W. maidd ‘whey’ < *meᵹ̑ẟ‑, met. for *meẟᵹ‑, Ir. medg ‘whey’, Gallo-Lat. mesga (s for ẟ? cf. § 96 ii (1)): Lat. mergo, Lith. mazgóti ‘to wash’, Skr. majjati ‘sinks’ < *mezɡ‑;—W. haidd ‘barley’ < *se‑zg‑, redupl. of *seg‑: Lat. seges;—perhaps W. twddf ‘a swelling’ for *tuẟᵹ- < *tuzg‑, s-stem of √teu̯ā- (: Goth. þūs‑) + ‑g- suff.: Lat. turgeo (Walde² rejects his first suggestion that this is from *tuzg- in favour of Solmsen’s *tūrigo, IF. xxvi 112ff., with ‑igo (: ago), though this is usually 1st conj., as navigāre).
W. gwẟf ‘throat’, N. W. dial. gwẟw, pl. gyẟfe, gyẟfa, S. W. dial. gwẟwg, pl. gyẟge, gythce, Bret. gouzoug, with ‑g for ‑ᵹ, § 111 vii (4), seems to require *guzg‑; ? g̑hu‑s‑, √g̑hēu‑, (: Lat. fauces) + ‑g‑, as in mwn‑g ‘mane’.
iv. Ar. ‑zb(h)- > Kelt. ẟb > Ir. db, W. ẟf. Thus W. oddf ‘knag, knot, nodule’, Ir. odb: Gk. ὀσφύς (< *ost-bhu‑?).
v. (1) The above groups are found only medially. Initially Ar. s- did not become z‑, but changed a following media to a tenuis ; thus sb- > sp‑, *sbh- > sph, etc., Siebs, KZ. xxxvii 277 fF. Hence the initial alternations b‑: sp- and dh‑: sth‑, etc., as in Germ. dumm, E. dumb < *dh‑: Germ. stumm, W. di-staw < sth‑, § 156 i (11).
(2) As s- could be prefixed or dropped in Ar. and for a long time after the dispersion, § 101 ii (1), Siebs l. c. holds that the above explains the initial alternation of a media and tenuis. In a large number of cases it undoubtedly does so. Where the media is general and the tenuis exceptional, it affords a satisfactory explanation, as in the case of the Kelt. t- in tafod ‘tongue’ corresponding to d- elsewhere (O. Lat. dingua), which is parallel to the t in taw! ‘be silent’ (s still kept in di-staw) corresponding to the *dh- which gives the d- of E. dumb. But it hardly explains the alternation when the tenuis is general and the media exceptional, as in W. craidd, Lat. cord‑, Lith. szirdìs, E. heart, Gk. καρδία < *k̑‑: Skr. hŕ̥d‑, Av. zərədā < *g̑h‑, since k̑ < sk̑h, without a trace of the s- in the whole of Europe, is improbable. But whatever the explanation may be, the fact of the alternation can hardly be called in question.
(3) As an example of the variety of forms produced by variable s‑, we may take √bhu̯erē‑, extd. *bhu̯erē̆‑ɡ‑/‑ɡh‑/‑q‑, orig. meaning 1. ‘hurl’, 2. ‘smite’; hence from 1. ‘sprinkle, cast (seed); roar, snore; rattle ; talk’; from 2. ‘break; crash, break out, burst; smell’. bh‑: W. bwrw ‘hurl, smite’, bwrw glaw ‘to rain’, bwrw had ‘to cast seed’ < *bhur’ɡ- (ur < u̯ₑr); Lat. frango < *bhrənɡ‑, frāgor < *bhr̥̄ɡ‑, frā-L
1402
