ŏi for o-ai not final, as in trŏis for tró-ais. On account of the long vowel gwnâi, trôi, etc. are generally sounded and often written gwnae, troe, etc.; but in the bards ‑âi rhymes with ai, see wnâi / ehedai g. 242. Both forms are seen in Ml. W. gwnai W.M. 25, 54, gwnay r.m. 237 (ae=ay, § 29 ii (1)).
(4) The vowel is long in o’i, a’i, da i̯, etc., § 33 v, of course only when accented. In Ml. W. o’i, a’i are written oe, ae or oy, ay.
§ 53. When the accent in a polysyllable falls on the ultima, the above rules apply as if the ultima were a monosyllable; thus, short, pahắm ‘why?’, penắig, § 41 iii (2), parhắu ‘to continue’, gw͡yrdrŏ́i ‘to distort’; long, Cymrā́eg, parhā́nt (for parhá-ant), gw͡yrdrṓi (for gw͡yrdro-ai) ‘he distorted’, penllā́d ‘summum bonum’.
In parhau, caniatau, etc., some recent writers circumflex the a, possibly a practice first intended to indicate the long vowel in the uncontracted form ‑ha-u, § 54 iii. When contracted the a is short. In D.D. and Bible (1620) it is not circumflexed. J.D.R. 144 writes cadarnháu. But see § 55 ii.
§ 54. In the accented penult—
i. (1) The vowel is short, if followed by two or more consonants, or by p, t, c, m, ng, ll, s; as hărddwch ‘beauty’, plĕntyn ‘child’, cănnoedd ‘hundreds’, by̆rrach ‘shorter’, ĕstron ‘stranger’, ĕpil ‘progeny’, ăteb ‘answer’, ămeu, ‘to doubt’, ăngen ‘need’, ăllan ‘out’, Iĕsu ‘Jesus’, glăndeg ‘fair’, glănw̯aith ‘cleanly’, tăni̯o ‘to fire’, ty̆bi̯af ‘I suppose’. There is no exception to this rule, though before m the vowel is sometimes wrongly lengthened in words learnt from books, such as trămor ‘foreign’, ămwys ‘ambiguous’.
Silvan Evans marks many obsolete words, such as amwg, amug with long ā, for which there is no evidence whatever; it merely represents his own misreading of Ml. W. ‑m‑, which always stands for ‑mm‑.
(2) The consonants above named are each double in origin. In Ml. W. t, c, s were usually doubled in this position, as atteb, racco or racko, messur; but ‑m- is generally written single, owing to the clumsiness of ‑mm- and its frequency; possibly ‑p‑, which is not very common, followed the analogy of ‑m‑; ll and ng being digraphs can hardly be doubled in writing. In early Bibles m and p are doubled; and G.R. wrote gaḷḷu, doubling ḷ (his ḷ = ỻ). As however each is etymologically double (except in borrowed words), the double origin
