The alleged early apocope of *-i in Celtic

THE ALLEGED EARLY APOCOPE OF *-i IN CELTIC* 1. Warren Cowgill has proposed an early loss of final *-i in the 3rd person verbal endings *-ti and *-nti in Insular Celtic (1975: 57). This development was morphologically conditioned because it did not affect e.g. Old Irish innuraid ‘last year’ < PIE. *peruti (Cowgill 1985: 109). It somehow affected the 3rd sg. but not the 3rd pl. relative form (Cowgill 1975: 59). His theory does not account for the 2nd sg. forms (cf. Cowgill 1975: 61, also fn. 13, and Kortlandt 1979: 36). Kim McCone has tried to broaden the range of the apocope in order to explain the short dat.sg. forms of the Old Irish consonant stems (1978), to my mind unsuccessfully (cf. Kortlandt 1994). It forces him to develop special explanations for such forms as muir ‘sea’ < *mori, dat.sg. déit ‘tooth’ < *danti, and inn-uraid ‘last year’. The gen.sg. forms anmae ‘name’ < *-ens and sléibe ‘mountain’ < *-esos rather suggest that we have to start from a zero loc.sg. ending in dat.sg. ainm and a long ending *-esi in sléib. As intervocalic *-s- was lost at an early stage in loc.sg. *tegesi ‘house’, the original nonzero ending of the s-stems merged phonetically with the original zero ending of the n-stems, leaving as its only trace the raising of the root vowel in dat.sg. tig (cf. Cowgill 1975: 57, 1985: 113, Kortlandt 1979: 39f.). McCone has failed to notice that the distribution of long and short dat.sg. forms of consonant stems in the glosses reflects a distinction between animate and inanimate nouns and may therefore continue a difference between dative and locative forms (cf. Kortlandt 1994: 63 and fn. 3). Peter Schrijver has tried to formulate an intermediate position between Cowgill’s and McCone’s, proposing that the apocope of *-i was limited to the position after a voiceless obstruent (1994: 164). Since his article may give rise to a number of misunderstandings with respect to my views, I feel the need to clarify my position here. * Études Celtiques 32 (1996), 91-97. 100 Italo-Celtic origins and prehistoric development of the Irish language 2. The lenition after 3rd sg. neuter object pronouns, e.g. in Wb 5b 5 ni-cheil ‘he does not conceal it’, shows that PIE. final *-t/d was lost in Celtic at an early stage (cf. Cowgill 1975: 52). This explains the adoption of perfect endings by the thematic aorist luid ‘he went’ < *ludhet. On general phonetic grounds it is probable that final *-t was also lost in clusters and could later be restored by analogy, e.g. in trícho ‘thirty’ < *trīkont (cf. Kortlandt 1979: 46). In my view, the 3rd pl. thematic present ending *-o was replaced by the secondary ending *-on(t) when the 3rd sg. thematic present ending *-e merged with the secondary ending *-e(t), with *-t before clitics, as in French a-t-il ‘has he’ beside il a ‘he has’. The athematic endings did not enter into this analogy. Schrijver objects that “it seems unlikely that *-t in *-ont would have been maintained before a clitic: cf. *ludet es > *lude es > *lude-h > OIr. abs. 3sg. pret. luid ‘went’, where *-t was obviously lost” (1994: 159, fn. 4). The objection is beside the point because the clitic *es probably had not yet arisen at this stage. Moreover, the thematic aorist adopted the endings of the perfect, so that there was no motivation for a restoration of *-t in this category. On the other hand, the 3rd pl. variant *-on-t- beside *-on may have given rise to a 3rd sg. variant *-e-t- beside *-e in the thematic present. In my earlier treatment I adduced Old Latin esed ‘erit’ as a parallel for the addition of secondary *-t to the 3rd sg. thematic present ending *-e (1979: 38). Cowgill objects to my reconstruction of the 3rd sg. thematic present ending *-e that “some Old Irish thematic 3rd sg. conjunct presents of roots ending in dental stops actually contain a relic of the *t of the ending *-et(i)”, e.g. tadbat ‘shows’ < “t-ad-wēd-e-t” (1985: 110), for which he assumes an additional irregular loss of the thematic vowel. Following Thurneysen (1946: 377), I rather assume that the athematic conjunct ending of -tét ‘goes’, prototonic -tet, spread to the semantically close verbs -fet ‘leads’, -rét ‘rides, drives’, *-ret ‘runs’, prototonic -at, -rat, and then to other verbs with a rootfinal dental stop such as ad-fét, -adbat ‘relates’, ar-nëat, -airnet ‘expects, sustains’ (cf. Kortlandt 1994: 66). Cowgill’s theory does not explain the alternative forms -feid, -réid, -reith beside -fet, -rét, -fét (cf. also Meid 1972: 350). Thus, I conclude that there is no evidence for an early apocope of final *-i in the 3rd person verbal endings while there is counter-evidence in the athematic 1st sg. conjunct ending -imm, e.g. -crenaim ‘I buy’ < *kwrinami The alleged early apocope of *-i in Celtic 101 (cf. Schrijver 1994: 161), and that there is no evidence in the dat.sg. forms of the consonant stems while there is counter-evidence in inn-uraid ‘last year’ < PIE. *peruti. Schrijver suggests that -uraid may reflect an accusative *erutam < *perut-m (1994: 162), but this is improbable because *peruti is an isolated case form in the attested Indo-European languages. Note that Sanskrit parút ‘last year’ does not occur as a loc.sg. form in old texts. Schrijver tries to produce additional evidence for the apocope of *-i in fri ‘against’ < *writ(i) and la ‘with’ < *(p)let(i) (1994: 158). It seems to me that no conclusions can be based on these prepositions, which may have suffered any number of analogical remodelings in prehistoric times. The pretonic preverb friss- instead of *frith- can easily have taken its -s from *eks, *uts, *ups (cf. Thurneysen 1946: 515). The conjugated prepositions friss ‘against him’ and leiss ‘with him’ probably took their -s from the emphatic forms frissom and leisom, where *-th had assimilated to the following -s-, cf. especially Ml 25b 6 faissine ‘prophecy’ beside usual -ths- with analogical -th-, Wb 5b 11 con-dositis ‘so that they should fall’ from *-ths-, Wb 1a 3 ro-cretsisi ‘ye have believed’ for -dsi (Thurneysen 1946: 88). Schrijver’s derivation of -s from *-t(i) leads him into major chronological difficulties (1994: 167, fn. 7 and 169f.). I conclude that there is no evidence for an apocope of *-i in fri and la while there is counter-evidence in imm ‘about’ < *ambi and ar ‘before’ < *pari (cf. Schrijver 1994: 161). 3. In his discussion of arimp ‘in order that it may be’, with -p < *bes ‘may be’, Schrijver refers to “Kortlandt’s reconstruction *beseti-s” (1994: 166, fn. 6). In fact, I have argued against such a reconstruction, which in my view would yield the wrong output **beïd, cf. 2nd sg. bee < *beses-es (Kortlandt 1984: 185). On the basis of an athematic paradigm, Schrijver “would expect absolute *bes-t-es > *besseh, which could never have been replaced by beith” (l.c.), a view which evidently requires no further justification. This is a far cry from my substitution of beith for *beeh < *bes-es like beraid ‘may carry’ for *beraeh < *beras-es (1984: 185). Schrijver derives 1st pl. -beram ‘carry’ from *beromosi (with early i-apocope), which he compares with Sanskrit -masi < *-mesi and Latin -mus, which “may reflect *-mosi” (1994: 171, fn. 10). This is certainly incorrect in view of Latin agere ‘to conduct’ < *-esi (cf. Kortlandt 1981: 18). In fact, the 102 Italo-Celtic origins and prehistoric development of the Irish language OIr. 2nd pl. absolute ending -the < *-tes-es, not **-thi, requires 1st pl. *-mos, not *-mosi, as a model for the addition of *-s to earlier *-te. The analogical ending of bermai ‘we carry’ < *-moih for *-moeh < *-mos-es was evidently taken from the other forms of the paradigm (cf. Kortlandt 1979: 46). Elsewhere I have argued that in Italo-Celtic the final *-ro of the 3rd pl. middle ending *-ntro was “reinterpreted as a voice marker and spread to the singular intransitive middle endings: 1st sg. *-ōro (thematic ending), 2nd sg. *-toro, 3rd sg. *-oro. Analogy created a 3rd sg. ending *-tro and a 1st pl. ending *-moro. The addition of *-ro to the 3rd sg. and pl. transitive middle endings yielded passive forms of transitive verbs in *-toro and *-ntoro” (1981: 17). This explains “the absence of palatalization in the absolute deponent endings 1st sg. -ur < *-ōro-s and 2nd sg. -ther < *-toro-s” and the “absence of reduction in the Old and Middle Welsh ending -tor < *-toro” (Kortlandt 1981: 19), as well as the different patterns of syncope in the deponent and the passive. Schrijver objects that it is “difficult to account for the palatal final of the absolute 3 sg. th(a)ir, which seems to reflect *-tor-es rather than *-toro-s” (1994: 171, fn. 12). My point is that the palatalization in the 3rd person endings can easily have been taken from the active paradigm whereas no such explanation is possible for the absence of palatalization in the 1st and 2nd sg. absolute deponent endings. Note that the pattern of syncope was also subject to the analogy of the corresponding active forms, e.g. do-formagar ‘is increased’ for *doformgar after do-formaig (Thurneysen 1946: 369). Moreover, Schrijver’s “absolute 3 sg. -th(a)ir” is incorrect: unlike the conjunct endings -thar, -ther, pl. -tar, -ter, the absolute endings -thir, -tir always have a palatalized obstruent in unsyncopated forms, both passive and deponent (cf. Cowgill 1983: 95). This is clearly the result of analogical influence from the active paradigm. Incidentally, Welsh gwelir ‘is seen’, not **gwylr (Cowgill 1983: 103), also points to *-ro, not *-r. In my earlier account of the Old Irish relative forms I rejected the traditional view that the ending -e reflects an uninflected particle *yo < *iod for a number of reasons (1979: 50f.): “First of all, the relative particle does not palatalize a preceding consonant, cf. sóeras ‘who delivered’, tías ‘who may go’, giges ‘who will pray’, and all of the passive and deponent forms. Palatalization is limited to those cases where the relative particle was preceded by a front vowel, e.g. téte ‘who goes’ < *tēxti-, luide ‘who went’ < The alleged early apocope of *-i in Celtic 103 *lude-, and the prepositions imme- ‘about’ < *embi- and are- ‘for’ < *ari-. Secondly, it is not clear how the PIE. relative pronoun *ios came to lose its inflection. When the antecedent is the subject of the relative clause, one would expect gemination rather than lenition if the relative particle is to be derived from *ios. Finally, the relation between *io and the relative prepositions such as cosa n- ‘to which’ remains to be explained. All these problems vanish if we identify the relative particle with the PIE. anaphoric pronoun *so, fem. *sā, and assume that it occupied the same position in the clause as the absolute particle *es, e.g. in fer téte ‘the man who goes’ < *sindos wiros steikti so ‘this man, he goes’.” This theory accounts for “a number of curious features which have never received any serious attention” such as the identity of conjunct and relative forms in passive and deponent paradigms which “is not explicable on the basis of any of the many theories which have been put forward to account for the absolute and conjunct endings”, as Greene put it (1977: 28). It also accounts for the coincidence between absolute and relative forms in the passive preterit, e.g. in fer brethae ‘the man who was carried’ < *sindos wiros britos est ‘this man, he was carried’, and for the substitution of absolute or deuterotonic for relative forms in nasalizing relative clauses, e.g. Wb 23d 25 hóre ni-ro-imdibed ‘because he had not been circumcised’ (Kortlandt 1979: 50). Schrijver objects to my theory that the “relative ending -thar cannot reflect *-tor-so, which would have yielded OIr. **-tharr” (1994: 172), but this only demonstrates that his reconstruction of the passive ending *-tor instead of *-toro is mistaken. He follows Cowgill’s view that the relative particle *yo “remained a separate word long enough to undergo the regular Irish loss of initial *y-” and “thus was attached to preceding elements in the shape -o” (1983: 78). This deprives him of the possibility to recognize the relative particle in the relative prepositions, e.g. cosa n- ‘to which’, frissa n- ‘against which’ (cf. Schrijver 1994: 172, fn. 14). In fact, there are several indications that we have to start from *kon-so > cos-, *in-so > as-, not from *kon-o > con-, *in-o > en-. As I had pointed out to Schrijver, we find as- for in-so- in Ml 48c 32 as-dloing ‘who cleaves’ beside Sg 15a 5 in-dlung ‘I cleave’, also Ml 18d 2 asidgrennat ‘who persecute him’ beside Ml 36d 2 a n-inda-greinn-siu ‘when thou persecutest them’ (cf. Thurneysen 1946: 520). It seems to me that these instances merit rather more serious attention than Schrijver’s easy dismissal (1994: 172). My theory actually explains why “at- 104 Italo-Celtic origins and prehistoric development of the Irish language of Class B is replaced by as-” in Class C (Thurneysen 1946: 258), e.g. Ml 54d 6 as-id-roillet ‘who deserve it’ beside Ml 61a 20 ad-id-roillifet ‘who shall deserve it’ (with restoration of ad-), similarly friss-id- instead of frit-. The relative prepositions *es < *in-so and *cos < *kon-so were evidently replaced by i n- ‘in which’ and co n- ‘so that’, which lack the suffixed -a of cosa n- ‘to which’ and frissa n- ‘against which’, and the extended forms *esa-d- and *cosa-d- by in-dand con-d-. This analysis accounts for the absence of a relative preposition **issa n- ‘in which’. The relative form beres ‘who carries’ is most easily derived from *bere-tso, with analogical *-t- before the clitic (see above). In my earlier treatment I rejected this possibility (1979: 51) for chronological reasons. The main point is that the absolute form with suffixed pronoun beirthi ‘carries him’ represents the athematic ending *-ti-s-en, not thematic *-t-es-en (Kortlandt 1979: 39, cf. Cowgill 1975: 59). If one accepts that *-t- before clitics was preserved long enough to play a part in the interaction between thematic and athematic paradigms (Kortlandt 1979: 45f.), this renders the distribution of primary and secondary endings outside the present indicative less easily understandable (cf. Kortlandt 1994: 65 and fn. 4). I therefore adopted the view that beres took its -s from the relative copula as < *es(a) < *est-so, which may be preserved in Breton so, zo (1979: 51). Schrijver objects to my theory that “an early OIr. form asa ‘which is’, as advocated by Breatnach,” would imply that “the reconstruction as < *esa < *est-so is incorrect” (1994: 174f.). However, a form asa < *esti-so beside as < *est-so is no more remarkable than the coexistence of is ‘is’ < *esti- with a nominal predicate and *-es < *est with a verbal predicate, for which I have adduced a parallel from Slavic (1979: 51f.). Schrijver’s own hypothesis of a phonetic development *beret(i) (y)o > *beres-o forces him to assume a semantically unmotivated restoration of *-i in téte ‘who goes’ < *-ti-o, also in 3rd pl. *beronti-o, and further analogical spread of *-i to the preterit bertae ‘who carried’, which he derives from *bersti-o, and even to the copula, where he assumes *essi-o, *senti-o beside *ess-o, *sent-o (1994: 175ff.). I conclude that there is no evidence for an early apocope of *-i and that the alleged development of *-ti > *-t > -s is a fallacy. Note that Middle Welsh gwyl ‘he sees’ and na welyd ‘that he sees not’ (Evans 1976: 119) are also derived more easily from a thematic form *wele which could be The alleged early apocope of *-i in Celtic 105 followed by a relative particle (cf. Schrijver 1994: 176, fn. 16). I think that the coexistence of *esti-so with a nominal and *est-so with a verbal predicate is also reflected in Old Breton isi(o) ‘which is’ beside Middle Breton so < *eso (cf. Hemon 1975: 203). There is no reason to assume an irregular loss of *-i- in the latter form. 4. Schrijver proposes to identify the absolute particle *es with the final obstruent of the Middle Welsh negative preverb nyt, e.g. nyt af ‘I do not go’, Middle Breton ned < Old Breton *nit (1994: 182). This is certainly incorrect, as is clear from the variant nend beside ned and Middle Cornish nyns < *nind (ibidem, fn. 21), which show that the dental stop did not immediately follow the negation, e.g. MBr. nenn d-aff a-dreff ‘I do not go back’, nen d-es ‘there is not’ (Hemon 1975: 281). Schrijver himself raises the objection that in Middle Welsh “the -t- also occurs after the relative negative na ‘that not’ preceding verbs beginning with a vowel”, e.g. nat erchis ‘who did not require’, which invites a comparison of the negative relative nat with its Old Irish equivalent nad. He rejects this objection because the Middle Breton negative relative is nac before vowels, e.g. an nep nac eu discret ‘whoever is not discreet’, but note the imperative nag-a ‘do not go’ and Old Breton nac erminom ‘we do not ask’ (Hemon 1975: 282). There is simply no evidence for Schrijver’s assumption of absolute -t versus relative -k in British (1994: 183). Equally unfounded is Schrijver’s supposition that the absolute particle *es “is in complementary distribution with *de and *kwe” (ibidem). His derivation of OIr. frita- ‘against + 3rd pl.’ from *writi-de-sons is mistaken because fri(th-) ended in a consonant (Thurneysen 1946: 258), so that we have to reconstruct *writ-es-de-sons, similarly 3rd sg. frit- < *writ-es-d-en, also cot‘with + 3rd sg.’ < *kon-s-d-en (cf. Kortlandt 1979: 49 on the reduction of *es to *-s after nasals), and at- ‘to, in, out, up + 3rd sg.’ < *ad-es-d-en, *in-s-d-en, *eks-es-d-en, *ups-es-d-en, Wb 5b 40 cotd-icc ‘he can do it’ < *kon-s-d-e-d-, with a second -d- to protect the infixed pronoun *-e- from elision. In nachitbeir ‘who does not carry you’ we do not have *ne-kwe- (thus Schrijver 1994: 184), which does not explain the relative meaning, but *na-so-kwe-, similarly MBr. nac ‘who (does) not’ < *na-so-k, but nag-a ‘do not go!’ < *na-k age, where Celtic *na- is a reduced form of PIE. *ne ‘not’. In Old Irish we find e.g. Ml 32d 5 nacham-dermainte ‘forget 106 Italo-Celtic origins and prehistoric development of the Irish language me not’ < *na-kwe-me-, but Sg 209b 27 naich ndeirsed ‘that he would not desert him’ < *na-so-k-en-, Wb 6c 18 nách-beir ‘who does not pass it’ < *naso-k-e-, Wb 25d 14 nachid-chualatar ‘who have not heard it’ < *na-so-kwed-e-, Wb 15b 14 nadid chreti ‘who does not believe it’ < *na-so-de-d-e-, Ml 97d 10 nanda-tiberad ‘that he would not give them’ < *na-son-de-sons-, with *-sons- replacing a dative. I therefore derive MW. nat from *na-so-, with -t from nyt < *nīh-d < *nēst de, cf. Slavic ně ‘is not’ < *nēst, similarly OIr. nítat ‘they are not’ < *nēst de senti, and interrogative in < *in(-est) na-so-, Ml 17b 17 innad-naccai ‘seest thou not?’ < *in(-est) na-so-de na-. Thus, I find no evidence for Schrijver’s *-ti > *-t > -s. It follows that his derivation of the absolute particle *es from *eti cannot be upheld. Such a derivation is unattractive anyhow because PIE. *eti ‘beyond’ is not a clitic and does not fit semantically. It seems to me that *es represents a focus particle *est ‘it is (the case that)’ (cf. Kortlandt 1994) and that its development as a pro-Verb cannot be separated from that of the anaphoric pro-Noun *so into a relative marker and, more generally, from the Insular Celtic restructuring of verbal syntax. The Gaulish evidence is difficult to interpret. As the Greek theta represents a fricative in 1st century AD Pompeian spellings (Allen 1974: 21), it seems probable to me that it denotes a (long) fricative in bueθ (Larzac), which may represent *bwes-so (cf. Lambert 1994: 67). It has recently been suggested that the form karnitus (Briona) represents 3rd pl. preterit *karnintu plus an enclitic particle -s from *es or *so (de Hoz 1995: 62f.). This raises a problem with respect to the distribution of *so and *yo in view of dugiiontiio (Alise) and toncsiiontio (Chamalières). However this may be, I think that there is a clue to the Insular Celtic redistribution of the two particles in the Middle Welsh relative forms, e.g. na welyd ‘that he sees not’ beside gwyl ‘he sees’ < *wele, which suggests that we have to reconstruct *na-so wele-yo, and in Old Welsh nit egid ‘goes not’ (Evans 1976: 119), which apparently reflects *nēst-de age-yo ‘it is not the case that he goes’. Note that the derivation of -yd from *-e-so is difficult in view of MW. tei ‘houses’ < *tegesa (cf. now Schrijver 1995: 391). The phonetic merger of *so and *yo after the athematic ending *-ti may have been instrumental in the further development. REFERENCES Allen, William Sidney 1974 Vox Graeca, Cambridge: University Press. Cowgill, Warren 1975 “The origins of the Insular Celtic conjunct and absolute verbal endings”, Flexion und Wortbildung, Wiesbaden: Reichert, 40-70. 1983 “On the prehistory of Celtic passive and deponent inflection”, Ériu 34, 73-111. 1985 “On the origin of the absolute and conjunct verbal inflexion of Old Irish”, Grammatische Kategorien: Funktion und Geschichte, Wiesbaden: Reichert, 109-118. Evans, Daniel Simon 1976 A grammar of Middle Welsh, Dublin: DIAS. Greene, David 1977 “Archaic Irish”, Indogermanisch und Keltisch, Wiesbaden: Reichert, 11-33. Hemon, Roparz 1975 A historical morphology and syntax of Breton, Dublin: DIAS. Hoz, Javier de 1995 “Is -s the mark of the plural of the preterite in the Gaulish verb?”, Hispano-Gallo-Brittonica, Cardiff, 58-65. Kortlandt, Frederik 1979 “The Old Irish absolute and conjunct endings and questions of relative chronology”, Ériu 30, 35-53. 1981 “More evidence for Italo-Celtic”, Ériu 32, 1-22. 1984 “Old Irish subjunctives and futures and their ProtoIndo-European origins”, Ériu 35, 179-187. 1994 “Absolute and conjunct again”, Münchener Studien zur Sprachwissenschaft 55, 61-68. 2007 Italo-Celtic origins and prehistoric development of the Irish language, Amsterdam: Rodopi. Lambert, Pierre-Yves 1994 La langue gauloise, Paris: Errance. McCone, Kim 1978 “The dative singular of Old Irish consonant stems”, Ériu 29, 26-38. Meid, Wolfgang 1972 “On two points of Celtic morphology”, Études Celtiques 13, 346-352. Schrijver, Peter 1994 “The Celtic adverbs for ‘against’ and ‘with’ and the early apocope of *-i”, Ériu 45, 151-189. 1995 Studies in British Celtic historical phonology, AmsterdamAtlanta: Rodopi. Thurneysen, Rudolf 1946 A Grammar of Old Irish, Dublin: DIAS.