ITALO-CELTIC ORIGINS AND PREHISTORIC DEVELOPMENT OF THE IRISH LANGUAGE Frederik Kortlandt CONTENTS Introduction.......................................................................................... ix The Old Irish absolute and conjunct endings and questions of relative chronology [1978].................................................................... 1 More evidence for Italo-Celtic [1980]................................................... 25 Phonemicization and rephonemicization of the Old Irish mutations [1981].................................................................................................... 51 Old Irish subjunctives and futures and their Proto-Indo-European origins [1983]........................................................................................ 65 Posttonic *w in Old Irish [1983]........................................................... 75 The origin of the Slavic imperfect [1983]............................................. 81 Lachmann’s law [1985]......................................................................... 87 Absolute and conjunct again [1993]...................................................... 91 The alleged early apocope of *-i in Celtic [1995]................................. 99 Thematic and athematic verb forms in Old Irish [1996]....................... 107 Old Irish ol ‘inquit’ [1996].................................................................... 113 On the relative chronology of Celtic sound changes [1997]................. 117 Lachmann's law again [1997]................................................................ 121 Three notes on the Old Irish verb [1999].............................................. 125 Old Irish feda, gen. fedot ‘Lord’ and the 1st sg. absolute ending -a in subjunctives and futures [2002]............................................................. 129 More on the Celtic verb [2006]............................................................. 133 Italo-Celtic [2006]................................................................................. 149 Appendix: Old Irish verbal paradigms.................................................. 159 References............................................................................................. 179 Index...................................................................................................... 195 INTRODUCTION The night after my first arrival in Dublin in 1978 I met David Greene at the Greek restaurant on Upper Baggot Street which no longer exists. During the following weeks we discussed various topics of common interest, in particular the development of verbal categories in Celtic. When I explained my ideas about the relative chronology of sound changes and about the role of the thematic inflexion in the verbal system (cf. Kortlandt 1979a, footnotes 15 and 18), David asked me to prepare an article for Ériu, which appeared the following year (1979b). This article is reprinted here as the first chapter of the present volume. In the summer of 1979 I visited Warren Cowgill at Yale University in order to exchange views about the Celtic verb. As it became clear that the work of Dybo and Illič-Svityč was practically unknown in the West, I decided to write another article for Ériu clarifying what progress had been made in Moscow (1981a). At the same time I felt that it was necessary to treat the development of the consonantal system in more detail, which resulted in my presentation at the International Conference on Historical Linguistics in Galway in April 1981 and in my following article in Ériu (1982b). During that conference and later that month in Dublin I had the opportunity to discuss many problems with David Greene, including the continuation of Celtic Studies in the Netherlands, which at that time faced major budget cuts. When I came back to Dublin in the summer of 1981, the sad news that David had passed away shocked me deeply. At that time Daniel Binchy, Ernest Quin, James Carney and my dear friends Heinrich Wagner and Proinsias Mac Cana were still alive. The economy was in bad shape in those days, and further budget cuts and administrative problems prevented me from attending the meeting of the Indogermanische Gesellschaft in Berlin (1983) where Warren Cowgill criticized my views (1985a, 1985b). When these contributions were published, Warren had passed away, which made it difficult for me to answer his objections. In the meantime I had written two more articles for Ériu, clarifying my views on the Indo-European origins of the Old Irish subjunctives and futures (1984) and on the development of x Italo-Celtic origins and prehistoric development of the Irish language posttonic *w (1986a). At the same time I wrote my article on the Slavic imperfect (1986b), which is of major importance for a correct understanding of the Old Irish ā-preterit, and later on the occasion of the 1985 Pavia conference my little contribution on Lachmann’s law (1989a), which is relevant to the problem of Italo-Celtic. In those days we had a bright young student named Peter Schrijver who specialized in Latin and was going to write a dissertation (1991a) under the inspiring guidance of my Indo-Europeanist colleague Robert Beekes, who had been teaching Old Irish since 1981. Of course, we did everything we could to stimulate Peter’s interest in Celtic, and I felt that I should refrain from publishing on this branch of Indo-European for a number of years and give him room to develop his own line of thought. When he had clearly gone his own way (1994), I resumed my series of publications on Celtic (1994, 1996a, 1996b, 1997a, 1997b, 2000) and wrote another little contribution on Lachmann’s law (1999). The present volume contains not only articles published earlier, which are reprinted here in the order in which they were written (as indicated in the table of contents), but also discussions of additional topics and some revisions of my earlier views. Patrick Sims-Williams’ analysis of feda, fedot in the Cambrai Homily (1999) has enabled me to simplify my account of the phonological and morphological development of Old Irish somewhat. I have added a chapter on the newest scholarly literature, dealing with infixed pronouns, athematic i-presents, original aorists and perfects, suffixed pronouns, phonological developments not discussed earlier, Continental Celtic data, middle endings, and points where I have changed my opinion. The final chapter provides a discussion of the Italic data which are essential to a reconstruction of Proto-Italo-Celtic. In the appendix I present my reconstruction of the Old Irish verbal paradigms given by Strachan (1949) and Thurneysen (1946). The publication of this volume owes a lot to David Greene, who asked me to start publishing on Celtic, to Proinsias Mac Cana, who welcomed the idea of putting things together in a single volume, to Heinrich Wagner, who was a great partner in discussions of wider issues, to Fergus Kelly, who granted me hospitality at the Dublin Institute for Advanced Studies, to my Leiden colleagues Rob Beekes and Sasha Lubotsky, who were always ready Introduction xi to discuss my views, and to my wife Annie, who supported me throughout the years. I am indebted to the publishers of Ériu (Royal Irish Academy, Dublin), Études Celtiques (CNRS Editions, Paris), Münchener Studien zur Sprachwissenschaft (J.H. Röll, München), Historische Sprachforschung (Vandenhoeck & Ruprecht, Göttingen), Fs. Bräuer (Böhlau, Köln), Fs. Beekes (Rodopi, Amsterdam), Fs. Lehmann (Institute for the Study of Man, Washington D.C.), and the Pavia volume (Mouton de Gruyter, Berlin) for permission to reprint my work, to Tijmen Pronk for editing the present volume, and to Heleen Plaisier for compiling the index. Frederik Kortlandt Leiden, November 18th, 2006 THE OLD IRISH ABSOLUTE AND CONJUNCT ENDINGS AND QUESTIONS OF RELATIVE CHRONOLOGY * 1. Introduction. 2. Cowgill’s theory. 3. Chronology. 4. Loss of *-i. 5. 2nd sg. 6. Thematic flexion. 7. Greek. 8. Baltic. 9. Slavic. 10. Tocharian. 11. Latin. 12. Irish. 13. u-diphthongs. 14. i-diphthongs. 15. *ē. 16. Shortening. 17. Palatalization. 18. Raising. 19. u-infection. 20. 1st sg. 21. Shortening. 22. 2nd sg. 23. 3rd sg. 24. Plural forms. 25. Lowering. 26. Apocope. 27. Syncope. 28. Subjunctive. 29. Secondary endings. 30. Future. 31. Passive preterit. 32. Relative forms. 33. Etymology. 34. Slavic je. 35. Slavic jest. 1. Recent years have brought a considerable improvement of our insights into the prehistory of the Celtic languages. Cowgill has decisively shown how the distinction between absolute and conjunct verbal endings came about (1975). Rix has clarified the historical relation between the s-subjunctive and the a-subjunctive (1977: 153). Besides, Greene has solved a number of unclear points in the historical phonology of Old Irish and established a relative chronology of the main developments from the rise of lenition up to the end of the Old Irish period (1974 and 1976a). In this article I intend to eliminate a number of difficulties which have remained after Cowgill’s discussion of the absolute and conjunct endings and to show their chronological implications for the history of the Celtic verb. 2. Elaborating a line of thought which had been developed by Strachan, Thurneysen, Dillon, and Boling (1972), Cowgill comes to the conclusion that “the endings of the Insular Celtic present indicative, conjunct as well as absolute, come entirely from the Indo-European primary endings, and the differences between the two sets derive solely from the placement of the particle *(e)s, following Wackernagel’s Law, second in its clause: after the verb, if that was the first word, otherwise after the first preverb” (1975: 56). I have the impression that those colleagues who have not been convinced by * Ériu 30 (1979), 35-53. 2 Italo-Celtic origins and prehistoric development of the Irish language Cowgill’s demonstration do not attach sufficient weight to the fact that analogic change requires not only a model, but also a motivation. The latter is conspicuously absent in the case of the absolute and conjunct endings, which are in complementary distribution: the choice between them depends entirely on the position of the verb in the clause. The massive analogic spread of a redundant morphological distinction is simply not credible. Since Cowgill has been quite explicit about this point, I shall not take it up here. 3. Accepting the view that a particle *(e)s was incorporated in the verb form, one may wonder if the fusion can be dated in relation to other developments which have been established for the Celtic languages. The following paradigm offers two chronological indications: fo-ceird ‘puts’ < *wo-s kerde fa-ceird ‘puts him’ < *wo-s-en kerde fom(m)-cheird ‘puts me’ < *wo-s-me kerde fot-cheird ‘puts you (sg.)’ < *wo-s-tu kerde fob-ceird ‘puts you (pl.)’ < *wo-s-swis kerde The retention of t in the form with 2nd sg. infixed pronoun shows that the phonetic law which changed PIE *st into Celtic *ss had ceased to operate at the time when the particle was incorporated. The presence of b in the form with 2nd pl. infixed pronoun shows that the cluster *ssw was simplified to *sw before the lenition. Moreover, this simplification must have been anterior to the assibilation of *k in the medial cluster of seisser ‘six men’ < *sweks-wirom. Thus, the rise of the difference between absolute and conjunct verb forms can be dated to the period between the progressive assimilation in *st and the regressive assimilation in *ks. 4. The weakest point in Cowgill’s analysis is the ad hoc assumption that there was an early loss of -i in third person verb forms. According to his theory, this Proto-Celtic apocope affected 3rd sg. and pl. conjunct forms (p. 57), but not the corresponding absolute forms (p. 59). This amounts to saying that the absolute form continues the primary ending and the conjunct form the secondary ending in third person verb forms: the only difference from the The Old Irish absolute and conjunct endings 3 traditional doctrine is the view that the redistribution of the endings came about as a result of a morphologically limited phonological process. The conjectured apocope is not supported by any additional evidence. Moreover, Cowgill suggests that the early loss of *-i affected the 3rd sg., but not the 3rd pl. relative form (p. 59). It seems preferable to say that the relative forms remain to be explained. 5. The 2nd sg. forms are not satisfactorily accounted for either. Conjunct -bir can phonetically represent both *bherei and *bheresi, as Meillet pointed out already (1908: 413). The latter reconstruction, which Cowgill adopts, leaves the endings of the present classes AI, AII, AIII, BII, BIV unexplained. While AIII conj. -taí ‘are’, -gní ‘do’, -soí ‘turn’ can be regular from *tāsi, *gnīsi, *sowesi, Cowgill is forced to regard abs. cíi ‘thou weepest’ either as an irregular spelling or as an analogical formation on the basis of the corresponding conjunct form (p. 61). On the other hand, he has to suppose that AII -léici ‘leave’ and BII -gaibi ‘take’ are levelled absolute forms, to be derived from *lēggīsi-s and *gabisi-s. The problem is even more considerable for the AI and BIV ending -(a)i: “The apparent contrast between -(a)i from *-asi in 2sg. pres. as-renai ‘impendis’ Ml 44a 6 and zero, preceded by vowel raising and consonant palatalization, from *-ai in tuil is hard to work into a plausible chronology. The solution requiring the least amount of analogic change seems to be to suppose that -renai is originally an absolute form, leveled into conjunct position also, and analogic for *rini < *rinīh < *rinais < *rinasi-s, with /en/ for */iń/ after the rest of the present indicative” (Cowgill 1975: 57, fn. 13). Thus, all sounds of -renai except the initial consonant are analogic. 6. The difficulties in Cowgill’s theory can be eliminated if we return to Meillet’s view that the difference between conjunct and absolute endings reflects in part the distinction between the thematic and the athematic flexion of the proto-language (1907). Since the thematic paradigm, with the exception of the 1st sg. form, adopted the athematic endings in Indo-Iranian, Italic, and Germanic, the evidence from these languages cannot be used for the reconstruction of the original thematic flexion. Such a reconstruction must necessarily be based upon Celtic, Baltic, Slavic, Tocharian, and Greek, all of which point to a 3rd sg. ending *-e. The combined evidence of these 4 Italo-Celtic origins and prehistoric development of the Irish language languages also points to a 2nd sg. ending *-eHi and a 3rd pl. ending *-o in the thematic paradigm. Moreover, the supposition that these endings once existed in Italic eliminates the necessity for an ad hoc assumption that *-i was lost in finite verb forms. 7. In Greek, the endings of the thematic present are: 3rd sg. -ei, 2nd sg. -eis, 3rd pl. -onti. The 3rd sg. ending is best explained as PIE *-e plus an additional i from the athematic flexion (cf. Chantraine 1967: 297). The motivation for the enlargement can be found in the obliteration of the distinction between primary and secondary endings as a result of the loss of final *t. The 2nd sg. ending is derived from *-ei plus an additional s from the secondary endings, which was also added in the athematic present. The additional -nti in the 3rd pl. ending was apparently borrowed from the athematic flexion on the basis of the secondary ending *-nt, which was common to both flexion types. 8. I shall be brief about the Baltic and Slavic material, which I have discussed in detail elsewhere (1979a). The Lithuanian endings are: 3rd sg. -a, 2nd sg. -ì, 3rd pl. -a. The remarkable correspondence of je/o-verbs in Baltic with e/o-verbs in Slavic and Sanskrit can be explained if we assume that the 3rd sg. ending has replaced earlier *-e. The 2nd sg. and 3rd pl. endings are phonetically regular. I cannot share the widespread view that the original 3rd pl. form was lost in Baltic. If the nt-endings once had the same extension here as in the southern and western Indo-European languages, their disappearance would be totally unmotivated. On the other hand, the addition of *-nti to an original 3rd pl. ending *-o in Indo-Iranian, Greek, and Germanic is a trivial innovation. 9. The Slavic material is complicated. The 3rd sg. ending *-e has been preserved in all dialects except those of western Macedonia, which include the dialect of the Old Slavic translation of the Gospel, and the Russian dialects on which the standard language is based. Its antiquity is evident from the Novgorod birch bark documents. The 2nd sg. form of the copula esi must be derived from *esei, where -ei represents the original thematic ending (cf. Van Wijk 1916: 111f.). The Old Bulgarian ending -ši resulted from a blending of the athematic and the thematic ending. The original 3rd pl. ending *-o was The Old Irish absolute and conjunct endings 5 enlarged with *-nti from the athematic flexion, as in Greek, but the earlier form can still be inferred from the chronology of the Slavic developments. The addition of *-nti must have taken place at a relatively recent stage because it was posterior to the generalization of the secondary ending in the so-called semi-thematic present. For the details I refer to the article mentioned above. 10. Tocharian preserves the 3rd sg. ending *-e in B āśäṃ (A *āśäs) ‘agit’, where -ṃ(-s) is an enclitic element (cf. Pedersen 1941: 142). The 3rd pl. form B ākeṃ ‘agunt’ contains the ending *-o before the clitic. The ending cannot be identified with PIE *-ont, which is preserved in kameṃ ‘came’ and lateṃ ‘went’, because the distinction between primary and secondary endings was not lost in Tocharian and *-nti is preserved in A -ñc. The latter dialect added *-nti to *-o in ākeñc, which therefore shows a deceptive similarity to the corresponding ending in Greek and Slavic, but preserved the original ending in a considerable number of instances, e.g. tāke next to tākeñc ‘will be’ (cf. Sieg c.s. 1931: 326ff.). The short forms are especially frequent in the Maitreyāvadānavyākaraṇa, which is archaic for other reasons as well: it still uses the śä-doublet, which was apparently eliminated in the other texts because of its resemblance with the ṣä-doublet and the ya-sign (Pedersen 1941: 19), and writes krañc and lāñc for kraṃś and lāṃś, also krañcän for krañcäṃ, and often ī, ū for i, u (Sieg c.s. 1921: viii). 11. The endings of Latin agit, agis, agunt cannot be derived from *-eti, *-esi, *-onti because *-i is not lost in this language, cf. mare, loc. pede, inf. amāre. The simplest assumption is that the secondary endings *-t (*-d), *-s, *-nt were added to the original thematic forms in *-e, *-ei, *-o. This hypothesis also accounts for the form esed ‘erit’ on the cippus from the Forum Romanum (circa 500 B.C.), where the final consonant remains unexplained in the traditional doctrine. When the athematic present endings lost their *-i on the analogy of the corresponding thematic (and secondary) forms, the 2nd sg. and 3rd pl. endings merged in the two paradigms. The 3rd sg. endings became confused in the fourth century. The theory advanced here may also explain the difference between Umbrian tiçit ‘decet’ < *-ēti and heri ‘vult’ < *-ie. 6 12. Italo-Celtic origins and prehistoric development of the Irish language A different development must be assumed for Celtic. The thematic 3rd sg. ending *-e is preserved in fil ‘there is’, as Watkins has convincingly argued (1969: 168). The corresponding absolute form, which represents *wele-s, is attested in Wb 11d 2 fil ní de as fír ‘there is something thereof which is true’. This form shows that the ending had no final *t and that the 3rd sg. relative form beres ‘who carries, that he carries’ cannot be derived from *beret-sa(n). The other thematic verbs inserted *-ti from the athematic flexion before the absolute suffix *-s, e.g. berid ‘carries’ < *bere-ti-s versus -beir < *bere. As Cowgill pointed out already (1975: 59), the absolute form cannot be derived from *beret-es because the latter reconstruction would yield the wrong final vowel in the form with 3rd sg. suffixed pronoun beirthi ‘carries him’ < *bereti-s-en. Since the reason for the insertion of *-ti before the absolute suffix *-s must be sought in the interaction of the thematic and the athematic flexion which originated from the shortening of long final vowels, I have to make a digression on the historical phonology of Irish here. I shall refer to the stages of Greene 1974 as G1-G11. 13. Earlier investigators have observed that the loss of intervocalic *s was anterior to the monophthongization of the Indo-European u-diphthongs (cf. Jackson 1953: 313 and Greene 1976a: 27), e.g. tauë ‘silence’ (Welsh taw) < *tawia < *tausiā. The loss of intervocalic *s was probably posterior to its reduction to *h as a result of the lenition. On the other hand, the rise of *ō2, which resulted from the monophthongization of the u-diphthongs, must have been posterior to the split of *ō1 (PIE *ō) into *ū in final syllables and *ā elsewhere. (Note that ó1 and ó2 of Greene 1976a: 28 correspond with my *ō2 and *ō4, respectively.) The development is similar to what we find in Slavic, where the u-diphthongs were monophthongized into *ō2 (later u) at a stage when *ō1 (PIE *ō) had become *ū (later y) before nasals in final syllables and *ā (later a) elsewhere, e.g. kamy ‘stone’ < *akmōn, dati ‘to give’ < *dōtei. Thus, we arrive at the following relative chronology: (1) Lenition (G2): rise of *h from PIE *s. (2) Loss of intervocalic *h. ANTE (3) Split of *ō1 into *ā and *ū. (3) Monophthongization of u-diphthongs: rise of *ō2. The Old Irish absolute and conjunct endings 7 14. There is no reason to separate the monophthongization of the idiphthongs chronologically from that of the u-diphthongs. In stressed syllables, *ē2 from *ei did not merge with *ē1 (PIE *ē), which was raised to *ī. The development is typologically comparable to what we find in certain varieties of Dutch, where ei is monophthongized to [ε:], while ee remains close [e:]. Stressed *ai and *oi were not affected by the monophthongization, which suggests that the u-diphthongs had merged into *ou before the rise of *ō2. In unstressed syllables, the i-diphthongs merged with *ē1 and *ī, e.g. nom.pl. fir ‘men’ < *wirī < *wiroi, dat.sg. tuil ‘will’ < *tolī < *tolāi. Since *-āi from *-āsi did not merge with PIE *-āi (see below), I assume that the latter had been shortened to *-ai before the loss of intervocalic *h. However, *-ōi did not merge with *-oi, e.g. dat.sg. fiur < *wirū < *wirōi. It is therefore reasonable to suppose that the shortening of long final diphthongs was posterior to the raising of *ō1 to *ū in final syllables. This hypothesis is supported by the Gaulish dat.sg. ending -ui. I find no evidence against the merger of *-ei and *-esi, cf. especially dat.sg. tig ‘house’ < *tegī < *tegesi. Since there is no evidence for a different treatment of prevocalic and preconsonantal *ei, the loss of intervocalic (consonantal) *i must be dated after the monophthongization. We now arrive at the following relative chronology: ANTE (1) Split of *ō1 into *ā and *ū. (1) Shortening of long final diphthongs. (2) Loss of intervocalic *h. (3) Monophthongization of i-diphthongs: rise of *ē2. POST (3) Loss of intervocalic *i. 15. I do not share the common view that *ē1 had been raised to *ī in Proto- Celtic times already. An early merger of *ē1 and *ī would have yielded a phonological system where the vowel height oppositions between the short vowels outnumbered those between the long vowels. Though such a system is by no means impossible, it is not probable that it would have remained in existence over a longer period of time. Moreover, Gaulish shows e for *ē1 in a number of instances, e.g. Dubno-rex. It seems better to connect the raising of *ē1 with the development of the i-diphthongs in the separate languages. 8 Italo-Celtic origins and prehistoric development of the Irish language The development of PIE *-oi and *-āi into *-ī suggests that *ē1 and *ē2 merged in unstressed syllables before the raising of *ē1 to *ī. I find no evidence for e from *ē1 in final syllables. In particular, carae ‘friend’ < *karēh < *karants does not contain *ē1. This *ē, which I shall write *ē3, is also found in fiche ‘twenty’ < *wikēh < *wikent and gen.sg. abae ‘river’ < *abēh < *abens. I conclude that the rise of *ē3 from *en and *an before a dental consonant was posterior to the raising of *ē1 and *ē2 to *ī in unstressed syllables. It was also posterior to the raising of *ē1 in stressed syllables because *ē3 merged neither with *ē1 nor with *ē2, e.g. cét ‘hundred’ < *kenton versus íasc < *peiskos: *ē3 was apparently lower than*ē2, just as the latter was lower than *ē1. The open character of *ē3 is not unexpected since *en and *an merged, e.g. géis ‘swan’ (Latin ānser). The long vowel of cét shows that the loss of the nasal in *nt (G1) cannot have been anterior to the lenition (G2). The nom.sg. athair ‘father’ for *aither < *patēr is easily explained as an analogic form. I assume that the word underwent palatalization metathesis so as to conform to the pattern of the i-stems. Thus, we can add: (4) Raising of *ē1 to *ī. (5) Loss of *n before dentals and velars: rise of *ē3. There is evidence for *ō3 (which apparently merged with *ō2) in trícho ‘thirty’ < *trīkont and cano ‘poet’ < *kanonts. It should be clear that final *ē3 and *ō3 cannot represent PIE *-ent and *-ont because final *t had been lost at an early stage, as is evident from the merger of the perfect with the thematic aorist. The restoration of final *t in the secondary 3rd pl. ending, where it had been preserved before PIE clitics, was apparently posterior to stage (5). Another source of *ē3 is found in the absolute 2nd pl. form beirthe ‘you carry’ < *beretēh < *beretes-es. 16. Greene assumes that unstressed long vowels were shortened except in final syllables ending in *h (G3). It is typologically improbable, though not impossible, that distinctive quantity was preserved during a considerable period of time in closed final syllables only. Moreover, the history of the verbal flexion is more easily accounted for if we assume that vowel length in medial syllables was preserved up to a later stage. Thus, I suggest that the early shortening of long vowels was limited to word-final position. The The Old Irish absolute and conjunct endings 9 raised vowel in dat.sg. tuil ‘will’ < *tolāi shows that the shortening was posterior to stage (4). There is no direct evidence for its chronological relation to stage (5) because word-final *ē3 did not arise phonetically. A cogent argument can be derived from the 1st sg. conjunct ending of the asubjunctive, e.g. -ber ‘I carry’. As will be pointed out below, this form must be derived from *berason. When *-an from PIE *-ām had merged with *-en, e.g. in acc.sg. túaith ‘people’ < *tōten < *teutām, earlier *-on developed into *-an. After the loss of intervocalic *s (2), the form contracted to *berān in the same way as *beretes-es yielded *beretēh. Since the latter contraction cannot have been anterior to the rise of *ē3 (5), the former must not be dated earlier either. When *n was lost before dentals and velars (5), the nasal mutation became a morphological process (G8c). Incidentally, this chronology explains why *n disappeared before initial *w: the latter was still a resonant at this stage. The shortening of the long vowel in *berā n- can now be identified with the general shortening of long final vowels, which is consequently posterior to stage (5). The resulting short vowel was apocopated at a later stage (G8a). I conclude that we can add: (6) Long final vowels were shortened. I also assume that final *e was lost after a long vowel, which can be viewed as a corollary of (6). This rule affected the 3rd sg. conjunct form of weak verbs, e.g. -marba ‘kills’ < *-ā < *-āe < *-āie and -rádi ‘speaks’ < *-ī < *-īe < *-eie. This loss of *-e, which requires the preservation of distinctive quantity in the prefinal syllable, must have been posterior to stage (6) because it reintroduced word-final long vowels. The loss of intervocalic *i must be dated between (3) and (6). 17. The rise of palatalization in Irish has largely been clarified by Cowgill (1969) and Greene (1974). I summarize their findings as follows: (7a) All consonants were palatalized between front vowels and before stressed front vowels. (7b) Dentals were palatalized before posttonic *i. (7c) Labials and velars were palatalized before posttonic *i unless they were preceded by a back vowel. 10 Italo-Celtic origins and prehistoric development of the Irish language Examples: (7a) -beir /b’er’/ ‘carries’< *bere, caíche ‘blindness’ < *kaixia < *kaikiā, (7b) -ráidiu ‘I say’ < *rādīu < *rōdeiō, tuirem ‘enumeration’ < *torīma, calad ‘hard’ < *kaleθah (Welsh calet), (7c) -gaibet ‘they take’ < *gabiot, gábud ‘danger’ < *gābiθuh, tugae ‘cover’ < *togia.1 As is clear from these examples, long *ā was a back vowel at this stage, whereas short *a was neutral with respect to the opposition between front and back vowels. Following Thurneysen and Cowgill, Greene assumes that a preceding short *u did not block the palatalization of dentals by a following *e (G5b). This assumption forces him to date the vowel height assimilation in stressed syllables before the rise of palatalization, cf. sonairt ‘strong’ < *sunertih and muinél ‘neck’ < *monixlah. On the other hand, the vowel height assimilation in unstressed syllables must be dated after the rise of palatalization, as is clear from the same examples. For the intermediate period, this chronology requires the simultaneous existence of a five-vowel system in unstressed syllables and a three-vowel system under the stress, which is a very unlikely reconstruction. Moreover, it does not account for the absence of palatalization in Iudei ‘Jews’, gen. Iudae. If this word had not yet been borrowed into the language at this stage, the unpalatalized obstruent would be all the less comprehensible in view of the rising tide of palatalization, cf. aiccent < Latin accentus, where e palatalized the preceding velar. Cowgill adduces two instances in support of the hypothesis that *u did not block the palatalization of a following dental by *e (1969: 35): do-fuisim ‘pours forth’ and tuisel ‘stumble’, which he derives from *to-uss-semet (with analogic f) and *t-uss-swelas, respectively. But the first word has evidently taken its palatalized obstruent from do-essim ‘pours out’ < *to-ess-seme, where it arose phonetically, and the etymology of the second word is probably incorrect because *-ssw- yields -b- in the 2nd pl. infixed pronoun, e.g. fob-ceird ‘puts you’ < *wo-s-swis-kerde. I conclude that the formulation of the palatalization rule given above is not only simpler and more natural, but also closer to the facts than earlier formulations. 1 Intervocalically, I write *t and *d where other authors use *d and *δ, or *dd and *d, respectively. I write single and double consonants for intervocalic lenis and fortis resonants, but single consonants in those positions where fortis resonants are not in phonemic opposition to lenis ones. The Old Irish absolute and conjunct endings 11 18. Certain vowel features diffused through the preceding consonant to the vowel of the preceding syllable. The opposition between high and mid short vowels was neutralized if the following syllable contained a high vowel (G4). When the phonemic contrast was reintroduced in stressed syllables, the product of the neutralization merged with the corresponding high vowel, e.g. biru ‘I carry’ < *berūh, muinél ‘neck’ < *monixlah. I see no evidence for a different treatment of stressed and unstressed vowels except for the fact that the raising of unstressed *e to *i was blocked by a preceding unpalatalized consonant, a situation which did not occur in stressed syllables, cf. sonairt ‘strong’ < *sunertih, where the lowering of *u to o shows that non-high *e had been preserved up to a later stage (see below). This is the origin of the difference between the vocalic alternation in cingid ‘steps’ < *kingeθih, 3rd pl. cengait < *kingatih and the constant vocalism of bongid ‘breaks’ < *bungeθih. The raising of *e and *o to i and u before high vowels was certainly posterior to the rise of *ē3 (5), cf. sét ‘way’ < *sentuh versus rind ‘star’ < *rendu. It was probably posterior to the rise of palatalization (7) because *i is likely to have palatalized a preceding consonant before it affected the vowel of a preceding syllable. If one accepts that the raising was not limited to stressed syllables, a cogent argument can be derived from sonairt, where the nasal would have been palatalized if the vowel of the medial syllable had been raised before the rise of palatalization. Thus, I add: (8) Raising of short *e and *o before a high vowel in the following syllable. 19. Not only the vowel height, but also the rounding of *u affected the vowel of the preceding syllable. The resulting u-infection became phonemically relevant in those instances where the conditioning factor was lost as a result of subsequent phonological processes after having sufficiently affected the preceding vowel. This was the case in gen.sg. caurad ‘warrior’ < *karuθah, where au was phonemicized as a result of the lowering of *u to *o (see below). It was also the case when the vowel of the prefinal syllable was short and final *u was apocopated at a later stage, e.g. in dat.sg. fiur ‘man’ and the conjunct form -biur ‘I carry’. The latter word suggests that the raising of *e to *i was anterior to the u-infection. When the final vowel was not apocopated, the infection was not phonemicized, e.g. acc.pl. firu ‘men’ < *wirūh and absolute biru < *berūh. As Greene has demonstrated (1976a: 29), 12 Italo-Celtic origins and prehistoric development of the Irish language intervocalic *w also produced u-infection, e.g. auë ‘grandson’ < *auweah < *awios. The word nuë ‘new’ < *nuweah < *nowios shows unequivocally that the u-infection was posterior to the raising of *o to *u before *i in the following syllable, cf. gáu ‘falsehood’ < *gouwa < *gowā, gen. gue < *guwiāh < *gowiās with *-iās replacing original *-ās. I therefore add: (9) u-infection (G7b). 20. Here I have to discuss the 1st sg. form of the consuetudinal present biuu, -bíu ‘am wont to be’. Thurneysen and Greene write bíuu, but the form in Wb 16d 8 biuu-sa, to which Thurneysen refers, is written without an accent mark in the Thesaurus. The vowel must originally have been short, as is clear from Welsh byddaf. In Irish, there is no reason to assume an intervocalic glide since the elimination of consonantal *i between stage (3) and stage (6), cf. above. At the time of u-infection, it is reasonable to suppose that a subphonemic u-glide developed before postvocalic *u, so that we can write *biuūh, *-biuu, also *-gnīuu ‘I do’ < *gnēiō. Both the fact that the glide did not merge with *w and the u-infection before *w suggest that *w became a fricative around this time. When final *u was apocopated (see below), the uglide in *-biuu and *-gnīuu became phonemically relevant in the same way as the u-infection in -biur. The regular lengthening of the vowel in the former word yielded the historical form -bíu, with the same vocalism as -gníu, cf. also clé ‘left’ < *kleah < *klios, dat.sg. clíu < *kliuu, and béu ‘living’ < *beuw < *biuwah < gwiwos, dat.sg. bíu < *biuw < *biwu. Thus, I agree with Boling (1972: 100) that the form -gníu is phonetically regular. The absolute form biuu differs from biru in the presence of u instead of r only and can hardly be analogic because there was no motivation for a morphological innovation. I see no evidence for a different treatment in posttonic syllables, cf. centarach ‘hither’, comparative centarchu < *k’enoθerax’u < *k’enoθerax’iuu < *kinoθerāxiūh: this word underwent the lenition at stage (1), the palatalization of *k and *x at stage (7), the rise of the u-glide at stage (9), the shortening of *ā at stage (10), the lowering of *i in the initial syllable at stage (11), the shortening of *ū at stage (14), the loss of the second *i at stage (16), the syncope of *o and *a at stage (19), and finally the delenition of *θ and the depalatalization of *x’. The word toimtiu ‘opinion’ < *tomet’u < *tomētiu < *to-mentiō underwent the raising of *ō before stage (1), the loss of *n and rise of *ē3 at stage (5), the shortening of final *ū at stage (6), the The Old Irish absolute and conjunct endings 13 palatalization of *t at stage (7), the rise of the u-glide at stage (9), the shortening of *ē at stage (10), the apocope of *u at stage (15), the loss of *i at stage (16), the palatalization of *m at stage (18), and the syncope of *e at stage (19), cf. below. In the same way, the 1st sg. abs. and conj. endings of weak verbs AI -u < *-āiō(-s) and AII -iu < *-eiō(-s) and the BII ending -iu < *-iō(-s) represent phonetically regular developments, cf. also -táu ‘am’ < *stāiō. 21. We now arrive at the shortening of posttonic long vowels in non-final syllables. This shortening must have been posterior to the u-infection because the latter did not affect comet ‘preservation’ < *komētuh < *komentus, cf. tomus ‘measure’ < *tomeus < *tomessuh, where *e was lost at stage (16). The u-flexion of comet is evident from Ml 55d 6 a-chometa ‘of his protection’. Thus: (10) Long vowels in medial syllables were shortened. As a result of this shortening, the thematic flexion of class AI (*-āie-) merged with the athematic flexion of class BIV (*-na-) in a number of forms. The thematic flexion of AII verbs (*-eie-), which had merged with the athematic flexion of AII verbs (*-ē-) in a number of forms as a result of the shortening of final long vowels at stage (6), now merged with the thematic flexion of BII verbs (*-ie-) in the remaining forms. The obliteration of the distinction between thematic and athematic flexion led to a reshuffling of the two sets of endings. 22. The PIE 2nd sg. ending has been preserved in BI -bir ‘carry’ < *beri < *berei, abs. biri < *-īh < *-ei-s, AI -marbai ‘kill’ < *-āi < *-āiei, abs. marbai < *-āīh < *-āiei-s, AII -rádi ‘speak’ < *-īi < *-eiei, abs. rádi < *-īīh < *-eiei-s. The latter endings replaced the athematic AII endings *-ī < *-ēsi and *-īih < *-ēsi-s after the shortening of long final vowels at stage (6) and merged with the BII endings *-ii < *-iei and *-iīh < *-iei-s when the medial long vowel was shortened at stage (10). The latter shortening may have evoked the analogical replacement of the BIV endings *-i < *-asi and *-īh < *-asi-s with the AI endings *-ai, *-aīh. More probably, however, the characteristic vowel *-a- of class BIV had already been reintroduced on the basis of the other athematic verbs at an earlier stage. The substitution of the thematic for the athematic ending was apparently total. The absence of 14 Italo-Celtic origins and prehistoric development of the Irish language raising in Ml 110d 9 do-eim ‘protectest’ shows that the form replaces an original athematic present *ēsi, with *ē3 from *en since stage (5). The forms cíi ‘weepest’ and -taí ‘art’ represent the regular development of the thematic ending. Only the copula at < *ē-tu with *ē2 < *ei < PIE *esi preserves the original athematic ending. 23. On the basis of the foregoing paragraphs we arrive at the following reconstruction of the Irish present tense at stage (7). 1st sg. abs. 2nd sg. abs. 3rd sg. abs. 1st sg. conj. 2nd sg. conj. 3rd sg. conj. 1st sg. abs. 2nd sg. abs. 3rd sg. abs. 1st sg. conj. 2nd sg. conj. 3rd sg. conj. *bere- ‘carry’ *marwāie- ‘kill’ *bina- ‘strike’ berūh berīh bereh beru beri bere marwāūh marwāīh marwāeh marwāu marwāi marwā binamih bin(a)īh binaθih binami bin(a)i binaθi *gabie- ‘take’ *rōdeie- ‘say’ *rudē- ‘redden’ gabiūh gabiīh gabieh gabiu gabii gabie rādīūh rādīīh rādīeh rādīu rādīi rādī rudīmih rudīīh rudīθih rudīmi rudīi rudīθi At this stage, the final *e of *gabie was apparently eliminated on the analogy of the weak verbs. The two types of ī-flexion merged through the generalization of 3rd sg. abs. *-īθih and conj. *-ī. The element *-θi was perhaps reinterpreted as a clitic, which was incompatible with the conjunct form. The athematic conjunct ending may have been preserved in co cóic séotu cingith ‘it extends to five chattels’ (cf. Binchy 1971: 160). When medial long vowels were shortened at stage (10), the absolute ending *-θih spread to the BII and AI verbs on the analogy of the AII and BIV verbs. The spread of *-θih to class BI may have taken place at a relatively recent stage: the original absolute ending has been preserved in Wb 11d 2 fil ‘there is’. The Old Irish absolute and conjunct endings 15 24. The plural endings cannot be reconstructed with the same chronological precision. The thematic 3rd pl. ending *-o was replaced with the secondary ending when the corresponding 3rd sg. endings had merged as a result of the early loss of word-final *t. The phonetic reflex *-on of PIE *-ont was later replaced with *-ot, which was the reflex of *-ont- before a clitic. The latter replacement was posterior to the rise of *ō3 < *-ont at stage (5), e.g. in trícho ‘thirty’ < *trīkont. It follows from this word, where the rise of final *t was posterior to the loss of PIE final *t, that the ending of -berat ‘they carry’ cannot be derived phonetically from PIE *-ont. The final consonant of dét ‘tooth’ < *dents, -bert ‘bore’ < *bert, do-r-ét ‘has protected’ < *dē-ro-ent (*em-), do-rósat ‘has created’ < *to-ro-uss-sent (*sem-) is also due to restoration. It is clear from these examples that the creation of the t-preterit, which was apparently posterior to the loss of interconsonantal *s (cf. echtar ‘outside’ < *ekster) and to the assimilation of *st to *ss, was anterior to the loss of final *t. The new 3rd pl. conj. ending *-ot spread to the athematic flexion in accordance with the general tendency toward generalization of the thematic endings. The original athematic ending has been preserved in the copula it < *ēti < *senti. In the absolute forms, the generalization of *-otih must be viewed in connection with the substitution of 3rd sg. *-θih for *-eh in the weak verbs. The model of 1st sg. *-mih, 2nd sg. -īh, 3rd sg. -θih, and 3rd pl. *-otih evoked the replacement of the 1st pl. ending *-moeh < *-mos-es with *-moih, e.g. bermai ‘we carry’. The latter development did not affect the 2nd pl. ending *-θēh < *-tes-es, which had received a long vowel at stage (5), e.g. beirthe ‘you carry’. The lenition after nídan ‘we are not’ suggests that we have to reckon with an earlier ending *-mo next to *-mos, the distribution of which can no longer be ascertained. 25. After the shortening of posttonic long vowels in non-final syllables, the opposition between high and mid short vowels was neutralized if the following syllable contained a non-high vowel (G4-6). When the phonemic contrast was reintroduced, the product of the neutralization merged with the corresponding mid vowel, e.g. fer ‘man’ < *wirah, cloth ‘fame’ < *kluθan, sonairt ‘strong’ < *sunertih. Here again, I see no evidence for a different treatment of stressed and unstressed syllables. The lowering was blocked by an intervening palatalized consonant, e.g. voc.sg. fir < *wire, fiche ‘twenty’ < *wixēh. This is the origin of the difference between the high vowel in cingid 16 Italo-Celtic origins and prehistoric development of the Irish language ‘steps’ < *kingeθih and the lowered vowel in 3rd pl. cengait < *kingatih, bongid ‘breaks’ < *bungeθih, conj. -boing < *bunge, nom.pl. coin ‘hounds’ < *kuneh. The palatalization in the latter words had not yet come into existence at this stage. (The argumentation of Kortlandt 1978b: 297, n. 18 cannot be maintained.) The lowering of *i and *u to *e and *o was obviously posterior to the rise of palatalization at stage (7), cf. aile ‘other’ < *aliah/-ia < *alios/-iā versus calad ‘hard’ < *kaleθah, acc.sg. máthair ‘mother’ < *māθeren. It was also posterior to the raising at stage (8), e.g. uile ‘all’ < *oliah/-ia, muinél ‘neck’ < *monixlah. The raised vowel was not lowered in these words because the medial vowel was not distinctively non-high at the time of neutralization and because the intervening consonant was palatalized. The lowering can even be dated after the shortening of long vowels in medial syllables (10) because it affected the medial vowel of the suffix *-tūt- before the gen.sg. ending *-ah and the acc.sg. ending *-en, e.g. oíntu ‘unity’ < *oinoθūh, gen. oíntad, acc. oíntaid. The absence of lowering in the first syllable of uilen ‘elbow’ < *olīna does not provide counter-evidence against this chronology because the word is of the same type as muinél. Thus, I add: (11) Lowering of short *i and *u before a non-high vowel in the following syllable. 26. Following the course of events we now approach the apocope. The loss of short final vowels was preceded by their merger into some kind of schwa, as a result of which long final vowels lost their distinctive quantity. The colour of the short vowels was partly preserved after their merger because front vowels palatalized the preceding consonant and the latent u-infection of a preceding short vowel became phonemically relevant. I conclude that we can add: (12) Palatalization of all consonants before *i and *e in final syllables (G7a). (13) Reduction of short vowels in final syllables: rise of schwa. (14) Shortening of long vowels in final syllables (G8b). (15) Apocope: loss of final schwa (G8a). Examples: ball ‘member’ < *ballah, gen. baill < *ball’i, dat. baull < *ballu, voc. baill < *balle, -cain ‘sings’ < *kane, canaid < *kaneθ’ih, luib ‘plant’ < The Old Irish absolute and conjunct endings 17 *lubih, nom.pl. lubai < *lubīh. The loss of final *h and concomitant rise of lenition as a grammatical process (G8c) can be dated anywhere between stage (6) and stage (14). I think that it was a gradual development. The lenition of initial *s was already grammaticalized simultaneously with the nasal mutation at stage (5) because eclipsed *s merged with unlenited *s, so that the choice between initial *h and initial *s was no longer dependent on the presence or absence of a preceding vowel. 27. Since the subsequent developments are of minor significance for the history of the absolute and conjunct endings, I list them here without comment and refer to Greene 1976a: 31ff. (16) Reduction of vowel sequences and coalescence of preverbs. Example: tomus ‘measure’ < *tomeus < *tomesuh. (17) Loss of fricatives before resonants and compensatory lengthening (G9): rise of *ē4 and *ō4. Example: muinél ‘neck’ < *mun’exl < *monixlah. (18) Reduction of short vowels in medial syllables to schwa with partial preservation of the vocalic timbre in the preceding consonant (G10). Example: gen.sg. toimseo ‘measure’ < *tomeso < *tomesōh. (19) Syncope: loss of schwa in weak syllables (G11). (20) Loss of intervocalic *w. (21) Diphthongal shift. (22) Reduction of hiatus. The palatalization assimilation in consonant clusters can be dated after stage (19). 28. Both the s-preterit and the t-preterit are most easily derived from 3rd sg. aorist forms, to which the primary thematic endings were added in order to supply the 1st and 2nd sg. forms (and also the plural forms of the s-preterit). The PIE secondary thematic endings have been preserved in the asubjunctive, which is historically identical with the s-subjunctive (cf. Rix 18 Italo-Celtic origins and prehistoric development of the Irish language 1977: 153). The characteristic vowel of the a-subjunctive represents the final laryngeal of seṭ-roots before the suffix of the s-subjunctive. Since the subjunctive mood had thematic endings, *s was lost between the reflex of the laryngeal and the thematic vowel at stage (2). The flexion of the s-subjunctive was evidently reshaped on the pattern of the s-preterit. The original 1st sg. absolute ending has been preserved in the s-future, where it was supported by the other future paradigms. 29. When final *t was lost in Proto-Celtic, the secondary 3rd sg. ending *-et merged with the primary thematic ending *-e. As a result of this merger, the absolute form *beraeh < *berase-s was replaced with *beraθih in the same way as *marwaeh was replaced with marwaθih after stage (10). The conjunct form *berae was replaced with *berā on the analogy of *marwā. The derivation of the 2nd sg. ending from both *-ases and *-ases-es presents no difficulties: after the regular development to *-aeh and *-aēh and the apocope, which yielded *-e and *-ae, the endings merged into -e at stage (16). The 1st sg. conjunct form -ber cannot be derived phonetically from *berām because *-ām yielded *-en, cf. acc.sg. túaith ‘people’ < *tōθen < *teutām. As I pointed out above, the form can represent the regular development of *berason, which was reduced to pre-apocope *bera n- as a result of the loss of intervocalic *s at stage (2), the lowering of *-on to *-an and its coalescence with the preceding *a into *-ān, the rise of the nasal mutation at stage (5), and the shortening of *-ā at stage (6). The suggestion that the absolute form bera originates from a reshaping on the basis of the conjunct form is not convincing because there is no motivation for such an analogic development. I consider it more likely that we have to start from the hypothesis that the absolute suffix was *s after vowels and nasals and *es after obstruents. This rule is typologically comparable with the elision of e in Latin -a est, -um est. The form *berason-s developed into *beraōs at stage (5), and the latter may have yielded the expected pre-apocope form *berāh at stage (6). 30. The future paradigm requires some discussion because the origin of the weak f-future has not finally been elucidated. According to the most plausible theory, -f- is the phonetic reflex of intervocalic *-bw- (Sommerfelt 1922). The suffix before the thematic ending must have been *-ibw- or *-ībw- in The Old Irish absolute and conjunct endings 19 view of the palatalization in such instances as Wb 14a 8 ainfa ‘I will stay’ < *anibwāh, cf. anaid ‘stays’ < *anaθih, which is to be compared with Skt. ániti ‘breathes’. In accordance with the rules given in section 17 of this article, the consonant was not palatalized in such forms as Wb 12d 3 -tucfa ‘he will understand’ < *to-ukībwā. The cluster *bw became palatalized before the front vowel of the 2nd and 3rd sg. endings, but not before the 1st sg. ending. When *w became a fricative around stage (9), the cluster merged with intervocalic *hw < *sw into fortis *w, which can be written *ww. The latter caused u-infection in the same way as lenited *w. If we assume that the ffuture had the same endings as the a-subjunctive, we arrive at the following reconstruction of the two paradigms at stage (10): 1st sg. 2nd sg. 3rd sg. abs. -iuwwōh -iuwwēh -iuwweh conj. -iuwwa n-iuwweh -iuwwe abs. -āh -aēh -aeh conj. -a n-aeh -ae It is probable that the 1st sg. ending *-iuwwōh was replaced with *-iuwwāh around this stage. The 3rd sg. ending *-iuwweh was replaced with *-iuwweθih when *bereh was replaced with *bereθih. This leads us to the following reconstruction at stage (16): 1st sg. 2nd sg. 3rd sg. abs. -’ufa -’uf’e -’uf’eθ’ conj. -’uf -’uf’ -’uf’ abs. -a -e -eθ’ conj. -Ø -e -a At this stage, the 2nd and 3rd sg. conjunct endings of the a-subjunctive, which were used in the reduplicated future and the ē-future already, replaced the zero endings in the corresponding forms of the f-future. There is no need to connect the u-infection in the 1st sg. conjunct form with the primary thematic ending. The 3rd sg. conjunct ending *-a was also restored in the subjunctive of all verbs, e.g. -lécea ‘leaves’, cf. the phonetic development in gen.sg. guide ‘prayer’ < *-iāh (pace Cullen 1972: 229). The 1st sg. endings are likewise due to restoration in the subjunctive of this class. 31. Thus far I have left the relative forms out of consideration. One of the most remarkable facts about the relative forms is their coincidence with the 20 Italo-Celtic origins and prehistoric development of the Irish language absolute form in the passive preterit, but with the conjunct form in the other passive and deponent paradigms. Greene has recently drawn attention to the interesting syntactic homonymy which results from this coincidence (1976b), e.g. digéni cummen cétaig ríthae friéladach, which can be translated either ‘Cummen made a coat which was sold to Éladach’ or ‘Cummen made a coat. It was sold to Éladach’. I cannot share Greene’s conclusion that the relative form came to be used in absolute position. On the contrary, I think that it supports Cowgill’s tentative etymology of the absolute suffix *es as an enclitic form of the copula *esti (1975: 66). When *es came to be used obligatorily in second position, its occurrence after a non-initial participial form received the status of a relative particle. Thus, to Cowgill’s examples brethae in fer ‘the man was carried’ < *britos est sindos wiros and ní-breth in fer ‘the man was not carried’ < *nēst britos sindos wiros we can add in fer brethae ‘the man who was carried’ < *sindos wiros britos est ‘this man, he was carried’. I conclude that the absolute form came to be used as a relative rather than the other way round. The hypothesis advanced here is supported by the possibility of substituting absolute for relative forms in nasalizing relative clauses, e.g. Wb 23d 25 hóre ni-ro-imdibed ‘because he had not been circumcised’, which is especially common in clauses containing the copula. 32. In this connection it seems appropriate to reconsider the other relative forms. There are several obstacles to the common view that the relative ending -e reflects an uninflected particle *io < PIE *iod. 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-,2 luide ‘who went’ < *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 2 I assume that *x was eliminated in téte on the analogy of the 3rd sg. abs. and conj. (and 2nd pl. conj.) forms, where it was lost phonetically in the position between a long vowel and a tautosyllabic t around stage (17). The Old Irish absolute and conjunct endings 21 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’. The nasalization in relative clauses where the antecedent is not the subject of the verb points to an acc.sg. form *san, which was created on the analogy of *sa. When *bereh was replaced with *bereθih after stage (10), the relative form *berea < *bere-so/-sā was replaced with *beresa on the analogy of the relative copula as < *esa < *est-so/-sā, cf. Breton so. The original thematic relative ending has been preserved in file ‘which there is’. The plural relative forms were apparently created on the basis of the 3rd sg. form, which was originally used for both numbers when the relative particle represented the subject, cf. Ml 124b 3 ní sní cet-id-deirgni ‘it is not we who have done it first’, where -deirgni is the 3rd sg. perfect form of dogní ‘does’.3 33. Cowgill has not gone into the original function of the absolute particle *es and the reason for its coexistence with the copula is < *esti. The etymological identity of the two is supported by the presence of a copula form in Bergin’s law constructions, e.g. ceso femmuin mbolgaig mbung ‘although I reap blistered seaweed’. Thieme has drawn attention to the similar co-occurrence of asti with a finite verb form in Sanskrit (1965: 90f.), e.g. pr̥cchati: asty atra kāṃcid gāṃ paśyasi ‘asks: is it (that) you see a certain cow here’. If we suppose that the absolute particle may have grown out of this type of usage, the bifurcation of the copula remains to be explained. In this connection I want to point to the comparable existence of two forms in Slavic, which can also be derived from *est and *esti. Here I shall list the 3rd sg. forms of the copula in the oldest Slavic texts, the Freising Fragments (unmarked) and the Kiev Leaflets (KL). These texts are of particular interest because they were written in a dialectal area where the coexistence of je and jest was more persistent than in the Bulgarian and Russian territories. 3 Cf. the comparable construction in Russian: te, kto ne xočet prinjat'sja za rabotu, mogut otdat' svoi rasčetnye knižki 'those who do(es) not want to get down to work, can (pl.) return their pay-books'. 22 Italo-Celtic origins and prehistoric development of the Irish language 34. The short form, which is written ie or ge in the Freising Fragments, is a clitic and is used with a verbal predicate (l-participle or infinitive): II 79 ese ge ... stuoril ‘quod fecit’, II 93 pozled ge pozstavv(il) ‘postremo constituit’, II 94 i ucazal ge ‘et monstravit’, I 8 da mi ie ... iti ‘mihi eundum esse’, I 9 imeti mi ie sivuot ‘mihi vita habenda est’, I 10 imeti mi ie otpuztic ‘mihi remissio accipienda est’, II 71 nu ge stati pred stolom bosigem ‘sed ante thronum Dei standum est’, II 86 nu ge pred bosima osima stati ‘sed ante oculos Dei standum est’. To these instances can be added two cases where je was deleted after the reflexive pronoun se (written ze): I 16 ese mi ze tomu chotelo ‘quod concupivi’, II 59 i nam ze modliti ‘et nobis exorandus est’. In the Codex Suprasliensis, which is the only Old Bulgarian text where the short copula occurs more than three times, it is used 5× with a verbal predicate, 1× in the construction jakože je podoba ‘ut decet’, 2× with a nominal predicate, and 8× after čto ‘quid’. The only examples of the short copula in the Codex Zographensis (2×), the Codex Assemanianus (2×), and the Savvina Kniga (2×) are found after čĭto (čto) ‘quid’. 35. The long form, which is written iezt, iest, gest in the Freising Fragments, is used with a nominal predicate (noun or n-participle): I 35 ese v(i) iezt ugotoulieno ‘quod vobis paratum est’, II 64 ese iest ugotouleno ‘quod paratum est’, II 90 ise gest bali ‘qui est medicus’, KL VI 7 äko balĭstvo estŭ ‘medicinam esse’. The Old Irish absolute and conjunct endings 23 Besides, the long form is used instead of the short form if there is no orthotonic word in the neighbourhood: II 75 i gest ze pred bosima osima ... izbovuedati ‘et ante oculos Dei confitendum est’. In the Old Bulgarian texts, the long form of the copula is used almost exclusively. ABSOLUTE AND CONJUNCT AGAIN* Lediglich Meillets Theorie bietet einen passenden Rahmen für -ō als konjunkte Endung (nicht jedoch für abs. -u). (Meid 1963: 17) Dagegen scheint mir, daß die Annahme, -s habe sich von irgend einem bestimmten Ausgangspunkt Endungen ausgebreitet, derselben aus der auf verschiedene absoluten Flexion Gestalt mancher die gut erklären würde. (Thurneysen 1914: 30) In 1978 David Greene asked me to submit an exposition of my views on the development of the Old Irish absolute and conjunct endings and their chronological implications. The resulting article appeared the following year (1979b). In 1983 Warren Cowgill criticized my views at a conference which I unfortunately could not attend. When the proceedings of the conference appeared (1985), the great American Indo-Europeanist had passed away and I did not feel like answering his criticism. As the publications which have come to my knowledge since I wrote the original article have not given me reason to change my opinion on the principal issues, I think that it is time to take the matter up again here.1 Let me first of all emphasize again (cf. 1979b: 35) that I cannot accept any theory which builds on an analogical differentiation between absolute and conjunct endings. If there was any interaction between the two sets of forms, the only result can have been the replacement of one by the other, as indeed happened in later Irish. The distinction between absolute and conjunct * 1 Münchener Studien zur Sprachwissenschaft 55 (1994), 61-68. I shall not discuss McCone's theory, which Cowgill has refuted in a conclusive way (o.c.), nor the variants proposed by Sims-Williams (1984) and Koch (1987), which are open to similar objections. 92 Italo-Celtic origins and prehistoric development of the Irish language endings must originally have been motivated semantically, as morphological distinctions always are. After the loss of the semantic element there can have been no such thing as the massive spread of a redundant morphological category. Following Thurneysen (1897, 1914), Cowgill hesitantly derived the difference between the two sets of verbal paradigms from the presence versus absence of an enclitic copula (1975: 66). I am more convinced now than I was before (cf. 1979b: 51, 1984: 182) that we have to start from an enclitic focus particle *est ‘it is (the case that)’ which distinguished the absolute and deuterotonic from the conjunct and prototonic forms. There actually seem to be traces of the original meaning in Archaic Irish. Following Mac Cana (1973), Greene has drawn attention to the difference between what he called Tmesis III and a cleft sentence (1977: 24f.): manip fri fasach fuirmider sceo fursantar fír Féine ‘unless the truth of Irish law be fixed and illuminated by precedent’ (lit. ‘unless be by precedent fixed and illuminated the truth of Irish law’), as opposed to *manip fri fasach fo-ruimedar sceo for-osnathar fír Féine ‘unless it be by precedent that the truth of Irish law is fixed and illuminated’ (lit. ‘unless be by precedent it is fixed and illuminated the truth of Irish law’). When the particle became a fixed constituent of initial phrases in statements, its absence was limited to responsive and cohortative (imperative, emphatic future) usage, e.g. Laumur ar dochondaib dílsi caille ‘Let me venture for (the benefit of) the immature (to state) the immune things of the forest’ (Binchy 1971: 157, Greene 1977: 18), as opposed to ‘(It is the case that) I venture (…)’. Cowgill assumes an early loss of final *-i in 3rd person verb forms (1975: 57, 1985b: 109). Even this restrictive formulation does not work because this early loss of *-i affected the 3rd sg. but not the 3rd pl. relative form (Cowgill 1975: 59) and does not account for the 2nd sg. forms (cf. Kortlandt 1979b: 36). McCone has tried to turn the rule into a general phonetic apocope of *-i (1978). It seems to me that neuter i-stems like muir ‘sea’, the dat.sg. form déit ‘tooth’, and especially the isolated form inn-uraid Absolute and conjunct again 93 ‘last year’, which must be identified with Gr. πέρυσι and Arm. heru, suffice to show that his view is mistaken.2 The gen.sg. forms anmae ‘name’ < *-ens and sléibe ‘mountain’ < *-esos suggest that we have to start from a zero loc.sg. ending in dat.sg. ainm and a long ending *-esi in sléib. Like Cowgill (1975: 57, 1985b: 113) I think that loc.sg. *tegesi developed via *tegī into dat.sg. tig ‘house’ because *-s- was lost at an early stage (cf. OW. tig, MW. ty). The loss of intervocalic *-s- must be dated before the monophthongization of the Indo-European u-diphthongs, as is clear from tauë ‘silence’ (W. taw) < *tawia < *tausiā (cf. Kortlandt 1979b: 39). It follows that the original nonzero loc.sg. ending of the neuter s-stems merged phonetically with the original zero ending of the neuter n-stems, leaving as its only trace the raising of the root vowel in tig. As a result, the two types of ending may have been subject to redistribution after stage (6) of my chronology (1979b: 41). When we reconsider the material presented by McCone (1978), it appears that the distribution of “short” and “long” dat.sg. forms of consonant stems in the glosses reflects a distinction of inanimate versus animate: on the one hand oíntu ‘unity’, toimtiu ‘opinion’, tíchtu ‘coming’ and other abstracts, further tene ‘fire’, cin (acc.) ‘fault’, traig (acc.) ‘foot’, cathair ‘city’, talam ‘earth’, brú ‘breast’, and on the other hand r- and nt-stems such as athair ‘father’ and carae ‘friend’, further coimdiu ‘lord’, fili ‘poet’, rí ‘king’, brithem ‘judge’, feichem ‘creditor’, fiada ‘witness’, also menmae ‘mind’. The absence of short dat.sg. forms of cré ‘clay’ and lie ‘stone’ seems to be accidental. From the Blathmac poems McCone cites the short forms cathir, talam, brú, crí ‘clay’, aitite ‘recognition’, also druí ‘magician’, and long forms of coimdiu, brithem, and trú ‘doomed man’. It follows that his material cannot be used as evidence for a general apocope of *-i.3 2 Note that the addition of inn- is recent, as it is in in-dé ‘yesterday’ (W. ddoe); cf. also the substitution of in fecht-so for ind-echt-so ‘this once’. 3 It could be argued that the “short” and “long” forms represent the locative and the dative, respectively, a possibility which McCone does not consider (cf. 1978: 35). This version of the theory would still be unacceptable to me in view of the objections stated above (muir, déit, inn-uraid). Since “it is hard to discern a powerful motivation towards the creation of a separate dat.sg. form” anmaimm ‘name’ (McCone 1978: 32, 94 Italo-Celtic origins and prehistoric development of the Irish language The early loss of intervocalic *-s- solves two problems which Thurneysen perceived already in the earliest beginnings of the particle theory. Firstly, “rucad neben ro hucad (mit bedeutungslosem h) Sg. 174a 1 zeigt, dass ro von u- durch kein gesprochenes h getrennt war” (1897: 3). This is an unfortunate example because “forms like ro-ucc ‘has brought’ are hardly ever elided to **r’ucc, whereas the preverb ro in ro-icc ‘reaches’ is often elided, giving r’icc” (Sims-Williams 1984: 143), but Cowgill admits that the latter type of elision is a real difficulty in his theory (1985b: 111). The reason is that he evidently limits the early loss of intervocalic *-s- to the position between unstressed vowels instead of viewing it as a general phonetic development which was obscured by the regularization of a morphophonemic alternation (cf. Kortlandt 1982b: 77). It is only natural that the restoration of ro- was earlier in the perfective particle, e.g. ro-ucc, where it carried a clear grammatical meaning, than in the preverb, as in ro-icc, where it expressed a lexical meaning in combination with the root, distinguishing it e.g. from doicc ‘comes’. Secondly, though the 3rd sg. copula is does not lenite, “doch ist eine vokalisch schliessende Grundform (*issi *essi aus *esti) nicht nur aus etymologischen Gründen wahrscheinlich, sondern wird, wie mir scheint, durch eine eigentümliche Wortverbindung direkt bewiesen” (Thurneysen 1897: 5), viz. is inse ‘it is difficult’ from *essi anassion, cf. ní anse ‘it is not difficult’ from *nīs anassion. The general absence of lenition after is, from which Thurneysen infers that “sich also etwa nach *nīh vor dem Wirken der Auslautsgesetze *issih für *issi gebildet hat” (1897: 6), suggests that the aphaeresis in *isi ’nasia was conditioned by the loss of intervocalic *-h- (my stage 2 in 1979b: 39f.). The absence of lowering in the initial vowel of inse points to a reanalysis as *is inase after my stage 11 (1979b: 47). Unlike Cowgill, I am convinced that there was in Indo-European a fundamental distinction between the thematic and the athematic present endings which is reflected in Indo-Iranian (Beekes 1981), Greek, Armenian (Kortlandt 1981b), Baltic, Slavic (Kortlandt 1979a), and perhaps all other major branches of the family. In Old Irish we expect thematic endings in BI fn. 27), I think that this is an original plural form (cf. Pedersen and Cowgill apud McCone). Absolute and conjunct again 95 berid, -beir ‘carries’, BII gaibid, -gaib ‘takes’, AI marbaid, -marba ‘kills’, AII rádid, -rádi ‘says’, and athematic endings in BIV benaid, -ben ‘strikes’, AII ruidid, -ruid ‘blushes’, also BI -tét ‘goes’ (see below), BI/II -said ‘sits’ (cf. Kortlandt 1990: 8), BI/III -cing ‘steps’ (cf. Kuiper 1937: 168), BIV/V -gnin ‘knows’. I reconstruct 3rd sg. *-e in the thematic and *-ti in the athematic flexion, after which *-s was added in the corresponding absolute forms. As a result of the lenition the regular 3rd sg. endings became BI *-e(h), BII *-ie(h), BIV *-aθi(h), AI *-āe(h), AII *-īe(h) and *-īθi(h), while *-ti was preserved in *tēxti(h) ‘goes’. This fairly transparent system collapsed when *-e was lost after a long vowel, which yielded a zero ending in AI *-ā and AII *-ī, but not in the corresponding absolute forms in *-āeh and *-īeh (cf. Kortlandt 1979b: 41, 45). The zero conjunct ending evidently spread from the weak verbs to BII *-i for *-ie, further to AII *-ī for *-īθi, and eventually to BIV *-a for *-aθi. Such a development could not take place in the absolute forms because there was no model. It is in my view essential that there was no interaction between absolute and conjunct endings because they were in complementary distribution after the loss of *-es as a clear meaningful element until the later Irish disintegration of the system of two sets of endings. The generalization of the athematic 3rd sg. present ending *-θih in the absolute forms was motivated by the merger of the present and preterit (sigmatic aorist) endings in the weak verbs. While the conjunct endings AII *-ī(e) and *-īh < *-īs remained distinct up to the apocope, the corresponding absolute endings merged into *-īeh as a result of the loss of intervocalic *-s- in the preterit. The present ending was therefore replaced by the available alternative *-īθih. In a similar vein I think that the absolute present ending AI *-āeh was replaced by *-āθih for differentiation from the subjunctive ending *-āeh < *-āses (cf. Kortlandt 1984: 182). When the functional distinction between primary and secondary endings was lost and after the shortening of long vowels in medial syllables (stage 10 of Kortlandt 1979b: 44), the subjunctive ending *-aeh was in its turn replaced by *-aθih for differentiation from the preterit ending *-aeh < *-ases (cf. Kortlandt 1984: 183). The replacement of the latter ending by *-aseh was probably motivated by the introduction of primary endings in 1st sg. *-asūh and 2nd sg. *-asīh on the analogy of the present tense. The absence of *-θih in the paradigm of gabsu, gabsai, gabais ‘I, you, he took’ suggests that this element was still absent in the present BII gaibid ‘takes’ 96 Italo-Celtic origins and prehistoric development of the Irish language when the primary endings were introduced into the preterit. It follows that the spread of *-θih to BI berid ‘carries’ was a recent development.4 It appears that the original athematic conjunct ending was preserved in -tét, Wb. -téit (Thurneysen 1946: 376) < *tēxti ‘goes’, where the root-final consonant was lost in the position between a long vowel and a tautosyllabic plosive after the apocope (Kortlandt 1979b: 50, fn. 2). Following Thurneysen (1946: 377), I assume that the athematic conjunct ending spread to -fet ‘leads’, -rét ‘rides, drives’, *-ret ‘runs’, prototonic -tet, -at, -rat, and then to other verbs with a root-final dental plosive such as ad-fét, -adbat ‘relates’, arnëat, -airnet ‘expects, sustains’. The root sed- of the latter verb probably had an athematic present *sediti ‘sits’, as in Germanic (Kortlandt 1990: 8). Note that the depalatalization in *-tēxt is regular, as it is in secht ‘seven’ and the oblique case forms of deacht ‘divinity’ (Thurneysen 1946: 101). There is no reason to assume an irregular syncope (Meid 1972: 351) or apocope (Cowgill 1985b) which does not explain the alternative forms -feid, -réid, -reith. Another athematic conjunct form may be attested in co cóic séotu cingith ‘it extends to five chattels’ (Binchy 1971: 157, Greene 1977: 18). The form cingith is actually an emendation of cingit (Binchy 1971: 153), which may represent *kingiti or *kinixti ‘steps’. The semantic affinity with *tēxti may have played a role in the preservation of the athematic ending. I shall be brief about the passive and deponent forms. As I have pointed out elsewhere (1981a: 18f.), I think that the conjunct forms ended in *-ro with the exception of 2nd pl. -d < *-dwe and that the relative and absolute forms were derived by adding the particles *so and *es, respectively. Since the latter particle took the shape *-s after a vowel and intervocalic *-s- was lost in the former before the shortening of long final vowels, the absolute 4 It is of course conceivable that the introduction of the new preterit endings was earlier, which would render the introduction of *-θih in the subjunctive less wellmotivated. This chronology seems less plausible to me because it makes the preservation of secondary endings in the subjunctive and the future difficult to understand (cf. Kortlandt 1979b: 48f. and 1984: 182). The absence of raising in the subjunctive and the preterit of thematic AII verbs (Thurneysen 1946: 385, 419) shows that these paradigms had adopted a different suffix at my stage 8 (1979b: 43). This yields a terminus ad quem for the generalization of *-θih in the AII presents. I withdraw my agreement (1979b: 38, 46) with Watkins’ view that the thematic 3rd sg. ending *-e is preserved in fil ‘there is’ (cf. Thurneysen 1946: 479). Absolute and conjunct again 97 forms ended in *-rah and the conjunct and relative merged into *-ra at stage (6) of my chronology (1979b: 41). This explains “a number of curious features which have never received any serious attention” and are “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). Thus, I think that we have a relative form in atáit secht fuili la Féniu fertar ‘there are seven bloods which are spilt in Irish law’ < *-ntoro-so and an absolute form in ó thestaib córaib cengar ‘one proceeds from proper witnesses’ and brechtaib ban mberar ‘he is taken by the spells of women’ < *-oro-s. The original absolute deponent ending was preserved in 1st sg. -ur < *-ōro-s and 2nd sg. -ther < *-toro-s. The final palatalization in the regular 3rd sg. and pl. endings -thir, -dir, -tir and 1st pl. -mir was evidently taken from the active paradigms after the apocope. THREE NOTES ON THE OLD IRISH VERB* “Il est étrange que, pour éviter d’admettre des traitements phonétiques qui ne contredisent aucun traitement connu des mêmes phonèmes placés dans les mêmes conditions, on ait recouru à des hypothèses analogiques qui sont ou arbitraires ou invraisemblables, comme si les difficultés morphologiques étaient, par nature, chose moins grave que les difficultés phonétiques” (Meillet 1914: 8). Here I shall give three examples where scholars have in utter despair proposed to assume suppletive formations because they could not get the sound laws right. 1. bá, boí ‘was’ < *bhōuAccording to the traditional view (Thurneysen 1946: 483), 3rd sg. boí “could go back to *bhowe, an unreduplicated perfect, or alternatively to *bhōwe (whence Celtic *bāwe); but this would leave unexplained the ā of the other forms, which show no trace of w.” I have argued (1986: 90-92) that these forms are compatible if we start from an original full grade root aorist *bhāw- < *bheH2u-, 1st sg. -bá < *bām < *bhāum, 3rd sg. -boí < *bau-e < *bhāut with added *-e on the analogy of the perfect. On the basis of Armenian boys ‘herb, plant’, busanim ‘grow’, I now reconstruct *bhōw- < *bheH3u- (note that the timbre of the laryngeal cannot be established on the basis of Germanic bō-, Slavic bav-, or Vedic bodhí ‘be!’). This renders the derivation of Welsh bu ‘was’ < *bōu < *bhōut straightforward but implies a difference between OIr. -bá < *bōm < *bhōum and cú ‘dog’ < *kuō, which shows the regular development of polysyllables (cf. Thurneysen 1946: 58). My theory has been disregarded by McCone, who wrongly attributes a quite different view to me (1991: 128): “Although phonetically unobjectionable on its own, Kortlandt’s derivation of OIr. 1/2sg. -bá ‘I was, you were’ < *bāwa(s) < full grade *bheh2w- is hard to square with 3sg. *bow(e) implied by MW. bu and OIr. boí,” as he puts it. On the contrary, I * Études Celtiques 34 (1998-2000), 143-146. 126 Italo-Celtic origins and prehistoric development of the Irish language would maintain that McCone’s reconstruction *bāwa(s) yields OIr. **báu > **báo > **bó (cf. now Uhlich 1995). The monosyllabic forms of the original root aorist have been preserved in the preterit of the copula 1st sg. -b(sa) < *bōm, 3rd sg. -bo < *bōu. McCone’s derivation of boí < *bowe < *buwe is phonetically unobjectionable on its own (cf. Kortlandt 1979b: 46f.) but does not explain the stem form bá- of the other persons in the paradigm. 2. níta, ní ‘am/is not’ < *nēst (de es-) This paradigm is evidently based on ní < *nīh < *nēst (cf. Thurneysen 1946: 487). I find it impossible to separate negative níta from positive -da, which must be derived from *d(e) es-, 3rd sg. -t < *-d(e)h < *d(e) est, thus níta < *nīh d’ e- < *nēst d(e) es-, 3rd pl. nítat < *nīh d(e) (h)ēt- < *nēst de senti, Welsh antevocalic nyt ‘not’ < *nīh d(e) < *nēst de (cf. Kortlandt 1996a: 96). While the athematic flexion is preserved in the absolute paradigm am < *esmi-s, at < *esi-s tu, is < *esti-s, the leniting 1st and 2nd sg. conjunct forms -da, níta suggest that the flexion was thematicized to *-eu, *-ei, as if from *esō, *esei (cf. Kortlandt 1979b: 45). We find complementary distribution between absolute *es and relative *so, and also between the connectives *de and *kwe, e.g. relative nád < *na-so-de, nách < *na-so-kwe, MW. nat < *na-so-d(e) with -t from nyt, MBr. nac ‘who (does) not’ < *naso-k but nag-a ‘do not go!’ < *na-k age (cf. Kortlandt 1996a: 96). Schrijver derives OIr. nítat ‘are not’ < *nent-, “whatever the ultimate origin of *-nt- may have been” (1997: 158), and identifies MW. nyt with OIr. ní, allegedly from *ne et(i). He thus separates the negative from the positive paradigm of the copula, which I find unacceptable. Apart from the objections to a derivation of the absolute particle *es < *eti which Schrijver discusses (1997: 156-158), the main points against his theory are that it requires massive analogical spread of apocopated *-i which cannot be motivated (cf. Schrijver 1994: 175-177 and Kortlandt 1996a: 95) and that it presupposes the absence of the particle in verbs with a telic Aktionsart (cf. Schrijver 1997: 123-128), which in my view disqualifies the theory in a fundamental way. These problems vanish if we do not derive *es from *eti but from *est, which is also preferable for functional reasons. As I pointed out a long time ago (1982b: 76-78), original intervocalic *-s- was lost at an early stage and *-h- was restored on the basis of the anteconsonantal reflex as the regular Three notes on the Old Irish verb 127 alternant of *-s-. As a result, the particle *es was lost phonetically before rootinitial vowels. Schrijver objects (1997: 123) that it is difficult to find a model for the restoration of *-h in such instances as a h-ech ‘her horse’. This is a consequence of his lack of chronological perspective: while the loss of intervocalic *-h- can be dated to stage 2 of my chronology (1979b: 39f.) and the rise of the nasal mutation to stage 5, final *-h was evidently preserved up to the apocope at stage 15 in such instances as nant ‘that it is not’ < *-d(e)h and arimp ‘in order that it may be’ < *-b(e)h, which leaves plenty of time for the restoration of *-h before initial vowels (see further Kortlandt 1982b: 79-82). Before the infixed object pronoun masc. *en, neuter *e, the particle *es was lost after a vowel (before class A pronouns) and reduced to *-e- after a consonant (before class B pronouns). Since the nonzero reflex merged with the object pronoun, the forms were disambiguated by the insertion of *d(e) before the infixed pronoun, as happened again to protect the object pronoun *-e- from elision before the root-initial vowel in Wb 5b 40 cotd-icc ‘he can do it’ < *kon-s-d-e-d-, similarly nachid- < *na-so-kwe-d-e beside nách- < *na-sok-e ‘who ... not it’ (cf. Kortlandt 1996a: 96). Thus, I reconstruct *-s- in class A, *-es-de- in class B, and *-so-de- in class C. Schrijver’s reconstruction of the masc. relative subject pronoun *yo beside object *en < *em and neuter *e < *ed (1997: 129) can actually be adduced in support of my view that we must reconstruct *so instead of *yo. The complementary distribution between absolute *es and relative *so may then support Pedersen’s derivation of *es from a resumptive subject pronoun (cf. Kortlandt 1984: 182). For the time being I stick to a derivation of *es from *est ‘it is (the case that)’ because there seem to be traces of the original meaning in Archaic Irish (cf. Kortlandt 1994: 62). 3. tíagu, téit ‘go(es)’ < *steighAccording to the usual view, 3rd sg. téit is a form of the root *ten‘stretch’, unlike the other forms of the paradigm, which clearly represent the root *steigh- ‘step’ (Bergin 1938: 227f., Thurneysen 1946: 473). Schrijver proposes to derive téit from a nasal present *stingh- (1993: 44). I would prefer to derive the whole paradigm from a single stem, which can only have been *steigh-, so that abs. téit, conj. -tét, Wb. -téit reflect *steighti(-s), 128 Italo-Celtic origins and prehistoric development of the Irish language thematicized in 1st sg. tíagu, -tíag < *steighō(-s). I have therefore suggested that *x was lost after a long vowel in *tēxt′ (1979b: 50). The problem with this chronology is that the voicing of the final dental stop is no longer automatic after the apocope (cf. Kortlandt 1982b: 78f.). It is therefore preferable to date the loss of *x between stages 5 (loss of *n before dentals and velars) and 15 (apocope) of my chronology (1979b), probably after the palatalization of the cluster [xt] between front vowels (my stage 7). As a result of this development, the 3rd sg. form *tēti(h) escaped the thematicization of the paradigm of *tēg- around stage 10 (cf. Kortlandt 1979b: 44-46). This explains the isolated character of the 3rd sg. present tense form. Note that the preservation of [x] in téchtae ‘proper, right’ suggests that the palatalization of [x′] was a prerequisite for its loss in téit, as might be expected on phonetic grounds. OLD IRISH feda, GEN. fedot ‘LORD’ AND THE 1ST SG. ABSOLUTE ENDING -a IN SUBJUNCTIVES AND FUTURES Patrick Sims-Williams has argued (1999), to my mind correctly, that the Old Irish forms nom. feda, gen. fedot ‘Lord’ in the Cambrai Homily represent an older stage of the later forms fíado, fíadat. This simplifies my account of Old Irish historical phonology (1979b) and morphology (1984). In my relative chronology of Old Irish sound changes, I distinguished between *ē1 < Indo-European *ē, *ē2 < *ei, and *ē3 < *en before *t/s, and between *ō1 < Indo-European *ō, *ō2 < *ou, and *ō3 < *on before *t/s, and argued that the more recent vowel was lower than the earlier one (1979b: 40). I followed the earlier view (cf. Greene 1976a: 27) that the loss of intervocalic *s preceded the monophthongization of the u-diphthongs and noted that there is no reason to separate the monophthongization of the i-diphthongs from the latter development. Stressed *ai and *oi were not affected by the monophthongization, which suggests that the u-diphthongs had merged into *ou before the rise of *ō2. In unstressed syllables, the i-diphthongs merged with *ē1 and *ī, e.g. nom.pl. fir ‘men’ < *wirī < *wiroi, dat.sg. tuil ‘will’ < *tolī < *tolāi. The split of *ō1 into *ū in final syllables and *ā elsewhere must have preceded the monophthongization of the diphthongs, e.g. dat.sg. fiur ‘man’ < *wirū < *wirōi. I do not share the common view that *ē1 had been raised to *ī in ProtoCeltic times already. An early merger of *ē1 and *ī would have yielded a phonological system where the vowel height oppositions between the short vowels outnumbered those between the long vowels. Though such a system is by no means impossible, it is not probable that it would have remained in existence over a longer period of time. It seems better to connect the raising of *ē1 with the development of the i-diphthongs in the separate languages. The development of Indo-European *-oi and *-āi into *-ī suggests that *ē1 and *ē2 merged in unstressed syllables before the raising of *ē1 to *ī. In final syllables we find *ī < *ē1 in the paradigm of méit ‘quantity’ (cf. Schrijver 1991a: 388f.). The *ē3 of carae ‘friend’ < *karēh < *karants is also found in 130 Italo-Celtic origins and prehistoric development of the Irish language fiche ‘twenty’ < *wikēh < *wikent and in gen.sg. abae ‘river’ < *abēh < *abens. Another source of *ē3 is found in the 2nd pl. absolute form beirthe ‘you carry’ < *beretēh < *beretes-es. I conclude that the rise of *ē3 from *en and *an before a dental consonant was more recent than the raising of *ē1 and *ē2 to *ī in unstressed syllables. It was also more recent than the raising of *ē1 to *ī in stressed syllables because *ē3 merged neither with *ē1 nor with *ē2, e.g. cét ‘hundred’ < *kenton versus íasc ‘fish’ < *peiskos. The open character of *ē3 is not unexpected because *en and *an merged, e.g. géis ‘swan’, Latin ānser ‘goose’. Thus, we arrive at the following relative chronology: (1) (2) Lenition and rise of *h from Indo-European *s. Loss of intervocalic *h. (3) (4) (5) Monophthongization of i- and u-diphthongs and rise of *ē2 and *ō2. Raising of *ē1 to *ī. Loss of *n before *t/s and rise of *ē3. At this stage, the nasal mutation became a morphological process (cf. Kortlandt 1979b: 41). It follows that the acc.pl. ending of the consonant stems -a cannot be the phonetic reflex of syllabic *-ns, which should yield *-ē3. The attested ending evidently has an analogical long vowel after which the nasal consonant was lost at an earlier stage. For the next developments I refer to my earlier work (1979b: 41-48, cf. also Schrijver 1991b: 23): (6) (7) (8) (9) (10) (11) Shortening of word-final long vowels. Palatalization. Raising. u-infection. Shortening of long vowels in medial syllables. Lowering. The apocope is dated to stage (15), the syncope to stage (19), and the loss of intervocalic *w to stage (20) of my chronology. In my earlier contribution (1979b: 40, 46) I stuck to the traditional view that *ō3 merged with *ō2 in trícho ‘thirty’ < *trīkont and cano ‘poet’ < *kanonts. It now appears that *ō3 yielded -a word-finally and -o- in medial syllables. I pointed out already that final *ē3 and *ō3 cannot represent IndoEuropean *-ent and *-ont because final *-t had been lost at an early stage, as Old Irish feda, gen. fedot ‘Lord’ 131 is evident from the merger of the perfect with the thematic aorist (cf. also Kortlandt 1996a: 91f. and 1997a: 135). The 3rd pl. thematic present ending *-o had been 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’. At this stage, the 3rd sg. ending was *-e in the thematic present and aorist and in the perfect, *-to in the imperfect and the imperative, *-(t)ro in the deponent, and *-toro in the passive, while the 3rd pl. ending was *-on(t) in the thematic present and aorist, *-r in the perfect, *-nto in the imperfect and the imperative, *-ntro in the deponent, and *-ntoro in the passive (cf. Kortlandt 1981a: 17-20). The elision of the first vowel in the 3rd pl. ending -atar of the suffixless preterit, which continues the thematic aorist and the perfect, suggests that *-on(t) was replaced by *-onto before the added -r (cf. Thurneysen 1946: 434), and the same replacement may be assumed for the 3rd pl. conjunct ending of the thematic present -at, which is -ot in the archaic forms tu-thegot ‘who come’ and tu-esmot ‘who pour out’ of the Cambrai Homily (cf. Sims-Williams 1999: 473, who mistakenly assumes *-i instead of *-o in these and other forms). This development can be dated before or to approximately the same stage as the reshuffling of thematic and athematic endings (cf. Kortlandt 1979b: 44-46 and 1997a). In my earlier studies I argued that the absolute endings originated from a particle *es which was reduced to *-s after vowels and nasals (1979b: 49 and 1984: 182f., cf. also 1994). The Indo-European secondary thematic endings 1st sg. *-om and 2nd sg. *-es have been preserved in the a-subjunctive, which is historically identical with the s-subjunctive of roots with a final laryngeal (cf. Kortlandt 1979b: 48 and 1984: 182f.). The conjunct forms 1st sg. *berahon < *-asom, 2nd sg. *beraheh < *-ases yielded pre-apocope *bera n- and *beraeh as a result of the loss of intervocalic *-h- at stage (2), the lowering of *-on to *-an and its coalescence with the preceding *-a- into *-ān, which became *-ā n- by the rise of the nasal mutation at stage (5), and the shortening of *-ā at stage (6) of my chronology, resulting in 1st sg. -ber and 2nd sg. -berae after the apocope. The absolute forms represent 1st sg. *berasom-s and 2nd sg. *berases-es, which became *-aō3h and *-aē3h at stage (5) and then developed into pre-apocope *berāh, *beraēh, yielding the historical forms bera, berae. Thus, the 1st sg. absolute ending -a is the regular outcome of *ō3 < *-ons and does not require any additional 132 Italo-Celtic origins and prehistoric development of the Irish language assumptions (cf. Kortlandt 1979b: 49). The same endings are found in the ffuture, where the suffix represents the future of the verb ‘to be’ *bwias-, as in the formation of archaic Middle Welsh 3rd sg. deubyd ‘will come’ (cf. Sommerfelt 1922 and Kortlandt 1984: 185f.). The u-infection in the 1st sg. conjunct form, e.g. -léiciub ‘will leave’, must be derived from *-ww- < *-bw-, not from a primary thematic ending for which there is no evidence and which is at variance with the 1st sg. absolute ending -fa. The u-infection subsequently spread from the f-future to the s-future, e.g. -gigius, -érus of guidid ‘prays’, do-érig ‘abandons’. The reconstruction of *ō3 and *ē3 in the absolute forms of the future of the substantive verb 1st sg. bia and 2nd sg. bie enables us to identify the formation with its Brythonic counterpart. APPENDIX: OLD IRISH VERBAL PARADIGMS <...> = lost by analogy; [...] = substituted or added by analogy. This is to be understood in the sense that the reflex of the indicated segment (not necessarily the segment itself) was lost or added at some stage in the development from Proto-Celtic to Old Irish. I have left out the delenition of *m in the 1st sg. and pl. endings and more often than not the restoration of lost segments (or their reflexes) in the reconstructions. The suffix of the ffuture is given as *-bwas- for *-bw<i>as- (cf. Kortlandt 1984: 185). Following the order and classification of Thurneysen 1946, I have adopted the format of Strachan 1949, with the absolute forms on the left hand side and the conjunct forms on the right hand side, followed by a formal reconstruction of the respective Insular Celtic endings: absolute < PIC * conjunct < PIC * Examples: marbaid ‘kills’, léicid ‘leaves’, berid ‘carries’, gaibid ‘takes’, benaid ‘strikes’, labrithir ‘speaks’, suidigidir ‘places’, midithir ‘judges’, -cuirethar ‘puts’, téit ‘goes’, guidid ‘prays’, ro-fitir ‘knows’, canaid ‘sings’, -gainethar ‘is born’, do-moinethar ‘thinks’. PRESENT STEM present indicative active AI 1 sg. 1 sg. 2 sg. 3 sg. rel. marbu marbaim marbai marbaid marbas -āiōs -ā[mi]s -āieis -ā[ti]s -ā[s]so -marbu -marbaim -marbai -marba -āiō -ā[mi] -āiei -āie 160 Italo-Celtic origins and prehistoric development of the Irish language 1 pl. rel. 2 pl. 3 pl. rel. marbmai marbmae marbthae marbait marb(a)te -āiomos[i]s -āiomoses -āieteses -āiontes -āionteso 1 sg. 1 sg. 2 sg. 3 sg. rel. 1 pl. rel. 2 pl. 3 pl. rel. léiciu léicim léici léicid léices léicmi léicme léicthe léicit léc(i)te -īōs -ī[mi]s -īeis -ī[ti]s -ī[s]so -īomos[i]s -īomoses -īeteses -īontes -īonteso 1 sg. 2 sg. 3 sg. rel. 1 pl. rel. 2 pl. 3 pl. rel. biru biri berid beres bermai bermae beirthe berait bertae -ōs -eis -e[ti]s -e[s]so -omos[i]s -omoses -eteses -ontes -onteso 1 sg. 1 sg. 2 sg. 3 sg. rel. gaibiu gaibim gaibi gaibid gaibes -i[ō]s -imis -isis -itis -i[s]so -marbam -āiomos -marbaid -marbat -āietes -āiont[o] -léiciu -léicim -léici -léici -īō -ī[mi] -īei -īe -léicem -īomos -léicid -léicet -īetes -īont[o] -biur -bir -beir -ō -ei -e -beram -omos -berid -berat -etes -ont[o] -gaibiu -gaibim -gaibi -gaib -i[ō] -imi -isi -i<ti> AII BI BII Old Irish verbal paradigms 1 pl. rel. 2 pl. 3 pl. rel. gaibmi gaibme gaibthe gaibit gaibte 161 -imos[i]s -imoses -iteses -intes -inteso -gaibem -imos -gaibid -gaibet -ites -int[o] -benaim -benai -ben -ami -asi -a<ti> -benam -amos -benaid -benat -ates -ant[o] BIV 1 sg. 2 sg. 3 sg. rel. 1 pl. rel. 2 pl. 3 pl. rel. benaim benai benaid benas benmai benmae bentae benait bentae -amis -asis -atis -a[s]so -amos[i]s -amoses -ateses -antes -anteso present indicative deponent AI 1 sg. 2 sg. 3 sg. rel. 1 pl. rel. 2 pl. 3 pl. rel. 1 sg. 2 sg. 3 sg. rel. labrur labrither labrithir labrathar labrimmir labrammar labrithe labritir labratar suidigim suidigther suidigidir suidigedar -āi[ō]ros -āietoros -āietr[e]s -āietroso -āi[e]mor[e]s -āiomoros -āie[tes]es -āi[e]ntr[e]s -āiontroso -i[mi]s -itoros -itr[e]s -itroso -labrur -labrither -labrathar -āi[ō]ro -āietoro -āietro -labrammar -āiomoro -labraid -labratar -āiedwe -āiontro AII/BII -suidigur -suidigther -suidigedar -i[ō]ro -itoro -itro 162 1 pl. rel. 2 pl. 3 pl. rel. 1 sg. 2 sg. 3 sg. rel. 1 pl. rel. 2 pl. 3 pl. rel. Italo-Celtic origins and prehistoric development of the Irish language suidigmir suidigmer suidigthe suidigitir suidigetar midiur mitter midithir midethar midimmir midemmar mitte miditir midetar -imor[e]s -imoros -i[tes]es -intr[e]s -introso -i[ō]ros -itoros -itr[e]s -itroso -imor[e]s -imoros -i[tes]es -intr[e]s -introso -suidigmer -imoro -suidigid -suidigetar -idwe -intro BII/AII -cuiriur -cuirther -cuirethar -ei[ō]ro -eietoro -eietro -cuiremmar -eiomoro -cuirid -cuiretar -eiedwe -eiontro present indicative passive AI 3 sg. rel. 3 pl. rel. marbthair marbthar marb(a)tir marb(a)tar -āietor[e]s -āietoroso -āi[e]ntor[e]s -āiontoroso 3 sg. rel. 3 pl. rel. léicthir léicther léc(i)tir léicter -īetor[e]s -īetoroso -īontor[e]s -īontoroso 3 sg. rel. suidigthir suidigther -itor[e]s -itoroso -marbthar -āietoro -marb(a)tar -āiontoro -léicther -īetoro -léc(e)tar -īontoro AII AII/BII -suidigther -itoro Old Irish verbal paradigms 3 pl. rel. suidigtir suidigter 163 -intor[e]s -intoroso -suidigter -intoro -berar -oro -bertar -ontoro -gaibther -itoro -gaib(e)tar -intoro BI 3 sg. rel. 3 pl. rel. berair berar bertair bertar -or[e]s -oroso -ontor[e]s -ontoroso 3 sg. rel. 3 pl. rel. gaibthir gaibther gaibtir gaibter -itor[e]s -itoroso -intor[e]s -intoroso 3 sg. 3 pl. mittir miditir -itor[e]s -intor[e]s 3 sg. rel. 3 pl. rel. benair benar bentair bentar -ar[e]s -aroso -antor[e]s -antoroso BII BII/AII -cuirther -eietoro BIV -benar -aro -bentar -antoro imperfect indicative active AI 1 sg. 2 sg. 3 sg. 1 pl. 2 pl. 3 pl. -marbainn -marbtha -marbad -marbmais -marbthae -marbtais -āiema[m] -āieto-āieto -āiemos[te] -āie[t]e-āiento[ste] 164 Italo-Celtic origins and prehistoric development of the Irish language BI 1 sg. 2 sg. 3 sg. -berinn -bertha -bered -ema[m] -eto-eto 1 pl. 2 pl. 3 pl. -bermis -berthe -bertis -emos[te] -e[t]e-ento[ste] imperfect indicative deponent 1 sg. 2 sg. 3 sg. 1 pl. 2 pl. 3 pl. AII/BII -suidiginn -suidigthea -suidiged -suidigmis -suidigthe -suidigtis -ima[m] -ito-ito -imos[te] -i[t]e-into[ste] imperfect indicative passive AI 3 sg. 3 pl. -marbthae -marbtais -āieto-āiento[ste] -berthe -bertis -eto-ento[ste] BI 3 sg. 3 pl. 3 sg. 3 pl. AII/BII -suidigthe -suidigtis -ito-into[ste] Old Irish verbal paradigms 165 imperative active AI 2 sg. 3 sg. 1 pl. 2 pl. 3 pl. marb marbad marbam marbaid marbat -ā<ie> -āieto -āiomo -āiete -āionto AII 2 sg. 3 sg. 1 pl. 2 pl. 3 pl. léic léiced léicem léicid léicet -ī<e> -īeto -īomo -īete -īonto 1 sg. 2 sg. 3 sg. 1 pl. 2 pl. 3 pl. biur beir bered beram berid berat -ō -e -eto -omo -ete -onto BI imperative deponent AI 2 sg. 3 sg. 2 pl. 3 pl. labrithe labrad labraid labratar -āieto[es] -āieto -āiedwe -āiontro 2 sg. 3 sg. 2 pl. 3 pl. suidigthe suidiged suidigid suidigetar -ito[es] -ito -idwe -intro AII/BII 166 Italo-Celtic origins and prehistoric development of the Irish language BII/AII 2 sg. 3 sg. 2 pl. 3 pl. cuirthe cuired cuirid cuiretar -eieto[es] -eieto -eiedwe -eiontro imperative passive AI 3 sg. 3 pl. marbthar marbtar -āietoro -āiontoro 3 sg. 3 pl. léicther léicter -īetoro -īontoro AII AII/BII 3 sg. 3 pl. suidigther suidigter -itoro -intoro 3 sg. 3 pl. berar bertar -oro -ontoro 3 sg. 3 pl. cuirther cuirter -eietoro -eiontoro BI BII/AII Old Irish verbal paradigms 167 SUBJUNCTIVE present a-subjunctive active AI 1 sg. 2 sg. 3 sg. rel. 1 pl. rel. 2 pl. 3 pl. rel. marba marbae marbaid marbas marbmai marbmae marbthae marbait marbaite -asoms -aseses -a[ti]s -asso -asomos[i]s -asomoses -aseteses -asontes -asonteso 1 sg. 2 sg. 3 sg. rel. 1 pl. rel. 2 pl. 3 pl. rel. léicea léice léicid léices léicmi léicme léicthe léicit léc(i)te -īsoms -īseses -ī[ti]s -īsso -īsomos[i]s -īsomoses -īseteses -īsontes -īsonteso -marb -marbae -marba -asom -ases -[ā] -marbam -asomos -marbaid -marbat -asetes -asont[o] -léic -léice -léicea -ī<som> -īses -īs[ā] -léicem -īsomos -léicid -léicet -īsetes -īsont[o] -ber -berae -bera -asom -ases -[ā] AII BI 1 sg. 2 sg. 3 sg. rel. bera berae beraid beras -asoms -aseses -a[ti]s -asso 168 1 pl. rel. 2 pl. 3 pl. rel. Italo-Celtic origins and prehistoric development of the Irish language bermai bermae berthae berait bertae -asomos[i]s -asomoses -aseteses -asontes -asonteso -beram -asomos -beraid -berat -asetes -asont[o] present a-subjunctive deponent 1 sg. 2 sg. 3 sg. rel. 1 pl. rel. 2 pl. 3 pl. rel. 1 sg. 2 sg. 3 sg. rel. 1 pl. rel. 2 pl. 3 pl. rel. AII/BII -suidiger -suidigther -suidigedar suidiger suidigther suidigidir suidigedar suidigmir suidigmer suidigthe suidigitir suidigetar -isōros -isetoros -isetr[e]s -isetroso -isomor[e]s -isomoros -ise[tes]es -isontr[e]s -isontroso labrar labrither labrithir labrathar labrimmir labrammar labrithe labritir labratar AI/BII/AII -asōros -corar -asetoros -coirther -asetr[e]s -corathar -asetroso -as[e]mor[e]s -corammar -asomoros -ase[tes]es -coraid -as[e]ntr[e]s -coratar -asontroso -isōro -isetoro -isetro -suidigmer -isomoro -suidigid -suidigetar -isedwe -isontro -asōro -asetoro -asetro -asomoro -asedwe -asontro present a-subjunctive passive AI 3 sg. rel. marbthair marbthar -asetor[e]s -asetoroso -marbthar -asetoro Old Irish verbal paradigms 3 pl. rel. marb(a)tir marb(a)tar 169 -as[e]ntor[e]s -asontoroso -marb(a)tar -asontoro -berthar -asetoro -bertar -asontoro BI 3 sg. rel. 3 pl. rel. berthair berthar bertair bertar -asetor[e]s -asetoroso -asontor[e]s -asontoroso 3 sg. rel. 3 pl. rel. suidigthir suidigther suidigtir suidigter -isetor[e]s -isetoroso -isontor[e]s -isontoroso AII/BII -suidigther -suidigter -isetoro -isontoro past a-subjunctive active AI 1 sg. 2 sg. 3 sg. 1 pl. 2 pl. 3 pl. -marbainn -marbtha -marbad -marbmais -marbthae -marbtais -asema[m] -aseto-aseto -asemos[te] -ase[t]e-asento[ste] -berainn -bertha -berad -bermais -berthae -bertais -asema[m] -aseto-aseto -asemos[te] -ase[t]e-asento[ste] BI 1 sg. 2 sg. 3 sg. 1 pl. 2 pl. 3 pl. 170 Italo-Celtic origins and prehistoric development of the Irish language past a-subjunctive deponent AII/BII -suidiginn -suidigthea -suidiged -suidigmis -suidigthe -suidigtis 1 sg. 2 sg. 3 sg. 1 pl. 2 pl. 3 pl. -isema[m] -iseto-iseto -isemos[te] -ise[t]e-isento[ste] past a-subjunctive passive AI 3 sg. 3 pl. -marbthae -marbtais -aseto-asento[ste] -berthae -bertais -aseto-asento[ste] BI 3 sg. 3 pl. AII/BII -suidigthe -suidigtis 3 sg. 3 pl. -iseto-isento[ste] present s-subjunctive active 1 sg. 2 sg. 3 sg. rel. 1 pl. rel. 2 pl. 3 pl. rel. tíasu tési téis tías tíasmai tíasmae téiste tíasait tíastae -s[ō]s -s[ei]s -ses -sso -somos[i]s -somoses -seteses -sontes -sonteso -gess -geiss -gé -som -ses -s -gessam -somos -gessid -gessat -setes -sont[o] Old Irish verbal paradigms 171 present s-subjunctive deponent 1 sg. 2 sg. 3 sg. 1 pl. 2 pl. 3 pl. messur messer mestir messimir meste messitir -s[ō]ros -[i]storos -str[e]s -s[e]mor[e]s -se[tes]es -s[e]ntr[e]s -fessur -fesser -festar -fessamar -fessid -fessatar -s[ō]ro -[i]storo -stro -somoro -sedwe -sontro present s-subjunctive passive 3 sg. rel. 3 pl. rel. gessair gessar gessitir gessatar -stor[e]s -storoso -s[e]ntor[e]s -sontoroso -gessar -storo -gessatar -sontoro past s-subjunctive active/deponent 1 sg. 2 sg. 3 sg. 1 pl. 2 pl. 3 pl. -gessinn -gesta -gessed -gesmais -gestae -gestais -sema[m] -seto-seto -semos[te] -se[t]e-sento[ste] past s-subjunctive passive 3 sg. 3 pl. -gestae -gestais -seto-sento[ste] 172 Italo-Celtic origins and prehistoric development of the Irish language FUTURE f-future active 1 sg. 2 sg. 3 sg. rel. 1 pl. rel. 2 pl. 3 pl. rel. léicfea léicfe léicfid léicfes léicfimmi léicfimme léicfide léicfit léicfite -ībwasoms -ībwaseses -ībwa[ti]s -ībwasso -ībwasomos[i]s -ībwasomoses -ībwaseteses -ībwasontes -ībwasonteso -léiciub -léicfe -léicfea -ībwasom -ībwases -ībw[ā] -léicfem -ībwasomos -léicfid -léicfet -ībwasetes -ībwasont[o] f-future deponent 1 sg. 2 sg. 3 sg. 1 pl. 2 pl. 3 pl. suidigfer suidigfider suidigfidir suidigfimmir suidigfide suidigfitir -ibwasōros -ibwasetoros -ibwasetr[e]s -ibwas[e]mor[e]s -ibwase[tes]es -ibwas[e]ntr[e]s -suidigfer -suidigfider -suidigfedar -suidigfemmar -suidigfid -suidigfetar -ibwasōro -ibwasetoro -ibwasetro -ibwasomoro -ibwasedwe -ibwasontro f-future passive 3 sg. rel. 3 pl. rel. léicfidir léicfider léicfitir léicfiter -ībwasetor[e]s -léicfider -ībwasetoroso -ībwas[e]ntor[e]s -léicfiter -ībwas[e]ntoroso -ībwasetoro -ībwas[e]ntoro secondary f-future active/deponent 1 sg. 2 sg. 3 sg. -léicfinn -léicfeda -léicfed -ībwasema[m] -ībwaseto-ībwaseto Old Irish verbal paradigms 173 1 pl. 2 pl. 3 pl. -léicfimmis -léicfide -léicfitis -ībwasemos[te] -ībwase[t]e-ībwasento[ste] secondary f-future passive 3 sg. 3 pl. -léicfide -léicfitis -ībwaseto-ībwasento[ste] reduplicated future active 1 sg. 2 sg. 3 sg. rel. 1 pl. rel. 2 pl. 3 pl. rel. cechna cechnae cechnaid cechnas cechnaimmi cechnaimme cechnaithe cechnait cechnaite -asoms -aseses -a[ti]s -asso -asomos[i]s -asomoses -aseteses -asontes -asonteso -cechan -cechnae -cechna -asom -ases -[ā] -cechnam -asomos -cechnaid -cechnat -asetes -asont[o] reduplicated future passive 3 sg. rel. 3 pl. rel. cechnaithir cechnathar cechnaitir cechnatar -asetor[e]s -asetoroso -as[e]ntor[e]s -asontoroso -cechnathar -asetoro -cechnatar -asontoro ē-future active 1 sg. 2 sg. 3 sg. rel. béra bérae béraid béras -asoms -aseses -a[ti]s -asso -bér -bérae -béra -asom -ases -[ā] 174 1 pl. rel. 2 pl. 3 pl. rel. Italo-Celtic origins and prehistoric development of the Irish language bérmai bérmae bérthae bérait bértae -asomos[i]s -asomoses -aseteses -asontes -asonteso -béram -asomos -béraid -bérat -asetes -asont[o] ē-future passive 3 sg. rel. 3 pl. rel. bérthair bérthar bértair bértar -asetor[e]s -asetoroso -asontor[e]s -asontoroso -bérthar -asetoro -bértar -asontoro s-future active 1 sg. 2 sg. 3 sg. rel. 1 pl. rel. 2 pl. 3 pl. rel. gigsea gigse gigis giges gigsimmi gigsimme gigeste gigsit gigsite -soms -seses -ses -sso -somos[i]s -somoses -seteses -sontes -sonteso -gigius -gigis -gig -s[ō] -ses -s -gigsem -somos -gigsid -gigset -setes -sont[o] s-future deponent 1 sg. 2 sg. 3 sg. rel. 1 pl. rel. 2 pl. 3 pl. rel. messur messer mïastir mïastar messimmir messammar mïastae messitir messatar -s[ō]ros -[i]storos -str[e]s -stroso -s[e]mor[e]s -somoros -se[tes]es -s[e]ntr[e]s -sontroso -fessur -fesser -fïastar -s[ō]ro -[i]storo -stro -fessamar -somoro -fessid -fessatar -sedwe -sontro Old Irish verbal paradigms 175 s-future passive 3 sg. rel. 3 pl. rel. mïastair gigestar gigsitir messatar -stor[e]s -storoso -s[e]ntor[e]s -sontoroso -fïastar -storo -gigsiter -sontoro 176 Italo-Celtic origins and prehistoric development of the Irish language PRETERIT s-preterit active 1 sg. 2 sg. 3 sg. rel. 1 pl. rel. 2 pl. 3 pl. rel. léicsiu léicsi léicis léices léicsimmi léicsimme -iss[ō]s -iss[ei]s -isses -isso -issomos[i]s -issomoses léicsit léicsite -issontes -issonteso -léicius -léicis -léic -iss[ō] -iss[ei] -iss -léicsem -issomos -léicsid -léicset -issetes -issont[o] s-preterit deponent 1 sg. 2 sg. 3 sg. 1 pl. 2 pl. 3 pl. -suidigsiur -suidigser -suidigestar -suidigsemmar -suidigsid -suidigsetar -iss[ō]ro -istoro -istro -issomoro -issedwe -issontro t-preterit active 1 sg. 2 sg. 3 sg. rel. 1 pl. 2 pl. 3 pl. rel. birt bertae bert(at)ar -tes -teso -tonto[ro]so -biurt -birt -bert -t[ō] -t[ei] -t -bertammar -bertid -bert(at)ar -tomo[ro] -tete -tonto[ro] Old Irish verbal paradigms 177 reduplicated preterit active 1 sg. 2 sg. 3 sg. 1 pl. 2 pl. 3 pl. cechan cechan cechain cechnammar -as -<t>as -es -amo[ro]s cechnatar -a[nto]r[o]s -cechan -cechan -cechain -cechnammar -cechnaid -cechnatar -a -<t>a -e -amo[ro] -ate -a[nto]r[o] reduplicated preterit deponent 1 sg. 2 sg. 3 sg. 1 pl. 2 pl. 3 pl. génar génar génair génammar -a[ro]s -<t>a[ro]s -[ar]es -amo[ro]s génatar -a[nto]r[o]s -ménar -ménar -ménair -ménammar -ménaid -ménatar -a[ro] -<t>a[ro] -[ar]e -amo[ro] -ate -a[nto]r[o] ā-preterit active 1 sg. 2 sg. 3 sg. rel. 1 pl. rel. 2 pl. 3 pl. rel. gád gád gáid gáde gádammar gádammar -as -<t>as -es -eso -amo[ro]s -amo[ro]s gádatar gádatar -a[nto]r[o]s -a[nto]r[o]so -gád -gád -gáid -a -<t>a -e -gádammar -amo[ro] -gádid -gádatar -ate -a[nto]r[o] preterit passive AII 3 sg. 3 pl. léicthe léicthi -itoses -itois -léiced -léicthea -itos -itās 178 Italo-Celtic origins and prehistoric development of the Irish language BI 3 sg. 3 pl. brethae brithi -toses -tois -breth -bretha -tos -tās REFERENCES Beekes, Robert Stephen Paul 1981 “The subjunctive endings of Indo-Iranian”, Indo-Iranian Journal 23, 21-27. Bergin, Osborn 1938 “Varia I”, Ériu 12, 215-235. Binchy, David 1971 “An archaic legal poem”, Celtica 9, 152-168. Boling, Bruce D. 1972 “Some problems of the phonology and morphology of the Old Irish verb”, Ériu 23, 73-101. Chantraine, Pierre 1967 Morphologie historique du grec, Paris. Cowgill, Warren 1969 “A note on palatalization in Old Irish”, Festschrift für Konstantin Reichardt, Bern, 30-37. 1975 “The origins of the Insular Celtic conjunct and absolute verbal endings”, Flexion und Wortbildung, Wiesbaden: Reichert, 40-70. 1985a “The personal endings of thematic verbs in Indo-European”, Grammatische Kategorien: Funktion und Geschichte, Wiesbaden: Reichert, 99-108. 1985b “On the origin of the absolute and conjunct verbal inflexion of Old Irish”, Grammatische Kategorien: Funktion und Geschichte, Wiesbaden: Reichert, 109-118. Cullen, John 1972 “Primitive Irish vowels in final syllables following i̯”, Ériu 23, 227-229. Dybo, Vladimir Antonovič 1961 “Sokraščenie dolgot v kel’to-italijskix jazykax i ego značenie dlja balto-slavjanskoj i indoevropejskoj akcentologii”, Voprosy slavjanskogo jazykoznanija 5, 9-34. Greene, David 1974 “The growth of palatalization in Irish”, Transactions of the Philological Society 1973, 127-136. 1976a “The diphthongs of Old Irish”, Ériu 27, 26-45. 1976b “Formy passivnogo preterita v drevneirlandskom jazyke”, Teorija jazyka, anglistika, kel’tologija, Moskva, 265-269. 1977 “Archaic Irish”, Indogermanisch und Keltisch, Wiesbaden: Reichert, 11-33. Illič-Svityč, Vladislav Markovič 1962 “K istolkovaniju akcentuacionnyx sootvetstvij v kel’toitalijskom i balto-slavjanskom”, Kratkie soobščenija Instituta slavjanovedenija AN SSSR 35, 63-72. Jackson, Kenneth Hurlstone 1953 Language and history in early Britain, Edinburgh: University Press. Koch, John T. 1987 “Prosody and the Old Celtic verbal complex”, Ériu 38, 143-176. Kortlandt, Frederik 1978b “On the history of the genitive plural in Slavic, Baltic, Germanic, and Indo-European”, Lingua 45, 281-300. 1979a “Toward a reconstruction of the Balto-Slavic verbal system”, Lingua 49, 51-70. 1979b “The Old Irish absolute and conjunct endings and questions of relative chronology”, Ériu 30, 35-53. 1981a “More evidence for Italo-Celtic”, Ériu 32, 1-22. 1981b “On the Armenian personal endings”, Annual of Armenian Linguistics 2, 29-34. 1982b “Phonemicization and rephonemicization of the Old Irish mutations”, Ériu 33, 73-83. 1984 “Old Irish subjunctives and futures and their ProtoIndo-European origins”, Ériu 35, 179-187. 1986a “Posttonic *w in Old Irish”, Ériu 37, 89-92. 1986b “The origin of the Slavic imperfect”, Festschrift für Herbert Bräuer zum 65. Geburtstag, Köln: Böhlau, 253-258. 1989a “Lachmann’s law.” The new sound of Indo-European: Essays in phonological reconstruction (ed. Theo Vennemann), Berlin: Mouton de Gruyter, 103-105. 1990 “The Germanic third class of weak verbs”, North-Western European Language Evolution 15, 3-10. 1994 “Absolute and conjunct again”, Münchener Studien zur Sprachwissenschaft 55, 61-68. 1996a “The alleged early apocope of *-i in Celtic”, Études Celtiques 32, 91-97. 1996b “Old Irish ol ‘inquit’”, Études Celtiques 32, 143-145. 1997a “Thematic and athematic verb forms in Old Irish”, Sound Law and Analogy: Papers in honor of Robert S.P. Beekes on the occasion of his 60th birthday, Amsterdam-Atlanta: Rodopi, 133-137. 1997b “On the relative chronology of Celtic sound changes”, Historische Sprachforschung 110/2, 248-251. 1999 “Lachmann’s law again”, Language change and typological variation: In honor of Winfred P. Lehmann on the occasion of his 83rd birthday, vol. I: Language change and phonology, 246-248. 2000 “Three notes on the Old Irish verb”, Études Celtiques 34, 143-146. Kuiper, Franciscus Bernardus Jacobus 1937 Die indogermanischen Nasalpräsentia, Amsterdam: NoordHollandsche Uitgeversmaatschappij. Mac Cana, Proinsias 1973 “On Celtic word order and the Welsh ‘abnormal’ sentence”, Ériu 24, 90-120. McCone, Kim 1978 “The dative singular of Old Irish consonant stems”, Ériu 29, 26-38. 1991 The Indo-European origins of the Old Irish nasal presents, subjunctives and futures, Innsbruck: IBS. Meid, Wolfgang 1963 Die indogermanischen Grundlagen der altirischen absoluten und konjunkten Verbalflexion, Wiesbaden: Reichert. 1972 “On two points of Celtic morphology”, Études Celtiques 13, 346-352. Meillet, Antoine 1907 “Sur l’origine de la distinction des flexions conjointe et absolue dans le verbe irlandais”, Revue Celtique 28, 369-373. 1908 “À propos de v.irl. beri”, Mémoires de la Société de Linguistique de Paris 14, 412-415. 1914 “De quelques finales slaves”, Rocznik Slawistyczny 7, 1-8. Pedersen, Holger 1941 Tocharisch vom Gesichtspunkt der indoeuropäischen Sprachvergleichung (= Historisk-filologiske Meddelelser 28/1), København: Ejnar Munksgaard. Rix, Helmut 1977 “Das keltische Verbalsystem auf dem Hintergrund des indoiranisch-griechischen Rekonstruktionsmodells”, Indogermanisch und Keltisch, Wiesbaden: Reichert, 132-158. Schrijver, Peter 1991a The reflexes of the Proto-Indo-European laryngeals in Latin, Diss. Leiden, Amsterdam-Atlanta: Rodopi. 1991b “The development of Primitive Irish *aN before voiced stop”, Ériu 42, 13-25. 1993 “On the development of vowels before tautosyllabic nasals in Primitive Irish”, Ériu 44, 33-52. 1994 “The Celtic adverbs for ‘against’ and ‘with’ and the early apocope of *-i”, Ériu 45, 151-189. 1997 Studies in the history of Celtic pronouns and particles, Maynooth: Dept. of Old Irish. Sieg, Emil & Siegling, Wilhelm 1921 Tocharische Sprachreste, Berlin-Leipzig: Walter de Gruyter. Sieg, Emil, Siegling, Wilhelm, Schulze, Wilhelm 1931 Tocharische Grammatik, Göttingen: Vandenhoeck & Ruprecht. Sims-Williams, Patrick 1984 “The double system of verbal inflexion in Old Irish”, Transactions of the Philological Society, 138-201. 1999 “Old Irish feda (gen. fedot): a ‘puzzling’ form in the Cambrai Homily and its implications for the apocope of /i/”, Studia Celtica et Indogermanica: Festschrift für Wolfgang Meid zum 70. Geburtstag, Budapest: Archaeolingua, 471-474. Sommerfelt, Alf 1922 “Le futur irlandais en -f-”, Mémoires de la Société de Linguistique de Paris 22, 230-233. Strachan, John 1949 Old-Irish paradigms and selections from the Old-Irish glosses, Dublin: Royal Irish Academy. Thieme, Paul 1965 “Īśopanisad (= Vājasaneyi-Saṃhitā 40) 1-14”, Journal of the American Oriental Society 85, 89-99. Thurneysen, Rudolf 1897 “Ueber einige Formen der Copula im Irischen”, Zeitschrift für Celtische Philologie 1, 1-6. 1914 “Review of H. Pedersen’s Vergleichende Grammatik der keltischen Sprachen 2 (1913)”, Anzeiger für indogermanische Sprach- und Altertumskunde 33, 23-37. 1946 A Grammar of Old Irish, Dublin: DIAS. Uhlich, Jürgen 1995 “On the fate of intervocalic *-w- in Old Irish, especially between neutral vowels”, Ériu 46, 11-48. Watkins, Calvert 1969 Indogermanische Grammatik 3, Formenlehre 1, Geschichte der indogermanischen Verbalflexion, Heidelberg: Carl Winter. Wijk, Nikolaas van 1916 “Zur slavischen und baltischen Präsensflexion”, Archiv für slavische Philologie 36, 111-116.