(3) The vowel of the stem, if mutable, is of course mutated when the ending is dropped in the 3rd sg.; thus cỿrchaf ‘I make for’, cɥrch; dỿgaf ‘I bring’, dwg ‘brings’; ceisiaf ‘I seek’, Mn. W. cais ‘seeks’, Ml. keis, § 81 iii (1). In many cases ‑o- is a mutation of ‑aw-, the latter appearing in the 3rd sg., thus toddaf ‘I melt’, tawdd ‘melts’; boddaf ‘I drown, or am drowned’, bawdd ‘drowns’; holaf ‘I ask’, hawl ‘asks’. But in disyllabic stems, when the unaccented ‑aw- in the ult. became ‑o-, it was in some cases treated on the analogy of original ‑o- and affected to y; thus adroẟaf ‘I narrate’, adrawẟ b.a. 1 ‘narrates’, later edryẟ r.p. 1253; halogaf ‘I defile’ (denominative from halawc, Mn. W. halog), helyc Ỻ.A. 34 ‘defiles’. In the 16th cent. dichyn was used, § 196 ii, but was supplanted later by the original form dichon ‘can’, Ml. W. dichawn. The substitution of ‑ɥ as in tery c.m. 32 for the affected ‑eu of tereu, etc., see (1), is due to the same analogy; see § 83 ii.
A similar analogy gave rise to gwerchyd D.G. 175 ‘guards’ from the v.n. gwarchod for gwarchawd metath. for gwar-chadw̯; the old 3rd sg. was gwer-cheidw̯.
v. (1) Verbs with stems in ‑ha- had the vowel unaffected in the 3rd sg.; the affected forms bwyty ‘eats’, pery D.G. 441 ‘lasts’ are late; the original 3rd sg. of bwytá-af ‘I eat’ is bwyta w.m. 456 ‘eats’, ef a vwytta Ỻ.A. 170, and of parhá-af is pár(h)a, as parha b.t. 40 ‘lasts’, ny phara r.p. 1046, w.m. 86 ‘does not last’, ni phara T.A. g. 236. The accent falls regularly on the penult; and the ‑h- after it was lost, § 48 ii, as in pára, but not before changing a media to a tenuis as in bẃyta ‘eats’ (: bwyd ‘food’).
(2) The ‑(h)a of the 3rd sg. is thus the unaffected stem-forming suffix, but it came to be mistaken for a personal ending; and as ‑ha- forms the stems of denominatives, ‑(h)a seemed to be a 3rd sg. ending of denominatives, and was used to form the 3rd sg. of denominatives generally. This may have originated in doublets like neshá-af, v.n. neshá-u and nés-af, v.n. nésu ‘to approach’ (: nes ‘nearer’); the 3rd sg. of the first is regularly nés-(h)a, which, being very naturally taken to be the 3rd sg. of the second, suggested a 3rd sg. ending ‑(h)a. For exactly the same reasons it became a 2nd sg. imperative ending, and is used as such in all verbs in which it appears in the 3rd sg. pres. ind. In older examples the form is ‑ha, the ‑h- hardening a media or remaining as an ‑h‑; such examples survive in Ml. W. side by side with others in which the ending has come to be regarded as ‑a simply. Thus we find gwata r.p. 1382 ‘denies’, oetta impv. r.p. 1254 ‘delay’, gwatta m.a. i 319a ‘denies’, ehetta do. 319b ‘flies’, tremycca Ỻ.A. 150 ‘despises’, poenha do. 28 ‘punishes’, dielwha do. 147 ‘ruins’ (makes worthless), gwyhwa do. 148 ‘withers’, cerẟha do. 168 ‘goes’, gweẟha do. 165 ‘beseems’, side by side with gwada r.p. 1256 ‘denies’, oeda impv. do. 1285 ‘stay’, llettya do. 1254 ‘lodges’, ogana ib. ‘satirizes’, a gylchyna m.a. i 319b ‘surrounds’, a boena Ỻ.A. 147, kerdda do. 165, kerẟa do. 167, gw̯eẟa r.p. 1272. In the last examples simple ‑a has become a 3rd sg. ending.
