The proper names occurring in classical writers, on inscriptions and coins, have been collected by A. Holder in his monumental Altceltischer Sprachschatz (Leipzig, 1896-1908). The inscriptions have been most recently treated by J. Rhys in the Proceedings of the British Academy, vol. ii. See also a paper in this volume entitled “Celtae and Galli” by the same author for the text of the Coligny and Rom inscriptions. The value of Gaulish for grammatical purposes is set forth by Whitley Stokes in a paper on “Celtic Declension” in the Proceedings of the London Philological Society (1885-1886). For the extent over which Gaulish was spoken, its relation to Latin and its influence on Romance, see E. Windisch’s article on “Keltische Sprache” in the section “Die vorromanischen Volkssprachen” in Gröber’s Grundriss der romanischen Philologie², vol. i. pp. 373 ff. Cf. further the introduction to J. Loth’s Chrestomathie bretonne (Paris, 1890); G. Dottin, Manuel pour servir à I’étude des antiquités celtiques (Paris, 1906); R. Thurneysen, Keltoromanisches (Halle, 1884).
II. Goidelic and Brythonic.—When the monuments of the Celtic dialects of the British Islands begin to appear, we find a wide divergence between the two groups. We can only mention some of the more important cases here. The Brythonic dialects have gone very much farther in giving up inflectional endings than Goidelic. In Irish all final syllables in general disappear except long vowels followed by s or r and u < ō preceded by i. But these reservations do not hold good for Brythonic. Thus, whilst O. Irish possesses five cases the Brythonic dialects have only one, and they have further lost the neuter gender and the dual number in substantives. In phonology there are also very striking differences, apart from the treatment of the labialized velar qv already mentioned. The sonant n appears in Brythonic as an, whereas in Goidelic the nasal disappears before k, t with compensatory lengthening of the vowel, e.g. I. E. *kmtom, Ir. cét, “hundred,” W. cant, Bret. kant; Prim. Celt. *jovṇko-, O. Ir. óac, Mod. Ir. óg, “young,” W. ieuanc, Bret, iaouank. t, k standing after a vowel and preceding l, n (and also r if k precede) disappear in Goidelic with compensatory lengthening of the vowel, e.g. Prim. Celt. *stātlā-, Ir. sál, “heel,” W. sawdl; Prim. Celt. *petno-, Ir. én, “bird,” O. W. etn, Mod. W. edn. Similarly b, d, g disappear in Goidelic when standing after a vowel and preceding l, r, n with compensatory lengthening of the vowel, but in Welsh they produce a vowel forming a diphthong with the preceding vowel, e.g. Prim. Celt. *neblo-, Ir. nél, “cloud,” W. niwl; Prim. Celt. *ogno-, cf. Lat. agnus, Ir. uan, “lamb,” from *ōn, W. oen; Prim. Celt. *vegno-, cf. Ger. Wagen, Ir. fén, “wagon,” O. W. guein, Mod. W. gwain. The Goidelic dialects have preserved the vowels of accented syllables on the whole better than Brythonic. Thus Brythonic has changed Prim. Celt, ā (= I. E. ā, ō) to ō (W. aw, Bret. eu); and Prim. Celt. ū to ī, e.g. Ir. bráthir, “brother,” W. brawd, Bret. breur; Gaulish dūnum, Ir. dún, “fort,” W. din. Already in Gaulish the I. E. diphthongs show a tendency to become simple long vowels and the latter are treated differently by Goidelic and Brythonic. In early times I. E. eu, ou both became ō and I. E. ei gave ē. In Goidelic ō, ē, in accented syllables were diphthongized in the early part of the 8th century to ua, ia if the next syllable did not contain the vowels e or i, whereas in Brythonic ō gave ü (written u) and ē became in W. ui (wy), and in Bret. oe (oue), e.g. Gaulish Teuto-, Toutius, Ir. tuath, “people,” W., Bret. tud; Brythonic Lēto-cētum, Ir. tuath, “grey,” W. llwyd, Bret, loued. Similarly in loan-words, Ir. céir, fial, W. cwyr, O. Corn. guil, from Lat. cēra, vēlum. Further I. E. ai, oi are preserved in Irish as ai (ae), oi (oe), Mod. Ir. ao, but in Welsh I. E. ai gave either ai or oe, whilst oi changed to ü (written u), Ir. toeb, “side,” W., Bret. tu; I. E. *oinos, Ir. óen, “one,” W., Bret. un; Prim. Celt. *saitlo-, cf. Lat. saeculum, W. hoedl, “age,” Bret. hoal. In Goidelic accented e changes to i before i, u in the following syllable, cf. Ir. fid, “wood,” gen. sing, fedo, O. H. G. witu, and i changes to e before a or o under similar conditions. In like manner u becomes o before a or o, whilst o changes to u before i, u, cf. Ir. muir, “sea,” Prim. Celt. *mori, gen. sing. mora. Of Brythonic finals which disappear, ā, ī, (ō), j alone influence preceding vowels, whilst an i (y) which received the stress in O. W. was also able to modify vowels which went before it. In Goidelic the combinations sqv, sv appear respectively as sc, s (medially, f), but in Brythonic they both give chw; Prim. Celt. *sqvetlon, Ir. scél, “story,” W. chwedl; Prim. Celt. *svesor, Ir. siur, “sister,” but mo fiur, “my sister” (whence Scottish piuthar by false de-aspiration), W. chwaer, Bret. c’hoar. In Brythonic initial s becomes h in the 7th century, but this is unknown in Goidelic, e.g. Ir. salann, “salt,” W. halen, Cornish haloin, Bret, holenn; Lat. sé-men, Ir. sil, “seed,” W. hil. Initial v gives f in Goidelic in the course of the 7th century, whereas in Brythonic it appears as gu, gw, cf. Lat. vérus, Ir. fir, W., Bret. gwir. We may also mention that in Goidelic initial j and medial v disappear, e.g. Gaulish Jovincillus, W. ieuanc, “young,” Bret, iouank, Ir. óac, óc; W. bywyd, “food,” Ir. biad. Post-consonantic j in Brythonic sometimes gives -id (Mod. W. -ydd, Mod. Bret, -ez), e.g. Gaulish nevio-, novio-, O. Bret, nowid, W. newydd, Bret, nevez, Ir. núe. I.E. -kt and -pt both appear in Goidelic as -cht but in Brythonic as -ith, cf. Lat. septem, O. Ir. secht, W. seith, Bret. seiz.
We unfortunately know very little about the position of the stress in ancient Gaulish. According to Meyer-Lübke in place-names the penult was accented if the vowel was long, otherwise the stress lay on the preceding syllable, e.g. Augustodūnum, O. Fr. Ostedun, now Autun; Cataláunos (Châlons), Trícasses (Fr. Troyes), Bitûriges (Fr. Bourges). In Goidelic the stress, which is strongly expiratory, is always placed on the first syllable except in certain cases in verbs compounded with prepositional prefixes. In Old Welsh and Old Breton, on the other hand, the final syllable, i.e. the primitive penult, received the stress, but in both languages the stress was shifted in the middle period to the penultimate. The Goidelic dialects, like the Slavonic, distinguish between palatalized and nonpalatalized consonants, according as the consonant was originally followed by a front (e, i) or back vowel (a, o, u), a phenomenon which is entirely unknown to Brythonic.
Finally, the two groups differ radically in the matter of initial mutation or, as it is often called, aspiration. These mutations are by no means confined to initial consonants, as precisely the same changes have taken place under similar conditions in the interior of words. The Goidelic changes included under this head probably took place for the most part between the 5th and 7th centuries, whilst in Brythonic the process seems to have begun and continued later. It is easier to fix the date of the changes in Brythonic than in Goidelic, as a number of British names are preserved in lives of saints, and it is possible to draw conclusions from the shape that British place-names assumed in the mouths of the Anglo-Saxons. In Goidelic, we find two mutations, the vocalic and the nasal. Initial mutation only takes place between words which belong together syntactically, and which form one single stress-group, thus between article, numeral, possessive pronoun or preposition, and a following substantive; between a verbal prefix and the verb itself.
1. When the word causing mutation ended in a vowel we get the vocalic mutation, called by Irish grammarians aspiration. The sounds affected are the tenues k (c), t, p; the mediae g, d, b; the liquids and nasals m, n, r, l, s, and Prim. Celt. v (Ir. f, W. gw). At the present day the results of this mutation in Irish and Welsh may be tabulated as follows. Where the sound is at variance with the traditional orthography, the latter is given in brackets. In the case of n, r, l in Goidelic we get a different variety of n, r, l sound. In Welsh in the case of r, l, the absolute initial is a voiceless r, l written rh, ll, which on mutation become voiced and are written r, l. In Irish s becomes h written sh and the mutation of f is written fh, which, however, is now silent. Examples:—Irish, cú, “hound,” do chú, “thy hound”; Welsh ci, dy gi (do, dy represent a Prim. Celt. *tovo); Irish máthair, “mother,” an mháthair, “the mother,” Welsh mam, y fam (the feminine of the article was originally *sentā, sendā).
| Original sound | k | t | p | g | d | b | m |
| Irish | χ(ch) | h(th) | f(ph) | Ʒ(gh) | Ʒ(dh) | v,w(bh) | v,w(mh) |
| Welsh | g | d | b | nil | ð(dd) | v(f) | v(f) |
2. When the word causing mutation originally ended in a nasal, we get the nasal mutation called by Irish grammarians eclipse. The sounds affected are k (c), t, p; g, d, b; Prim. Celt. v (Ir. f, W. gw). In mod. Irish and mod. Welsh the results are tabulated below. Irish f becomes w written bh, whilst W. gw gives ngw. Examples:—Irish bliadhna, “year,” seacht m-bliadhna, “seven years,” cf. Latin septem, Welsh blynedd, saith mlynedd; Irish tir, “country,” i d-tir, “in a country,” Welsh tref, “town,” yn nhref, “in a town,” cf. Latin in.
| Original Sound | k | t | p | g | d | b |
| Irish | g | d | b | ng | n | m |
| Welsh | ngh | nh | mh | ng | n | m |
3. In Welsh k (c), t, p undergo a further change when the word causing mutation originally ended in s. There is nothing corresponding to this consonantal mutation in Goidelic. In this case k (c), t, p become the spirants χ (ch), th, f (ph), e.g. lad, “father,” ei thad, “her father,” ei represents a primitive *esiās. In the interior of words in Brythonic, cc, pp, tt give the same result as initial k, t, p by this mutation.
The relation in which the other Celtic dialects stand to this system will be mentioned below in dealing with the various languages. It will be noted from what has been said above that, with the exception of the different treatment of the labialized velar qv, and the nasal sonant ṇ, the features which differentiate the Brythonic from the Goidelic dialects first appear for the most part after the Romans had left Britain. At the beginning of the Christian era the difference between the two groups can only have been very slight. And Strachan has shown recently that Old Irish and Old Welsh agree in a very striking manner in the use of the verbal particle ro and in other syntactical peculiarities connected with the verb.