CC BY-NC-ND
42
Jadranka Gvozdanovic University of Heidelberg
On the linguistic classification of Venetic
Venetic and its variants in prehistoric and early historic Europe are still enigmatic to historical linguists. Traditionally, it is assumed that Venetic of Armorica was probably Celtic, but Venetic of the Northern Adriatic, for which we have written evidence, is assumed to have constituted a separate branch of Indo-European. The least evidence is known for East European Venetic. This paper discusses the available evidence, particularly on Venetic of the Northern Adriatic, in the light of linguistic data on other branches of Indo-European, such as Celtic and Slavic. Special attention is paid to the relations between Venetic and Continental Celtic, revealing some new aspects.
Keywords: Venetic language, Celtic languages, Indo-European linguistics, areal linguistics.
0. Introduction
By historical evidence, at times preceding and following the turn of the first millennium AD, there were ancient tribes called Veneti along the Vistula up to the Baltic Gulf in the East, in Armorica in the West, and along the Northern Adriatic in the South of Europe. Of these, only the Northern Adriatic Veneti left a set of mainly votive and funerary texts, classified by Le-jeune (1974: 21) with some approximation into periods called Archaic (550-475 BC), Ancient (475-300 BC), Recent (300-150 BC), and Veneto-Latin (150-100 BC). The boundary between the first and the second period was marked by the introduction of syllabic punctuation; the boundary between the second and the third period — by laryngeal changes (hi > h2); and the transition between the third and the fourth period by a change of alphabet. The Venetic alphabet of the older periods was based on Etruscan, but with an additional provision for writing the voiced consonants /b/, /d/, /g/ (i.e. the mediae) not known to Etruscan.1
There is no doubt that Venetic was an Indo-European language. This was recognized by Carl Pauli (1885) on the basis of case endings and weak/strong Ablaut alternations. Beeler (1949) classified Venetic as Indo-European by using predominantly morphological considerations, such as:
a) inflectional endings -os, -oi, -on (like the nominative, dative and accusative singular), and a set of inflectional endings -a, -as, -ai (like the nominative, genitive and dative singular);
b) the ending -to, presumably 3rd person singular of the middle voice (cf- Greek -to);
c) derivational suffixes, such as -o-, -no-, -so-, -tor-;
d) lexical correspondences, such as Venetic rXo — Latin ego, Venetic meXo — Gothic mik etc.;
e) characteristically Indo-European vowel alternations.
1 Following the rules of syllabic punctuation, represented from ca. 350 BC on, punctuation was used to separate final consonants from the preceding phonological units, but it was never present between a consonant or a consonant group (consisting of an obstruent and a sonorant) and a following vowel (either a monophthong or a diphthong). Thus, punctuation reveals a syllable template of the form C(C)V(C), which underwent changes due to weakening and loss of h- in archaic Venetic (e.g. *ho.s.ti.s. > *.o..s.ti.s. 'guest') and syncope of i preceding a final fricative (i.e. *.o..s.ti.s. > .o..s.t..s., cf. Lejeune 1974: 39) in ancient Venetic.
Journal of Language Relationship • Вопросы языкового родства • 7 (2012) • Pp. 33-46 • © Gvozdanovic J., 2012
These morphological and phonological similarities are more straightforward than lexical peculiarities, but together, all the characteristics enabled a clear classification of the language as Indo-European (cf. also Campbell — Poser 2008: 83ff.).
Within Indo-European, Venetic is predominantly grouped with the western languages, but details still remain to be clarified (cf. also Untermann 1980). Venetic was a centum language (cf. Venetic ke < IE *k'e 'and', and IE *kwe > Venetic -kve 'and'). As a prominent and relatively specific phonological feature, there was an a-reflex of IE laryngeals between consonants with parallels in Italic and Celtic, e.g. Venetic vha.g.sto 'made (an offering)', Latin facit, Oscan fakiiad from zero-grade of *dhehi with -k- extension (IE *dhhi-k-) (cf. Wallace 2008: 126).
In morphology, Venetic probably shared with Latin and Celtic the o-stem genitive singular ending -i (e.g. keutini 'of Keutinos').2 There was also a third-person-singular ending in -r, similar to the medio-passive in Hittite, Latin and Celtic, used either in agentless passive sentences or as a deponent (allowing for an agent and a transitive usage), e.g. Venetic tuler donom 'brings/brought (?) a gift (as an offering)'. In addition, there was a medium ending -to, according to Untermann (1980: 292), attested in Greek and Old Indian, but also — we should add — in Celtiberian.
Prominently discussed in the literature was the change of the word-initial mediae aspiratae into fricatives, i.e. > f, *dh > f, *gh > h, shared between Venetic, Oscan-Umbrian and Latin. As subtly stated by Lejeune (1974: 171-3), if this trait were to be taken as a defining trait of Italic, then Venetic would have to be grouped with Italic.3 But Lejeune (1974: 171) also allowed for the possibility that Venetic may have been a transitional type between Celtic and Italic, with the same treatment of sonorants as in Celtic, and the same treatment of the mediae aspiratae and the voiceless occlusive labial (*p) as in Italic.
We shall take up the hypothesis about Venetic as a transitional language between Celtic and Italic and examine it in more detail here. Our approach will re-examine Lejeune's and others' arguments for classifying Venetic and re-evaluate them in view of additional linguistic evidence.
1. Venetic and Italic
Discussions about the linguistic classification of Venetic among the western Indo-European languages gear towards two major solutions: either treating Venetic as a relatively archaic Indo-European language with some similarities, but not unequivocally attributable to any other western Indo-European group (cf. especially Krahe 1950 and Polome 1966), or assuming a close connection with Italic (especially Beeler 1949, or Euler 1993, who considers Venetic to be closely connected with Italic, but with archaic morphological traits). Essentially, all researchers point to significant morphological differences between Venetic and Italic, in contrast to several specific phonological and lexical similarities.
The assumption of a close genetic relation between Venetic and Italic is based on considerations of historical phonology. In Venetic, as in Latin, IE and *dh generally became voiceless fricatives in word-initial position (with some exceptions, to be mentioned below), and voiced stops medially (cf. Ven. vha.g.s.to, Lat. facio, Ven. vhratere 18, Lat. frater from initial *dh
2 Untermann (1961) proposed to analyse such forms as ending in -ei, which would make them datives.
3 Lejeune stated that Venetic had a significant majority of shared traits with Western Indo-European and none exclusively shared with Eastern Indo-European (even Venetic murtuvoi, 75 ter, cognate with Slavic murt-v-, also had a cognate in Latin mort-u-us), although possible examples would be the toponyms Opitergium and Tergeste, which are related to the Slavic root in the meaning of 'market(-place)'.
and *bh, resp., but medially, e.g., Ven. lo.u.derobo.s. dat.pl. from an IE form containing *dh and *bh, respectively). However, in contradistinction to Latin, the Italic languages Faliscan and Osco-Umbrian had the same reflexes initially and medially (cf. Lejeune 1974: 166). IE *gh yielded Ven. and Lat. h, e.g. Ven. ho.s.ti-havo.s 187, Lat. hostis.
The labiovelar *kw was preserved in Venetic and in Latin (e.g. in the enclitic conjunction Ven. -kve, Lat. -que 'and'), the palatal labiovelar was depalatalized (Ven. e.kvo.n. Lat. equuum 'horse, gen.pl.'), and the voiced labiovelar lost the stop component (e.g. *g™iwos, Ven. vivoi 'alive, dat.sg.', cf. Lat. vivus).4 The same development of labiovelars was attested in Celtiberian and Goidelic, as opposed to labialization of labiovelars in Osco-Umbrian, Ligurian, Lepontic, Gaulish and Brythonic. There was, however, some variation in Gaulish, and the Gaulish Calendar of Coligny has eqvos 'horse-month', exhibiting the same development as Venetic. Further variation can be observed concerning word-final merger of nasals (-m, -n > -n) in Venetic (e.g. re.i.tia.n., Este 30, cf. Lejeune 1974: 140) with parallels in Gaulish and Insular Celtic (also in Messapic and Germanic), but different from Latin-Faliscan, Osco-Umbrian, Lepontic and eastern Celtiberian.
The phonological evidence points to similarities with Celtic next to similarities with Latin which are not fully shared by the other Italic languages. In addition, as concluded by Beeler (1981: 70) concerning Venetic and Latin, "the close agreement in the phonological area is not matched in the lexicon and morphology", whereas — as he correctly added — "when we look at these areas in language families of moderate time depth — say 1500 — 2500 years, such as Romance, Germanic, and Slavic — what we find is a substantial measure of sameness in all areas".
Virtually all the specific morphological features of Venetic differ from Latin, and those which would seem similar, exhibit a different distribution of the relevant forms.
Superficially, the -*bhos ending of the dative plural seems to be a common feature of Venetic and Italic. However, this dative plural ending (e.g. lo.u.derobo.s.) occurs only in the paradigm of the IE -o- stems in Venetic, which is precisely the paradigm in which (cf. Polome 1966: 74) it never occurs in Italic (in Messapic, there are several occurrences of -bas < *-bhos in -a- stems).
The Venetic genitive singular of -o-stems in -i (cf. examples such as ENONI) evoked some doubt concerning a possible reading as *ENONEI (cf. Polome 1966: 74f.), but if accepted, it is paralleled by the same ending in Celtic (e.g. Old Irish) and in Latin, though not in Osco-Umbrian (cf. Beeler 1981: 66), with no equivalent elsewhere.
The Venetic patterning of the accusative singular of the 1st person pronoun on the nominative (i.e. acc.sg. mego, nom.sg. e.go) is found also in Germanic (cf. Gothic ik, mik), but not shared by Italic.
In the verbal system, Venetic had a preterite in -s- (cf. Ven. vha.g.s.to, Lat. fecit, Ven. do.na.s.to, Lat. donavit/ donabat, Ven. do.na.sa.n, Lat. donaverunt), to be compared with the Latin perfect, but not with the Latin -w- or imperfect -ba- suffixes (cf. Beeler 1981: 66).
Venetic had a 3rd person singular ending in -to (e.g. do.na.s.to, assumed to have been combined into a single paradigm with a 3rd person plural ending -nt, cf. Ven. donasan < *donasnt). In do.na.s.to, there is a denominative suffix -a- preceding -s- and -to, which is unknown to Latin, but attested in Celtic, Slavic and Greek (cf. Untermann 1980: 294). Venetic had a 3rd person singular middle voice with reduced grade, comparable to Gathic Avestan baxsta, Sanskrit ayukta (*ks-t > kt, cf. Polome 1966: 75) and also Celtic and Slavic.
The Venetic medio-passive 3rd person singular ending in -r (e.g. tole.r/tola.r 'sacrificed + object, cf. Lejeune 1974: 76' <Gurina/Lagole>, dido.r 'gave') has correlates in Celtic, Umbrian,
4 In the Upper Piave valley, however, Venetic retained the distinction between -n and -m.
Anatolian, Tocharian (cf. Beeler 1981: 66), but it is somewhat enigmatic, since it combines with the present stem, yet occurs in texts where a past tense would be expected. In comparison with Latin, there is a significant difference, because (as shown by Beeler 1966: 66) -r is added in Latin to a verbal segment already inflected for person, whereas Venetic adds -r to the verbal stem, not to the already inflected verb. Moreover, the Venetic forms with -r seem to have been used as either deponent or active forms, which led Untermann (1980: 292f.) to assume a promiscuous usage in Venetic.
Lexical evidence is inconclusive. As mentioned by Polome (1966: 76), Venetic a.i.su 'god' may be an Etruscan word, and Venetic lo.u.dero may be a result of Roman influence, transmitting the Latin word liberi for 'children'. Other lexical isoglosses do not point to clear correspondences with Italic either. Patterning of the adjective for 'dead', Ven. murtuvoi ('dead', cf. Lat. mortuus) on the adjective for 'alive' (IE *gwi-wo-) is also found in Old Church Slavic mritvu. The reflexive form SSELBOISSELBOI 'ipsi' has a parallel in Old High German selbselbo, but none in Latin. Venetic teu.ta 'people' lacks parallels in Latin, Slavic and Greek (cf. Beeler 1981: 67), but has a clear Gaulish correlate in teuta, touta 'tribe, people' (cf. Delamarre 2003: 295).
In the end, only the mediae aspiratae and the preservation of *p remained as arguments for a Venetic-Italic genetic relationship. Beeler (1981: 70) assumed that these similarities can hardly be attributed to diffusion or borrowing, because "at the opening of the historical period, the two languages are not spoken in contiguous territories".
But can these changes be dated to the opening of the historical period?
It is not likely that the mediae aspiratae would be decisive, because Italic is not homogeneous in this respect. Even worse: not even Venetic itself is fully homogeneous. As pointed out by Polome (1966: 72), there is also onomastic evidence of initial IE *bh having yielded b- as in Bo.i.iio.s., in addition to such doublets as Lagole FVTVS: butialos. Such attestations give rise to the hypothesis that these similarities may have arisen as contact phenomena between Venetic and Latin, not shared by the other Italic languages (especially in view of non-differentiation between initial and medial position in Osco-Umbrian discussed by Euler 1993: 98). At the same time, doublets lead to the assumption of language contacts.
Venetic preserves IE *p (e.g. in PATER 'father') in contradistinction to Celtic, where new labials result from labialization of labiovelars in parts of Gaulish and Brythonic (e.g. in IE *penkwe 'five' > kwenkwe (due to assimilation) > Middle Breton pemp, Middle Welsh pymp, Middle Cornish pimp), but not so in Goidelic (e.g. OIr. coic). 5
The loss of IE *p in Celtic is assumed to have followed the *p...kw > kw...kw assimilation attested in Celtic and Italic, and preceded the labialization of the labiovelars in parts of Celtic and Italic. Venetic occupies an intermediate position by preserving *p and having not only the assimilation *p...kw > kw...kw, but also its mirror image, namely *kw...p > p...p, attested in Venetic e.kvopetari.s > e.kupetari.s (EQVPET) > e.p.petari.s > e.petari.s 'cavallier' (Lejeune 1974: 121). This shows that the velarity of the labiovelar was dismissible in Venetic, whereas the labiality was always preserved and thereby basic. By this phonological hierarchy, Venetic exhibits similarity with Continental Celtic, in spite of preserving IE *p.
Concerning the time depth of the emergence of the relevant phenomena discussed above, especially the treatment of the original mediae aspiratae *bh, *dh, *gh, *gwh, Kortlandt (2007: 150f.) assumed that they became fricatives when *b, *d, *g, *gw lost their glottalization in early Proto-Italic. He dates the Latin merger of initial voiceless fricatives into f- "probably not before the internal voiced fricatives merged with -b-, -d- and -w- after the separation from Faliscan". It is thus likely that the crucial development of the initial mediae aspiratae happened as an areal
5 Labialization of labiovelars was attested in Gaulish, Lepontic, Brythonic, Ligurian and Osco-Umbrian.
phenomenon on the Apennine Peninsula, and, thus, at a time when Venetic could participate in this areal change. Doublets which run counter the common isoglosses between Venetic and Latin such as FVTVS: butialos (Untermann 1961: 90) or the tribal name Bo.i.iio.s. (from the root *bhey 'to strike' according to Polome 1966: 72) make it plausible to assume that the phonological similarities between Venetic and Latin may also have arisen due to areal contacts. This is underlined by the fact that similarities of Venetic with Latin are more far-reaching than with the other Italic languages.
2. Venetic morphology in relation to Celtic
The Venetic dative plural of -o-stems in -(o)bos (e.g. lo.u.derobo.s, Este 26 'to/with (the) children'), for which Latin has -is (or -eis in castreis), Oscan has -ois and Umbrian has -es, er (e.g. vesicles 'vasculis'), has direct parallels in Celtiberian arecoraticubos (Luzaga inscription), Le-pontic uvltiauiopos (Prestino) and Gaulish atrebo aganntobo (Plumergat), as already mentioned by Euler (1993: 100f.).
Also, the Venetic genitive singular of -o-stems in -i (i.e. -i) has direct correspondences in Celtic (cf. Euler 1993: 104), e.g. Gaulish Segomari, Dannotali (PN) as well as in Early Irish MAQI (Old Irish maic 'filii').
The Venetic accusative singular of the 1st person pronoun mego was patterned on the nom.sg. ego with parallels outside Italic (cf. Gothic ik, mik, Hittite uk, ammuk etc.), whereas Italic presumably preserved the old forms (cf. Lat. me, te, se). Also, the reflexive form SSEL-BOISSELBOI has a parallel in Old High German selbselbo in Notker's translation of psalms (77,54 and 4,9), whereas Lat. ipse has Italic correspondences in Oscan es(s)uf and Umbrian esuf 'ipse'. Both of the innovations in Venetic are justifiable on language-internal grounds as natural processes, and parallels with Germanic may, but need not, point to contacts.
In the realm of the verb, similarities with Celtic are quite significant in the -s-preterite (e.g. Ven. vha.g.s.to 'made', dona.s.to 'gave', usually in the formula dona.s.to.dono.m. 'gave (the) gift'), which has a correlate in the Celtic -s-aorist (e.g. Old Irish 3rd sg. car (< *-a-s-t), 3rd pl. carsat (< *-a-s-nt), Ven. kara.n.mn.s. (cf. Lejeune 1974: 168), found also in Slavic. The occurrence of the -s-preterite also with -a-conjugation verbs connects Venetic with Celtic, whereas Latin and Oscan have here -s-perfects (e.g. Lat. donavit, Oscan (d)uunated (cf. Euler 1993: 102).
For the Venetic medio-passive voice in -r (e.g.. Ven. tole.r dono.m 'brought a gift'), the closest parallel is probably the Celtic deponent in -Vr. Concerning the Venetic use as either deponent or active forms, signalized by Untermann (1980: 292f.), there are also Celtic parallels, since, e.g., the Old Irish deponent flexion fell together with the active flexion in the imperfect indicative, past subjunctive and secondary future, in the 2nd person plural of all tenses and moods, and in the 3rd person singular imperative (cf. Thurneysen 1946: 328).6
According to Euler (1993: 105), it is morphology which gives the clearest answers concerning relations among languages. Our analysis presented above leads to the conclusion that Venetic is not only relatively archaic, but, on the basis of morphology, significantly similar to Celtic. On the phonological side, Venetic occupies an intermediate position, but the similarities with Italic may well have arisen as areal phenomena.
The question of Celtic and Venetic was raised by Eska & Wallace (1999) in connection with the Venetic inscription *Oderzo 7. Oderzo was a Venetic locality east of Padua; the in-
6 There is a further correspondence in Umbrian ferar, in contrast to (the later development of) Latin forms in -ter, -tur (cf. Euler 1993: 103).
scription was found on a monument written in the Venetic alphabet. It consists of three names, the first two in the nominative, the third probably a genitive (relatively unusual for Venetic funerary inscriptions). The Oderzo 7 inscription runs as follows:
padros. pompeteguaios kaialoiso
By virtue of the absence of syllabic punctuation, this funerary inscription from the Venetic area was dated by Lejeune (1989: 71) as being not later than the middle of the fifth ct. BC. Eska & Wallace (1999) show that padros is in all likelihood a Gallicized borrowed name from Latin (possibly connected with the root for '4', because of kw- > p and medial -dr-). Pompeteguaios is taken to be a fully Celtic epithet in the meaning of 'quinquelingual'. Kaialoiso was analysed on the basis of the Lepontic suffix -alo-, employed in the formation of anthroponymic adjectives, and the ending -iso as a Lepontic genitive of o-stems. Thus padros betrays a Cisalpine Celtic sound substitution, pompeteguaios is broadly Celtic, kaialoiso is a Lepontic genitive, and the entire onomastic formula fits a well-known Celtic pattern (cf. Eska & Wallace 1999: 133).7
There were apparently areal contact phenomena between Latin and Celtic, and this inscription, identified as Celtic, was found in the traditionally Venetic area, written in the Ve-netic script.
3. East European Venetic and Armorican Venetic
3.1. East European Veneti
Next to the Northern Adriatic Veneti, there were also Veneti in Armorica and along the Vistula river in East Europe. They traded amber across Europe and into Asia since the second millennium BC, as attested archaeologically by Gimbutas (1965: 40), and the distinct tribes that went by the same name must have been at least well aware of each other.
Specifically, along the upper Vistula basin and in a wider region east of Dacia, Eastern European Veneti were attested by the Tabula Peutingeriana, a Roman road map from the beginning of our era, preserved in a copy from the third or fourth century AD. On the Tabula Peutingeriana, Veneti (by pronunciation rule of that period, Venedi) were attested west of where the Slavs must have had their prehistoric dwelling on the basis of evidence from indigenous lexical denotations.
No textual heritage can provide insights into this Venetic language, but there are indirect traces worth investigating. Specifically, Roman historians systematically mention Veneti and Slavs migrating together in the sixth and seventh centuries AD. In 551 Iordanes, the Gothic historian of the Roman Empire, wrote about Veneti and Slavs:
"ab una stirpe exorti, tria nunc nomina ediderunt, id est Venethi, Antes, Sclaveni" (XXIII).
"though off-shoots from one stock, have now three names, i.e. Veneti, Antes and Slavs".
The Roman army general Prokopios Kaisareos wrote soon afterwards in De bello gothico about Veneti, Antes and Slavs that they "even speak the same, quite barbarous language" (XXVI).
7 According to the authors, Padros was probably a Venetos who lived in the Lepontic area, where he became known by the five languages he spoke, and is now buried in the Venetic area beside his father, of entirely Celtic origin.
SEOHENTUM vm
Map 1: Tabula Peutingeriana, a 3/4th ct. Roman road map (Prontera 2003).
Among later historians, let us mention Fredegarius Scholasticus from the middle of the seventh century, who composed a chronicle of the Merovingian times (probably between 658 and 661). His chronicle repeatedly refers to Winidi (i.e. Veneti in the pronunciation of his time) concerning their encounters with the (allegedly Frankish) leader Samo.8 For the year 623 there are three references to Slavs, twice explained as Sclavi cognomento Winidi, and the third reference mentions Slavic war activities against the Avars, immediately following the first explanation of Slavs cognomento Winidi. At the same time, not a single reference to Veneti (i.e. Winidi, in the entire manuscript mentioned 29 times) characterizes them as *cognomento Sclavi.9 This asymmetry is a clear indication of the fact that Slavs and Veneti were distinct and hierarchically different: the Veneti were perceived as superordinate.
3.2. Phonological changes of the Slavic migration period
The migration period was a time of violent changes which reshaped the Slavic phonological system and remained in effect until the end of the first millennium AD. These changes affected syllables and introduced the bisyllabic domain which was not relevant for phonological rules before. The main changes were governed by the following principles and regularities: the principle of rising sonority within the syllable (resulting in open syllables), so-called syllabic synharmony by which a high-tonality adjustment occurred between syllable onsets and rhymes (resulting in palatalizations and umlaut), asymmetrical lenition of voiced palatalized consonants and the voiced velar (enabling the assumption that the consonant system was patterned as tense vs. lax), introduction of the bisyllabic domain for phonological processes during the third palatalization, and introduction of the trochaic pattern which came to the fore in the apocope and syncope of the yers.
8 Fredegar's chronicle writes: Samo natione Franco, de pago Senonago; if Samo's place of origin, Senonago, is to be equated with Sens, south-east of Paris, then Samo may have been of Gaulish or Gallo-Romance origin.
9 Source: http://www.intratext.com/IXT/LAT0785/_P2.HTM. Including the preface, the text contains 11 references to Winidi, 4 to Winidis, 9 to Winidos and 5 to Winidorum.
Palatalization developed as a complex, pluriphasal process by which velars were fronted when immediately adjacent to following front vowels, followed by a similar adjustment of other consonant classes to a following j. During the first palatalization, velars in syllable onsets were adjusted to front vowels or the glide j in syllable nuclei, i.e. k > c, g > j, x > s before e, e, i, i, j (e.g. wilke > wilce 'wolf', voc.sg.); as a result, front vs. non-front (i.e. acute vs. non-acute) syllables emerged. As a mirror image of palatalization, umlaut produced adjustments of syllable rhymes to high-tonality onsets. During the initial stage, back vowels (i.e. a, a, aN, u, u, uN) acquired fronted variants (i.e. a, a, aN, u, u, UN) following the front glide j (e.g. juga 'yoke').
In the aftermath of the first palatalization, the voiced affricate /3/, which emerged as an effect of the palatalization, became lenited. Lenition of the voiced affricate j > z introduced asymmetry viz. /c/, which remained unchanged. This change was blocked by a preceding dental fricative (i.e. [z]). There was no system-internal reason for the rise of the asymmetrical j > z lenition (which yielded a counterpart to /s/, yet eliminated the counterpart to /c/).
In accordance with the principle of rising sonority, diphthongs which ended in -i and -u, nasal diphthongs and diphthongs which ended in a laryngeal were monophthongized. The diphthongs which ended in -i yielded new front monophthongs (i.e. ai > e (e), ei > e/i (e/i)), capable of causing further palatalizations (e.g. kail- > cel- 'whole'). The second palatalization of velars produced new palatal consonants: k > c, g > 3', x > s' before the new front vowels which had emerged from the monophthongization (also sk > sc and zg > ¿3' before the new front vowels).10 The second palatalization cannot be shown to have been operative in kw, gw, xw sequences in West Slavic and the south-western East Slavic Kiev-Poles'e dialects (concerning the data cf. Shevelov 1964: 301), from where the first phase of the migrations apparently had already started when the second palatalization took place.11
Through the third palatalization (which was in part contemporaneous with the second palatalization), velars became palatalized (i.e. k > c, g > 3', x > s') immediately following a high front vowel (i.e. i, i, < iN) of the preceding syllable nucleus (e.g. atik- > otic- 'father'), unless followed by another consonant or a u-type vowel (u, u, uN) in the following syllable.12
The third palatalization was the first instance of a progressive palatalization and for the first time, palatalization could be blocked by the following syllable (e.g. Germanic stig- > Slavic stiga > stb3a > stbza 'pathway', but mbgla 'haze', lbgynji 'releif', blocked by the second consonant in the onset of the second syllable, or an u-like vowel in the second syllable, resp.). The introduction of the bisyllabic domain for palatalization was an innovation, in itself not motivated in any system-internal way (cf. also Shevelov 1964: 347 etc.). Because of its novel direction and domain, the third palatalization was a type-modifying phenomenon in Slavic.13
Another important innovation of the migration period was the asymmetrical lenition, first of j > z (unless blocked by a preceding z) in connection with the first palatalization, then of 3' > z' in connection with the second and third palatalizations. In addition, there was a lenited
10 Eastern South Slavic, Kiev-Poles'e and eastern West Slavic have traces of treating these sequences differently in word-initial and internal positions, cf. Shevelov (1964: 300).
11 Next to West Slavic, northern Krivici dialects of Russian also preserve these sequences next to generally lacking the second palatalization, cf. Zaliznjak (1995: 38).
12 The third palatalization did not affect /x/ in the northern Russian dialect of Novgorod (cf. Zaliznjak 1995: 38), and there were other exceptions discussed a.o. by Shevelov (1965).
13 The relative time of the second and the third palatalization is difficult to establish. Shevelov (1964: 351) wrote that the third palatalization started at the peak of the second palatalization and ended after the extinction of the latter. The absolute timing is given by the borrowing of Germanic panning > penning in the umlauted form into Slavic pen^ji, with the third palatalization.This word denoted Frankish denars whose coinage started ca. 650 AD (cf. Shevelov 1964: 350).
reflex of the voiced palatal dental dj>j found in western South Slavic, i.e. in western Stokavian, Cakavian, south-western Kajkavian (dialects of Croatian), in Slovene. In West Slavic, lenition dj>z is found in Slovincian, Upper and Lower Sorbian, Bohemian, almost all of Moravian and south-western Slovak. In all of these areas, lenition applied only to the voiced palatal and left the voiceless palatal unaffected. Finally, there was also an asymmetrical lenition of the voiced velar (i.e. g > y) in Ukrainian, Belorussian, southern Russian, Slovak, Czech, Upper Sorbian, northern and western dialects of Slovene, and in some north-western Cakavian dialects of Croatian (cf. Andersen 1969). This was also an asymmetrical process which left the voiceless velar stop /k/ and the voiceless velar fricative /x/ unaffected.
The bisyllabic domain, which arose during the migration period, continued to determine phonological and prosodic phenomena of the post-migration period as well. Most clearly, this came to the fore in the phenomena of apocope and syncope connected with the reduced vowels (so-called yers) in Slavic, by which, counting from the end of the word regressively, uneven yers were weak and even yers in adjacent syllables were strong and yielded full vowels (allowing for language-specific variation next to a resonant in the syllable rhyme). The fall of the yers revealed the trochaic pattern within the bisyllabic domain.
All these phenomena developed along the westward migration routes and either remained most pronounced there (e.g. the lenition g > y which remained confined to the westward migration routes, from the starting point in the Ukraine to West Slavic and the western fringes of South Slavic) or they started in the migration areas and spread from there (e.g. the trochaic pattern of apocope and syncope, by which weak yers were elided and strong yers strengthened). All these facts show that the source for these changes must have been present during the Slavic migrations towards the west, exactly those for which historians provided evidence of Slavs migrating together with Veneti, the East European Veneti. In the absence of direct evidence on the language of East European Veneti, we can only turn to the language of Northern Adriatic Veneti and ask the question of a possible similarity.
3.3. The phonological type of Venetic
Can Venetic of the Northern Adriatic be assumed to have been a language with palatalization, lenition of voiced consonants (i.e. of the mediae), especially the voiced velar, and the trochaic pattern starting from the word-final edge which provided the frame for apocope and syncope? In other words, did Northern Adriatic Venetic originally share the properties which were introduced in Slavic phonology during the migration period as innovative changes? If so, then Northern Adriatic Venetic would fit the type of the language supposed to have exerted influence on Slavic, and we could conjecture that Northern Adriatic Venetic fitted the type of the East European Venetic.
Indeed, all the typological characteristics introduced in Slavic during the migration time were in a nutshell present in the historically attested Venetic of the Northern Adriatic. In spite of the difficulty to extrapolate from Northern Adriatic Venetic of the last centuries BC to East European Venetic of the middle of the first millennium AD, sufficient knowledge of the historically attested Northern Adriatic Venetic (especially since Lejeune 1974) is present for establishing the following relevant features:
a) the unmarked pattern of open syllables (CV, exceptionally CCV), as the corresponding orthographic sequences were singled out as syllabic units by means of punctuation;
b) the loss of final consonants (e.g. donasnt > donasan, Lejeune 1974: 148), also contributing to rising sonority;14
14 According to Lejeune (1974: 158), a Venetic word could end in a short vowel (e.g. the enclitic -kve), a long vowel (e.g. instr.sg. -a, .a.tra, 125), a diphthong (e.g. dat.sg. -a.i), a sonorant (-r, e.g. augar, 248, -n, e.g. donasan, 123,
c) prothesis before a word-initial vowel, fitting the principle of rising sonority;
d) the consonant opposition of tense vs. lax consonants, as revealed by the orthography and the presence of a so-called lax voiceless velar, demonstrating that tense vs. lax was a relevant feature opposition in the consonant system;
e) lenition of the velar /g/, found also in Gaulish, where g had a tendency to vanish inter-vocalically (cf. Lambert 2003: 46);
f) presence of palatalization (in accordance with the principle of syllabic synharmony), resulting in a dental palatal;
g) non-palatalized sequences of velar + w preceding a front vowel, similar to West Slavic.
These typologically coherent properties of Northern Adriatic Venetic match closely the properties of the language which must have triggered the innovations in Slavic phonology and for which we have historical evidence that it was Venetic, the East European Venetic (cf. Gvozdanovic 2009). By this indirect evidence, East European Venetic belonged to the same phonological type as Northern Adriatic Venetic.
The Armorican Veneti were (since Caesar's De bello gallico) viewed as Celts. Although we have no direct attestations of this language from the first millennium AD, historical reconstruction of the Vannetais dialect of Breton, which stands out among the Breton dialects, shows significant typological similarities with Venetic.
Among the dialects of Breton, Vannetais has some outstanding features, observed as such throughout its history:
• The accent in Vannetais has, in general terms, since the apocope been on the final syllable (whereas in the other dialects it stays on the penultimate syllable, e.g. mer'cedes > mer'xed > merxed 'girls' through accent retraction in the eleventh century AD); this points to the right-edge trochaic pattern in the second half of the first millennium AD; in the history of Vannetais there was a general monophthongization of diphthongs (whereas the other dialects preserved them): au > o (e.g. potr 'boy'), ei > e in most of the southern parts, ai > e (written x, e.g. xr 'snake' in Guillevic and Le Goff, Jackson 1967: 71);
• Vannetais still has a striking distinction between tense and lax consonants; tense consonants are long and lax ones are short; in Low Vannetais only long vowels occur in monosyllables preceding a lax consonant;
• Vannetais stands out among the Breton dialects by having much more extensive palatalization, found with all the consonants (except m and r), as a regular feature preceding a front vowel or j (not only with dentals and exceptionally velars, as e.g. in Tregor-rois) or following a closed front vowel; the usual realization of palatalized k and g in Vannetais appears to be either as palatal [k'], [g'], or affricate [ts] and [dz] (Jackson 1967: 76); the bisyllabic palatalization pattern following an i-like vowel is clearly reminiscent of the third palatalization in Slavic;
• Occasionally, Vannetais g' is realized as j (Jackson 1967: 77); this points to asymmetrical lenition of the palatalized voiced/lax velar, also clearly reminiscent of the asymmetrical lenitions in Slavic.
We can see that the historical phonology of the Vannetais dialect of Breton also matches both the phonological type of the Northern Adriatic Venetic and the conjectured phonological type of East European Venetic, in spite of the time gap between them.
-l, e,g, .e.n.to.l, 148, exceptionally -m in Lagole), -s resulting from dissimilation, -t after apocope of final -j, -k after apocope of final -e (e.g. .e..i.k., 242a), or -ts old or after apocope of -j (e.g. .o.s.t.s., 125).
3.4. Celtic parallels
Palatalization, lenition, the bisyllabic domain, apocope and syncope within the trochaic pattern were also prominent changes in the history of Celtic around the middle of the first millennium AD. These changes affected both Goidelic and Brythonic Celtic, but with a somewhat different relative chronology and partially different rules (discussed in Gvozdanovic 2009). Among other facts, apocope and syncope in Goidelic depended on vowel length, whereas this was not the case in Brythonic. The mutations had more far-reaching effects in Brythonic than in Goidelic. But perhaps the most interesting are the differences in palatalizations.
The Goidelic palatalizations, represented by Old Irish, occurred in three phases (cf. McCone 1996: 118f.). The first palatalization affected single consonants and consonants with homorganic nasals; according to Kortlandt (2007: 9f.), (i) all consonants were palatalized between front vowels and before stressed front vowels (e.g. bere > -beir /b'er'/ 'carries'), (ii) dentals were palatalized before posttonic i (e.g. rodeio > rûdîu > -râidiu 'I say'), (iii) labials and velars were palatalized before posttonic i unless they were preceded by a back vowel (e.g. gabiot > -gaibet 'they take').
The second palatalization in Old Irish affected initial consonants or groups of consonants followed by e/ë or i/î (e.g. mliy'eO'ï > m'ï'iy'eO'ï > mligid 'milks'). The third palatalization came about after the truncation or loss of final syllables, when the front vowels i, e and u fell together as palatal schwa ï with a palatalizing effect "in a non-final internal syllable directly after the first syllable bearing the main stress and (where applicable) after the third syllable probably bearing secondary stress. Like its counterpart produced before the apocope, this sound palatalized any preceding consonant(s)" (McCone 1996: 119).
We can see that the Old Irish palatalizations referred to the bisyllabic domain for the rule description, but did not share all the properties with the corresponding Slavic rules. Of all the speakers of Brythonic languages, Bretons migrated to Armorica by the middle of the first millennium AD, where they presumably encountered Continental Celtic population. It is possible, and in view of the typological analysis presented above likely, that Vannetais is also a remnant, but of Venetic — as a transparently Celtic language. Next to phonological properties shared with Continental Celtic, Vannetais stands out by the strength of its palatalizations (cf. also Jackson 1967: 76-7). Not only are there no palatalized dental phonemes in Breton outside of Vannetais (Jackson 1967: 376), but palatalization of k and g preceding e and i is practically unknown in NW Breton, yet relatively regular in Vannetais, commonly pronounced as [ti] and [dz]; in a small SE corner of Vannetais, kw and gw also palatalize when preceding -oi/oi, e.g. coid 'wood' > Vannetais koed [k'w'et]. Palatalization of x appears to be practically unknown in the Léonais, Trégorois and northern Cornouaillais dialects of Breton.
The exact rule description of the Vannetais palatalizations offers an even clearer parallel with the conjectured regularity of East European Venetic, because palatalization occurs in Vannetais preceding a front vowel (e.g. dragenai > drezin > drein 'thorns') or following -i- (except -igi plural), but not following -e- (e.g. kerzeg 'horses'). Palatalization is also caused by front diphthongs which, as a result of i-affection, were fronted and raised, i.e. ai > ei, oi > ei. When the velars k and g are palatalized, this results in the corresponding affricates [ts] and [dz], but only the reflex of the palatalized voiced velar g' undergoes far-reaching lenition of the type g' > j. All these features are clearly reminiscent of what happened in Slavic. Moreover, these features are type-consistent in Vannetais, whereas their introduction had a type-modifying effect in Slavic.
Linguistically, Vannetais stands out also by its similarities with Cornish which are lacking in the other dialects of Breton (discussed by Fleuriot 1980: 67 etc.). All these properties show that Vannetais is a separate linguistic entity within Celtic. Its outstanding combination of palatalization and lenition makes it a likely candidate for a surviving link between the various varieties of Venetic.
3.5. Lenition in Slovene
When Slovenes, the most westerly South Slavs, settled the territories which formerly pertained to Veneti, they developed lenitions of voiced stops, and the isoglosses clearly point to the centre of this change in the Venetic area.
Map 2: Lenition in Slovene (Greenberg 2000: 140); west of 1) is a stop, west of 2) all the voiced stops were lenited, west of 3) the reflex of was lenited, north of 4) is a glottal stop.
As shown by Greenberg (2000: 140), the lenition isoglosses in Slovene have a westward orientation and the lenition area of the voiced velar (i.e. > y) is essentially broader (to the west of isogloss 3 on the map) than the lenition area of all voiced stops (to the west of isogloss 2 on the map). This shows that the centre of the lenition phenomena must have been in the west and south-west of the Slovene area, exactly in the historical Venetic area. These isoglosses enable a reconstruction by which lenition spread from leniting the voiced velar to leniting all the mediae in a way reminiscent of Celtic (discussed in more detail in Gvozdanovic 2009).
By analyzing critically the arguments proposed for the classification of Northern Adriatic Venetic, this study has reached the conclusion about its significant Celticity. The basis for this statement is provided by the morphological inventory and the functional distribution of the morphological elements. As to the phonology, areal phenomena have been observed in the similarities between Northern Adriatic Venetic and the neighbouring Italic. This confirms the increasing awareness among Celtic specialists (e.g. Sims-Williams 2007) that the classificatory parameters for Celticity are not as clear-cut as traditionally assumed. Our analysis has reached exactly this conclusion concerning areal phenomena shared between Venetic and Italic with the result that Northern Adriatic Venetic occupies an intermediate position on the traditional classi-ficatory parameters, but belongs to the Celtic type on the basis of the phonological patterning (including lenition in the consonant system based on the tense vs, lax opposition). This fact, combined with morphology which can be identified as Celtic, leads to the conclusion that Northern Adriatic Venetic was a variety of Celtic which was subject to areal common developments with Italic in the phonology and the lexicon, but remained entirely Celtic in the grammar.
4. Conclusion
The present investigation of East European Venetic has shown that its conjectured phonological type fits the phonological type of both Northern Adriatic Venetic and — even more so — of Vannetais as a surviving variety of Armorican Venetic. The Armorican Vannetais, originally surrounded by Celtic, has never raised any doubt about its Celticity. The fundamental similarity of Northern Adriatic Venetic, the conjectured East European Venetic and of Vannetais on the level of the phonological type, and the specific similarity of both of the latter varieties in the realm of palatalization combined with asymmetrical lenition, enable the conclusion that these were genetically related varieties of Venetic within the realm of Celtic that later became differentiated due to areal phenomena.
Literature
Andersen, Henning (1969) Lenition in Common Slavic. In: Language 45/3, 554-575.
Beeler, M. S. (1949) The Venetic language. Berkeley — Los Angeles: University of California Press.
Beeler, M. S. (1981) Venetic revisited. In: Arbeitmann, Y. & Bomhard, A. (eds.) Homo Homini Donum. Essays in Historical Linguistics in Memory of J. Alexander Kerns, 65-71. Amsterdam: Benjamins.
Campbell, Lyle & William Poser (2008) Language Classification. Theory and Method. Cambridge University Press.
Delamarre, Xavier (2003) Dictionnaire de la langue gauloise. Paris Editions Errance.
Eska, J. F. & R. Wallace (1999) «The linguistic milieu of *Oderzo 7», Historische Sprachforschung 109, 122-136.
Euler, W. (1993) Oskisch-Umbrisch, Venetisch und Lateinisch. In: H. Rix (ed.) Oskisch-Umbrisch. Texte und Grammatik. Wiesbaden: Reichert.
Falc'hun, François (1962() Celtique continental et celtique insulaire en breton. In: Annales de Bretagne 70: 426-454.
Fleuriot, Léon (1967) L'importance du dialecte de Vannes pour l'étude diachronique et comparative du breton armoricain. In: Wolfgang Meid (ed.) Beiträge zur Indogermanistik und Keltologie, Julius Pokorny zum 80. Geburtstag gewidmet, 159-170. Innsbruck: Innsbrucker Beiträge zur Sprachwissenschaft 13.
Fleuriot, Léon (1978) Brittonique et gaulois durant les premiers siècles de notre ère. In: Etrennes de septentaine. Travaux de linguistique et de grammaire comparée offerts à Michel Lejeune par un groupe de ses élèves, 75-83. Paris: Klincksieck.
Fleuriot, Léon (1979) L'inscription gauloise du Thiaucourt et le problème des désinences verbales en -seti et -setu du celtique antique. In: Ètudes Celtiques 16, 123-139.
Fleuriot, Léon (1980) Les origines de la Bretagne. Paris: Payot.
Fleuriot, Léon (1988) New documents on Ancient Celtic and the relationship between Brittonic and Continental Celtic. In: Proceedings of the First North American Congress of Celtic studies, 223-30. Ottawa: Ottawa University Chair of Celtic.
Gamkrelidze, T.V. & V.V. Ivanov (1984) Indoevropejskij jazyk i indoevropejcy. Tbilisi: Izd. Tbilisskogo Universiteta.
Gimbutas, Marija (1965) Bronze Age cultures in central and eastern Europe. Paris etc.: Mouton.
Gimbutas, Marija (1997) The Kurgan Culture and the Indo-Europeanization of Europe. Washington: Institute for the Study of Man.
Greenberg, Marc L.(2000) A historical phonology of the Slovene language. Heidelberg: Winter.
Greene, David (1956) Gemination. In: Celtica 3: 264-289.
Greene, David (1966) The spirant mutation in Brythonic. In: Celtica 7, 116-119.
Greene, David (1974) The growth of palatalization in Irish. In: Transactions of the Philological Society 1973, 127-136.
Greene, David (1976) The diphthongs of Old Irish. In: Ériu 27: 26-45.
Gvozdanovic, Jadranka (2009) Celtic and Slavic and the Great Migrations; Reconstructing Linguistic prehistory. Heidelberg: Universitätsverlag Winter.
Jackson, Kenneth H. (1967) A historical phonology of Breton. Dublin: The Dublin Institute for Advanced Studies.
Katicic, Radoslav (1976) Ancient languages of the Balkans. The Hague: Mouton.
Kortlandt, Frederik (1981) More evidence for Italo-Celtic, Ériu 32, 1-22.
Kortlandt, Frederik (2007) Italo-Celtic origins and prehistoric development of the Irish language. Amsterdam: Rodopi.
Krahe, Hans (1950) Das Venetische, seine Stellung im Kreise der verwandten Sprachen. Heidelberg: Carl Winter Universitätsverlag.
Lejeune, Michel (1974) Manuel de la langue venete. Heidelberg: Carl Winter Universitätsverlag. Mallory, J. P. (1989) In search of Indo-Europeans; Language, archaeology and myth. London: Thames and Hudson. McCone, Kim (1996) Towards a relative chronology of ancient and medieval Celtic sound change. Maynooth: Dept. of Old Irish.
Pauli, C. (1891) Die Veneter und ihre Schriftdenkmäler. Leipzig: Barth.
Polome, E. (1966) The position of Illyrian and Venetic. In. H. Birnbaum & J. Puhvel (eds.) Ancient Indo-European
dialects, 59-76. Berkeley — Los Angeles: University of California Press. Schrijver, Peter (1995) Studies in British historical phonology. Amsterdam — Atlanta: Rodopi. Shevelov, G. Y. (1964) A Prehistory of Slavic. Heidelberg: Winter.
Sims-Williams, P. (2007) Common Celtic, Gallo-Brittonic and Insular Celtic. In: Lambert, P.-Y. & Pinault, G.-J.
(ed.) Gaulois et Celtique Continental, 309-354.Geneve: Librairie Droz S.A. Thurneysen, R. (1946) A grammar of Old Irish. Oxford: Oxford University Press. Untermann, J. (1980) Die venetische Sprache. In: Glotta 58: 281-317.
Wallace, R. B. (2008) Venetic. In: R. D. Woodard (ed.) The Ancient Languages of Europe. Cambridge: Cambridge
University Press. 124-140. Zaliznjak, A.A. (1995) Drevne-novgorodskij dialekt. Moskva.
Венетский язык и его разновидности на территории доисторической Европы до сих пор остаются одной из загадок исторического языкознания. Традиционно принято считать, что венетский язык в Арморике, скорее всего, был кельтским, в то время как венетский язык на северном побережье Адриатического моря, от которого остались письменные памятники, обычно считается отдельной ветвью индоевропейской семьи. Меньше всего известно про венетский язык, на котором говорили в Восточной Европе. Статья посвящена общему обсуждению данных по венетскому языку (в особенности по северно-адриатическому варианту) в свете текущих представлений о сравнительном историческом развитии различных индоевропейских ветвей (в частности, кельтской и славянской). Особое внимание уделено связям между венетским и континентальными кельтскими языками.
Ключевые слова: венетский язык, кельтские языки, индоевропеистика, ареальная лингвистика.
