoj. (Fur. 263 A3): mo/rocqoj / mo/roxoj. 5b was discussed above. Note that it is no surprise that there is no * pt (5a.), as there is no regular metathesis *tp > pt in Greek; from * tp the t may have been simply lost. [However, a metathesis tp > pt may have operated in pto/lij.] Exx. (Fur. 301 A2): a'musge/la / a'mugda/lh, (Fur. 279, 319): di/skoj / di/ktu(on). 6a, b. (y/sp, x/sk). Furne!e (393) simply considered the interchange as due to metathesis, which is of course possible. * sp, *sk can represent *cp, ck. Exx. (Fur. 393): a'spi/nqion / a'yi/nqion, o'sfu~j / yu/ai. Exx. (Fur. 393): isci/on / i'xu~j, fou~skoj / foxo/j. *c became ss. 7b. x/ss. If x represents *kc, the k may have disappeared in other cases (which did not give x) after which

Exx. (Fur. 130 n. 59): krixo/j / krisso/j, (Fur. 317): si/bda / xi/mba; (Fur. 286 n. 72): trau/xana, trw/xanon / trau/sanon, i'xa/lh / i'sa/lh (i'sse/la, i'tqe/la). Ou'li/xhj / 'Odusseu/j. Exx. (Fur. 300): uriscoj / u`ri/ssoj. 8b. sk/ss can be explained parallel to 7b: * ck > sk, or with loss of the k, *c > ss.

I added 8c. for da/lagcan = qa/lassa. We have dalac- (with prenasal.) / qalass-, and I suppose that it had a * ky (this is a case where a cluster seems improbable). 9. t/ss. This is the well-known element that gave ss/tt. Furne!e does not dicuss it under that heading, because his system is to dicuss one phoneme (`letter') and its variants; thus he discusses sk/kt under k/kt. Also the situation is different as here we have a distribution among the Greek dialects. This is the only time, I think, where we can attribute the different rendering of these loan words to Greek dialects. But the fact that a foreign element was rendered in different ways is the same as with all other phenomena we discussed. F urne!e has the heading (253) t, d, q / s(s), z; I think it can better be stated as t (d, q), tt (tq) / s (z), ss, i.e. t with its usual variants d, q; or the geminated tt (with its expected variant tq, which is the Greek form of geminated qq) interchanging with s or ss; if the z was [sd] it does not fit in well. As to its interpretation, it could represent single * c, which was rendered tt or ss, or single s, t ( the variant z would then fit in, but one would also expect a variant st). Here again, however, it is

difficult to decide between * c or *ky or *ty. Exx. (Fur. 253ff): kitto/j / kisso/j, kro/tion / kro/ssofqon, mu/rti/nh / mursi/nh, teu~tlon / seu~tlon, ti/lfh / si/lfh, ga/doj / ga/zaj, a'smwlei~n / a'dmwlh\. 10. t/st may be from * ct giving st, or with loss of the t, *c > ss.

Exx. (Fur. 301ff): ballwth\ / ballau/stion [does this point to a c"?], ma/trulloj / ma/strulloj, mu/tij / mu/stax, pati/lh / pasti/lh, 11. ss(tt)/st can be *ct > st, or with loss of the t, *c > ss.

Exx. (Fur. 304 A1, 329): yitta/kion / pista/kion. As we saw it is very difficult in each case to indicate exactly what happened; on the other hand it is clear that with few assumptions probably all variations can be understood. 5.6 Velar/labial/dental: labio-velars There is limited evidence for variation velar and labial, velar and dental, and between labial and dental; and between all three classes (Fur. 388ff.). We find: k/p,b g/b c/f k/t, d g/d p/t b/d f/q g/b/d c/f/q

It is remarkable that mostly the variants agree in voice/aspiration (this shows that there is a large degree of regularity in these variations). Exx. k/p: kla/nion / plani/j; g/b: bra/kalon / r`o/palon, gle/pw / ble/pw; ca/lij / falikro/n. k/t: a'ska/ndhj / a'sta/ndhj; g/d: ga/latmon / a'dalto/mon. p/t: bapai/nei / batai/nei; b/d: sa/mbalon / sa/ndalon; f/q: gnufai/ / gnu/qoj. g/b/d: ge/fura / be/fura / de/fura; It is tempting to assume labio-velars in these cases, but some cases may have a different origin (bra/kalon / r`o/palon could be just dissimilation in the first word). On the existence of labio -velars see above on the phonemic system. 5.7 Dentals / liquids There are some instances of variation between dentals (including n) and liquids ( l, r), and between liquids. These variations are incidental. We find: 1. d, t, q, n/l; 2. d, n/r and also 3. l/r 1. Exx. (Fur. 387f.). d/l: ablaroj / bdaroi/ (330 n. 27); da/fnh / la/fnh; 'Odusseu/j / 'Olusseu/j. Cf. Myc. gen.

dapu2ritojo /daburinthojo / - labu/rinqoj; kala/minqa / Myc. kadamita. [d/l and the fact that Mycenaean has signs for la, le, li etc., which Lejeune explained by assuming a specific, unusual sound d, might point to a dental fricative, &.]
q/l: qa/pta / la/tta; n/l: ni/tron / li/tron. 2. d/r: si/bda / xi/mbrai. n/r: blh~cnon / blh~cron. 3. l/r: a'zhri/j / a'zhli\j, kri/banoj / kli/banoj, krw~max / klw~max. 5.8 Simple / geminate

Beside a few isolated cases we find this interchange in n/nn but notably in l/ll. On t/tt and s/ss see above 5.5.

Fur. 386f. Exx. n/nn: anhqon (t) / annhqon (t), thmeni/j / th/benna. Note the suffix -unn-. Note gei~s(s)on, sa/ri sa / sa/rissa. Note the case of (' Aqh\nh) - 'Atqi/j - 'Attiko/j. 5.9 s-/ zero l/ll: bal(l)h/n, quli/j / qulli/j, spe/leqoj / spe!llhxi (dat. pl.), make/lh / ma/kella (so this will be -alya-).

We discussed s / zero before consonant, under ` s mobile', above section B 4. An -s- from Pre-Greek is normally maintained. The only instance where it may have disappeared I know of is (Fur. 241) su/ricoj, su/rissoj / u`rico/j (-i/skoj, -i/scoj, -i/ssoj). Perhaps `Ella/j beside Selloi/ belongs here. Another instance seems apion, which is cognate with Lat. pirum which points to -pis-. [Prof. Ruijgh points to Li/gu-eh < *Ligus-ej, cf. Ligustiko/j.] 5.10 K-, T- / zero There are instances where a velar or a dental may be absent in initial position (Fur. 391, and 131 n. 59). Dentals may also be absent in inlaut. Exx. k / zero: ka/ndaroj / anqrax, kalinde/omai / a'li/ndw, ko/gcnai / ogcnai, kanqh/lion / a'nqh/lion. g / zero: gi/nnoj / i'nno/j 'but this form may be a late notation/development). As an explanation one could think here of a uvular, q. t / zero: ta/gcouroj / agcouroj, th/ganon / hganon, ti/fuon / ifuon (with i ); d / zero: deledw/nh / e`ledw/nh (e'-). Loss of the dental in inlaut: ne/twpon / ni/wpon, i'qouli/j / iouli/j, a'si/daroj / a'si/aroj. (This is reminiscent of Dutch neder > neer etc, which was a normal sound law.) 5.11 n-, l- / zero n- and l- can also be absent (Fur. 391f): na/fqa / afqa (a-). laiyhro/j / ai'yhro/j, lamph/nh / a'ph/nh, latmenei/a / a'tmh/n. Of a different type is gna/mptw / ka/mptw Perhaps it concerns palatal ny, ly, which are pronounced very `light'. 5.12. Metathesis, aspiration shift There are instances of metathesis. It mostly concerns r, sometimes l.The consonant changes to the other side of the vowel or the consonant: kirso/j / krisso/j, krixo/j, te/rminqoj / tre/miqoj. Cf. Termi/lai / Tremi/lai. arpix / a'pri/x, ke/dropa / ke/rdopa, na/qrax / na/rqhx. It is mostly unknown what was the original configuration. (In a case like erbwj / eurwj, where b may stand for - or continue - #, I would think that the # was anticipated. It may concern an original r".) The cases of sp/y and sk/x are discussed in 5.5. above. Aspiration shift is sometimes found: qrigko/j / trigco/j, a'qrage/nh / a'ndra/cnh. In the case of fa/tnh / pa/qnh the metathesis seems to have occurred late in the history of Greek (Beekes in Bammeberger -Venneman, Languages in Prehist. Europe ).

5.13 Secondary phonetic developments We may assume secondary phonetic developments, either in Greek or perhaps already in the original language. One might consider: bd > br: bde/llion, bre/llion (Fur. 308) bd > bl: bdaroi/, ablaroi. For this case cf. 5.7b d/l. gd- > d-: gdou~poj, dou~poj dn > gn: dno/foj, gno/foj km- > m-: kme/leqron / me/laqron, y- > sp-? yendul-? / spondu/lh. See 5. above. y- > s-: ye/faj / sei~fa, yi/ttakoj / si/ttakoj; cf. Yapfw!, Sapfw/ 5.14 Other variations. There are a few instances with - isolated - puzzling variations. I mention just one, the word for `night', where we have ye/faj, kne/faj, dno/foj, zo/foj. I think that in some of these cases the solution may be found in a cluster. E.g. Carian allows an initial cluster kbd-. Such clusters would have been simplified in Greek. (In IE we have the parallel of Lat. pecten, Gr. ktei/j, supposed to continue * pkt-.) If we assume in our example a cluster * kdn-, it may have been reduced to kn- or, with loss of the first consonant, to dn-. (The process is of course the same as the reduction gd- > d-, above 5.13.) Such variant simplifications are typical of loan words. In this way we could connect two of the words; but I see no way to connect the other two. 6. Variations: vowels 6.1 Single vowels (timbre) The vowels show many variations. I will discuss them in the following order: first a, then e and o; and within each of these groups first the short vowel, then the diphthongs, then the long vowel (and the long diphthong s, but these hardly occur). Note that a variation x/y is not repeated under y. 1. the vowel a. 1a. a/e has 80 occurrences in Furne !e's material (347). Exx. agcousa / egcousa, arusoj / erusoj, ga/linqoi / ge/linqoi, zakelti/j / zekelti/j, kai/ata / kaie/taj, ka/mpoj / ke/mpor, ka/cruj / ke/gcroj, sa/ndux / sendou/kh 1b. a/o. These vary also very often. Fur. 339 mentions that he found 80 instances. Exx. a'xouggi/a / o'xu/ggion, a'rrwde/w / o'rrwde/w, gra/bion / gobri/ai, h'pi/aloj / h'pi/oloj, ka/bax / ko/baktra, kalu/bh / ko/luboj, lukayo/j / lu/koyoj. 1c. a/ai Fur. 336ff. a'kraifnh/j / a'krapnh/j, a'su/fhloj / ai'su/fioj, la/gmata / lai/gmata. 1d. a/au. Fur. 302 n.37. kanau/stron / ka/nastron, mna/sion / mnau/sion; alox / aulax. 1e. a/w kla/doj / klw~nax. 1f. ai/ei. Fur. 352 A4; 339 A2. Exx. kairi/a / keiri/a, ku/pairoj / ku/peiroj, laiai/ / lei~ai. 1g. au/eu. Fur. 353 A5. laukani/h / leukani/h, pe/tauron / pe/teuron; aulax / eu'la/ka. 1h. au/w, o. Fur. 301 n. 32. kasau/ra(j) / kaswri/j, qau~ma / qw~ma, sau~sax / sw/sikej, bau/kalon / bw~koj, kalau~roy / kollw/robon / kollo/robon. 1i. a /ai Fur. 338. lh/qargoj / lai/qargoj, lhka/w / laika/zw, ph/ganon / fai/kanon. 1j. v/a : lvdoj (lV/dion) / la~doj (lh/dion)

2. the vowel e. 2a. e/a see under a. 2b. e/i. Fur. 355ff. bli/tux / ble/tuej, e'bi/skoj / i'bi/skoj, de/paj / Myc. dipa, entubon / intuboj, kello/n / ki/llix,

killi/baj / kelli/baj, ku/tesoj / ku/tisoj, le/sfoj / li/spoj (f). 2c. e/i/u. Fur. 354 n. 55. kecra/moj / ki/cramoj (kigkra/maj) / ku/cramoj. 2d. e/eu. Fur. 115. argetoj / arkeuqoj 2e. ei/ai see ai. 2f. ei/h. Fur. 339 A2. kei/qion (cei/tion) / kh/qion, ceiramo/j / chramo/j. 2g. eu/e see e/eu. 2h. eu/au see au. 2i. e/h. Fur. 358 n. 42. enustron / hnustron, me/rmeroj / me/rmhra, ya/kelon / ba/khlon, mh/dea / me/dea (me/zea); Phlago/nej / Pelago/nej. 2j. h/i. Fur. 171 n. 114. blh~ton / bli/ton, skh~noj / ski/nar, yhmu/qion / yimu/qion. 3. the vowel o. 3a. o/a see a. 3b. o/i. Fur. 191 n. 37. akonoj / akinoj, i'bri/kaloi / o'bri/kala, Onoglin / oniglin. 3c. o/u. Fur. 358ff. olonqoj / olunqoj, skolobre/w / skolubro/j, sku/th / -ko/tta, kudw/nion / kodw/nea, kurse/aj / korsi/j, pru/tanij / pro/tanij, topei~on / stuppei~on. 3d o/ou: Fur. 359 bro/koj / brou~koj, kolote/a / kolou/tea (lu, lw). 3e. o/w. Fur. 279 gnote/ra / gnwte/ra, kollw/robon / kollo/robon, fasi/wloj / fasi/oloj (-ouloj), wruggej / orux -goj; w'scoi/ / osch. 3f. oi/u. Fur. 127 cramadoi~lai / a'cradamu/la (a'kramu/la). 3g. oi/ou. Fur. 358 + n. 65. kolouti/a / koloiti/a (kolote/a), you/dion / yoi/qhj? 3h. ou/u Fur. 120 n. 29. ktu/poj / gdoupe/w, krou~nai / gru no/j. 3i. ou/w. Fur. 133 mwka/omai / moukh/zei, 148 lou/phj / lw/bhx, fasi/ouloj / fasi/wloj (-oloj). 3j. w/h. qrw~nax / a'nqrh/nh. 3k. w/u Fur. 302 n. 35 zw/gioj / zu/ggioj, usswpoj / i``ssu/poj, lwbeu/w / luba/zein. 3l. o/e. gorgura / gergura 4. i/u. There is some variation between i and u. I do not know how to interpret this. Exx. Fur. 364ff. ai'sumna/w / ai'simna/w, a'nqri/skoj / anqruskon, bi/dhn / budoi/, briko/j / bruko/j, zu/gastron / si/gistron, kinw/peton / kunou~pej, ku/besij / ki/bisij, ma/rsippoj / ma/ruppoj. [Prof. Melchert points out to me that an interchange i/u is sometimes found in Anatolian; as in kiklu/ipa- 'steel', i&qaru/ih$); see his Anat. Hist. Phon. 178.] 5. u/e. gurgaqo/j / gergaqoj The behaviour of the diphthongs may be summarized as follows: ai - ei au - eu, w further oi - u, ou ou - u, oi, w and (vice versa) ei - ai eu - au

All these variations are perfectly understandable in terms of adaptation from a three-vowel system. 6.2. Long/short I doubt whether Pre-Greek had a distinction of long and short vowels (see B 1). We do find h and w, but not often, and the latter has several variants. On the other hand, the variations w/o and h/e are not very frequent (in this case also the difference in timbre may have been important, depending on the Greek dialect). Variation between long en short i and u is frequent, especially in suffixes: gh/quon / ga/qia, ku/besij / ki/bhsij, qi~bij / qi/bij, kri/mnon / kri~mnon, qri~nax / qrina/kh; yhmu/qion / yimu !qion, spondu !lh / spondu/lion. Cf. ka !raboj / kara/mbioj (cf. khrafi/j), fena ki/zw / phnhki/zw `deceive'; Wghn(-)/ Wgen(-). gnote/ra / gnwte/ra. There is some evidence for short + CC : long + C: mu~koj / mu/skoj; La/risa / La/rissa; and see B 1 on -ix, -ux. 6.3. Single vowel / diphthong There are several instances where a diphthong varies with a single vowel. They can be found above (6.1). Most frequent is a/ai. We further find a/au, e/eu, and ou/u and oi/u. In two cases we find diphthong / long vowel: ai/a , ei/h. Examples were given above. 6.4. Rising diphthongs? Relatively frequent are sequences of a more closed followed by a more open vowel, sequences that are not found in IE. They would be rising diphthongs if they formed one syllable, but in fact we may have to do with two syllables: ea: sea gw/n (si-, su-) ia: batia/kh, qi/asoj, qri/amboj, si/alon, fia/lh, fiaro/j. Note sia gw/n (se-, su-). iu: i'ugh/ ua: bruali/zwn, gu/alon (gue-), ku/amoj, pu/aloj, pu/anon, suagri/j, ue: gue/lion (gu/alon), pu/eloj (pua-) Remarkable is also the sequende -wu-: wu: pw~u(g)x; mwu/j? 6.5. Secondary vowels (or elision) Sometimes words show a vowel which is absent in near identical forms. It mostly concerns vow els between a stop and a resonant. It is often not clear whether the vowel is secondary, or its absence. Fur. 378-385. Exx. bragcia / baragcia, ske/rboloj / ske/rafoj, knu/za / (s)ko/nuza, sko/rodon / sko/rdon, tonqoru/zw / tonqru/zw, 'Arepui/a / Arpuia, knw/y / kinw/peton / kunou~pej, Koru/bantej / Ku/rbantej.

C. MORPHOLOGY 1. Reduplication Some forms seem to have reduplication (often we cannot demonstrate that it is reduplication). Most frequent is partial reduplication, where only th e first consonant + a vowel is repeated. The vowel is mostly e or i. Exx. be/brax, be/(m)broj, ga/ggamon, gaggli/on, gaggrai/na, gi/garton, gi/gglumoj, ki/-kuboj, se/-suf-oj / Si/suf-oj (cf. a'-su/f-hloj); me-mai/kulon (mi-); neni/hloj; se/seli(j); si/sura (-urna); membra/j?; perhaps

ki/kumoj, ki/cramoj (ke-, ku-, kigk-); dendru/w. Cf. Ke/kroy; Pepa/rhqoj, Titarh/sioj; here also Le/-leg-ej? With prenasal. tenqrhd-w/n, tenqrh/n-h cf. a'nqrhdw/n, qrw~nax. Different vowels in: la/lamij (cf. lai~lay); ko-kru/j; also ghgh~lix?. Reduplication of a syllable in: mo/rmoroj (mormurai/a), marmarugh. More difficult are: ge/lgij : a'gli~j (< *ge-gl- : a'-gl-?), ke/rka : akrij (< *ke-kr- : a'-kr-?). Also Membli/aroj : Bli/aroj (cf. membra/j)?; Membli/j = Me/loj, also Mimalli/j. A completely different type perhaps in a'm-a/maxuj, cf. a'maxi/j; also a'mamiqa/dej? 2. Suffixes Introduction It seems to me that most suffixes have the same structure. They contain a consonant; i f this is a stop, it can be prenasalized, i.e. b - mb, q - nq, etc. The stop, of course, has its usual variants, b/p/f etc., though mostly one of these is dominant. Then the group is preceded by one of the three vowels of the language, i.e. a, i, u. In this way we find e.g. agg - igg - ugg, anq - inq - unq etc. A different structure have the suffixes with n (+ vowel) following a consonant: e.g. ku/dnoj, pisa/kna, mo/lucnon, fenakni/j, satarni/j. In this way the groups rn, dn, kn, mn will have arisen. With mn we find again the three vowels: -amn-, -imn-, -umn-, so this is almost certainly a/i/um-n. The well known groups -mn- and -rn- then can

be explained in this way as essential elements of this language. (See also on the suffix -rn-.) These groups are very important as they are found in Etruscan, which further shows little agreement with our language; mn is found as far as Cappadocian (Beekes BiOr 2002, 441f.). Did the groups ann, inn, unn also arise in this way? Also other sounds are found in this position: -r-, -d-, -g-, -l- (rarely): yudro/j, ku/riqra, panagri/j, fa/lakroj; skape/rda; la/qargoj; oniglin. (Note mo/lubdoj, which seems to continue moliwd-, Beekes). Probably the character of the consonant can be seen. Thus -ain- could render -any-, while aly seems to have

resulted in all (or -ell- with colouring). Thus eir could come from ary, air becoming eir. A nice confirmation could be aur, if this represents ar" (cf. au'rosca/dej beside a'rasca/dej, if this form had * ar"-). Cf. B 1.

Another type of suffix has s followed by a dental: ka/nasqon (-stron), laistro/n or another stop enqruskon, au'rosca/j, kannabi/ska; these forms may have been partly adapted to Greek suffixes ( -tron). See below on the suffix -st-. *-aut- from -at"? Cf. -aiu!- in elaion, where we may suspect ay" or awy (but it may be part of the root). See on B 1. A form like -eut- is deviating; we do not often find a diphthong before the consonant. Does it stand for

is perhaps be that length was not phonemic. In the case of u r one might think again of ur' > uir, though ry is a rare phoneme (like my). The Material The examples are mostly taken from Furne !e, to whom I refer for details. Words can also be checked in GED. In brackets variants are given. I added geographical names (GN) from Fick, Vorgr. Ortsnamen (+ more material, w. ref. to pages). 1. -ab- (Fur. 107) a'gra/kaboj, a'l(l)a/bhj, araboj, a'ska/laboj, a'tte//le/aboj, ka/kkaboj, me/s(s)abon, latrabo/j, ma/ttaboj. GN Katta/bioj (Rhodes 47), Ka/staboj (Caria) ka/nnaboj, ka/raboj, ko/llaboj,

Not seldom we find long and short vowel with a suffix (= consonant), e .g. iq - i q, uk - u k. The explanation

a`rpag-; cf. Cha. Form. 397ff. 3. -aggsfa/raggoj. 4. -ada'rasca/dej. 5. -aqa'spa/laqoj, gurgaqo/j. GN `Urna/qion (Epidauros) 6. -ai-/-e(i)- (See also 6b.) There are words in -aia/-e(i)a, like grume/a/grumei\a (or grumei~a)/grumai/a (note the hesitation in the accentuation). Frisk notes "die Bildung hat kein na #heres Gegenstu#ck." I suggest that the suffix was -ay-(a), which was pronounced [-a'y-a, -ey-a] (we saw that ei often varies with ai); the suffix was identified with Gr. ai or ei (before vowel), but the -y- could also be lost; in this way the three variant forms can be explained. Further we (kw/duia/kwdi/a are not clear to me; but cf. Ama/quia/Ama/qeia). have koloite/a/kol(o)ute/a; korcure/a (korko/drua H. is prob. an error); kw/deia/kw/dea [note the short a] I wonder whether this can give the solution for gh~ (<ga )/gai~a: from *gaya the i was retained or lost (as in

2. -ag-

-ea above), which gave ga#. (The i-less rendering, which gave -aa- > -a , being as old as the rendering gai~a; this means that the a became h Attic, not a as in the result of later contractions Another question that might be solved in this way is Athena's name Aqhnai/h. I always found this form, supposedly derived from the adjective, rather strange, as well as the coexistence with Aqh/nh (Homer uses both forms). The last form is often explained as a development from the form in -aia, according to an Ionoic-Attic rule ai > a ; however, this rule is unclear: "Aucune explication satisfaisante n'a e !te! donne! jusqqu'ici de ces faits." Lejeune, Phone t. 1972, 247. And it is impossible, as the rule, supposed to be Ionic-Attic, i.e. post-Mycenaean as Mycenaean already has At(h)a#na# (potnia) [the interpretation as gen. Atha#na#s is to be rejected]. Note that Ga is also already found in Mycenaean (in the Thebes tablets, Ma Ka). [I hesitate to add Ma~ - Mai~a, as one generally considers Ma~ as primary.] I further think that this * ay-a is the same suffix as -eia which makes feminine names, Ama/lqeia,

Phnelo/peia, Ifime/deia. Ruijgh also assumed that this suffix was Pre-Greek (Etudes $ 212 [Prof. Ruijgh wrote me however that he abandoned this view.]). ( Note that in Myc. Ipemedeja the -j- is preserved; cf. Ruijgh, El. Ach. 155 n.3.) - There are of course many place names in -eia: Kadmei/a, Kalau/reia, Keru/neia, Mi/deia, Skelerdei/a, Leba/deia etc. ai~oj (Chantr. Form. 91, type a'nagkai/h; [Chantr. speaks of "Le suffixe fe !m. -i a" but that has a short a]); cf. bruktai/a, dirkai/a, sibai/a, We also find -ei/a used in nouns: daurei/a, zalei/a, kouluba/teia, ... may be found in Wke/anoj < *-kay-an- (there was probably no /e/, and the hiatus is also remarkable; note the forms Nouns with -eo- are very rare; we find: gwleo/j, ei'leo/j, koleo/n, niku/leon, sufeo/j(?), fwleo/j. Further it Often the final was adapted to -ai/a (with long a) after the dominant type, derived from the adjective in -

Wghn, Wgen-). Beside -aia, eia we may expect -ai-oj/n; we find it e.g. in di/rkaion, sph/laion, yifai~on; grayai~oj, (*skarabaioj reconstr. by Fur. 169), ... (See also 6b.) 7. -ai(#)-o- (Fur. 233 n. 22, 255 n. 32. Partly from -ai#o-; it is often impossible to establish whether a form had a # or not.) See also 6. 'Acai#o/j. GN A'stupa/laia (58) 8. -aibGN Perraiboi/ (Thess.) 9. -aiqGN Su/maiqa (Thess.), Peraiqei~j (Arc. deme), Kelaiqei~j (Thess. deme), Kunaiqei~j (Arc. deme) 10. -ain- (Fur. 171 n. 117) akaina/on, boli/taina, ga/ggraina, kolu/bdaina (-umb-), koru/faina, mu/raina, smu/raina, tri/aina, fa/zaina. fa/llaina, fw/kaina. 11. -air-oku/pairoj (-eiron, -hrij, -eroj). 12. -ak- (Fur. 158 n. 64) a'burta/kh, aulax, batia/kh, e'riqa/kh, qrina/kh (qri~nax), kauna/kh, qu la/kh, pista/kh, fa/rmakon. GN Za/rax, -hx (Lac.) 13. -al(l)-o- (Fur. 254 n. 28) Exx. a'ru/balloj, ai'gi/qal(l)oj, bu/ssaloi, ko/kkaloj, koru/dal(l)oj (-o/j), pa/rdaloj. [Prof. Melchert suggests that these words may have been taken from IE Anat. languages, where -alla- is very productive. I think that it was PrGrAn., as it is frequent in Greek; cf. on tolu/ph in the introduction.] GN Kastali/a (Phoc. source), Fa/rsa loj, Stu/mfa loj (Arc.), 14. -amb- (Fur. 184) di qu/ramboj, qri/amboj, iamboj, kara/mbaj, sh/ramboj. 15. -amnsfe/ndamnoj. GN Se/damnoj (Crete), 15. -am-oartamoj. a'kulai~on, a'raio/j, bagai~oj, balaio/n, di/rkaion, elaion (Myc. era3/rawo), ma/taioj, messai~on, si/raion;

GN Ki/s(s)amoj (Kos), Pe/rgamon, Kw/gamoj (Lydia), Ku/amon (M Kydon.), `Udramoj (Kydon.) 16. -anGN 'Ia/rdanoj (R Crete, Elis), 'Apidano/j (R Thess.), 'Hridano/j (R), Ka/ntanoj (Crete), #Andanoj (18), Dra///kanon (Kos, Ikaria), 17. -a ngentianh/ 18. -anda'sga/ndhj. 19. -andrgelandro/n GN Th/landroj (51), Tu/mand(r)oj (Pamph.), Mui/and(r)oj (53), Fole/gandroj, 20. -anq/t- (Fur. 191 n. 35; 216 n. 71. -ant- unless otherwise stated) a'li/baj, a'ska/nqhj (ska/nqaj), killi/baj (but kelli/bat-), o'kri/baj, fa/lanq/toj, Pei/ranq/t-, Uanq/t-, Gi/gantej, Koru/bantej. GN Babra/ntion (Chios), Amuklant-, Lh/lanton (Eub.), ' Eru/manq/t- (Arc.) 21. -anntu/rannoj. 22. -axGN Kurta/rpaxon (Crete), Dana/rpaxon (Crete), 23. -ap-o- (Fur. 235 n. 31) arnapoj, gausapo/j, mo/napoj. GN Messap-ioj (Crete 24) 24. -ar (Fur. 134 n. 75: mostly neuters) Abantej, Me/lanq/t-,

iktar, ku/dar, ne/ktar, ski/nar, su~far; afar(?); adj. ma/kar; anim. oar, da/mar (gen. -rtoj; Myc. dama / duma).

25. -ar- (Fur. 257 n. 36) askaroj, bassa/ra, ba/ssaroj, ga/daroj, gi/gglaroj, ki/sqaroj, ku/ssaroj, ku/ttaroj, lesca/ra, fa/lara. Also si/da roj? GN Aptara (Crete, Lycia), Pa/tara (Lycia), Pi/nara (lycia), Me/gara (75), 'Allari/a (Crete), Kuari/a (Caria) 26. -as-a/o- (Fur. 157 n. 57) ka/rpasoj, krau/gasoj, kamaso/j (ka/baisoj), pa/gasa. GN Ku/rbasa (Crete), Me/dmasa (Crete), Ph/dasa (Mess.), Pu/rasoj (Thess.), Pagasai/ (Thess.)

27. -assGN Rutiasso/j (Crete), Kruasso/j (Crete), Muka/lhsso/j Tafiasso/j (32) 28. -atGN Kai/ratoj (Crete), Mi/latoj (Crete 27), 29. -aur-a/o(a')fau~roj, flau~roj, (a')mauro/j, aglauroj, kasau/ra (-aj), la/stauroj, pe/tauron (eu); GN 'Epi/dauroj. 30. -acbo/tracoj, ku/mbacoj. 31. -gda aprigda (adv.). 32. -gr- (cf. on -r-) panagri/j. 33. -edGN Te/nedoj, Le/bedoj. (Lakedai!mwn?) 34. -ez-a see -is-. 35. -eir-oaigeiroj, ku/peiron, sa/beiroj (sape/rdhj); Ka/beiroi. 36. -ell-a/o- (Cf. 28 -el-a/o-) a'kro/spelloj, ba/tella, bde/lla, pa/tella, pe/lla. 37. -el-a/o- (Cf. the foregoing) Exx. a'musge/la, a'sfo/deloj, bri/keloj, dru/yela, (e'pi)za/feloj, qike/lion, i'tqe/la, ku/bela, stufelo/j (stu/floj), duspempeloj? 38. -em-o- (Fur. 151 n. 42) i'a/lemoj (a ), koa/lemoj (a ), p(t)o/lemoj (or IE?), qelemo\n. 39. -emn-o- (Fur. 151 n. 44) a`mfi-ke/lemnon, Kartemni/dej. GN Se/lemnoj (95)

40. -enn-a (I wonder whether ny could give nn) th/benna. Cf. ble/nnoj. Cf. Lat., from Etruscan, ( doss-)ennus, Porsenna. 41. -er-a difqe/ra, a'ske/ra (-hra). GN 'Wleroj (Crete) 42. -et-o- (Fur. 115 n. 4) kaieto/j, kai/petoj, ma/speton, ne/petoj, thlu/getoj? GN Xupe/th (Att.), Taleto/n (Lac.), Taugetoj, 43. -eur- see -aur-. 44. -eut- (Fur. 173; 181 n. 7) baskeutai/, krateutai/. 45. -hbku/rhboj. GN Ka/ndhba, Te/ndhba, Torrhbo/j all in Lydia. 46. -hqGN Pepa/rhqoj, Sesa/rhqoj (67), Kiku/nhqoj (Pagas.), Ka/nhqoj (M Chalkis), Ainhqoj (M 68), Pa/rnhj, -hq(Att.). Cf. -aq-. 47. -hk-, -hc- (Fur. 199; 245 n. 70) bh/rhx, da/ndhx, ku/bhx, lw/bhx, na/rqhx, ph/lhx, tra/fhx, fh/lhx; ku/mhx / ku/mhca. 48. -hl- (Fur. 115 n. 5) a'bro/khloj, a'bu/rbhloj, arbhloj, a'su/fhloj, ba/khloj, qa/rghloj, ka/bhloj, ka/phloj, ki/bdhloj, kich/la, neni/hloj, tra/chloj, fa/shloj. GN Kudrhl-eioi (Kos), `Udrhla (Caria) 49. -h/n (Fur. 172 n. 118) a'dh/n, a'tmh/n, a'ttagh/n, au'ch/n (amfhn), bal(l)h/n, doqih/n, e'ssh/n, kamash/n (-aso\j), kufh/n, seirh/n, swlh/n, tagh/n, tibh/n. Qhrh/n (Crete ++31, 63), Troizh/n, 'Aradh/n (Crete) 50. -hnglabrh/nh 51. -hr

spi/nqhr; Ai'gla/hr? GN Poqhreu/j (Crete), Ku/qhra 52. -hr- (Fur. 204 n. 10) abdhra, andhra, ai'yhro/j, a'ska/lhron, a'skhra/ (-e/ra), mu/khroj; i'mbhrij, ku/phrij, lebhri/j. 53. -hs(s)- Cf. -ass-. Ma/rphssa (M Paros), Mukalhsso/j (Boe. +80); ' Ardhtto/j (Att.), `Umhtto/j (Att. +85) 54. -ht- (Fur. 172 n. 118)

a'l(l)a/bhj, ka/nhj, kw/nhj, le/bhj, ma/sqlhj, mu/khj, ta/phj (Myc. tepa). Here also: an(n)hton (-qon, -son)?

GN Ma !shta (71), Mu/hj, -htoj (Caria) 55. -htt- see -hss56. -hyGN Aidhyo/j (Eub.), Galhyo/j (Thrac.) 57. -qoj

See Cha. Form. 368; cf. -nqoj. bre/nqoj, kanqo/j, spe/leqoj, mi/nqoj, 58. -qrku/riqra, ma/raqron; GN Li/bhqra R. On -aqron Fur. 303 n. 39: ka/n(n)aqron, me/laqron, spa/laqron. Cf. on -str-on. 59. -ib-, -i be'ruqibh 60. -i gmasti g61. -igg-, -igk-, -igcastliggaj, brutiggoi/, elmiggoj, qri/ggoj (-igk-, -igc-), qw~migx -ggoj, sa/lpigx. 62. -i d-; cf. -ind- (Fur. 324 n. 7) balbi/j, gelgi/j, krhpi/j, sfra gi/j 63. -id-na This will be a combination of two suffixes. Cf. on -n-. a'ra/cidna (cf. arakoj, -c-)

64. -i q-, -iq- (cf. -ind-) aglij -i q-, ai'giqaloj (short i), ai'giqoj, a'spi/qion, ga/liqoi, ka/liqoj. 65. -ik-; cf. -ic- (Fur. 226 n. 102) ka/likon, ku/rnika, le/(i/)krika, nw/rikon, sw/sikej 66. -il-, -i l- (indicated) ai'gi/lwy, mari/lh, (s)pati/lh (-i l- = pasti/lh GN Skandi/lh (Kos) 67. -illargilloj, amilla, arilla, asilla 68. -imn- (Fur. 246 n. 71) me/dimnoj (i/), me/rimna, si/drimnon 69. -ina'po/linon, ba/kinon, gossu/pinon. GN Mu/rina (Lemn.), Si/kinoj (Cycl.), 'Arpina (Elis) 70. -i nku/minon, puti/nh, r`hti/nh, se/linon, foxi~noj, fori/nh; Salami !n71. -ind-; cf. -inq- and -id-, -it-. ku/bind-ij, kumi/nd-ala, alindon; cf. mi/ndij; GN Krausi/ndwn (R), Pu/rindoj (Caria), Brugi/ndara (Rhod. +46) 72. -inq-; cf. -ind-. aiginqoj, a'spi/nqion, kalami/nqh, labu/rinqoj, li/minqej, mh/rinj GN Kh/rinqoj (Eub.), Ko/rinqoj (+74), Sh/rinqoj (?) 73. -isarpisa (-eza), ku/tisoj GN La/risa, Kedriso/j, Khfiso/j (-i soj = -issoj 25, 61), Tuliso/j (Crete) 74. -it-; cf. -id-, -iq- (Fur. 163) ba/rbitoj, bo/l(b)iton, po/rfiton GN Su/brita (Crete) 75. -ic-; cf. -ikarsicoj

76. -kn- (Probably a combination of -n- with a preceding sound; see on -n-) abarkna, doru/knion, pisa/kna, fenakni/j, fida/knh 77. -m-oGN La/tmoj (Caria), Pa/tmoj 78. -n- (Fur. 132 n. 65.) Sometimes a preceding velar beco mes aspirated a'ra/cnh, daucna-, ke/rknoj, ku/dnoj, kuli/cnion, mo/lucnon, peli/cnh, satarni/j, sfa/gnoj, yu/dnoj; Ka/barnoj. GN Ku/qnoj (Cycl.) 79. -xamaxa, a'tra/faxuj, koti/xij, kuni/xeij, moroxo/j (cq), saba/xaj (kt), so/rnixa 80. -op- (Fur. 107; often there is a variant with -ab-) el(l)oy, kalau~roy, -poj (-o/fij), ko/lloy, ska/loy GN Koro/ph (Thess.), Kassio/ph (Corc.), Panopeu/j (Phoc.) 81. -or- (See also on word end) acora (-ura), le/porij 82. -oss-a, -ott-a GN `Ermwnossa (Chios), Adrotta (Lydia), ' Aziott-hnoj (Lydia), Molssi/a? 83. -oul-ofasi/ouloj (-wloj) 84. -ourindouroj, ka/bouroj, liggou\rion (lo-, lu-), pali/ouroj, pa/rouroj, pa/ndoura, ta/gcouroj, ski/ouroj. GN Luko/soura (Arc., the oldest town of all; +93) 85. -ous(s)-a (Fur. 197 n. 55) agcousa (e-), aiqous(s)a (aidwssa), ka/dousa, nh/qousa; 'Aki/dousa, Empousa, Khlou~sa (Kh/lwssa M Fur. 197), Me/dousa. 86. -pn- (This may rather be a suffix -n- after a stem) qera/pnh, ompnh 87. -pt- (this suffix will have consisted of one phoneme, so py? marupto/n, pe/ssu(m)pton, sa/naptin 88. -r- (Fur. 124 n. 37; 215 n. 62)

ba/lagroj, gh/ligroj, si/grai, ta/randroj, ca/landron; 'Idagroj (Lyc. ida~kre?).

Compare also on -rn-, -rg- and -gr-. 89. -rgla/qargoj (ai, h) 90. -rdGN Ku/arda (Caria) 91. -rn- (Fur. 48 n. 126; 215 n. 62) a'karna/n (a'ka/rnax), akorna, kuberna/w, liperne/w (lif-). We find variants without the -n-: So probably the cluster arose through the addition of the suffix -n-. Note that -rn- is found in Etruscan and already in Cappadocian; Fur. 48 n. 126. See also on (single) -r-. GN Fala/sarna (Crete), 'Alasa/rna (Kos); Le/rna; `Ali/kurna (Aet.) 92. -sku`ri/skoj (c, ss) 93. -s-o- (Fur. 254 n. 27. In several cases this seems not a suffix but the end of a root; cf. -as-, -is-, -us-) alsoj, ka/baisoj (as), mu\soj, pi~soj, mi~soj, fa/rsoj GN Pri/ansoj (Crete), S(a)ranso/j (Crete), 'Ialuso/j (Crete) 94. -ssku/passij, kupa/rissoj, sa/rissa 95. -stsi/surna : si/sura, kuberna/w : kumerh~nai, satarni/dej : satari/dej, ki/sirnij : ki/ssirij.

a'la/bastoj, qemist- (cf. Myc. temitija / timitija [Ruijgh]), lepasth/, plata/nistoj; GN Ka/rustoj, Lu/kastoj (Crete), Faisto/j, 'Ogchsto/j (R Thess.) 96. -stron; cf. -qr-. a'la\bastron, de/pastron (l-), enustron (h-), zu/gastron, ka/nastron, lai(s)tro/n, si/gistron de/paj is of Luwian origin; see Melchert in FS Manaster Ramer.] 97. -t-oasfaltoj, atraktoj, aflaston 98. -tt- (see 5.5 on tt/ss) kurittoi/, proko/tta; Fe/refatta 99. -ub[Prof. Melchert writes me that he thinks that the suffix may be Luwian ( -as-tar-ra/i-), as in de/pastron, where

entubon, qo/ruboj, intuboj (oub), si\lluboj, si/ttuboj, sko/luboj (m), skolu/bra (-ob-), ca/luy; cf. o'cqoibo/j 100. -uggla/rugx, pi/suggoj (i ), fa/rugx 101. -uda'mu/j -doj, korudo/j 102. -udna GN Kaludna/ (Kos) 103. -uq- with long and/or short u a'gnu/j (u ), lh/kuqoj (short u) 104. -ui-a aguia, kw/duia; Arpuia GN Kadui/h (18, 24), Kindui/a (Crete, also Kindu/h 18, 24) 105. -ukampux, idux GN Na~rux (Locris), Babu/ka (Lac.) 106. -u kkh~rux, -ukoj 107. -ul- (Fur. 205 n. 14) a'rbu/lh, batu/lh, da/ktuloj, kanqu/lh, kotu/lh, krwbu/lh, mimai/kulon (me-), sfo/nduloj (sp-) with short u. 108. -u l(s)kordu/lh; sfondu/lh (sp-) , GN Kardamu/lh (Mess.) 109. -umge/rsumon, gi/gglumoj GN Si/duma (? 33), Ka/rumai (Crete), Lw/ruma (Kos) 110. -umbiqumboj, ko/lumboj 111. -umn- (Fur. 243 n. 66 on -umn- in Etruscan and Cappadocian) ai'sumna/w, basumnia/ta j (a ), si/gumnoj

GN `Ri/t/qumna(Crete), Ma/qumna (Lesb. +28), La/rumna (Locr.) 112. -u n-. See also on -unn-. bo/qunoj, sigu/nh (cf. -unn-), la/gunoj GN Go/rtun (Crete) 113. -und-; cf. -unq/tBereku/ndai GN Kamund-ioj (Rhodes), Ka/lunda (Caria) 114. -unq/tbo/lunqon, koloku/nqh, olunqoj; Bereku/nq/tai, GN 'Ama/runqoj, Za/kunqoj(+88), Ti/runj 115. -unnsi/gunnoj, Di/ktunna See on -u n-. 116. -uxGN 'Olo/fuxoj (Athos) 117. -upissupoj (older usswpoj), ma/rsuppoj, oi'su/ph, tolu/ph. 118. -ura'h/suroj, acura (-ora), laiqur-a/zein, maukuro/n, o'nur-i/zetai, satu/rion, si/sura GN Eluroj (Crete). Aisuroj (Bith. ++18), Aiguron (+30), Koi/nura (Thrac.), Te/gura (Boe.), N/i/suroj (Kos) 119. -u ragkura, a'na/guroj ( o'no/-), ge/rgura (go/rg-), pi/turon, 120. -usarusoj 121. -utGN Lagina/puton (Crete), Kolluto/j (Crete) 122. -ucbo/(s)trucoj GN Mo/suc-lon (Lemn.) 123. -fq-

krosso/fqon, la/kafqon, mo/lofqoj, na/(s)kafqon, sarrufqei~n 124. -f-o- (on -af-o- Cha. Form. 269) a'rge/lofoi (il), mastrofo/j (-po/j), se/rifoj, se/sufoj 125. -wkGN Koqwk-i/dai (Att. +70) 126. -wla'pofw/lioj?, ma/gdwloj, fasi/wloj (-ouloj, -oloj), fa/skwloj GN Ki/mwloj (Cycl.) 127. -wmba/rwmoj 128. -wn- (Fur. 303 n. 39) a'lkuw/n, h'i#w/n, sandw/n, sindw/n, scadw/n 129. -wp-; a variant is -oupai'gi/lwp-, ku/nwy, kinw/peton (kunou~pej). GN Eu'rwpo/j/a (Crete), Kassw/ph (Epir.) 130. -wr- (Fur. 211 n. 50) a'cw/r, a'ma/nwr, bia/twr, i'cw/r (i ), lei/twr GN Pi/lwroj (Chalc. +22), Khskw/ra (Crete), Asswron (M Samos) 131. -wssa see -ous(s)a GN Dirfwsso/j (Eub.), Pidwsso/j (Caria +26) 132. -wt- (Fur. 283 n. 83; 384 n. 132) a'skalabw/thj, ballwth\, -kaudwto\n, ki bwto/j, kra/mbwton, mo/nwtoj, oi'spw/th GN Qesprwt-oi/ Add. -inna (Cha. Form. 205), 3. Word end Word end is interesting as some original finals of the Pre-Greek language may have been preserved. Of course, Greek endings must be removed, notably -oj, -on. Thus -ion, -uon may often continue original -i, -u. (Cf. Myc. dunijo beside duni.) The words in -vqoj have almost displaced those in -nq (Ti/runq- etc.). 1. in vowel Pre-Greek short -a. (It is often difficult to see whether an -a is short or long; the material must be furthe r studied.) 1a. -a. A short -a can in Greek only result from *-ya < *-ih2; in other cases we may have to do with an original,

Exx. abarkna, a'ga/nna, a'ge/rda, adalta?, aikouda, abdhra, akara, akorna, a'ko/stila, a'kta/ra, alara, a'moucra/, arda, a'staga/na, afqa, brou/ka, go/da, go/la, da/xa, da/rda, qa/pta, mo/da, r`o/mixa, sa/tta, so/rnixa, sou~a, ke/dropa (c-), etc. Note forms in -ua: arua,...and -enna. 1b. -i. IE words (i.e. neuters) in -i are very rare in Greek. Exx. za/kti, ko/ri, ta/guri (tagu/rion), a'kari/. We can safely assume, I think, that words in -ion, -uon originally ended in -i, -u. -ij is also frequent. Note that this situation is only to be expected if the language had only the vowels a, i, u.

1c. -u. a'baru/, ko/ndu, mw~lu, ... See the foregoing on -uon. -uj is also found several times: agduj, arpuj, a'tra/faxuj, bi/qun, ble/tuj, i'luj (i ), -me/nduj, mi/markuj, mwu/j, phlamu/j, r`a/puj. 1d. -euj. Though the ending may also be an IE heritage, in many words the ending is clearly of Pre -Greek origin. basileu/j (Myc. qasireu); 'Acil(l)eu/j.

1e. -w. Gellw/ koqw/, ko/rqw, motw/, ti tw/, ...

The suffix makes fem. names in -w/: Lhtw, Sapfw/. Words in

-wj are masculine: a'carnw/(j), hrwj; Mi/nwj, Ta/lwj. 2. -in -r 2a. in -ar asar(on)?, a'kca/ntar, ba/kar, e'li/mar, ku/dar, ne/ktar, nw~kar, seli/ar 2b. -ir Exx. [sullir is Lac. < -ij] 2c. -or Exx. a'digo/r, a'kko/r, kakko/r, ke/mmor (mp), pi/sor. 2d. -ur Exx. yi/qur, (Dor.) ma/rtur, 2e. -wr Exx. acwr, i'cw/r, ke/lwr, yo/qwr 3. Several words end (in the nom.) in -x or -y. 3a. -x, stem in -k-, is found quite often: -ax: abax, a'ka/rnax, ambax, a'ndra/fax, anqrax, bu/stax, ma/lbax, sau~sax. anax has a stem in -kt-. -a x: fe/nax, r`a/x, sfh/x, -hx: da/ndhx, bh/rhx, -ix: addix, a'nqe/rix, a'pri/x, ko/lix, arpix, -ox: be/brox, bi/rrox,

-oux: brou/x, -ux: bli/tux, go/rtux; Pnu/x, Stu/x. Note acc. ba/lleka; and acc. bri/gka. 3b. -y have: la/tray, lai~lay, aliy, ko/riy, ai'gi/poy, ko/lloy, me/roy, gu/y, mo/nwy. Monosyll.: cre/my. 4. in -n: (Kuiper 217) ballh/n, karba/n; Wgh/n. ki/ndun, mo/ssun, r`w/qun-ej, 5. Several words end in -aj (a -stems ): a'ba/j, ablaj, aqraj(?), a'mfi/aj, a'skalw/paj, a'skwndaj, a'ttaga~j, bada/j (bata~j), baska~j, bu/aj, kalaba/j, kasa~j; 'Aqa/maj, 'Asgelataj (a ). With a stem in -ant-: a'li/baj (nt), luka/baj (-nt-) etc.; see on the suffix. With stem in ad-: a'cra/j, bounia/j, prhmna/j; see on the suffix.

D. THE UNITY OF PRE-GREEK I think that the material itself shows that we have largely to do with one language, or a group of closely related dialects or languages. Of course, we cannot in every case demonstrate that the words that are non -Greek belong to this same language. And it is a priori probable that there are loanswords from other sources, but the bulk of the known non-Greek words seems to fit the general picture we have of `Pre-Greek', the Pre-Greek substratum. E.g. ko/tquboj / ko/sumboj shows not only the element ss/tq, well-known from geographical names, but also the suffix -ub- with prenasalization. kro/ssion / kro/tion also shows the ss/t, but kro/ssofqon has a suffix that is also typical for this language. The word dala/gcan beside qa/lassa (-tta) has again the suffix ss/tt but also the prenasalization. astligx / ostligx has both the typical (prenasalized) suffix -igg- and variation a/o. In mh/rinqoj / smh/rinqoj we have the `mobile s-' and the well known suffix, while me/rmij, -iqoj has the unprenasalized variant, and smh/riggej has another Pre-Greek suffix. In a(m)bruttoi / bru/ttoj (bru/ssoj) we have prothetic vowel and prenasalalization combined. Other languages may well have existed in the area. Thus it is not certain that Hieroglyphic Minoan expresses the same language as Linear A. Further, Eteocretan has not yet been connected with other elements and seems isolated. Another matter is that (non-Indo-European) loan-words from old Europe may have enter ed Greece cf. Beekes, 125 Jahre Idg., 2000, 21-31. And these may have been adopted already in Pre -Greek, as is suggested by e're/binqoj. Still another category are Anatolian (?) loanwords that entered Greek, and sometimes also other IE languages very early, like pe/lekuj. However, I think that it is methodologically better to start from the assumption that non -Greek words are Pre-Greek; and we have now a set of criteria to confirm this. Only when there is reason to assume that they have a different origin we should accept this possibility. E. NON-INDOEUROPEAN Our knowledge of Indo-European has grown so much, especially in t he last thirty years with notably the

growth of the laryngeal theory, that we can in some cases say that an Indo -European reconstruction is impossible. A However, a preform * gnh2dh- would have given Gr. * gna q- with a long a. One might think the assuming * h2e would remedy the problem, but * gnh2edh- would give * ganaq-, so we would have another problem. The conclusion is that no Indo-European proto-form can be reconstructed, and that the word cannot be Indo -European. There is no problem in assuming a Pre-Greek word (though the word has no typical characteristics of Pre -Greek). - Another example is the word krhmno/j `overhanging bank', for which a connection with kre/mamai `hang (up)' seemed evident. However, we now know that long vowels cannot be postulated at random, and here it is simply impossible: there is no formation type that would allow a long vowel. The objection is confirmed by the fact that there is no trace of the expected a < *h2 (as in kremamai < *kremh2-). Positively one can say that features of the landscape are often loanwords from a substratum. The inevitable conclusion is that the word is P re-Greek. The more we know about Indo-European, the less is possible. As our reconstructions become more and more precise, they have to conform to all the rule we have established by now. This holds for all etymological work: in a way, then, it becomes more difficult. This also regards Pre-Greek, as indicated: for some forms an Indo European origin is no longer possible. good example is the word gna/qoj. To explain the -a- of this word we need introduces a `second laryngeal' ( h2).

One question that rises when one adduces geographical names is to how far in the east such names can be adduced. I have simply followed Fick. I noted that his names go as far as Ciclicia. I noted the following sites in Cilicia mentioned by him: 9 Kwru/kion 24 Kidrolla~j 32 Mallo/j 37 Knw/j 43 Ka/sai, Sa/roj, Ka/rpaqoj(?) 47 Kasta/bala, Lae/rth 59 Dalisando/j 64 Kalu/kadnoj 70 Ke/skoj 72 S(a/)ra/nsoj, Kele/nderij, 74 Ku/inda 77 Pindenisso/j 79 Kastali/a 84 Adana

F. Geographical names