speech; v. Indogerm. Forschungen, 4, 246; Wharton's Etyma Latina recognise the three rows c, k, q; cf. Zupitza's treatment of the gutturals. In Gadelic the velar and the palatal series have fallen together, but there is a distinct treatment of the labio-velar.
21. Contamination may have been at work here. But although the Cymric cognate is daigr, and Old Latin shows dacruma, O. H. German, zahar, O. Icelandic, tár, Germ., zähre, in view of the Gadelic forms, we may take the pre-historic form to have been *dṇkru, which developed on the Brythonic side into a proto-Celtic *dakru. Compare Dr Walde's Lateinisches Etymolvgisches Wörterbuch, p. 319, also p. 5, where L. acer is given as cognate with Irish Gadelic ér, high.
22. méith should be mèith, as in the Dictionary, with long open è; this is diphthongized in the Northern dialect as mīath—a case of diphthongization of long open è where there has been no compensatory lengthening.
23. See Zeitschrift für Celtische Philologie, Band 3, 264, 275, 591.
24. See Zupitza on i, j in Celtic, in Zeitschrift für Celtische Philologie, 2, 189-192.
25. See Foy in Indogerm. Forschungen, 6, 337, on Celtic ar, al = Indogerm. ṝ, ḹ; and Zupitza on ṛ, ḷ in Celtic, in Kuhn's Zeitschrift, 35, 253.
