¹ CIL 1².581, hereafter referred to as SCdB; datable in 189 B.C.

² Trustworthy evidence is not abundant; there are not very many inscriptions which are datable with certainty before 150 B.C., except those dated by linguistic evidence; and to accept such is arguing in a circle. The Scipionic Epitaphs are all suspected of being later compositions, or at least of being later copies, like the Inscription of the Columna Rostrata. Many of the oldest inscriptions come from country districts and may show dialectal peculiarities.

³ Cf. F. Sommer, Handbuch der lateinischen Laut- und Formenlehre² 338.

⁴ For the ei, there is sometimes found the writing e as a sign of progress toward the monophthongal ī of the city and ē of the country; cf. Sommer Hdb.² 73.

⁵ Cf. Sommer Hdb.² 338; F. Stolz and J. H. Schmalz, Lateinische Grammatik ⁵ 268-9, revised by M. Leumann and J. B. Hoffmann (referred to hereafter as Stolz-Leumann).

⁶ Cf. Sommer Hdb.² ib.; Stolz-Leumann 269.

⁷ Cf. Sommer Hdb.² 347, 350.

⁸ Annaei CIL 3.6374; Annai CIL 6.11670, 9.4558.

⁹ Also ~ĭus; length is retained optionally, but seemingly only by the analogy of the dative; yet cf. Sommer Hdb.² 443-6.

¹⁰ For this text and interpretation, cf. American Journal of Philology 32.283, 289; and for illī, AJP 34.320-1, and H. Ehrlich, Untersuchungen über die Natur der griechischen Betonung 77. I must again uphold the accuracy of Lucilius's orthographic rules for the writing of i and ei; cf. AJP 32.272-93, 34.315-21, Glotta 4.299-302.

¹¹ I write ei without the macron; because it represents either a contracted monosyllable ei (diphthongal) from an earlier eī, or a dissyllabic ěī, kept from contraction by the influence of the other dissyllabic case-forms. Both values are metrically demonstrable in Plautus, as well as spondaic ēī, with pronunciation ei-jī.

¹² Cf. Sommer Hdb. ² 445-6.

¹³ For a good general discussion, cf. Sommer Hdb.² 593-5; for fuller bibliography, cf. Stolz-Leumann 328-9.

¹⁴ CIL l².4; cf. Lang. 2.218-9, where the problem of rhotacism in this word is dealt with.

¹⁵ Fest. ap. Paul. 48.19 ThdP = 68.13 M: dasi ‘dari’.

¹⁶ Occurrences listed in F. Neue Formenlehre der lateinischen Sprache ³ 3.224-35, revised by C. Wagener.

¹⁷ For other theories, cf. Stolz-Leumann 328-9.

¹⁸ Cf. Sommer Hdb.² 343.

¹⁹ Cf. Sommer Hdb.² 385, 392.

²⁰ Cf. Sommer Hdb.² 375-7; on loss of -d, id. 273.

²¹ Cf. Sommer Hdb. ² 346.

²² Cf. Sommer Hdb.² 373.

²³ CIL 1². 1.

²⁴ Cf. Sommer Hdb.² 410.

²⁵ Cf. Sommer Hdb.² 413.

²⁶ Cf. Sommer Hdb.² 347, 350.

²⁷ So Ehrlich Untersuch. 71; but Sommer Hdb.² 340 declares for a dissyllabic locative -iī in -io- stems and therefore for old Latin -ei, not -ī, on the basis of Brundisiī Enn. var. 37, Sūni (ī) Ter. Eun. 519. But Brundisiī may be an early remade form metri gratia, and Sūn(ī) with elision is just as good as Sūni(ī) in the Terentian passage.

²⁸ Cf. Sommer Hdb.² 574. The earliest occurrence is petiei alongside Accvmvlavi, genvi, optenvi CIL 1².15, which is one of the Scipio Epitaphs, but is later than 150 B.C. The second person singular likewise ends in -ī in classical Latin, getting the long vowel by analogy of the ending of the first person; the oldest example is gesistei ‘gessisti’, CIL 1².10, about 175-165 B.C., which would point rather to original -ai except that it is in one of the Scipio Epitaphs, and is therefore unreliable; cf. ftn. 2.

²⁹ Published by G. Herbig, Indogermanische Forschungen 32.86. There seems no reasonable doubt of his interpretation, though the final letter of the word is damaged.

³⁰ Occurring in CIL 1².60, 560; cf. also Lang. 5.22.

³¹ C. D. Buck, Grammar of Oscan and Umbrian, No. 19; R. S. Conway, Italic Dialects, No. 130; R. von Planta, Grammatik der oskisch-umbrischen Dialekte, No. 128.

³² It is interesting to note that L. H. Gray, in Lang. 6.230, on purely theoretical grounds, posits accented -ei and unaccented -oi as endings of the first person perfect middle, appearing in (active) Latin vīdī, OChSl. vědě, as well as in the Faliscan and Praenestine forms cited above. Now -ei would given Praen. -i, but either -oi or -ai would more easily give OChSl. -ě (K. Brugmann Grundriss d. vergl. Gram. d. idg. Spr. 1².190-2); and Fal. -ai requires -ai: Gray posits use of - in thematic presents middle. Possibly the -a of the active (cf. Greek) had in some Italic dialects an influence on the middle ending before the latter entirely supplanted it. Latin presumably, but not certainly, agrees with Praenestine, as the nearer dialect linguistically as well as geographically.

³³ As already given; cf. ftnn. 13, 14, 15, 17, supra.

³⁴ Verba auf -io 46 (not accessible to me; cited by Sommer Hdb.² 593); so also A. Fick Göttingische Gelehrte Anzeigen 1883.586.

³⁵ Hdb.² 593; Kritische Erläuterungen 173.

³⁶ Stolz-Leumann 328.

³⁷ See Excursus I, infra.

³⁸ According to the listings in W. D. Whitney, The Roots, Verb Forms, and Primary Derivatives of the Sanskrit Language.

³⁹ A similar particle ā was added to many locatives in Iranian as a formative, but this was the long vowel and not the short one, and the Iranian locatives thus formed were regular paradigmatic forms.

⁴⁰ Data from Whitney, op. cit.

⁴¹ So H. Hirt, Indogermanische Grammatik 3.51-2; so also previously, inter alios, A, Meillet and J. Vendryes Traité de Grammaire Comparée des Languages Classiques 421-2, 475. Perhaps the belief in dative -ai is a hangover from the time when the Skt. vocalism a (without separate o and e) was believed to be original; then when the separate existence of e o a was established for the primitive speech, the a remained as a general symbol for any vowel of the group whose exact quality was difficult or impossible to determine.

⁴² Op. cit. 333-4.

⁴³ So Stolz-Leumann 269.11-2, Ehrlich Untersuch. 67.10-2, etc.

⁴⁴ So N. van Wijk, Der Nominale Genetiv Singular, Zwolle, 1902; cf. H. Hirt Idg. Gram. 3.145.

⁴⁵ So all recent authorities, seemingly, except Ehrlich.

⁴⁶ So Sommer Hdb. ¹ 371.

⁴⁷ So J. Wackernagel Mélanges de Linguistique offerts à F. de Saussure 125-52.

⁴⁸ Cf. Sommer Hdb.² 340-1 and Krit. Erl. 103-5; Stolz-Leumann 269 init.; J. Schrijnen De Latijnsche Genitief Singularis van de O-Stammen, cf. Lang. 5.196; Meillet-Vendryes 408.

⁴⁹ Untersuch. 66-79. The Celtic genitive in -i may come from -ei as well as from -ī, cf. Brugmann Gdr. 1².239. The evidence of Messapian and other slightly known languages is too dubious to be used against Ehrlich's theory.

⁵⁰ Untersuch. 76-8.

⁵¹ On the other hand, Latin extended the genitive -ī of -o-stems to -ā- and -ē-stems.

⁵² The Oscan infinitives deicum ‘dicere’, acum ‘agere’, etc., are similarly ambiguous, being either accusatives of -o-stems or of consonant-stems, since the -o-stem ending of this case was in Oscan transferred to consonant-stems.

⁵³ Idg. Gram. 3.50.

⁵⁴ Verbal communication of E. H. Sturtevant.

⁵⁵ A similar lack of differentiation is seen in Oscan dat. altrei and loc. alttrei, both with the ending -ei.

⁵⁶ Cf. C. D. Buck Greek Dialects 81.

⁵⁷ So H. Jacobsohn Philologus 67.524-7; cf. W. Schulze Zur Geschichte lateinischer Eigennamen 456; Sommer's doubts, Hdb.² 117, 100-1, Krit. Erl. 28, are not justified.

⁵⁸ So F. Muller Altitalisches Wörterbuch, s. vv.; but A. Walde Lateinische-Etymologisches Wörterbuch,² s. vv., admits only a transferred -igō for castīgō and fatīgō.

⁵⁹ A difference between the development of postconsonantal -iyo- and that of -yo- in Latin, is hardly thinkable, despite Sommer Hdb.² 101 init.

⁶⁰ The nominatives abiēs ariēs pariēs may be taken as ending in -iess for -iets, the final syllable being closed and therefore long by position, but the vowel being short. If this be correct, then it is likely that ie did not contract when followed by two consonants, i.e., when the e was in a closed syllable. The pre-form of mulier was of course *muliēs, with a long vowel.

⁶¹ F. Skutsch de Nominibus latinis suffixo -no- ope formatis 14³ f. (not accessible to me; cited by Sommer, Krit. Erl. 28), Archiv f. lat. Lexikographie 12.206.

⁶² Or because the e stands in a closed syllable, cf. ftn. 60.

⁶³ If so correctly spelled; the testimonia are: ambegni bos et verbix appellabantur, cum ad eorum utraque latera agni in sacrificium ducebantur, Fest. ap. Paul. 4.7-8 ThdP = 4.16 M (ubi vide app. crit.); ambiegna bos apud augures, quam circum aliae hostiae constituuntur, Varro de Lingua Latina 7.31.

⁶⁴ CIL 1².583. The normal development of jaciō in compounds was to -iciō, cf. Transactions of the American Philological Association 43.38-40; also Sommer Hdb.² 99, 133-5, 486.

⁶⁵ Idg. Gram. 5.76.

⁶⁶ So F. Solmsen, IF 4.240-52, esp. 250, with references; against this view, Sommer Hdb.² 591. The Skt. infinitive of this type is found in 19 roots, uncompounded, and all used in Vedic texts, according to Whitney, op. cit.

⁶⁷ So E. W. Hopkins, ‘The Vedic Dative Reconsidered’, in TAPA 37.87-120.

⁶⁸ Idg. Gram. 3.50.

⁶⁹ F. Solmsen, 1.c.

⁷⁰ Sommer Hdb.² 591; Stolz-Leumann 328.

⁷¹ G. M. Lane, A Latin Grammar for Schools and Colleges, revised edition by M. H. Morgan, gives some examples, §1331, §§1344-5; thus Plaut. Cas. 110 rūre ‘in the country’.

⁷² So Sommer Hdb.² 340, after F. Skutsch Jahrbücher f. klass. Philologie, Suppl. 27.97 ff.

⁷³ I am not denying the short syllables and short vowels resulting from iambic shortening, which however is not the phenomenon maintained by Hirt.

⁷⁴ Idg. Gram. 3.50-1, 5.75-6. Meillet-Vendryes 333-4 by implication deny the development of (unaccented) -ai to -ӗ in Latin.

⁷⁵ Idg. Gram, 5.72-4; cf. Lang. 5.263.