place in Brit. So in some cases the ‑i- of the spv., as in Ml. W. nessaf < *ned’samos § 148 i (1).
Disyllabic and compound prefixes are treated like the first element of a compound; thus Kelt. *ari- > Brit. *are- > W. ar‑; Brit. *kanta- > W. cannh- § 156 i (6), (7); *kom-(p)ro- loses its ‑o- and gives cyfr- as in cyfr-goll; so *u̯or-en-sed- loses its ‑e- and gives gorsedd ‘high seat’, as if from *u̯ore-ssed‑.
(3) The inscribed stones (5th to 7th cent.) do not throw much light on the above changes. The ogam inscriptions are Goidelic, and those written in Roman letters are in bad Latin, while many of the names even in the latter are Goidelic in form. In some cases a name has the Lat. nom. ending ‑us, as Catamanus Rhys no. 6 (LWPh.² 364), Aliortus no. 14, Veracius g, but most have the Lat. gen. ending ‑i, as Cunogusi hic jacit 5 ‘[the body] of C. lies here’. The names and the following maqu̯qu̯i of the ogams show that ‑i is gen., and not a debased form of the Brit. nom. ‑os. (The ogam ‑i is the Kelt. gen. suffix *‑ī, being the Italo-Kelt. gen. of ‑o- stems.) As a rule the Lat. fili agrees, but often does not, thus Dervaci filius Justi ic jacit 37. Fem. nouns end in ‑e, which is doubtless the ordinary late Lat. ‑e for ‑ae, though the noun in apposition stands in the nom., as Tunccetace uxsor Daari hic jacit 77, et uxor eius Caune 20. A nom. in ‑a appears in Avitoria filia Cunigni Eglwys Cymun insc. Possibly a Brit. nom. ending in ‑o for ‑os occurs in Aliortus Elmetiaco hie jacet Rhys 14 (the only stone with jacet) and Vitaliani Emereto 76. In a few cases no ending occurs: Etterni fili Victor 71, in which the legend is complete, and Victor is gen.; Velvor filia Broho 32. These and the false concords seem to indicate that the case endings were lost in the spoken language.
The stem-vowel u appears as ‑u- and ‑o‑, as Catu-rugi Rhys 60, Cato-tigirni 47; and ‑o- appears as ‑o‑, ‑u‑, ‑e‑, ‑i‑, as Cuno-gusi 5, Vendu-magli 45, Vinne-magli 21, Vende-setli 12, Venni-setli 67, pointing to ‑ə- for which the Roman alphabet has no symbol. The form ‑a- for ‑u- or ‑o‑, as in Cata-manus 6, is Goidelic; cf. in bilingual stones Cuno-tami in Roman characters, Cuna-tami in ogam 75; Trene-gussi in Roman, Trena-gusu in ogam 73. In some cases the stem-vowel was preserved, and forms containing it survive beside forms in which it is lost; thus Dumnagual beside Dumngual both in gen. v. That the former is not merely an archaic spelling of the latter is shown by the survival of both in the Mn. language:
- Mal mab i Ddyfnwal Moel-mū́d
- Yw Phylip braff i olud …
- Mae yn llaw hil Dyfnawal
- Yr erwi maior a’r aur mâl.—L.G.C. 209.
