¹ Abbreviations: TH¹ = Rudolf Thurneysen, Handbuch des Alt-Irischen, 1. Teil: Grammatik; Heidelberg, 1909. TH² = the same, translated and edited by Osborn Bergin and D. A. Binchy (in preparation). VG = Holger Pedersen, Vergleichende Grammatik der keltischen Sprachen, 2 vols.; Göttingen, 1908–11. LP = Lewis and Pedersen, A Concise Comparative Celtic Grammar; Göttingen, 1937. A figure after one of these abbreviations refers, unless otherwise indicated, to a numbered section and subsection.

² In Brittonic, initial s remains in a very limited number of words (Welsh sil ‘spawn, fry’ as against hil, epil ‘posterity, offspring’, Ir. síl ‘seed’; Welsh saith, Corn. syyth, Breton seiz ‘seven’, Ir. secht). See VG 34.

In adopting the form Brittonic instead of the older Brythonic, I follow the usage of T. F. O'Rahilly, Early Irish History and Mythology 434 n. 2 (Dublin, 1946).

³ In Gaulish names and words the -os always appears as -os or in its Latinized form -us; in the ogham inscriptions it is usually written -as, less commonly -a (TH¹ 86, TH² 90.4). In this paper I shall occasionally represent a final aspiration subject to sandhi laws by an h raised above the line.

⁴ ‘Aus -esa erklärt sich ungezwungen auch c. -oedd.‘

⁵ According to Bergin, also Skt. rahas ‘lonely place’ etc.; see Êriu 12.222.

⁶ In Cornish and Breton, the corresponding forms are os and out; it is generally impossible to distinguish between Brittonic oi and ui in these languages. When, however, wy- appears for expected oe- in this case, it is rather to be considered as due to initial lengthening of e-, as in wyneb (as against Cornish and Breton enep, eneb; cf. VG I, p. 38), than to the following short -i. One thus finds *ēi- (< *ēs- < *es-), as in the case of hwy (see under the following heading, below).

⁷ The final -dd in oedd (cf. Corn. o, Breton oa) and the suffix -oedd (not reflected in Cornish and Breton) is peculiar to Welsh, and probably dialectal or of later origin; it may represent a semivowel (j) once produced between the i and the following vowel.

⁸ The last three forms are difficult to explain, but they evidently represent cases in which -s for some unknown reason (originally before a consonant, in sandhi?) did not produce a diphthong but remained as -h, eventually reinforced to -ch (dropped in Cornish and written -c'h in Breton).

⁹ For the various forms of the name Owein, see VG I 73 and Baudiš, Grammar of Early Welsh, Part 1, p. 46). It is important to note that one may here have an alternative compound with *iwo- ‘yew.‘

¹⁰ Cf. also Marstrander, Une correspondence germano-celtique 23 (Skrifter utgit av Videnskapsselskapet i Kristiania, 1924), where an identical theory, founded on other evidence, is advanced.

¹¹ Intervocalic -es-, as commonly supposed (cf. Pokorny, Altirische Grammatik, [Samml Göschen, No. 46], and LP 24.3), develops into -i- in Old Irish, as stated above. However' although Thurneysen maintains (TH¹ 74) that OIr. nime is derived from Celtic *nemesos and that the dat. sg. nim has its vowel from the gen. sg., I am inclined to believe in the opposite process. According to my view, OIr. nime is not from Celtic *nemesos (which would have become *naime, cf. taige), but has its vowel from nim, the regular evolution of *nemes. I think such a process is quite as likely; for the dative (innim ‘in heaven‘) must have been more frequent than the gen. sg. The tendency to make the gen. sg. conform to the dat. still prevails in later Irish; cf. OIr. lám(a)e from lám, dat. láim, as against Mod. Ir. láimhe, Scottish Gaelic làimhe, etc.

¹² All examples given by Thurneysen (TH¹ 76) of the change of a to u after a labial consonant show the influence of -i (not -e) in the following syllable.

¹³ Some of these, moreover, were originally neuters: OWelsh cam, pl. cemmein ‘step’; anu, pl. enuein ‘name’; dant, pl. deint ‘tooth’ (cf. Corn. dans, pl. dyns; Breton dant, pl. dent). According to Pedersen's explanation (VG II, p. 113) they are formed by analogy with the o-stems.

¹⁴ Speaking of IE *-ejes: ‘possibly also in Ir. -i’.

¹⁵ Cernyw is the only example quoted by Baudiš, p. 47, and is derived by him from a Celtic *kornowī (cf. VG II, p. 16). The most common of the later Latinized forms, however, is Cornubia, evidently for Cornuuia (Holder, s.v. Cornovii). A Celtic *kornuwī would unquestionably lead to Welsh Cernyw, Breton Kerneo, etc. The stem Cornuu-, instead of Cornou-, is further supported by the derivative Cernyweg ‘Cornish language,’ which cannot easily be explained from *kornow-; cf. Welsh cneuen, Breton knaouen ‘nut’, Welsh cnewyll ‘kernels’ (< Celtic *know-i-, VG II, p. 100); Welsh newydd, Breton nevez ‘new’ (< Celtic *now-i-); for other similar examples, see VG I, p. 61.

¹⁶ ‘If we assume that -es became -is‘ (TH² 560).

¹⁷ E.g. do eim ‘thou protectest’ (Ml 110 d 9); see TH² 78.1.

¹⁸ The vocalism of the latter form may be based on the conjunct 2d sg. pres. ind. bir; the combination -ert is not affected by a following -i, cf. neirt, gen. sg. of nert. The independent pres. ind. biri (for which further see Bergin, Ériu 12.135), may, however, be original (*beresi > *berihi > biri).

¹⁹ No distinction is made in Early Welsh between absolute and conjunt forms in the 2d person sg.

²⁰ Pedersen assumes (LP 449, p. 283) that IE *eje-verbs (cf. Skt. -aya-ti) survive in this suffix. Should this be correct, we still have an example of the narrowing of e to y, which, according to Pedersen, is due to an IE form *-ej-ei.

²¹ The i- affection of the root vowel is a characteristic of the Brittonic verb. It took place regularly in the 3d sg. of the ī-verbs, which ended in *-īt in early Celtic (> OIr. -i, cf. Latin fīnit); e.g. Welsh dyd from dod ‘put, place’, gesyd from gosod ‘put, set’, Middle Welsh treing from trengi ‘die’ (cf. Ir. trécimm 'abandon, give up, fail, fade'). The same change took place also in some thematic verbs, such as gwŷl (cf. OIr. feil, unstressed fil) from gweled ‘see’, ettyb (cf. OIr. in coissig) from atteb ‘answer’, etc.; see Morris-Jones, A Welsh Grammar, 173.4. This fact might be explained by analogy. As in early Celtic the 2d and 3d persons sg. of the ī-verbs (originally ending in *-īs and *-īt respectively) became identical (in -ī), so the early Celtic -ei^h (from *-es), being the 2d sg. of the thematic verbs, might similarly have attracted the 3d sg. (originally in *-et) in the Brittonic languages, with the result that after the 2d person had assumed a distinct termination (cf. above), the i-affection actually appears in the 3d sg. of many verbs. Cf. a similar evolution in Early English: bears from OE bires, originally the 2d sg., and bearest from OE birest, a later formation (cf. Luick, Hist. Gramm. der Engl. Sprache, 698; Sievers–Brunner, Altenglische Grammatik 357, Anm. 1).

²² In this connection I take the opportunity of revising and completing my observations on this word in Études Celtiques 3.83.

²³ By assuming a dissyllabic form in Celtic, the vowel alternation represented in tig, taig, tige, taige might be partly explained; for according to what I have tried to prove elsewhere (Études Celtiques 3.71 ff.), the a would be due to an original trisyllabic form stressed on the penult (*tagésos < *tegésos).

²⁴ An additional example might be quoted here, which was omitted in my paper in Études Celtiques: duliu, dulem (same vocalism as in lugu < laugu), comp. and superl. of dil ‘dear’ (TH¹ 49); cf. Welsh del 'pretty, neat¹, OE til 'good, valiant', tela ‘well’.

²⁵ They are essentially the same, the pronunciation generally being gah in either case.

²⁶ Cf. OE gár, ON geirr (which are possibly borrowed from Celtic), as well as Greek χα ος, χα ον ‘shepherd's staff’, sometimes identified with the Celtic word (Walde–Pokorny 1.528). Every fact points to an original o-stem in Celtic.

²⁷ I am indebted to Dr. Osborn Bergin for the latter reference, as well as in general for references to TH².

²⁸ With this may be compared OIr. bói, bái ‘was’ (< *bowe, *bawe?), with original vocalic termination.

²⁹ The Irish glé might quite as well be an original *gleiso-, but the Welsh form would demand internal *-es-.

³⁰ According to Holthausen, this is, however, a Frisian form and could thus have Germanic a.

³¹ It is fairly certain that neither of the words in which this w appears is an u-stem.

³² Cf. Sievers-Brunner, Altenglische Grammatik 122.2, 126.2.

³³ Surviving, however, in the sandhi h mentioned above (cf. TH¹ 241, p. 150; LP 201).

³⁴ OIr. fo dí ‘twice’ might be based on IE *dwis (Latin bis, Greek δίς), though according to what has been stated above the vowel should rather have been short.