(3) The change of i̯ to *ẟ in the above cases took place before the Roman period, for there is no example of it in any word borrowed from Lat. The alteration was therefore earlier than the period of vowel affection, and the *ẟ could not affect; hence arẟaf, not *eirẟaf, etc.
The fact that the change does not take place initially corroborates the view that it did not happen before an accented vowel. All forms that occur can be explained under this supposition; thus all- < *áli̯o‑, but ail ‘second’ < *ali̯ós, etc.; see § 165 vi.
iv. Ar. ‑mi̯- became ‑ni̯- in Pr. Kelt.; as W. dyn ‘man’, Ir. duine < *g̑hðomi̯o‑, § 98 i (3), § 121 i;—W. myned, ‘to go’, Ml. Bret. monet, Corn. mones < *momi̯- for *mami̯- § 65 v (2), by assim. for *bam‑i̯- < *ɡu̯ₑm‑i̯‑, √ɡu̯em‑: Lat. venio, Gk. βαίνω both < *ɡu̯ₑmi̯ō, Goth. qiman, E. come. The ‑i- disappeared before the ‑e- of the suffix; the suffix may have been ‑at‑, § 203 ii, which following the accent would become ‑et- after i̯, see § 65 vi (1). The i̯ was lost in the compounds an-fon, dan-fon ‘to accompany, send’, prefix § 156 ii (1).
v. In some cases metathesis of i̥ took place in Brit. Thus Ir. suide ‘soot’ comes from *sodi̯o‑, but W. hudd- in huddygl ‘soot’ implies *soi̯d‑; O. E. sōt, Lith. sů́džiai ‘soot’ have L°-grade; so W. suddaf ‘I sink’ < *soi̯d- < *sodi̯- beside W. soddaf ‘I sink’, sawdd ‘subsidence’ < *sōd‑, √sed- § 63 ii.—W. drum ‘ridge’ < *droimm- < *drommi̯- < *dros‑mi‑: Ir. druimm < *drommi- (i-stem): Lat. dorsum < *dr̥s-so‑m, Gk. δειράς < *ders-ad‑, Skr. dr̥ṣ-ád ‘rock, millstone’, √deres‑;—W. turi̯o ‘to delve’ < *toirg- < *torgi̯‑: Lat. porca § 101 iii (1);—W. ar-o-fun ‘intend’, dam-(f)un-aw, dym-un-o ‘desire’, with ‑fun- < *moin- < *moni̯‑: Lat. moneo, √menēi̯‑, extension of √men- ‘mind’;—W. ulw ‘ashes, powder’ < *oi̯lu- < *polu̯i̯‑: Lat. pulvis < *polu̯is;—W. Urien, O. W. Urb-gen § 25 i < *oirbo-gen- < *orbi̯o‑: Gaul. Orbius ‘heir’, Lat. orbus, Gk. ὀρφανός;—W. wyneb ‘face’, in comp. wynab- r.m. 30 < *einep‑, *einap‑ < *eni̯-əqu̯- (§ 65 vi (1)): Skr. ánīkam ‘face’ < *eni-əqu̯-, √ōqu̯‑; the un-metathesized form is seen in O. W. einepp, where ein- is from *en(i̯)- § 70 v, since old ei had then become ui ≡ Mn. wy; O. W. enep, Corn. eneb Bret. enep, Ir. enech show i̯ lost, which occurs before e in Brit., see vi below, and cf. § 35 ii (2), and is usual in Ir., cf. i above;—W. wybr, wybren ‘cloud’ Ỻ.A. 104, 91, ‘sky’, O. Corn. huibren gl. nubes < *eibbr- < *embhri- § 99 vi (1): Lat. imber gen. imbris (i-stem) < *embhri-
