The 'Hittite' ending -teni (e.g. jatteni 'ye go') seems to be a secondary termination corresponding to that found in Vedic vádathana 'ye speak' etc., and parallel with the secondary -ueni, -meni of the first person plural.62 In Kuchean triçcer 'ye will sin' the -r is apparently an affixed particle.63 Some of the Modern Iranian and Indian forms are quite obscure, as Afyān -aī, Kuhrūdī -ige, -ike, and Bhašgalī -r (with nasalisation of the preceding vowel); Ossetic has -etä in the western dialect and -ut in the eastern.64

VIII. Third Person Plural

The original endings of this person are universally regarded as *-nti for primary and *-nt for secondary tenses. The overwhelming majority of forms actually found seem to support this view: Sanskrit bháranti, Avesta -barenti, Kuchean weske, 'they say', Turfanian şeñcä 'they are', 'Hittite' janzi 'they make', West Ossetic fä(v) úncä, East Ossetic fäúnc 'they are', Armenian beren, Doric pépovт (Delphic ἀνατίθεντι, Phocian ἀφίεντι, Elean μετέχοντι, Locrian φυλάσσοντι, Boeotian xaλéove with 0 on the analogy of the middle endings -μea and -σ0), Latin ferunt, Old Latin tremonti 'they tremble' (?), Umbrian furfant 'they purify' (?), Gaulish cartaont 'they cleanse' (?), 67 Old Irish berit, Gothic bindand, Old High German bintant < *bindandi, Old Saxon, Anglo-Saxon bindad < *bindánpi, Old Russian berat 68 for the present, and Sanskrit ábharan, Avesta barən, Armenian berin, Greek pe pov, Latin fere-b-ant, Old Irish -berat, Old Church Slavic pleta, 'they wove' for the imperfect. Here, too, Attic pépovoι (Les

66

it from *berete ues, *ues being the personal pronoun of the second plural. The first person singular is, however, lenited in other Celtic dialects (Middle Welsh, Cornish caraf, Middle Breton caraff 'I love'; but note Old Breton rannam, 'partior').

Geiger, in Grundriss 1. 2. 220, 396; Linguistic Survey of India 8. 2. 37; Miller 69-70. For the Lithuanian and Lettish reflexive and the Old Prussian forms, of minor interest in the present connexion, see Brugmann, Grundriss 2. 3. 626; Trautmann 274.

65 Brugmann, Grammatik 400–4; Hirt 488-90; G. Meyer, Griechische Grammatik 343-5; Meillet-Vendryes 304-6.

"This form is, however, not beyond suspicion (see especially Lindsay 530). 67 Dottin 122, 243.

❝ On -tŭ beside -tt in Slavic see Vondrák 2. 140-1; Leskien Altbulgarische Sprache 191; Wiedemann 23-5; Meillet, in MSLP 18. 232-8, and Slave 272-4.

bian agioto) is reckoned, to which should be added Arcadian Totevo (cf. the Arcadian subjunctives κελεύωνσι, κρίνωνσι, παρετάξωνσι).

There are, however, a couple of forms which are difficult to explain on this hypothesis. Of these the most striking is the Osco-Umbrian secondary ending -ns. The material here is as follows:69

(a) Primary: Oscan present se(n)t, amfret, fiiet, staíet, stahint, eestínt; future censazet; second future tríbarakattuset, angetuzet; Umbrian sent, furfant, furfa0; future furent; second future benurent, dersicurent, fakurent;

(b) Secondary: Oscan imperfect fufans; perfect uupsens, fufens, prúfattens, teremnattens; present subjunctive deicans, pútíans; imperfect subjunctive patensíns, h]erríns; perfect subjunctive tríbarakattíns; Paelignian coisatens; Marrucinian amatens; Volscian sistiatiens; Umbrian perfect eitipes; present subjunctive sins, dirsans, neiřhabas, etaians.

This -ns is explained by Buck as for -n < *-nd < *-nt + s from the plural of nouns (for a similar phenomenon cf. Armenian beremk', berēk' 'we, ye, bear' on the analogy of bank' 'words' etc.). Another hypothesis may, however, be even more probable, especially as it avoids analogy with nouns and is in closer harmony with what seems to have been the Indo-European inflection of the present and imperfect tenses. The writer would suggest that this Osco-Umbrian -ns was derived from *-nts.

The combination *-nts and *-nt-su (or *-nt-si) obviously occurred in the nominative singular and locative (Greek dative) plural of the present active participle.70 For the nominative we have, for instance, Sanskrit bhára,s, bhárantas, Avesta žvans, Greek ǎeis <*afeVTS, φέρων < *φεροντς, φέροντος, Latin ferens, ferentis, Umbrian zeref, serse, kutef, restef, reste, frehtef," Gothic frionds, Old High German friunt, Lithuanian vēžāns (dialectic -ans, -us), Old Prussian sidans, Old Church Slavic znaja,, cf. also 'Hittite' dan 'giving', Kuchean Ikāṣṣeñcañä (nominative plural) 'seeing'.72

For the locative or dative plural we have: Sanskrit bháratsu, Avesta fšuyasū, Greek pé povo, and the following dialectic forms:73 -001

"Von Planta 2. 280–2, 290 sqq., 315 sqq.; C. Buck, Grammar of Oscan and Umbrian 71-3, 80-1, 152, Boston, 1904. Lydian vqbapent (formerly read vqbahe,nt) 'may they destroy' may be a similar formation (cf. Littmann 37, 69).

70 See, in general, Brugmann, Grundriss 2. 1. 130-1, 255-6, 298-9. 71 For this -f < *-nts see Buck 107, 112; von Planta 1. 508-10.

72 Hrozný 82, 89; Meillet, in Indogermanisches Jahrbuch 1. 9.

73 Bechtel 1. 180, 2. 120, 250, 400, 1. 353, 2. 484, 61-2, 342, 25, 427.

in Thessalian (KATOLKÉVTEσσL), Phocian (ie pouvaμóveσσi), and Corinthian (νομιζόντεσσι, νικώντεσσι); -ασσι in Tarentine and Herac lean (ποϊόντασσι, πρασσόντασσι, Κυπαρχόντασσι); -ονσι in Arcadian (πολιτεύονσι), Cretan (ἐπιβάλλονσι), Argive (έπαγγέλονσι), and Argolic (Ovovac); -OVTOLS in Aetolian (VIKEóvTOLS, etc. this termination taken over in Laconian); and -ovvTOLS in West Locrian (ἐπιτελούντοις, ἐκλογευόντοις). As the dative plural pé povo shows, the third plural pépovo may have come from *bherontsi just as plausibly as from *bheronti, as is usually assumed; but the OscoUmbrian -ns can scarcely be explained by a secondary pre-form *-nt, whereas it may very well be derived from *-n(t)s. We may, therefore, say that in Osco-Umbrian secondary final *-nt-s (*bherent-s) > -f (Umbrian frehtef, etc.), but that original final *-nts > -ns (Oscan fufans). It is possible that *-ntsi may explain the Apabhra,sa Prakrit termination -ahi, (e.g. vattahi, ordinary Prakrit vaṭṭanti, Sanskrit vartanti), which Pischel regarded as of doubtful origin, and which can scarcely be derived from an original *-nti. One may, however, suggest the following development: -ahi, <*-ahin < *-asin < *-atsin < *-antsi, the evolution being influenced by analogy with the first and second persons plural, vaṭṭahu,, vaṭṭahu.

và

=

The termination *-nts may survive not only in Osco-Umbrian fufans, etc., but also in the Modern Greek vernacular third plural present ending -ουν(ε) (e.g. θέλουν(ε) θέλουσι, δένουν(ε) ‘δέουσι'), which is carried analogically into the aorist subjunctive active and passive (và déσovv(e) ‘δήσωσι, νὰ δεθοῦν (ε) δεθῶσι)." This form is traced by Hatzidakis 76 to the Ancient Greek -ovol. A more probable explanation, however, seems to be that dévovv (e) is for *SEVOVT <*SEVOVTS, with τουν < * οντς like the Classical φέρουσι < *φεροντσι. This finds an analogue in Old Lithuanian forms like gina, 'they defend, ward off';77 and in some Old Lithuanian texts the forms with and without nasals occur side by side, as: kurie gárbina, vgni . . . kurie žinauian, búre,, nuodiian, álwu* yr waškú lāia* 'who worship fire . . . who divine, conjure, poison, cast tin and wax'.78 Though the writing an,

74 Grammatik 323-4; for h < s cf. Apabhransa Prakrit nisarahi, Sanskrit niḥsarasi 'goest forth' (ib. 183); see also Beames 3. 103-4; Hoernle 336-7; Bloch 234-5.

"The vernacular imperfect termination (derav, édévave 'édeov') is aoristic in origin (έδεσαν, έδεσανε ἔδησαν).

76 Einleitung in die neugriechische Grammatik 110-12, Leipzig, 1892.

77 For the material see Bezzenberger 197.

78 Catechism of Daukšas (1595), ed. E. Wolter 24, Petrograd, 1886, and Lietuviška Chrestomatija 58, do. 1903-04.

* The 1 of this word should be a crossed letter.

etc., is still employed etymologically, it is now pronounced as a, etc., but we know from the testimony of Praetorius" that a, was spoken with nasalisation as late as the end of the seventeenth century. It is not altogether correct, then, to say, as does Brugmann,80 that in Lithuanian the third singular has replaced the third plural; the fact is, rather, that the loss of nasalisation in the third plural has led this to coincide with the third singular, which, as already noted, was probably an injunctive in origin.

Returning to the Modern Greek form, we observe that in certain dialects both -ovσ and -ovv occur, as in the Southern Sporades, where the ending is used with or without the pronoun, but the ending -σ only without it, the exact opposite to the rule in East Crete.81 It is also worth noting, in this connexion, that -o- and -ou- often interchange in the Modern Greek inflection of the verb, e.g. present active dévo(v)μe; present passive δένο(υ)μαι, δενο (ύ)μαστε; imperfect passive ἐδένουμου(ν), èdevóμovv (e), etc.; and in some dialects, especially of Northern Greece, one even has such forms as dévov 'déw'.82 The - in dévovv(e) is probably added by analogy with δένο(υ)με δέομεν, etc.

Finally, one may suggest that the rule in Sanskrit sandhi that a sibilant is added to words ending in a nasal when they are followed by words beginning with a palatal or dental, and that final n is doubled before a word beginning with a vowel, often really points to an original termination *-nts in the verb such as demonstrably exists in the participle, thus explaining not only gacchans ca, icchans tatra, but also abhara,s tataḥ, and possibly abharann iha.

The theory here proposed may be tabulated thus:

79 Deliciae Prussicae, oder preussische Schaubühne, ed. W. Pierson 141, Berlin, 1871 ('wie an, doch dass das n nicht deutlich sey, sondern gleichsam durch die Nase gezogen').

80 Grundriss 2. 3. 615-6, 637, note.

81 R. M. Dawkins, Modern Greek in Asia Minor 53, cf. also 179, Cambridge, 1916. 82 Thumb, Modern Greek 151-2, 156.

83 W. Whitney, Sanskrit Grammar 70, Boston, 1891; Thumb, Handbuch des Sanskrit 130-1, Heidelberg, 1905; cf. Macdonell 68, 69; R. Gauthiot, La Fin du mot en indo-européen 148–51, Paris, 1913.

IX. Summary

If we summarise the views advanced in this study, we may formulate our reconstruction of the Indo-European personal endings in the present and imperfect active in the following table. Here the italicised forms are those in which I deviate more or less from the current views, my difference in the thematic dual and first and second persons plural being that I consider the forms in *-si to have been originally athematic, and those in *-(s), without *-i, to have been originally thematic.

The origin of the distinction between the athematic and the thematic inflections is very problematical. In any case, the theory of H. Zimmer that Old Irish preserves the ancient differentiation by employing conjunct (thematic) forms after preverbs, but absolute (athematic) forms when no preverb precedes (e.g. di-an-beir 'to whom he carries': berid 'he carries'), is scarcely tenable. 85 It seems possible to suggest, however, that the basis of the difference was accent, since the athematic types, as shown especially by Sanskrit and Greek, show varying accent, whereas accentuation is unchanging in the thematic types (Sanskrit émi, imás: Greek ei, tuev [<*luév, cf. είμι, ΐμεν κ*ϊμέν, cf. also εσμέν, Sanskrit

34 KZ 30 (1890). 119-20 (note).

"Meillet, in Revue celtique 28. 370-1; Brugmann, Grundriss 2. 3. 587-9; cf. also Thurneysen 327, and, for an entirely different theory, deriving the absolute from the conjunct by suffixing pronouns (e.g. berid < *bheret is), Pedersen 2. 340-1.

96 Brugmann, Grundriss 2. 3. 60; Meillet Introduction, 151-2. In Teutonic and Balto-Slavic athematic verbs analogy has destroyed the apophonic alternations.