¹ See note 6.

² In Lang. 9.30 ff. I was misled by this Toch. -e < IE -ai, erroneously interpreted as primary ending only, to conclude that the single middle paradigm of Hittite and Tocharian was due to merging the IE primary and secondary endings. Recognition of -ai as also the latter, invalidates that conclusion, and we must assume that these two languages show no trace of alleged IE primary endings.

³ So, e.g., Streitberg, Urgerm. Gram. 322. Brugmann (647) is undecided between -ai and -o.

⁴ For the perfect see below.

⁶ The -āi in the subjunctive, e.g. Skt. yájāi Av. yazāi, is due to contraction, cf. Thumb, Handb. d. Skt. 293.

⁶ These endings are marked with the asterisk because deduced as the necessary basis for the ri-presents like ya-ha-ri from ya- ‘go’ and for preterites like es-ha-t from es- ‘sit’. Actually existing tapar-ha ‘I governed’ and the like cannot be cited as direct evidence for the existence of this form. They are marked with the ‘Glossenkeil’, and clearly are Luwian preterites borrowed in their entirety; cf. Sturtevant, Gramm. 264. Nevertheless the Luwian -ha must be ultimately identical with the Hitt. *-ha, but the former apparently did not develop a secondary distinction between presents without final -t and preterites with it.

⁷ Influence of aorists in -ην is also suggested by Gray, Lang. 6.231, but otherwise his interpretation of the ending -μην differs widely.

⁸ Also in the OLat. form in -rus, e.g. experīrus < *-so-s, made by adding the -s of the active before the change of -so to -re, cf. Brugmann 647.

⁹ That Goth. here retained the old special form in -za opposed to the 1. sing. and the 1. and 2. pl., which adopted the forms of the 3. person, is explained by Van Helten, IF 14.89, as due to the support -za (according to him -zai) had in the -zi of the 2. sing. active.

¹⁰ Even more uncertain is the identification of the -dēs of the 2. sing. of Germanic weak preterites, as Goth. nasidēs, with IE -thēs; so, e.g., Streitberg, op. cit. 339 f., for primary verbs.

¹¹ Toch. kälpā-te won out over the older *kälpā-t because the latter had become identical with the 3. sing. < IE -to.

¹² That the addition of -e took place after the separate existence of the Tocharian language, is proved by failure of the t to become palatalized to c. Contrast, e.g., ce-m = IE *toi with added -m.

¹³ Even without the evidence of Tocharian, Güntert, Idg. Ablautsprobl. 12, concluded that Skt. -makē must have been a more recent creation than -mahi.

¹⁴ After IE -medhə has been definitely established by Toch. -mät, it is unnecessary to take further cognizance of older opinions which held that Gr. -μϵθα could not have been old. So Holger Pedersen, KZ 36.80; Hirt, IF 17.67.

¹⁵ The first a in place of IE e is a case of vowel assimilation, cf. JAOS 54. 161 ff.

¹⁶ Cf. Lang. 9.31.

¹⁷ According to Wackernagel, KZ 33.57, the entire ending -σθϵ is an analogical creation beside act. -τϵ.

¹⁸ The same form functions also as first and second person. It has been suggested that in the former use it came from IE -medhə with syncopation of the middle syllable and change of m to n before the dental. However, the assumption of such a syncope is dubious, and somewhere in the process analogy must have played a part, cf. Van Helten loc. cit. (note 9, above).

¹⁹ See Brugmann-Thumb 409.

²⁰ Sturtevant, e.g., Hitt. Gram. 266, supposes that the w belonged to the original plural (instead of dual) ending.

²¹ Since unchanged -σι in the 2. sing. is found only in ἔσ-σι, and -τι only in West-Greek, the whole process must have been completed in the earliest stages of the Greek language and before its dialectic differentiation.

²² For Av. -yē = -vē see Reichelt, Av. Elementarb. 72.

²³ Cf. Sommer, BoSt. 7. 22 ff.

²⁴ Cf. Sturtevant, Gram. 126.

²⁵ So Meillet, BSL 32.1 ff.; E. F. Claflin, AJP 48.157 ff.; Lang. 5.232 ff.

²⁶ For details see Lang. 9.14.

²⁷ Cf. also the amazingly subtle and complicated reconstruction of IE active and middle forms made by Roberts, Lang. 11.220 ff., to account for the OHG active 1.pl. on the basis of the IE middle. To me the difficulties of his theory exceed many times those of some of the others which he has rejected. It is particularly hard to see any serious objections to the interpretation of Brugmann, KVgl. Gr. 591, who thinks of the long ē of beramēs as being derived from the secondary -mē. That this existed in Germanic with a long vowel as well as a short one, is shown by the Goth.1.pl. optative nimaima (< -mē or -mō) beside the pret. indicative nēmum (< -mě or -m⊖). If now we assume that the older form of -mēs was mēsi, we can derive the latter by a proportional analogy going back to the time when -mě and -mē existed alongside of each other: -mě: -měsi = -mē: -mēsi. The ending -měsi with short e is identified with -masi, e.g. in Ved. Skt. s-mási, bhárā-masi, Av. -barā-mahi, and with the ending of Ir. ammi < *es-mesi. If one balks at the notion that this old form should have persisted only in Old High German of all Germanic languages, it is easy to assume that it was created independently as a dialectic pre-Germanic form in what later became OHG territory. The analogy by which -i slipped from the 3.pl. in -nti and the 1.sing. in -mi to the 1.pl. in -mes to change it to -mesi, is so easy that it could have been performed repeatedly in different communities without geographical contact.

²⁸ Thus according to the above exposition the supposed old ablaut between ai and i in Gr. δίδονται: Dor. δίδοντι and Skt. dádhate: dádhati has turned out to be not IE ablaut, but a late analogical growth in Aryan and Greek.

²⁹ This means, of course, ablaut in the inflectional endings. Since the words themselves largely belonged to older periods, it is inevitable that they should show ablaut variations of all kinds in their radical parts.

³⁰ [One fact seems to have been overlooked. It is true that in Greek before the use of the augment became obligatory φέρϵσϑον, φϵρόμϵϑα, φέρϵσϑϵ were wide forms, indifferent to the present-past distinction. The same, however, is true for the corresponding forms of the active φέρϵ ον, φέρομϵν, φέρϵτϵ. In these forms a present-past distinction is valueless, since all designate an action performed by the person addressed, who consequently needs not to be told about its time. Then a similar wide form is to be expected for the second person singular. In the middle there are only narrow forms φέρϵαι: φέρϵο; but the active with its use of -ς both as a ‘primary’ and ‘secondary’ ending points to the existence at one time of wide forms for this function too. Indeed τίϑης, δίδcως, ὄμνς, φής, φέρϵς (Dor. Cypr.) are just what is to be expected (contrast Brugmann-Thumb 398), provided that the earliest types survive without analogical replacements.

I do not intend to attempt to trace the various ways in which the present-past distinction received formal expression. But the wide forms are evidently archaic, survivals from a time when the language, as Hirt (Idg. Gram. 6.209) believed, did not distinguish by verbal forms present and past tense. The differentiation began, no doubt, in forms of the third person, where the need for it was greatest. Whether the speaker means ‘I am carrying’ or ‘I was carrying’ must frequently have been obvious from the context; and Hirt is probably right in his suggestion (Idg. Gram. 4.138) that *-mi is analogical to -ti, -nti.

One corollary may be mentioned: *esi is not in some obscure way a phonetic continuant of *es-si; both are analogic remakings of *es(s) < *es-s. GMB]