31

objection, what we have seems best interpreted as a contamination, to avoid such confusion, of *bir with the ber-forms of the plural (-beram, -berid, and -berat), these, like -biur and -bir, being regular in their vocalisation. Following Bezzenberger, one might consider here such Old Lithuanian forms as pa-vystai 'withers', džustai 'becomes dry', plaūkiai 'swims', and possibly even Old Prussian pīdai 'carries', pogaunai 'receives', postānai ‘becomes', dīlinai 'works', and niswintinai ‘unhallows'. This-ai may, however, be regarded as a particle;32 and if these Baltic words belonged here, they would point to a termination in *-oi(t) rather than in *-ei(t). The usual Lithuanian termination in -a (e.g. vēža 'carries') is generally regarded as an injunctive. 33 Albanian del' 'comes forth' is ambiguous, being derivable equally well from *dalet or *daleit. Kuchean wessa, 'says' beside yapi 'enters' apparently receives its nasal from an enclitic pronoun.34 Gaulish legasit and cariedit may, or may not, be verbs.35 Peside the usual ending -zi (jezzi 'makes') 'Hittite' hasi (e.g. dāi ‘gives', ezzai 'eats' [beside ezzazi], tašuwaḥḥi 'injures' [beside tašuwaḥzi], halzai 'calls', hatrai 'writes' [but contrast batrāmi], šipanti 'sacrifices'). Hrozný-36 could reach no satisfactory conclusion; Marstrander37 denies any connexion with the Greek -a, and considers it comparable with the third person singular indicative middle (Sanskrit bhárate, Greek pépera). I myself fail to see the force of his objections, and am inclined to compare ezzai directly with the ede 'eats' of Iliad 15. 636.

The athematic termination (Sanskrit bhárati, Armenian berē < *bhereti, Doric didwr [Attic didwo] 'gives',38 Latin agit 'does', δίδωσι δίδωσι] Oscan staít 'stands',39 Old Irish crenaid 'buys', Gothic bairip 'bears',

31 Litauische Sprache 197.

32 E. Berneker, Preussische Sprache 212, Strasbourg, 1896; R. Trautmann, Altpreussische Denkmäler 280, Göttingen, 1910, takes it as characteristic of bases in -ăi.

33 e.g. Brugmann, Grundriss 2. 3. 615; Meillet, Introduction 192.

34 Cf. Meillet, in Indogermanisches Jahrbuch 1. 8, 9; Lévi and Meillet, in MSLP 18. 8. Lydian fennsλibid (for λ instead of un cf. Buckler, Sardis 6. 2. xiii) 'destroys' is perhaps an optative, cf. varbtokid 'may he take vengeance' (see E. Littmann, Lydian Inscriptions Sardis: Publications of the American Society for the Excavation of Sardis, 6. 35, 40, Leyden, 1916).

=

35 See Dottin 37, 45, 122, 242, 265.

36 Sprache 161.

37 Caractère 91.

38 On the relation of -r and - cf. Brugmann, Grammatik 118; MeilletVendryes 61-2.

39 Lindsay 526; Sommer 489.

Old Russian budeti 'becomes' beside bude40) calls for no special remark in the present connexion.

V. The Dual

The dual number in the verb has survived only in Sanskrit, Avesta, Greek, Teutonic (Gothic, Runic Norse, and possibly in the stereotyped Old Saxon wita 'let us'41), Lithuanian, and Old Church Slavic. The scanty material available may be tabulated as follows, the two forms in brackets being restored analogically:

These terminations are apportioned thus:

Primary athematic: 1-Gālā Avesta usvahi 'we two wish';

thematic: 1 a-Sanskrit bhárāvas, Gothic baírōs; 1 b-Lithuanian sukava (reflexive súkavo-s[i]), Old Church Slavic berevě,43

2 a Sanskrit bhárathas, Gothic baírats; 2 bLithuanian sukata (reflexive sukato-s[i]), Old Church Slavic bereta;

3 a-Sanskrit bháratas, Avesta barato; 3 b— Old Church Slavic bereta, berete.44

Secondary: 2-Greek pépeтov <*еретoμ (cf. Sanskrit ábharatam). (In the present pépeтov has also assumed the functions of the third person, though the original distinction is maintained in the imperfect, so that ἐφέρετον : ἐφερέτην [ef. Boeotian ἀνεθέταν] : : Sanskrit ábharatam: ábharatām.)

40 Cf. also O. Wiedemann, Beiträge zur altbulgarischen Conjugation 11-23, Petrograd, 1886; for Slavic -tu beside -ti see Meillet, 'La Désinance -tu du vieux slave', in MSLP 18. 232-8, and Slave 270-2.

"Brugmann, Grundriss 2. 3. 638.

"Originally there may also have been forms in *-ue, *-uō if we may draw an inference from the corresponding plurals *-mē, *-me, *-mō, *-mo.

43 The final -ě is probably influenced by the pronoun ve 'we two' (Brugmann, Grundriss 2. 3. 639).

"See Wiedemann 25.

It would also appear that, unlike the first and third persons, the second person dual and plural have had a strong reciprocal influence, though from or to what point is wholly obscure. This inter-relation becomes clear from the following table:

The first person plural seems, for the most part, to be derived from forms so similar to those of the first person dual-except that the plural is characterised by m as the dual is by u-that we may again resort to tabulation:

(a) -mési: Sanskrit smási, Old Irish ammi < *esmési, Kuchean ymäşä 'we go';

(b) '-mosi: Sanskrit bhárāmasi, Avesta barāmahi, Albanian kemi, Old Irish bermi < *bherómosi;

(c) -mé(s): (i) Sanskrit imás, Doric pépoues, Old High German berames (?); (ii) Czech neseme;

(d) '-mo(s): (i) Sanskrit bhárāmas, Turfanian kalkāmas ‘we go’; (ii) Latin ferimus, Albanian pú0(e)me, púðime, Gaulish priavimo (?),45 Old Irish -beram, Gothic bindam, Serbian pletemo.

There are also some forms which do not fit in this scheme.

(1) Apabhra, sa Prakrit vaṭṭahun (ordinary Prakrit vaṭṭāmo, -mu, Sanskrit vartāmas), for which Pischel could offer no explanation, probably receives its -u, by analogy with the first person singular, while the h may be by another analogy with the second person plural, where h legitimately corresponds to Sanskrit th. The Pāli ending -ma (labhāma 'we seize') is taken from the imperfect;47 the Afyān -ū is of uncertain

45 Dottin 122.

46 Grammatik 257, 323; cf. also Hoernle 335-6; Beames 3. 105-6. Sindhi has its plural in -ūn, Marāṭhī in -un or -on, and Uriyā in -un; Hindi shows -ên (<*-ahin?), Panjabi has -ye, Gujarātī -(i)ye, and Bengāli -i

47 Geiger, Pali 107.

origin;48 and the Balūči -ūn (-in) is probably influenced by the -n of the first singular, as is also West Ossetic -än (first singular, -un) as contrasted with East Ossetic -äm (first singular, -in).

(2) Greek pépoμev has a curious parallel in 'Hittite' -wen (i), -meni (e.g. jaweni 'we make', waḥnummēni we surround')." Before 'Hittite' had been deciphered, Brugmann50 advanced the suggestion that Greek -μev was formed on the analogy of a vanished first dual *-fev (so that the Indo-European imperfect dual should be reconstructed as *-ye, -uo, -yen [-yon (?)]; -tom; -tām); but however this may be, in the 'Hittite' forms we have an alternation between m and w not uncommon in that language.51

(3) Old High German tuomēs 'we do', berames 'we carry', etc., used both as indicatives and as adhortatives, can scarcely be derived from an Indo-European termination *-mēsi, of which we find no evidence elsewhere. All the explanations hitherto advanced52 are quite doubtful, perhaps the most plausible being that -mēs receives its ē by analogy with the subjunctive (e.g. berēs, berēt, berēn), especially in view of its frequent use as an adhortative, a possible parallel being found in the Balūči first person singular in -ān. On this hypothesis beramēs would be for a pre-Teutonic *berames, Indo-European *bheromes. One also finds a few instances of the ending -ēm (e.g. bittēm 'we ask'), which later regularly becomes -en (e.g. nëmën 'we take'), this being the subjunctive used as an indicative.53

(4) Old Church Slavic nesemu 'we bear' (beside forms in -mi, -my, and once -mo),54 though usually regarded as very obscure in formation, is probably best explained as for *-mos (cf. Old Church Slavic vléků,

48 Geiger, in Grundriss 1. 2. 220.

49 Hrozný 155; cf. Marstrander 91.

50 Grundriss 2. 3. 618; cf. also Hirt 488; Meillet, Introduction 141-2, 193; Meillet-Vendryes 309.

51 Marstrander 151-2.

52 Brugmann, Grundriss 2. 3. 621; W. Braune, Althochdeutsche Grammatik 253-4, Halle, 1911; W. Streitberg, Urgermanische Grammatik 321, Heidelberg, 1896; F. Hartmann, in F. Dieter, Laut- und Formenlehre der altgermanischen Dialekte 505-6, Leipzig, 1900; O. Behagel, Geschichte der deutschen Sprache 266-7, Strasbourg, 1911; F. Kluge, Urgermanisch3 186, do. 1913; R. Kögel, in Beiträge zur Geschichte der deutschen Sprache und Literatur 8 (1882). 126-34. The rare forms in -mas represent an internal dialectic change, those in -mus (found five times) are probably miswritings, and those in -men apparently arise from a wrong. interpretation of the final m- in Old High German manuscripts (Kögel 130-1). 63 Braune 254.

54 Wiedemann 7-8.

55 Brugmann, Grundriss 2. 3. 622; Vondrák 1. 138, 2. 138; Meillet, Slave 265-8.

Lithuanian vilkas 'wolf': *ulgos, dative plural Old Church Slavic vlŭkomu: Lithuanian vilkamus, vilkáms).56

VII. Second Person Plural

The thematic form of this termination usually appears as *-t(h)e, e.g. Sanskrit bháratha, Gala Avesta išala 'ye seek', Armenian berēk' < *berete- ?,57 Greek pépere, Latin (imperative) agite, Old Irish -berid, Gothic bindip, Old High German bintet,58 Lithuanian vēžate, Old Church Slavic berete. Here belongs also Albanian delne 'ye go forth' for *dalnet(h)e, the n being taken over from presents in -n.59

On the other hand, *-t(h)es seems to be the source of the ordinary Latin form, as in agitis, this explanation apparently being more plausible than an interpretation from *agite with -s added on the analogy of agimus, or as being to agis what agite is to age, or as an old dual.60

One may suggest, furthermore, that an athematic *-t(h)esi underlies the Old Irish absolute berthe < *bertsi < *bheret(h) esi, this being lenited as one would have expected to be the case in the first persons singular and plural berim(m), bermi (*bheremi > *berimh; *bher ómosi > *berimhi), except that they were influenced by the analogy of the copula am, ammi, an influence which is lacking in the second person plural adi(b), idib.61

Whether the Old West German dative Vatvims, Runic Norse borum R* 'to the sons' really belongs here, or whether we have an instrumental used as a dative, is uncertain (Brugmann, Grundriss 2. 2. 257-8, 262; Streitberg 232; R. Bethge, in Dieter 546-7; Kluge 197; A. Jóhannesson, Grammatik d. urnordischen Runeninschriften 49, 52, Heidelberg, 1923).

57 Cf. Meillet, Esquisse 89. For an interesting explanation of -k' as a plural ending both in nouns and in verbs (< *-tva-, i.e. a neuter abstract used as a plural for nouns and then transferred analogically to verbs) see A. Zanolli, 'Qualche osservazioni sulla formazione del plurale nell' antico armeno', in Huschardzan 275-8, Vienna, 1912.

58 On the Old High German forms, including the difficult Monsee type of zimbrit 'ye build' (perhaps by analogy with the second and third persons singular), see Brugmann, Grundriss 2. 3. 625-6; Bethge, in Dieter 332-3; Braune 255; Kögel 135-9.

59 G. Meyer, Kurzgefasste albanesische Grammatik 32, Leipzig, 1888.

6o Cf. Brugmann, Grundriss 2. 3. 625; Sommer 489-90; Lindsay 529; F. Stolz and J. Schmalz, Lateinische Grammatik 305, Munich, 1926; R. von Planta, Grammatik der oskisch-umbrischen Dialekte 2. 283, Strasbourg, 1897; J. Wackernagel, in IF 31 (1913). 260, note. Meillet-Vendryes 310, 326 also hold that Latin -tis is for *-tes, which may have been Indo-European.

61 Thurneysen 340 offers no explanation for this form; Pedersen 2. 344, because of the Brythonic forms (Middle Welsh carewch, Cornish careugh ‘ye love'), derives

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