rally as h‑, sometimes as s‑. Examples: Ir. samail ‘likeness’, W. hafal ‘like’ < *sₑmₑl‑: Lat. similis, Gk. ὁμαλός, √sem- ‘one’.—Ir. sam, W. haf ‘summer’: O. H. G. sumar, E. summer, Skr. sámā ‘year’.—Ir. sen, W. hên ‘old’: Lat. senex, Gk. ἕνος, Skr. sána‑ḥ ‘old’, Lith. sẽnas ‘old’.—W. had: Lat. satus § 63 vi (1).—W. hun ‘sleep’: Lat. somnus, Gk. ὕπνος, § 63 viii (1).—W. hynt ‘way’, Ir. sēt, § 65 iii.—W. hîr ‘long’, Ir. sīr: Lat. sērus, § 72.— — Ir. secht n‑, W. saith ‘seven’: Lat. septem, Gk. ἑπτά, etc. < Ar. *septm̥ § 86 ii (1).—W. sīl ‘progeny, seed’, beside hīl < *sē‑l‑, √sē- § 63 vi (1).—W. serr, Ir. serr, § 86 i (5).—W. saer, Ir. sāer < *sapero‑?: Lat. sapio.—W. sugnaf, Ir. sūgim ‘I suck’: Lat. sūcus, sūgo, O. E. sūgan, sūcan ‘suck’.
ii. Medially between vowels Ar. s remained after the separation of the P and Q divisions; and is found in Gaulish, as in Isarno‑. In Ir. and W. it became h, and generally disappeared, except where it became initial by metathesis, as in W. haearn, though it is in some cases still written in Ml. W.; thus W. eog, Ml. W. ehawc, Ir. eo, gen. iach < Kelt. *esāk- < *esōk‑, Lat. esox < Kelt. The reduction of vowel-flanked s gave rise to new diphthongs in Brit., which developed largely like original diphthongs; see § 75 i, ii, vi, vii, § 76 ii (3).
iii. The change of s to h differs from the soft mutation; in the latter a voiceless consonant becomes voiced, thus t > d; the corresponding change of s would be to z. But s did not become voiced; it remained voiceless, but was pronounced loosely, and ultimately became h. It must have been loosened already in the Roman period, for Lat. intervocalic s introduced at that period remains, as in caws < cāseus. Now Lat. explosives undergo the soft mutation; the loosening of Brit. s is therefore earlier, and so the interchange s/h does not enter into that system. Before such a system of interchanges was organized it was natural to choose one or the other sound for the same word; and the postvocalic reduced s was chosen for most in Brit., the postconsonantal full s for others. It is quite possible that the two forms persisted in many words for a considerable period, so that we have e.g. W. Hafren beside Brit. (‑Lat.) Sabrĭna. There is only one certain example of Lat. initial s- giving h‑; that is hestawr < sextarius; this either was a trade term borrowed early, or has followed the analogy of words like Hafren. Possibly a transition stage is represented by Ixarninus, Isxarninus beside Isarninus Rhys LWPh.² 418. (The Ir. reduction of s is independent, and is included in the Ir. system of initial mutation.)
