SUBSTRATUMI GJUHËSOR PARAGREK

 Les Études classiques 79 (2011), p. 257-283. THE PRE-GREEK LINGUISTIC SUBSTRATUM A Critical Assessment of Recent Theories * Résumé. — Un précédent article, paru dans Les Études classiques, avait présenté diverses théories récentes sur le préhellénique (théories dites « pélasgique », « anatolienne », « égéenne » et « kartvélienne ») ; elles sont ici comparées et évaluées sur la base d’une sélection de problèmes étymologiques (notamment, les étymologies de pölekuj, de t⁄mboj, de si -gª/si %wpª, d’¢gaq“j et du suffixe -nqoj). Ces sont les théories « égéenne » et « kartvélienne » qui s’avèrent les plus productives, bien qu’elles posent encore divers problèmes méthodologiques. Les mots les plus problématiques sont ceux qui peuvent avoir été adoptés d’une langue sémitique et les gloses d’Hésychios. En outre, dans quelques cas, une explication à l’intérieur du grec ou une interprétation indo-européenne semble préférable au postulat d’une origine préhellénique. Introduction In a previous volume of LEC, I published a first article in which I gave an overview of the linguistic research into the Pre-Greek substratum of the last thirty years 1 . While my approach there was mainly descriptive, the present article confronts and evaluates the theories presented there through a selection of etymological problems. For a great part, I shall comment on the research by R. S. P. Beekes, probably the most important linguist studying the Pre-Greek substratum at present. In order to understand the evaluation of the selected etymologies, a brief summary of the main Pre-Greek theories should be given. A first theory, founded by V. I. Georgiev and A. J. Van Windekens, and in the 1980s and 1990s mainly represented by E. P. Hamp, regarded Pre-Greek as a hypothetical Indo-European language, dubbed ‘Pelasgian’. A second group of scholars (viz. L. R. Palmer and Margalit Finkelberg) identified the substratum as Anatolian (more precisely Luwian). According to a third * I would like to thank professor L. Isebaert and professor H. Seldeslachts for their significant comments and suggestions. I am also grateful to my former fellow students and friends Liesbeth Schulpé and Lore Willemse for correcting and improving my English text. 1. G. VERHASSELT (2009). 258 LES ÉTUDES CLASSIQUES theory, which we may call ‘Aegean’, the substratum was non-IE and nonSemitic and extended over (a large part of) the Mediterranean. At present, the main proponent of this theory is R. S. P. Beekes, who has developed his ideas on Pre-Greek within the framework of ongoing substratum research in Leiden. Finally, at the end of the 1970s, a fourth theory was developed by E. J. Furnée (a former adherent of the ‘Aegean’ theory), viz. the Kartvelian theory. His main thesis was that the Pre-Greek substratum consisted of two important components: an East-Mediterranean substratum (as in the Aegean theory) and a genuine Kartvelian substratum (termed ‘Pelasgian’). Currently, the main advocate of this Kartvelian theory is the Georgian scholar R. Gordeziani. Having established this brief survey of the main theories, we can move on to the etymological case-studies, viz. pölekuj, t⁄mboj, si - gª/si %wpª, ¢gaq“j and the suffix -nqoj 2 . 1. pölekuj Traditionally, pölekuj “(battle-)axe” (already attested in Homer) is less closely associated with the Pre-Greek substratum than the words that will be discussed in the next sections. Usually, it is traced back to PIE *peleKu-: cf. OI paraSú- “axe” (epic párSu-), Osset. färät “id.” (Toch. A porat “id.”, B peret “id.” is an Iranian loanword). In his doctoral thesis, E. J. Furnée 3 mentioned the following Greek variants: pölekkon “axe-handle” (Il. 13, 612), ¹mipölekkon “half-axe, one-edged axe”, bölekkoj (Hsch.) 4 , pölekra (Hsch.) 5 and pölux “a kind of axe”. According to him, pelek£n “pelican” too is related to this set of words. He supposed that pölekuj and paraSú- represent a pre-IE cultural term borrowed by both the Greeks and the Indo-Iranians. bölekkoj was interpreted as a younger Pre-Greek loanword, showing the Pre-Greek variations p/b and k/kk 6 . Hellenistic pölux would be a later borrowing from Asia Minor. In her review of E. J. Furnée’s work, Françoise Bader 7 criticised the Dutch scholar for treating loanwords of various origin and Pre-Greek substratum words indiscriminately. For pölekuj, she considered a 2. I confine myself to recent literature (i.e. starting from 1977) on each word discussed, combined with the etymological dictionaries of P. CHANTRAINE (1968) and H. FRISK (1972) and with the doctoral thesis of E. J. FURNÉE (1972). 3. E. J. FURNÉE (1972, p. 150-151). 4. bölekkoj· ‘spri“n ti ômferùj laq⁄rJ mögeqoj ôrebÖnqou úcon (Hsch.). Cf. Aristoph. Ran. 755; POxy. 1801, 21. 5. pölekra· ¢xÖnh (Hsch.). 6. According to P. CHANTRAINE (1968, p. 875), kk goes back to *k#. 7. Françoise BADER (1975, p. 104). THE PRE-GREEK LINGUISTIC SUBSTRATUM 259 borrowing of PIE date, directly inherited by Greek and Indo-Iranian. A good candidate for a donor language would be Semitic. Thus, G. Takács 8 connected pölekuj with Akk. pilakku “spindle”, a derivative of the Semitic root *plq- “split”. R. A. Brown 9 (another proponent of the Aegean theory) too assumed pölekuj and paraSú to be Near Eastern loans. R. S. P. Beekes 10, conversely, stressed that a word occurring in several IE languages could still be an early borrowing (although the IndoIranian group is usually excluded in such cases). In his etymological dictionary 11, he followed H. Frisk 12, who had rejected the connection with Akk. pilakku on semantic grounds, since the Semitic word does not mean “axe” but “spindle” 13. On other occasions too, R. S. P. Beekes stated that in doubtful cases, “it is better to consider such words as Pre-Greek, and to define them as loanwords [...] only when there is reason to do so” 14. What those reasons may be, we are not told. This rather peculiar methodological principle illustrates R. S. P. Beekes’ attempt to keep his corpus of PreGreek words as large as possible. Pace E. J. Furnée and R. S. P. Beekes, pölekuj is a less convincing example of Pre-Greek vocabulary (in R. S. P. Beekes’ use of the term, i.e. belonging to a non-IE, non-Semitic substratum). The parallel with the Semitic language family (despite the minor semantic change) makes an early loanword in PIE plausible, although Semitic might have borrowed it from PIE as well. Alternatively, it could be explained as a so-called Wanderwort. Nevertheless, a PIE reconstruction *peleKu- remains plausible, even though the root structure is unusual for PIE 15. For the problematic form bölekkoj, the safest option is to assume a secondary evolution (rather than a Pre-Greek variation), since this variant is only attested from Aristophanes onwards, whereas the ‘standard’ form is already used by Homer. Still, this development remains puzzling. Perhaps the b was influenced by the word böloj “missile” (later used of any weapon). Of course, caution is advised with this type of argumentum ex silentio, given the fragmentary preservation of Greek literature: many words may be unattested by sheer coincidence. Thus, the later attestation of bölekkoj 8. G. TAKÁCS (1998, p. 144). 9. R. A. BROWN (1985, p. 81). 10. R. S. P. BEEKES - A. H. KUIPERS (1975, p. 78). Cf. R. S. P. BEEKES (1996, p. 215). 11. R. S. P. BEEKES (2010, p. 1166-1167). 12. H. FRISK (1972, p. 497). 13. M. MAYRHOFER (1996, p. 87) concluded the same for the Old Indic words in his etymological dictionary of Old Indo-Aryan. 14. R. S. P. BEEKES (2010, p. xv). 15. T. V. GAMKRELIDZE - V. V. IVANOV (1995, p. 771). 260 LES ÉTUDES CLASSIQUES does not mean that it did not exist yet in Homer’s time. That is exactly why the Hesychian glosses, despite their late attestation, are so important for E. J. Furnée and R. S. P. Beekes. It was precisely the lexicographers’ purpose to collect rare and perhaps even archaic forms of various origins. Nevertheless, nothing guarantees the Pre-Greek origin of such words. At best, they can point to a non-Greek (not necessarily Pre-Greek) origin. Moreover, since Hesychius is usually the only source for the word in question, the possibility of scribal errors cannot be excluded. Since it is not even certain when these words entered the Greek language, some glosses may well be Byzantine loanwords. Therefore, the examination of PreGreek features should primarily be based on (relatively) early attested words, whereas material from Hesychius should be treated more carefully. This does not imply that the latter material is useless, but such an approach is methodologically sounder. Finally, the form bölekkoj invites a remark on E. J. Furnée and R. S. P. Beekes’ interpretation of the Pre-Greek stop variations. E. J. Furnée 16 considered most of these cases expressive variants (cf. the expressive geminate in Iuppiter), a hypothesis which proves to be untenable, since most words showing Pre-Greek variations belonged to the technical rather than the expressive vocabulary. In the debate on E. J. Furnée’s work, several other explanations were suggested, the most popular one explaining these vacillations as resulting from the different nature of the Pre-Greek phonological system. According to R. S. P. Beekes, Pre-Greek lacked a phonological opposition between voiceless, voiced and aspirated stops 17. A possible explanation is that they were allophonic variants differing from one word to another. Yet in several cases (e.g. b“lbiton/bolbÖdion/b“lbiqoj “cow-dung”), these variations occur within almost identical words. The suggestion by A. Heubeck 18 in his review of E. J. Furnée’s doctoral thesis, viz. that Pre-Greek had a phonological opposition between fortes and lenes, is a reasonable explanation for the velars (where variation between voiced and aspirated stops is rare) but not for the labials and dentals 19. One could assume that there were two velar phonemes, whereas labials and dentals lacked this 16. E. J. FURNÉE (1972, p. 108). 17. The real phonological opposition, according to R. S. P. BEEKES (2010, p. xvi), is the opposition between plain, labialised and palatalised consonants. 18. A. HEUBECK (p. 276). 19. The variation g ~ c occurs only 5 times and the variation k ~ g ~ c only 7 times, whereas k ~ g is attested 55 times and k ~ c 51 times. For the labials and dentals, by contrast, the following proportions are found: b ~ f (28), p ~ b (53), p ~ f (43), p ~ b ~ f (11), d ~q (16), t ~ d (27), t ~ q (26), and t ~ d ~ q (6). See R. S. P. BEEKES - A. H. KUIPERS (1975, p. 72). THE PRE-GREEK LINGUISTIC SUBSTRATUM 261 phonological contrast. Still, were aspiration and voicing attributed randomly in the borrowing process? A final possibility is that Pre-Greek had a voicing pattern different from that of Greek. J. Clackson 20, for example, mentioned with P. Ladefoged and I. Maddieson five types of voicing attested in the languages of the world: breathy voice, slack voice, modal voice, stiff voice and creaky voice. Pre-Greek might have had an unusual type of voice, which the Greeks sometimes perceived as voiced, sometimes as voiceless and sometimes as aspirated. 2. t⁄mboj t⁄mboj “burial mound, grave” is one of the stock examples of the PreGreek substratum research. The pivotal question is whether it is related to t£foj “grave, tomb” and q£ptw “bury”, the latter two being derivatives of PIE *d hembh - “dig”: cf. Arm. damban/dambaran “tomb, grave” 21 , OPr. dambo “ground” 22, Goth. faur-dammjan “dam up”, Rum. dîmb “bank” 23, Av. daxma- “grave”. In his controversial book on the ‘AfroAsiatic roots of classical civilisation’, M. Bernal 24 rejected the PIE root *d hembh -, assuming that t£foj and q£ptw are Egyptian loanwords, cf. »tpHt “hole, cavern”, an improbable hypothesis, given the parallels in other IE languages. 2.1. Genetic relationship between t⁄mboj and t£foj According to V. I. Georgiev 25 , t⁄mboj and t£foj are genetically related. The latter was interpreted as the regular Greek outcome of the PIE proto-form, whereas the former showed a ‘Pelasgian’ development characterised by the following sound laws: dissimilation *d h -bh > *d-bh , vocalisation *m > um, *d > t and *b h > b. In his re-examination of the Pelasgian theory, K. Strunk 26 distinguished two evolutions of a nasal + *b h : a Greek evolution, viz. mf (e.g. ¢mfÖ “around” < PIE *h2mb h í “around”: cf. Lat. amb- “around”, Ved. abhí “on both sides”, OHG umbi “around”) and a Pelasgian evolution, viz. mb or f (e.g. k“rumboj “top of a 20. J. CLACKSON (2007, p. 48). 21. J. CLACKSON (1994, p. 120-121), however, interpreted damban and dambaran as loanwords, given the late attestation of these words (from the 10th century onwards). In the earliest Armenian, the words to designate a tomb or grave were gerezman and sirim. 22. J. POKORNY (1969, p. 248-249) only accepted the Armenian words as derived from PIE *d hembh -, explaining that OPr. dambo had to be corrected into daubo, a derivative of PIE *d he2b- “deep”. 23. According to J. CLACKSON (1994, p. 121), Rum. dîmb is a Dacian survival. 24. M. BERNAL (2006, p. 428-429). 25. V. I. GEORGIEV (1981, p. 103). 26. K. STRUNK (2004, p. 94-96). 262 LES ÉTUDES CLASSIQUES hill”, korufª “top, summit, peak of a mountain”). The former group shows parallels in other IE languages, whereas the latter does not, thus implying a non-IE origin for this group. K. Strunk, furthermore, suspected that the b was of an allophonic nature rather than an independent phoneme. I. Hajnal 27, by contrast, attempted an inner-Greek explanation for t⁄mboj. According to him, t⁄mboj and t£foj are etymologically related, though not through Pre-Greek sound laws. t£foj would derive from the zero-grade *d hmb h -os, whereas t⁄mboj was explained as reflecting the fullgrade *d hómbh -os. In I. Hajnal’s view, PIE *O developed into U in accordance with Cowgill’s rule 28 (cf. n⁄x “night” < PIE *nokwt- “id.”), whereas the loss of the aspiration could be explained by Miller’s rule: in the combination of a nasal and a voiced aspirated stop, the stop lost its aspiration if the accent was on the previous syllable. As I. Hajnal himself acknowledged, the problem with this reconstruction is that the expected form would be †q⁄mboj. An early dissimilation of the aspirates is excluded, since Miller’s rule operated before Grassmann’s law 29. The only possible explanation was analogy, although I. Hajnal was not clear about its model. 2.2. No genetic relationship between t⁄mboj and t£foj Other scholars have rejected the etymological relation between t⁄mboj and t£foj. In his complementary Greek etymological dictionary, A. J. Van Windekens 30, who had been a staunch defender of the Pelasgian theory in his previous work, suggested an inner-Greek explanation, a tendency also found in the rest of his dictionary. He explicitly rejected V. I. Georgiev’s analysis, claiming that t⁄mboj could not derive from PIE *d hmb h -, since in that case, it would mean “what is dug out”. He also mentioned the variant t‡moj (found in inscriptions from Corcyra and Eretria), which he connected with Lat. tumulus “hill, sepulchral mound” and tumere “swell”, as had been proposed before him by H. Frisk 31. If the relation with these Latin words, which must derive from the PIE extended root *te2-m- “swell” (cf. OE duma, OHG dumo “thumb”), is accepted, the b in t⁄mboj becomes problematic. The ad hoc solution suggested by A. J. Van 27. I. HAJNAL (2005, p. 197-198). 28. See A. L. SIHLER (1995, p. 42-43). Note, however, that Cowgill’s rule is generally restricted to a position between labial and resonant. 29. According to I. HAJNAL (2005, p. 197), the order is as follows: loss of the aspiration (Miller) > devoicing of the PIE voiced aspirated stops > dissimilation of the aspirates (Grassmann). 30. A. J. VAN WINDEKENS (1986, p. 223). 31. H. FRISK (1972, p. 943-944). THE PRE-GREEK LINGUISTIC SUBSTRATUM 263 Windekens was that t⁄mboj resulted from a contamination of t‡moj and k“rumboj “top, summit, peak of a mountain”. A Pre-Greek interpretation of the word-pair t‡moj - t⁄mboj was proposed by E. J. Furnée 32, who analysed the fluctuation between m and mb as reflecting two Pre-Greek sound variations, viz. b/µ and b/mb (‘nasalisation’ or, in R. S. P. Beekes’ terminology, ‘prenasalisation’). In this interpretation, a form with the single labial (*tuboj) must be reconstructed, a variant of which is found in to‡foj (Hsch.) 33. E. J. Furnée associated t⁄mboj / t‡moj not only with Lat. tumulus but also with MIr. tomm “little hill”, Arm. t‘umb “bank” and OWN þuf “little hill”, assuming a pre-IE Wanderwort. R. S. P. Beekes 34 considered the possibility that t⁄mboj derives from a nasalised variant of the PIE extended root *te2H-bh - “hump, bump, knag” 35 (cf. t⁄fh “plant used for stuffing cushions and beds”), with loss of the aspiration after the nasal. Like A. J. Van Windekens, he rejected the connection with t£foj on semantic grounds: in his opinion, t£foj denotes the grave or pit, whereas t⁄mboj primarily indicates the hill, later more specifically the burial hill, and finally the grave. 2.3. Conclusion It cannot be excluded that t⁄mboj and t£foj are related, although popular etymology may be involved. Some other word might have been remodelled to t⁄mboj under the influence of t£foj, an association that may have been stimulated by their semantic proximity. However, if t⁄mboj is related to Lat. tumulus, tumere, OE duma and OHG dumo, the connection with PIE *d hembh - is impossible, since the Latin and Germanic words cannot derive from this root. The pivotal question is whether this set of words continues a PIE root (viz. *te2-m- “swell”). By connecting t⁄mboj with the Latin and Germanic words and interpreting t⁄mboj as a Pre-Greek word, R. S. P. Beekes implied that the Latin and Germanic words are nonIE as well. K. Strunk, on the other hand, proposed viewing mb as an allophone of m (i.e. with a weakly pronounced b). If a Pre-Greek origin is assumed, one could reconstruct a phoneme *mb , although caution is advised, given the limited number of cases showing the variation m/mb. 32. E. J. FURNÉE (1972, p. 287-288). 33. to‡foj· t£foj (Hsch.). 34. R. S. P. BEEKES (2010, p. 1517-1518). 35. This was also suggested by P. CHANTRAINE (1968, p. 1144). 264 LES ÉTUDES CLASSIQUES 3. si - gª - si %wpª 3.1. si - gª Although si - gª “silence” shows parallels in other IE languages, viz. OHG swigen “be silent”, OSax. swigon “id.”, OE swigian/sugian/suwian “id.”, an IE reconstruction for Greek is problematic, since the initial *sshould have developed into an aspiration in Greek 36. The proponents of the Pelasgian theory derived si - gª from the same PIE root as the Germanic words and considered it a Pelasgian loanword (i.e. a word belonging to a hypothetical IE substratum). They did not agree, however, on the exact reconstruction of the PIE root. V. I. Georgiev 37 posited a root *s2ig h -, with preservation of the initial s- and an evolution of PIE *g h into Pelasg. g. K. Strunk 38 adjusted this into *s2iHgh -, the long i resulting from the combination of *i with a laryngeal. E. P. Hamp 39, on the other hand, suggested a reconstruction *s2e1g h - with a Pelasgian development of *ei to i. Incidentally, V. I. Georgiev 40 suggested the same phonetic evolution for sãtoj “grain, food” but did not invoke this rule to explain the i - of si - gª. The problem with V. I. Georgiev’s reconstruction is that the assumed PIE root is of an unusual shape. R. S. P. Beekes 41 originally considered si - gª a ‘European’ substratum word (i.e. belonging to a prehistoric language that extended over Central Europe and left traces in several IE languages), thus explaining the similarity with the above-mentioned Germanic words. On the basis of ·Öga (Hsch.) 42, probably a writing error for *#Öga (by confusion of <#I> and in majuscule script 43), he concluded that the original Greek form must have been *sw-. #- can easily derive from *s#- but s- cannot. In his etymological dictionary, by contrast, R. S. P. Beekes 44 considered the etymology of si - gª unknown. 36. Because of this irregularity, M. BERNAL (2006, p. 307) suspected that the Greek word is an Egyptian loanword, viz. sgr(i) “silence”, a seemingly rash conclusion. 37. V. I. GEORGIEV (1981, p. 100). 38. K. STRUNK (2004, p. 91). 39. E. P. HAMP (1983, p. 148). 40. V. I. GEORGIEV (1981, p. 102). 41. R. S. P. BEEKES (1996, p. 233-234). 42. ·Öga· siËpa (Hsch.). 43. P. CHANTRAINE (1968, p. 1001) and K. STRUNK (2004, p. 91). 44. R. S. P. BEEKES (2010, p. 1327). THE PRE-GREEK LINGUISTIC SUBSTRATUM 265 A final explanation was suggested by G.-J. Pinault 45. This scholar derived si - gª from the interjection sãga “hush”, based on an onomatopoeia /ss/, which was syllabified by the insertion of an /i/. 3.2. si %wpª P. Chantraine 46 assumed that the word-pair si - gª - si %wpª went back to an onomatopoeia and that si %wpª was an expressive variant. This expressive interpretation was rejected by G.-J. Pinault 47, who reconstructed the verb si %w-p£w (from which si %wpª would be derived) as a compound with *peh2 - “protect, keep” as its second member 48. Thus, he believed, the assumed Homeric meaning of siwp©n, viz. “garder le silence” (as opposed to sig©n “être silencieux”) could be explained 49. For the first element of the compound, he considered two explanations: an interjection /si/ “hush” or a reduced form of the imperative si - gÓ > si %w- (the loss of the intervocalic g being explained on the basis of a similar evolution in —lÖoj < —lÖgoj “little, small”) 50. The former explanation, however, fails to account for the w, whereas the latter is implausible, since vocalic contraction and weakening of intervocalic g are both relatively late developments, not yet operating in Homeric Greek, and since the variation of long and short i remains unexplained. According to A. J. Van Windekens 51, the original form of si %wp£w/ si %wpª was swp£w/swpª: cf. seswpamönon (Pind.) “kept silent”, diaswp£somai (Pind.) “I shall be silent”, swpª (Call.) “silence”, eŸswpÖa (Hsch.) 52. He explained this form as a Pelasgian loanword derived from the PIE lengthened grade *s2op- “sleep” (cf. Lat. sopire “lull to sleep”); more specifically, he assumed a Pelasgian dialect showing the development PIE *s2- > Pelasg. s- but not the consonant shift PIE *p > Pelasg. f. The form si %wp£w, in his opinion, arose through contamination of swp£w either with si - g£w (despite the contrast between i % and i - ) or with an unattested Pelasgian word *si % -, deriving from the PIE zero-grade *Kiof the root *Ke1- “lie down” (cf. Gr. keã-tai “he lies down”, OI Sé-te “id.”) and showing the development of PIE *K into Pelasg. s. 45. G.-J. PINAULT (1994, p. 503). 46. P. CHANTRAINE (1968, p. 1008). 47. G.-J. PINAULT (1994, p. 503). 48. G.-J. PINAULT (1994, p. 518-519). 49. G.-J. PINAULT (1994, p. 516). 50. G.-J. PINAULT (1994, p. 521). 51. A. J. VAN WINDEKENS (1986, p. 216). 52. eŸswpÖa· ¹sucÖa (Hsch.). 266 LES ÉTUDES CLASSIQUES Originally, R. S. P. Beekes 53 also connected si %wpª with si - gª, the variation velar/labial, the ablaut i - /i % and the w indicating, in his view, a non-IE origin. He did not rule out the possibility of a loanword from a lost IE language, although, in his opinion, the variation labial/velar rather points to a ‘European’ (i.e. non-IE) borrowing. si %wpª would be derived from *swip-, which developed into *siwp/siup- through metathesis, then into *siop- (because Greek had no diphthong iu) and finally into *siop-, where the long o arose through popular etymology under the influence of words in -wp-. This whole explanation seems somewhat far-fetched, however. Later, R. S. P. Beekes 54 withdrew his aforementioned interpretation, abandoning the etymological relation between si %wpª and si - gª and explaining the former as a genuine Pre-Greek substratum word that can be reconstructed as *s yup- with a palatalised sibilant *s y . This palatal element was normally rendered with i, but in some cases, it was completely ignored, which would explain the above-mentioned variants without i. Note, however, that these forms are poetic variants, which may derive from a secondary synizesis. In any case, in this new interpretation by R. S. P. Beekes, the variation labial/velar, one of the main reasons for explaining si - gª as a ‘European’ loanword, was dropped. Note, moreover, that this scholar neglected to explain the development of *u into o in *s yup- > si % wpª, for which he could not assume an original form *s yop-, since, in his opinion, Pre-Greek originally only had the vowels a, i and u. 55 Recently, R. S. P. Beekes has revised his system of the Pre-Greek vowels, now assuming a system consisting of five vowels. 56 3.3. Pre-Greek labiovelars Even though R. S. P. Beekes now rejects an alternation between velar and labial for si - gª-si %wpª, this pair invites a comment on the vacillation between labial, velar and dental stops (cf., for instance, göfura, Boeot. böfura, Lacon. döfura “[Hom.] dam; [later] bridge”), for which he usually reconstructs a Pre-Greek labiovelar 57, the existence of which is attested, for instance, by Myc. qasireu /gwasileus/. Apparently, this 53. R. S. P. BEEKES (1996, p. 233-234). 54. R. S. P. BEEKES (2008, p. 52). 55. One explanation could be to assume analogical influence of words in -wp-, as R. S. P. Beekes suggested in his former reconstruction. 56. See R. S. P. BEEKES (2010, p. xix-xx). Incidentally, R. GORDEZIANI (2008, p. 33) suggested that i and u were vocalic allophones of 1 and 2 respectively. 57. R. S. P. BEEKES (1995/1996, p. 12-13), R. S. P. BEEKES (2002, p. 17) and R. S. P. BEEKES (2010, p. xxviii). THE PRE-GREEK LINGUISTIC SUBSTRATUM 267 phoneme developed differently from the PIE labiovelars 58, which is only plausible if the Greeks borrowed the substratum word after the evolution of the PIE labiovelars to labials or dentals, since the Linear B tablets show that Mycenaean Greek still knew the labiovelars. This implies that in PreGreek words that were borrowed into early Greek, the Pre-Greek labiovelars must have developed in the same way as the IE ones. Accordingly, words showing the variation labial/velar/dental, are not likely to have previously been borrowed in Mycenaean Greek. An enigmatic case is xÖfoj “sword”, attested in Mycenaean in the dual form qisipee, which unmistakably shows a labiovelar. R. S. P. Beekes 59 assumed that the cluster *k ws normally developed into y (e.g. M“yoj < Myc. Moqoso /Mokwsos/), which implies, of course, that qisipee is read as /kwsiphee/ (i.e. with a dummy i) and not as /kwisiphee/ (as R. S. P. Beekes suggested in his etymological dictionary 60), unless a (Pre-Greek?) syncope is assumed. According to R. S. P. Beekes, the labial element was lost through dissimilation against the following f. Alternatively, the initial consonantal cluster may have been reduced to ks for articulatory reasons. 4. ¢gaq“j The etymology of ¢gaq“j is highly controversial. The etymological explanations largely fall into two categories: one assumes the word to be of non-IE origin, whereas a second regards it as a compound based on IE elements. The following section explores both tendencies. 4.1. Non-IE interpretation 4.1.1. Aegean substratum word In his dissertation, E. J. Furnée 61 mentioned several Greek variants of later date, viz. ¢kaq“j (Hsch.) 62 , ¢kht“n (Hsch.) 63 , *¢kat“j ('AkatÖdhj) and ¢geq“j (Cypriot inscription 64). In these variants, he recognised the following Pre-Greek sound variations: g/k 65 , q/t and a/e. 58. The PIE labiovelars developed in Greek into labials (*k w > p, *g w > b, *g wh > f), except before front vowels, in which case they developed into dentals (*k w > t, *g w > d, *g wh > q). In contact with u or w, they were delabialised at an early stage. For the Pre-Greek labiovelars, these restrictions are lacking. 59. R. S. P. BEEKES (2002, p. 17). 60. R. S. P. BEEKES (2010, p. 1036-1037). 61. E. J. FURNÉE (1972, p. 124; p. 347). 62. ¢kaq“n· ¢gaq“n (Hsch.). 63. ¢kht“n· kr£tiston (Hsch.). 64. H. VAN HERWERDEN (1910, p. 9). 65. E. J. FURNÉE (1972, p. 106) interpreted this variation as an assimilation of g-q to k-q. 268 LES ÉTUDES CLASSIQUES In his review of this work, R. S. P. Beekes 66 added the gloss c£sioj (Hsch.) 67, which shows the variations g/c and q/s. The latter variation, however, is more likely to have resulted from Greek assibilation of the dental before i rather than from a Pre-Greek consonantal variation. 4.1.2. European substratum word ¢gaq“j resembles words of similar meaning in other IE languages, viz. Germ. *goda- “good, fitting” (Goth. gods “good”, OHG guot “id.”, MLG gaden “fit”), OCS god#n$ “pleasant”, Russ. gódnyj “useful”, which point to an underlying root *g hadh - “fit”. According to R. S. P. Beekes 68 , this reconstruction presents us with two problems. First, *g hadh - cannot develop into ¢gaq“j through regular Greek sound laws, since the expected outcome would be †k£qoj < *c£qoj (in accordance with Grassmann’s law). Second, this root contains the supposedly non-IE phoneme /a/. In the aprioristic conception by the Leiden IE School of the phoneme /a/, a Greek a can only go back to a PIE laryngeal or vocalised liquid. Although few words require the reconstruction of a PIE phoneme /a/, this does not mean that the phoneme did not exist at all. R. S. P. Beekes was right, however, to point out other irregularities in the Greek forms, viz. variation of the stops (g/k, q/t, g/c), variation of a/a and the prothetic vowel. Since these variations are ‘typical’ of the ‘European’ substratum 69, he traced ¢gaq“j with its IE parallels back to a European substratum word with the root *(a)ghadh -/*(a)ghad h - “good”. He justified the consonantal alternation by assuming that the ‘European’ aspirated stops were of a different quality than the PIE stops. In his etymological dictionary 70, by contrast, he was more cautious about the origin of ¢gaq“j, acknowledging that it can be either of IE origin or a substratum word. 4.1.3. Kartvelian substratum word In his later work, E. J. Furnée 71 identified ¢gaq“j (and its variants ¢kaq“j and ¢geq“j) as a ‘Pelasgian’ substratum word, i.e. a substratum word of Proto-Kartvelian origin. More precisely, he reconstructed the PreGreek word as *agat-/*aget-, analysing it as a combination of the Kartvelian prefix *ag- “upwards” and an adverbial formation in -ad-/-at-. This reconstruction seems rather random, however. While E. J. Furnée 66. R. S. P. BEEKES - A. H. KUIPERS (1975, p. 79). 67. c£sioj· ¢gaq“j, crhst“j (Hsch.). 68. R. S. P. BEEKES (1996, p. 227-230). 69. Incidentally, vowel prothesis does not belong to the ‘European’ features as enumerated by R. S. P. BEEKES (2000, p. 23-24). 70. R. S. P. BEEKES (2010, p. 7). 71. E. J. FURNÉE (1986, p. 131). THE PRE-GREEK LINGUISTIC SUBSTRATUM 269 supposed that the basic form was an adverb derived from a prefix, all other scholars have attempted to connect it with a (predominantly verbal) root. A Kartvelian hypothesis was also advocated by R. Gordeziani 72, who connected ¢gaq“j with Geo.-Zan k.et “good, kind”. The irregularity in this reconstruction is that Kartv. k. usually corresponds to Pelasg. k 73, unless some secondary development k > g is assumed (although in Greek, the form in k appears to be secondary, given its late attestation). A second problem is that R. Gordeziani did not explain the origin of the initial ¢-. Vowel prothesis, in the system of E. J. Furnée 74, on which R. Gordeziani’s relied, was a feature typical of Aegean (i.e. the non-IE, nonKartvelian substratum layer in the eastern part of the Mediterranean). 4.2. IE interpretation 4.2.1. Haplology: *aga-ghadh -oM. Harari 75 explained ¢gaq“j as developed from *¢ga-kaqoj through haplology. The first element, the intensifying prefix *¢ga-, is also found in such words as ¢ga-kleªj, ¢ga-kleit“j, ¢ga-klut“j “very famous”, ºg£-qeoj, Dor. ¢g£-qeoj “most holy”, ¢g£-nnifoj “much snowed on”, ¢g£-rrooj “strong-flowing”, etc. The second element developed through regular Greek sound laws from the PIE root *g hadh - “seize”. Apart from the above-mentioned Germanic and Slavic words, M. Harari also connected this root with OI gádhyaH “which has to be held on to”, a-gadhitaH “seized”, OSax. gigado “one’s equal”, Latv. gads “supply”, OCS god$ “right time”, u-goditi “please”. Thus, the meaning of *kaqoj evolved from “tangible” to “useful” and finally “good”. M. Harari interpreted these semantic shifts on the basis of the primitive hunter-gatherer culture. R. S. P. Beekes 76, however, found this reconstruction “most improbable” and rejected the connection between the Old Indic words (meaning “seize”) 77 and the Germanic and Slavic words (meaning “fit”). 4.2.2. Dissimilation: *sm-ghadh -o- or *n-ghadh -oIn his article on Grassmann’s law, A. Pârvulescu 78 also interpreted ¢- gaq-“j as a derivative of the verbal root *g hadh - “fit”. The first element was not explained as an intensifying prefix *¢ga- but as derived from PIE 72. R. GORDEZIANI (2008, p. 35). 73. E. J. FURNÉE (1979, p. 56). 74. E. J. FURNÉE (1979, p. 15; p. 32). 75. M. HARARI (1979). 76. R. S. P. BEEKES (1996, p. 230). 77. According to M. MAYRHOFER (1993, p. 461), the Old Indic words (related to Lat. prehendo “seize” and praeda “booty”) derive from PIE *g he(n)d-. 78. A. PÂRVULESCU (1993, p. 63-65). 270 LES ÉTUDES CLASSIQUES *sm- “with” or *n- “together with”, the combination of both elements resulting in *sm-ghadh - or *n-ghadh - “fitting together”. That g in ¢gaq“j should go back to PIE *g h is confirmed, in his view, by the variants c£sioj (with PIE *g h > Gr. c) and ¢kaq“j (with PIE *g h > Gr. c > k according to Grassmann’s law), both of which show the expected outcome. Consequently, A. Pârvulescu had to assume a dissimilation of the aspirates that already operated in PIE before Grassmann’s law, a phenomenon which apparently only left traces in Greek. R. S. P. Beekes 79 rightfully thought it unlikely that this early, ‘Pre-Grassmannian’ dissimilation would only have operated in Greek and not in Germanic. Moreover, there are practically no other instances of such a sound law in Greek 80 . 4.2.3. ghqöw: *sm-gh2 d h -oIn his complementary etymological dictionary, A. J. Van Windekens 81 also regarded ¢gaq“j as a derivative with the prefix *sm- (with intensifying value). The second element, however, he connected with ghqöw, Dor. ga - qöw “rejoice” (which he derived from PIE *geh2 d h -; cf. Toch. AB katk- “be happy”), assuming -gaq- to be derived from the zero-grade *gh2 d h -. *sm-gh2 d h -, he explained, developed into *°gaq“j and then into ¢gaq“j through dissimilation of the aspirates. The original meaning was thus “about which one rejoices”. Contrary to most of A. J. Van Windekens’ proposals, this reconstruction was applauded by M. E. Huld, one of Van Windekens’ critics 82. Note that it implies that the Germanic words are not related to ¢gaq“j, since in Germanic, PIE *g developed into k in accordance with Grimm’s law. 4.2.4. *h2 gadh -oLike A. J. Van Windekens, Rosemarie Lühr 83 also connected ¢gaq“j with Toch. AB katk-. The PIE root was reconstructed as *h2 gadh -, however, probably in order to explain the so-called prothetic vowel in 79. R. S. P. BEEKES (1996, p. 229). 80. A. PÂRVULESCU (1993, p. 65) saw parallels for this phenomenon in fe⁄gw “flee” < PIE *b he2g h -, qug£thr “daughter” < PIE *d hughh2 têr and qigeãn “touch” < PIE *d h iG h -. R. S. P. BEEKES (2010, p. 561; p. 1564-1565), however, reconstructed the first two words as *b he2g- and *d hugh2 -ter- respectively (i.e. without a second aspirated stop). qigeãn probably took the g from the present tense qigg£nein, where *g h might have lost the aspiration after a nasal: see R. S. P. Beekes (2010, p. 549). 81. A. J. VAN WINDEKENS (1986, p. 1). 82. M. E. HULD (1988, p. 467). 83. Rosemarie LÜHR (2000, p. 119-120). THE PRE-GREEK LINGUISTIC SUBSTRATUM 271 Greek. Contrary to A. J. Van Windekens, she did not mention a connection with ghqöw 84 . 4.2.5. mögaj: *mGh2 -dhóG.-J. Pinault 85 explained ¢gaq“j as a combination of the zero-grade of mögaj “big” and a Caland suffix *-d ho-. Apart from this suffix, he found PIE *meGeh2 - also connected with the following suffixes belonging to the Caland system: *-i- (Hitt. mekki- “big” < PIE *meGh2 -i-), *-es- (Ved. máhas- = Av. mazah- “size” < PIE *meGh2 -es-) and *-ent- (Ved. mahánt- = Av. mazant- “big” < PIE *meGeh2 -ent-, Gr. •gan “too much” < PIE *mGeh2 -nt-, Lat. ingens “enormous” < PIE *mGh2 -nt-). 4.2.6. mögaj: *mGh2 -d hh1 -óLike G.-J. Pinault, several other scholars have also identified the first element as the prefix *¢ga- (interpreted as the zero-grade *mGh2 - “big”), though now taking the second part as the zero-grade of the verbal root *d heh1 - “put, do” 86. J. W. Poultney 87 was the first to propose this reconstruction in a short article read at the Bopp-Symposium of 1992. According to C. J. Ruijgh 88, a similar compound is found in Latin, viz. magnificus < *magno-fak-o-s (with *fak- derived from PIE *d heh1 -k, cf. Gr. ú-qhk-a “I put”). Among the scholars who adopted this reconstruction, there has been discussion concerning the meaning of the compound, however. According to J. W. Poultney, C. J. Ruijgh, S. Scarlata and P. Ragot, ¢gaq“j originally had an active sense, viz. “große Taten wirkend” 89 , “Großes leistend” 90, “qui accomplit de grands actes” 91 , “dont les actes sont grands” < “qui s’applique à des actes de grande allure” 92 . O. Panagl and S. Neri, by contrast, assumed that it had a passive sense, viz. “hochgestellt”, “groß gemacht” 93. For both interpretations, reference was made to Homer’s use of ¢gaq“j, in which the word is 84. Recently, O. HACKSTEIN (2002, p. 8) explained ghqöw and Toch. AB katk- as compounds of *geh2 - “brilliance” and *d heh1 - “put”, thus, originally meaning “in Glanz versetzen”. According to R. VIREDAZ (2003, p. 115), however, an active compound in *d heh1 - is incompatible with the intransitive meaning of the verbs. 85. G.-J. PINAULT (1979; 1991). 86. The laryngeal h1 was lost between consonant and vowel. 87. J. W. POULTNEY (1994). 88. C. J. RUIJGH (1996, p. 379). 89. J. W. POULTNEY (1994, p. 210). 90. S. SCARLATA (1999, p. 260). 91. P. RAGOT (2006, p. 340). 92. C. J. RUIJGH (1996, p. 378; p. 388; p. 393). 93. See O. PANAGL (1995, p. 235) and S. NERI (2003, p. 48). Irene BALLES (2003, p. 16; 2006, p. 223) remained neutral as to the exact semantics and mentioned both interpretations. 272 LES ÉTUDES CLASSIQUES applied to noble heroes. According to O. Panagl, the word in Homer is still felt as a compound, since there are no compounds with ¢gaqo- yet in the Homeric epics. C. J. Ruijgh, on the other hand, claimed that already at an early stage, ¢gaq“j was no longer recognised as a compound: ¢ga- was no longer identified as the zero-grade of möga- and -q“- was no longer associated with the verb tÖ-qh-mi “I put”. As proof of this hypothesis, C. J. Ruijgh argued that already in Homer, compounds were made with mega- (e.g. meg£-qumoj “great-hearted”). 4.2.7. •gw: *aGn-dhh1 -oR. Anttila 94, finally, accepted the reconstruction of the second element as the verbal root *d heh1 - “put, do” but derived the prefix ¢ga- from the PIE root *h2 eG- “drive” instead of the zero-grade *mGh2 - “big” 95. In his opinion, the aga were games or contests (cf. ¢gËn “contest”), so that an adjective such as ¢gaklut“j “very famous” originally meant “famous with respect to the aga, the games”. In the derived superlative meaning, the prefix was also combined with words that were not connected the games (e.g. ¢g£nnifoj “much snowed on”). In ¢gaq“j, R. Anttila explained, the first element goes back to *aG-n “drove, herd”, an agrarian term which, he believed, was applied to people at a certain time (cf. the similar use of poÖmnh, ¢gölh, pÓu “herd”). Consequently, the original meaning of the compound *aGn-dhh1 -o-s was “supporting the aga, upholding the (social) unit”. Like J. W. Poultney, C. J. Ruijgh, S. Scarlata and P. Ragot, R. Anttila interpreted the verbal root in an active sense. The ¢gaqoÖ were the drivers or leaders of society, doing what was useful and therefore good for society. Thus, ¢gaq“j originally had a primarily social meaning. Via this social meaning, R. Anttila made the connection with the games (aga): the contester did not act as an individual but represented his group or family. 4.3. Conclusion A. Pârvulescu’s reconstruction (viz. an early dissimilation that only worked in Greek and that chronologically precedes Grassmann’s law) is the least convincing of all proposals, since it relies on controversial sound laws. Occasionally, the variants ¢kaq“j, ¢geq“j and c£sioj are adduced (viz. by E. J. Furnée, R. S. P. Beekes and A. Pârvulescu). Other linguists, however, ignored these forms and based their reconstructions solely on 94. R. ANTTILA (1996) and R. ANTTILA (2000, p. 70-75). 95. R. BRACCHI (1999, p. 89) accepted the reconstruction of the second element as PIE *d heh1 - but hesitated between *mGh2 - and *h2 eG- for the reconstruction of the first element. THE PRE-GREEK LINGUISTIC SUBSTRATUM 273 ¢gaq“j itself (e.g. M. Harari, A. J. Van Windekens, O. Panagl and R. Anttila). Indeed, the question arises whether these words are all related. Especially in the case of c£sioj, this claim seems doubtful. This word, which probably developed from *c£qioj (with assibilation of the dental before i) may be derived from PIE *g hadh - “to fit”, an interpretation which would enable a connection with the Germanic and Slavic words. If the root is reconstructed as *g heh2 d h - (in an attempt to avoid reconstructing a phoneme /a/, as the Leiden researchers are eager to do), the Slavic words cannot be related for phonological reasons 96 . ¢gaq“j itself is most likely a compound (*mGh2 -dhh1 -ó-s or *aGn-dhh1 -o-s), with *mGh2 -dhh1 -ó-s (either meaning “made great” or “doing great things”) as the most likely reconstruction from a semantic point of view. The variant ¢geq“j is probably a secondary form, arisen by (epichoric) dissimilation of the vowels. ¢kaq“j, finally, might be a contamination of ¢gaq“j and an unattested form *kaq“j (which, like c£sioj, might continue PIE *g hadh -). This last suggestion should be treated cautiously, of course, given the great risk involved in working with unattested forms. Although R. S. P. Beekes’ hypothesis of a ‘European’ substratum word showing several variants is often a reasonable suggestion, we should be careful not to label words as substratum words too easily. Indeed, E. J. Furnée’s dissertation was often criticised for listing several words that could be alternatively explained within Greek or PIE. In fact, the reconstruction *mGh2 -dhh1 -ó-s appears to be the common opinion on ¢gaq“j in comparative linguistic research 97 . 5. The suffix -nqoj A final case-study concerns the recent theories about the suffix -nqoj, which was used to form both place names (e.g. K“rinqoj, Z£kunqoj, 'Er⁄manqoj) and appellatives (e.g. ôröbinqoj “chick pea”, kol“kunqoj “gourd, pumpkin”). From the beginning of the substratum research, it was considered a Pre-Greek suffix, since it could not be derived from PIE through regular Greek sound laws. An important question is whether words in -nqoj form one homogeneous group and whether the suffix is related to the Anatolian suffix -nda (-nda), with which it is often connected. 96. In Balto-Slavic, a laryngeal is never vocalised between consonants; eh2 would give a. Therefore, in OCS god#n$ “pleasant” and Russ. gódnyj “useful” a laryngeal is excluded. See R. S. P. BEEKES (1996, p. 228). 97. See the recent works of M. JANDA (2000, p. 242) and Dagmar S. WODTKO - Britta IRSLINGER - Carolin SCHNEIDER (2008, p. 108; p. 470-471). 274 LES ÉTUDES CLASSIQUES 5.1. Identity between -nqoj and -nda A first group of scholars identified the suffixes -nqoj and -nda with each other but disagreed on their origin. L. R. Palmer 98 and Margalit Finkelberg 99 assumed these suffixes to be of Anatolian origin, considering Luwian as the most probable donor language for appellatives and place names. Y. Duhoux 100, though sceptical of any hypothetical Pre-Greek language, thought of one or more Anatolian (i.e. not necessarily Luwian) languages. According to him, -nda could be connected with PIE *-nt-, like -nqoj (with aspiration of the dental), although he neglected to explain the origin of the aspiration. In her monograph on the suffix -nqoj, Adriana Quattordio Moreschini 101 also concluded it to be of Anatolian origin. More specifically, she assumed the suffix -nq- to be derived from two different Anatolian suffixes, viz. -(a)nt- (used to build plurals and collectives) and -(a)nda/-(a)nta (a place name suffix). Because the singular morpheme -oj was added, the Greeks were no longer aware of the original collective meaning in appellatives. Thus, ôröbinqoj, for instance, should originally have denoted a collective of chick peas or the plant name in general and only later a single chick pea, an interpretation that seems needlessly complicated. Adriana Quattordio Moreschini, moreover, failed to explain the discrepancy between Anatolian d and Greek q. Should we assume that the Greeks perceived the Anatolian voiceless dental in -ant- as an aspirate? V. I. Georgiev 102 accepted the IE origin of both -nqoj and -nda but considered the former suffix to be of Pelasgian instead of Anatolian origin. He added that the suffixes -anqoj, -inqoj and -unqoj derived from different PIE forms, viz. -anq-oj < PIE *-ont- (with PIE *o > Pelasg. a and PIE *t > Pelasg. q), -inq-oj < PIE *-ent- (with PIE *e > Pelasg. i before nt without stress) and -unq-oj < PIE *-nt- (with PIE *n > Pelasg. un), whereas in Anatolian, all these variants developed into -ant-/-and-. This last remark cannot apply to PIE *-ent-, however, since in Hittite, PIE *e developed into e/i (not a). The proponents of the Aegean theory associated -nqoj and -nda as well but assumed both of these suffixes (i.e. also Anatolian -nda) to be of nonIE origin 103. Thus, R. A. Brown 104 considered -nqoj to be a prenasalised 98. L. R. PALMER (1980, p. 11-12). 99. Margalit FINKELBERG (2005, p. 52). 100. Y. DUHOUX (2007, p. 228). 101. Adriana QUATTORDIO MORESCHINI (1984, p. 107-108). 102. V. I. GEORGIEV (1981, p. 106). 103. H. HAARMANN (1996, p. 5) and K. STRUNK (2004, p. 89) agreed on the nonIE origin of -nqoj. 104. R. A. BROWN (1985, p. 12). THE PRE-GREEK LINGUISTIC SUBSTRATUM 275 variant of the Pre-Greek suffix -a - toj. R. S. P. Beekes 105 elaborated on this theory and connected -nqoj with a set of other not prenasalised suffixes. Thus, he connected -anq-/-inq-/-unq-, -and-/-ind-/-und- and -ant-/-unt106 with -aq-/-iq-/-uq-, -ad-/-id-/-ud- and -at-/ -it-/-ut-. In other words, the basic suffix had the form VC, which could be prenasalised and could show the Pre-Greek variation t/q/d. The proponents of the Kartvelian theory also accepted the association of the suffixes -nqoj and -nda. According to E. J. Furnée 107, these suffixes were of East Mediterranean origin (i.e. belonging to a non-IE, nonKartvelian substratum that left traces in both Greek and Kartvelian). More specifically, he connected -nq- with Georgian -nd- and supposed a secondary, ‘Aegean’ aspiration. Later, he considered the East Mediterranean -int- a secondary, ‘nasalised’ variant of -it-, corresponding to Kartvelian -et-/*-etj- 108, whereas Anatolian -ind- was explained as assimilated from *-int-. A. Uruschadse 109, on the other hand, connected the suffixes -nth-, -nd- and -nt- with Georgian -ian-ta, -an-ta and -ta (without the nasal element), which are used to build plurals and to express possession. 5.2. -nqoj and -nda: two different suffixes Other scholars have kept -nqoj and -nda etymologically separated, most of them agreeing on the genuine Anatolian nature of the suffix -nda (derived from PIE *-nt- and *-2ent-). Anna Morpurgo Davies 110 doubted that -nqoj could be a reflex of this Anatolian suffix, since the expected outcome would be -nt- and -nd- (not -nq-). C. Renfrew 111 too kept the two suffixes separated, explaining -nda as an Anatolian suffix and -nqoj as a Pre-Greek and Pre-Anatolian but not Pre-IE word. F. Lochner von Hüttenbach 112 considered the suffixes -wanda, -anda and -assa to be of IE origin, but since there are no traces of a Hittite-Luwian migration to Greece, he preferred to keep -nqoj separated from the similarly sounding suffixes in Asia Minor. In his contribution to Der neue Pauly, J. L. García-Ramón 113 considered three possibilities: these suffixes could derive from a common pre-IE substratum, go back to an extension of (IE or non105. R. S. P. BEEKES (2010, p. xxxiv). 106. The suffix *-int- is not attested. 107. E. J. FURNÉE (1982, p. 29). 108. E. J. FURNÉE (1986, p. 67-68). 109. A. URUSCHADSE (1984, p. 102-104). 110. Anna MORPURGO DAVIES (1986, p. 120). 111. C. RENFREW (1998, p. 254). 112. F. LOCHNER VON HÜTTENBACH (1985/1986, p. 7). 113. J. L. GARCÍA-RAMÓN (2002, p. 334). 276 LES ÉTUDES CLASSIQUES IE) linguistic material from Asia Minor, or reflect different developments of the same basic form. In contrast to the previous scholars, who frequently did not go beyond a negative conclusion (viz. that the Greek and Anatolian suffix could not be related), Françoise Bader 114 in her review of Adriana Quattordio Moreschini’s work suggested a positive etymological explanation of the suffix -nqoj. She believed that it derives from an inherited combination of a nasal and an aspirated dental (parallel complex suffixes being *-nt- and *-nd-). The dental could be connected not only with a nasal but also with *-i- (e.g. ‘rn-i - -q- “bird”) or *-u- (e.g. k“r-u-q- “helmet”). Consequently, Anatolian -nd- shows a different, though parallel, combination of a nasal with a dental. Another explanation of the suffix was proposed by M. Bernal in his controversial Black Athena 115, where the suffix was assumed to have various origins. First, it would go back to “simple introduction of a nasal before a dental” 116 (i.e. [pre]nasalisation), an explanation which, as J. H. Jasanoff and A. J. Nussbaum 117 argued in their review of the linguistic evidence in Black Athena, cannot be invoked as a regular Greek sound change without further justification. Second, M. Bernal assumed that some instances of -nqoj were renderings of Eg. -n»tr “holy”, which, in his view, was also borrowed as nÖtron “sodium carbonate”, •nqoj “flower” (with vowel prothesis and loss of final r), xanq“j “yellow” (< sn»tr “make holy”, with a causative s- and transcription of an ‘uncertain’ Egyptian sibilant as x), k£nqaroj “dungbeetle” (< *k n»tr “holy spirit”) and s£turoj “satyr” (< sn»tr, where the n was dropped). This second explanation does not seem likely either, since no traces of the so-called original meaning of -nqoj are preserved in the Pre-Greek toponyms and appellatives. 5.3. Conclusion Most scholars agreed on the Pre-Greek origin of -nqoj, except for M. Bernal and Françoise Bader. As the Anatolian suffix -nda can be explained on a PIE basis, it seems safer to keep both suffixes separated when discussing their etymology. If one wants to associate them, Françoise Bader’s explanation (i.e. IE origin for both suffixes, although the formations are not completely identical) seems the most likely suggestion. The strongest objection against the identification of both suffixes is the fact 114. Françoise BADER (1987, p. 232-236). 115. M. BERNAL (2001, p. 126-132). 116. M. BERNAL (2001, p. 128). 117. J. H. JASANOFF - A. J. NUSSBAUM (1996, p. 187). THE PRE-GREEK LINGUISTIC SUBSTRATUM 277 that in toponyms, they are never attested with the same root (for instance, we do not find Anatolian *Korinda besides K“rinqoj) 118 . 6. General conclusion A first conclusion to be drawn is that the Anatolian theory has proven to be the least influential. Its main proponents concentrated on the suffixes -nq- and -ss-. The other three theories (Pelasgian, Aegean and Kartvelian) had a wider scope, examining more data. Second, several words not only have competing Pre-Greek interpretations (as presented in my first article) but have also received explanations within PIE or Greek (e.g. t⁄mboj, si - gª and ¢gaq“j). For other words (e.g. pölekuj), an alternative borrowing could be considered (e.g. from Semitic) rather than substratum influence. Third, the Pelasgian theory, though influential in the past, failed to explain a large part of the data collected by E. J. Furnée. For most of the terms discussed in the present article, a Pelasgian suggestion could be made, but this is far from true for the rest of E. J. Furnée’s corpus (e.g. göfura - böfura - döfura). A fourth conclusion concerns the Kartvelian theory, an innovation in the substratum research. The proponents of this approach pointed to a few more or less systematic sound correspondences between Pre-Greek and Kartvelian, although the question of the original donor language often cannot be resolved. Moreover, not every Pre-Greek word can be explained as a Kartvelian substratum word, as becomes evident from the elements examined: there are no Kartvelian parallels for pölekuj, t⁄mboj and si - gª. Therefore, E. J. Furnée did not abandon the Aegean theory but incorporated it into his theory of Pre-Greek as consisting of two substrata, viz. a non-Kartvelian and a Kartvelian one. Still, the great chronological gap between ‘Pelasgian’ (which, according to E. J. Furnée 119, goes back to the third millennium BC) and the historically attested Kartvelian languages (with Georgian, attested from the fourth century BC onwards, as the only Kartvelian language with a historical tradition) invites us to be cautious. Unsurprisingly, K. H. Schmidt 120 was sceptical of the extent to which the Proto-Kartvelian material could still be reconstructed on the basis of the Kartvelian languages. Research on the basis of the Aegean theory (as currently practised by R. S. P. Beekes) seems the most fruitful approach. For several instances, it 118. A possible exception may be the pair Labraunda - Lab⁄rinqoj. 119. E. J. FURNÉE (1979, p. 14). 120. K. H. SCHMIDT (1979, p. 96). 278 LES ÉTUDES CLASSIQUES can be supplemented with the Kartvelian theory. Still, a few critical remarks are in order, first concerning the criteria used by the linguists of Leiden in the identification of substratum elements 121, which are the following: (1) Absence of a good IE etymology (2) Limited geographical distribution: the word in question only occurs in one language (group) (3) Unusual word formation: suffixes unknown in PIE (4) Meaning: loanwords for certain local phenomena (5) Phonological irregularity: ablaut patterns impossible in PIE Objections can be made to each of these criteria. First, the limited geographical distribution of a certain word does not necessarily point to a non-IE origin, since it may have been lost in other languages. This might be the case for pölekuj, which is only attested in Greek and Old Indic, although these languages may have borrowed the word separately from Semitic. Second, certain phonological irregularities can be explained by assuming assimilation, dissimilation, onomatopoeia, taboo or contamination. An onomatopoeic value may be present in si - gª, whereas contamination and dissimilation may have given rise to ¢kaq“j and ¢geq“j respectively (two variants of ¢gaq“j). Third, a so-called non-IE suffix may be a complex IE suffix cluster, for instance in the case for -nqoj (as Françoise Bader suggested), although most scholars agreed on its PreGreek nature. Fourth, the semantic argument is not sufficient to label a certain word as non-IE either, since our knowledge of the Indo-European culture is still fairly limited. Finally, the presence of the so-called non-IE phoneme *a is also a debatable issue. It is one of the main reasons for R. S. P. Beekes to assume that ¢gaq“j is of non-IE origin. Most modern scholars, however, do accept the existence of this phoneme in PIE. Although each of these criteria can thus be challenged, a non-IE origin is usually made plausible not by simply one of these criteria but by the combination of these features, as P. C. H. Schrijver 122 argued. Another problem is that R. S. P. Beekes’ methodology can often be questioned. He adduced many Hesychian glosses as so-called proof of the Pre-Greek sound variations, although nothing guarantees the antiquity of these glosses: some of them may even be recent (perhaps Byzantine) loanwords. Furthermore, many borrowings that presumably came from a different source (e.g. Semitic) are subsumed under the Pre-Greek 121. For a more detailed discussion of these criteria, see P. C. H. SCHRIJVER (1997, p. 293-296). See also R. S. P. BEEKES (1999, p. 14) and R. S. P. BEEKES (2010, p. xxiii). 122. P. C. H. SCHRIJVER (1997, p. 296). THE PRE-GREEK LINGUISTIC SUBSTRATUM 279 vocabulary. Therefore, as a suggestion for future research, a stricter methodology seems in order, in which a distinction should be made at least between early attested Pre-Greek words, Hesychian glosses and loanwords that may be of a different origin. One last crux is the unity of Pre-Greek, which was assumed by R. S. P. Beekes. The wide geographical distribution of the Aegean substratum, however, makes a linguistically diverse continuum (perhaps even comprising IE elements) more plausible. Moreover, given the chronological gap between the various attested words, diachronic variation too remains possible. Finally, dialectal differentiation within Pre-Greek cannot be a priori excluded. Any speaker of a natural language will agree that the dialects of his language can show extreme differences. Retrieving these Pre-Greek dialects, however, may forever be beyond our grasp. Gertjan VERHASSELT KU Leuven - Faculty of Arts Research Unit: Ancient History Blijde Inkomststraat 21 box 3307 3000 Leuven (Belgium) gertjan.verhasselt@arts.kuleuven.be 280 LES ÉTUDES CLASSIQUES Literature Raimo ANTTILA (1996): “'Agaq“j Again”, HS 109, p. 237-239. Raimo ANTTILA (2000): Greek and Indo-European Etymology in Action. ProtoIndo-European *aG-, Amsterdam, Benjamins. Françoise BADER (1975): Review of E. J. FURNÉE (1972), RPh 49, p. 102-106. Françoise BADER (1987): Review of Adriana QUATTORDIO MORESCHINI (1984), BSL 82, 2, p. 227-236. Irene BALLES (2003): “Die lateinischen Adjektive auf -idus und das Calandsystem”, in Eva TICHY, Dagmar S. WODTKO & Britta IRSLINGER (ed.), Indogermanisches Nomen. Derivation, Flexion und Ablaut. Akten der Arbeitstagung der Indogermanischen Gesellschaft, Society for IndoEuropean Studies, Société des Études Indo-Européennes. Freiburg, 19. bis 22. September 2001, Bremen, Hempen, p. 9-29. Irene BALLES (2006): Die altindische Cvi-Konstruktion. Form, Funktion, Ursprung, Bremen, Hempen. Robert S. P. BEEKES (1996): “Ancient European Loanwords”, HS 109, p. 215- 236. Robert S. P. BEEKES (1999): Indo-Europees en niet-Indo-Europees in het Nederlands, Leiden, Onderzoekschool CNWS. Robert S. P. BEEKES (2000): “European Substratum Words in Greek”, in M. OFITSCH & C. ZINKO (ed.), 125 Jahre Indogermanistik in Graz, Graz, Leykam, p. 21-31. Robert S. P. BEEKES (2002): “Hom. göfura, and Arm. kamurj ‘bridge’”, Glotta 78, p. 12-21. Robert S. P. BEEKES (2008): “Palatalized Consonants in Pre-Greek”, in A. LUBOTSKY (ed.), Evidence and Counter-Evidence. Essays in Honor of Frederik Kortlandt. Volume 1. Baltoslavic and Indo-European Linguistics, Amsterdam - New York, Rodopi, p. 45-56. Robert S. P. BEEKES (2010): Etymological Dictionary of Greek, 2 vol., Leiden - Boston, Brill. Robert S. P. BEEKES - Aert H. KUIPERS (1975): Review of E. J. FURNÉE (1972), Lingua 36, p. 69-85. Martin BERNAL (2001): Black Athena Writes Back. Martin Bernal Responds to His Critics, Durham - London, Duke University Press. Martin BERNAL (2006): Black Athena. The Afroasiatic Roots of Classical Civilization. Volume III. The Linguistic Evidence, New Brunswick - New Jersey, Rutgers University Press. Remo BRACCHI (1999): “Turdus e i suoi corrispondenti. L’uccello ‘che cova nel fango inaridito’”, Athenaeum 87, p. 79-92. Raymond A. BROWN (1985): Evidence for Pre-Greek Speech on Crete from Greek Alphabetic Sources, Amsterdam, Hakkert. Pièrre CHANTRAINE (1968): Dictionnaire étymologique de la langue grecque. Histoire des mots, Paris, Klincksieck. THE PRE-GREEK LINGUISTIC SUBSTRATUM 281 James CLACKSON (1994): The Linguistic Relationship between Armenian and Greek, Oxford, Blackwell. James CLACKSON (2007): Indo-European Linguistics. An Introduction, Cambridge, University Press. Yves DUHOUX (2007): “Pre-Greek Languages. Indirect Evidence”, in A.-F. CHRISTIDIS (ed.), A History of Ancient Greek. From the Beginning to Late Antiquity, Cambridge, University Press, p. 223-228. Margalit FINKELBERG (2005): Greeks and Pre-Greeks. Aegean Prehistory and Greek Heroic Tradition, Cambridge, University Press. Hjalmar FRISK (1972): Griechisches etymologisches Wörterbuch, Heidelberg, Winter. Edzard J. FURNÉE (1972): Die wichtigsten konsonantischen Erscheinungen des Vorgriechischen mit einem Appendix über den Vokalismus, The Hague - Paris, Mouton. Edzard J. FURNÉE (1979): Vorgriechisch-Kartvelisches. Studien zum ostmediterranen Substrat nebst einem Versuch zu einer neuen pelasgischen Theorie, Leuven, Peeters. Edzard J. FURNÉE (1982): Beiträge zur georgischen Etymologie. Faszikel 1. Georgisch-vorgriechische, georgisch-vorromanische und georgischvorindogermanische Materialien, Leuven, Peeters. Edzard J. FURNÉE (1986): Paläokartvelisch-pelasgische Einflüsse in den indogermanischen Sprachen. Nachgewiesen anhand der spätindogermanisch-griechischen Reflexe urkartvelischer Sibilanten und Affrikaten, Leiden, The Hakuchi Press. Thomas V. GAMKRELIDZE - Vjaceslav V. IVANOV (1995): Indo-European and the Indo-Europeans. A Reconstruction and Historical Analysis of a ProtoLanguage and a Proto-Culture. Part I. The Text, Berlin - New York, De Gruyter. José L. GARCÍA-RAMÓN (2002): “Vorgriechische Sprachen”, in H. CANCIK & H. SCHNEIDER (ed.), Der neue Pauly. Enzyklopädie der Antike 12, 2, Stuttgart - Weimar, Metzler, p. 331-334. Vladimir I. GEORGIEV (1981): Introduction to the History of the Indo-European Languages, Sofia, Bulgarian Academy of Sciences. Rismag GORDEZIANI (2008): Mediterranea-Kartvelica IV. Summary. Abbrevations. Indices. Addenda et Corrigenda, Tbilisi, Logos. Harald HAARMANN (1996): “Aspects of Early Indo-European Contacts with Neighboring Cultures”, IF 101, p. 1-14. Olav HACKSTEIN (2002): “Uridg. *CH.CC > *C.CC”, HS 115, p. 1-22. Ivo HAJNAL (2005): “Das Frühgriechische zwischen Balkan und Ägäis. Einheit oder Vielfalt”, in G. MEIER & O. HACKSTEIN (ed.), Sprachkontakt und Sprachwandel. Akten der XI. Fachtagung der Indogermanischen Gesellschaft. 17.-23. September 2000. Halle an der Saale, Wiesbaden, Reichert, p. 185-214. Eric P. HAMP (1983): “Prehellenica. 3. stifr“j, stãfoj”, A 33, 2, p. 147-148. Maurizio HARARI (1979): “Nuova proposta di etimologia per ¢gaq“j”, Athenaeum 57, p. 150-153. 282 LES ÉTUDES CLASSIQUES Henricus VAN HERWERDEN (1910): Lexicon Graecum suppletorium et dialecticum, Leiden, Sijthoff. Alfred HEUBECK (1974): Review of E. J. FURNÉE (1972), IF 79, p. 272-277. Martin E. HULD (1988): Review of A. J. VAN WINDEKENS (1986), AJPh 109, 3, p. 463-467. Michael JANDA (2000): Eleusis. Das indogermanische Erbe der Mysterien, Innsbruck, Universität Innsbruck. Institut für Sprachwissenschaft. Jay H. JASANOFF - Alan J. NUSSBAUM (1996): “Word Games. The Linguistic Evidence in Black Athena”, in Mary R. LEFKOWITZ & G. MACLEAN ROGERS (ed.), Black Athena Revisited, Chapel Hill, University of North Carolina Press. Fritz LOCHNER VON HÜTTENBACH (1985/1986): “Hellenen und Athener. Überlegungen zu Sprachwissenschaft und Frühgeschichte”, GB 12/13, p. 1-20. Rosemarie LÜHR (2000): Die Gedichte des Skalden Egill, Dettelbach, Röll. Manfred MAYRHOFER (1992): Etymologisches Wörterbuch des Altindoarischen, I, Heidelberg, Winter. Manfred MAYRHOFER (1996): Etymologisches Wörterbuch des Altindoarischen, II, Heidelberg, Winter. Anna MORPURGO DAVIES (1986): “The Linguistic Evidence. Is There Any?”, in G. CADOGAN (ed.), The End of the Early Bronze Age in the Aegean, Leiden, Brill, p. 93-123. Sergio NERI (2003): I sostantivi in -u del gotico. Morfologia e preistoria, Innsbruck, Institut für Sprachen und Literaturen. Abteilung Sprachwissenschaft. Leonard R. PALMER (1980): The Greek Language, London - Boston, Faber and Faber. Oswald PANAGL (1995): “Griechisch ¢gaq“j. Ein etymologischer Versuch”, in H. HETTRICH, W. HOCK, P.-A. MUMM & N. OETTINGER (ed.), Verba et Structurae. Festschrift für Klaus Strunk zum 65. Geburtstag, Innsbruck, Universität Innsbruck. Institut für Sprachwissenschaft, p. 229-235. Adrian PÂRVULESCU (1993): “IE. dhugh5têr ‘daughter’ and Grassmann’s Law. A Phonetic and Semantic Analysis”, IF 98, p. 50-91. Georges-Jean PINAULT (1979): “Grec ¢gaq“j”, MSS 38, p. 165-170. Georges-Jean PINAULT (1991): “Grandeur et excès. Avatars du morphème ¢gadans le lexique et le discourse”, RPh 65, p. 195-218. Georges-Jean PINAULT (1994): “Les deux formes du silence homérique et l’origine du verbe siwp£w”, in D. CONSO, Nicole FICK & B. POULLE (ed.), Mélanges François Kerlouégan, Paris, « Les Belles Lettres », p. 501-526. Julius POKORNY (1969): Indogermanisches etymologisches Wörterbuch, Bern, Francke. James W. POULTNEY (1994): “Die Etymologie von ¢gaq“j”, in R. STERNEMANN (ed.), Bopp-Symposium 1992 der Humboldt-Universität zu Berlin, Heidelberg, Winter, p. 208-211. Adriana QUATTORDIO MORESCHINI (1984): Le formazioni nominali greche in -nth-, Rome, Edizioni dell'Ateneo. THE PRE-GREEK LINGUISTIC SUBSTRATUM 283 Pièrre RAGOT (2006): “¢gaq“j”, RPh 80, p. 340. Colin RENFREW (1998): “Word of Minos. The Minoan Contribution to Mycenaean Greek and the Linguistic Geography of the Bronze Age Aegean”, CAJ 8, 2, p. 239-264. Cornelis J. RUIJGH (1996): “L’étymologie de l’adjectif ¢gaq“j”, in A. RIJKSBARON & F. M. J. WAANDERS (ed.), Scripta Minora ad linguam graecam pertinentia, II, Amsterdam, Gieben, p. 378-395. Salvatore SCARLATA (1999): Die Wurzelkomposita im Àg-Veda, Wiesbaden, Reichert. Peter C. H. SCHRIJVER (1997): “Animal, Vegetable and Mineral. Some Western European Substratum Words”, in A. LUBOTSKY (ed.), Sound Law and Analogy. A Festschrift for Robert S. P. Beekes, Amsterdam - Atlanta, Rodopi, p. 293-316. Andrew L. SIHLER (1995): New Comparative Grammar of Greek and Latin, Oxford, University Press. Klaus STRUNK (2004 [= 2003]): “‘Vorgriechisch’/‘Pelasgisch’. Neue Erwägungen zu einer älteren Substrathypothese”, in A. BAMMESBERGER & T. VENNEMANN (ed.), Languages in Prehistoric Europe, Heidelberg, Winter, p. 85-98. Gábor TAKÁCS (1998): “Afro-Asiatic (Semito-Hamitic) Substratum in the ProtoIndo-European Cultural Lexicon?”, Lingua Posnaniensis 40, p. 141-172. Gertjan VERHASSELT (2009): “The Pre-Greek Linguistic Substratum. An Overview of Current Research”, LEC 77, p. 211-239. Rémy VIREDAZ (2003): “ghqöw”, RPh 77, p. 115. Albert J. VAN WINDEKENS (1986): Dictionnaire étymologique complémentaire de la langue grecque, Leuven, Peeters. Dagmar S. WODTKO - Britta IRSLINGER - Carolin SCHNEIDER (2008): Nomina im indogermanischen Lexikon, Heidelberg, Winter.