aspirates were merged into the mediæ themselves, so that b and bh appear in Celtic as b, d and dh as d, g and gh as g, and ꬶ and ꬶh as g. The Balto-Slavonic, in this matter, shares the peculiarity of the Celtic.
All the explosives, when intervocalic, are “aspirated” in Gaelic—p to ph, b to bh (= v), t to th (= h), d to dh (= y), c to ch, g to gh, (= y); the corresponding Welsh changes are the tenues to mediæ, and the mediæ to f, dd, and nil in the case of g. Intervocalic preserved explosives in Gaelic arise from a doubling of the explosive, the cause of which in many cases is obscure. The following are the leading cases and causes of intervocalic G. mutes:
(1) Doubling of the explosive in the course of inflection or word-building.
- a. Inflection. The participle passive in ‑te preserves the t or d of the root as t; thus [caith gives caithte,] bàth (for bàdh) gives bàite, ràdh gives ràite, etc.
- b. Word-building. The prepositional prefixes which end or ended in a consonant preserve the succeeding explosive; even vowel-ending prepositions like air (*are), aith- (*ati) do the same, if the accent is on the preposition. Thus—abair is for ad-ber, aitreabh is for ad-treb, aidich is for ad-dam, faic for ad-ces‑, agair for ad-gar. In the way of affixes, we have ruiteach from rud‑t and ruicean from rud‑c, creid from *cred-dhô; compare the compounds boicionn, laoicionn, and craicionn.
(2) After sunk n or m. Thus deud comes from dṇt, and so with ceud, teud; ceud, first, from *cento‑, so seud; eug from ṇko‑, etc.
(3) After sunk spirant z. This is assured for zd, as in brod (*broz-do‑, Norse broddr), cead, gad, maide, nead; but zg giving g is doubtful—eagal seems for *es-gal or *ex-gal‑, beag for ꬶvezgo‑s (Lat. vescus), mèag for mezgo‑.
(4) Cases corresponding to double explosives in other languages: cat and Lat. catta (borrowing?), cac and Gr. κάκκη. Compare also slug.
(5) Doubtful cases. Many of these cases can be satisfactorily explained as due to suffixes immediately affixed to consonant-ending roots. Thus brat may be for brat-to‑, trod for trud-do‑, ìoc for *yak-ko‑, breac for mṛg-ko‑. Even suffixes in ‑bho- and ‑ꬶo- (Eng. k in walk) are not unknown, and they might account for reub (*reib-bo‑, *reib-bho‑, Eng. reap, rip), slug for slug-go‑, etc. Dr Whitley Stokes has given a different theory founded on the analogy of a Teutonic phonetical law, stated
