^{page 297 note 1} Primitive forms which contradict ordinary grammatical analogies hold their ground best (against the levelling tendency) in words that are in hourly use; fert, est, it, are practically the only non-thematic verbs left in Latin, and in Greek it is the well-worn that have preserved the intricate inflexion of the verbs in μι.

^{page 297 note 2} The -sis of -ensis is difficult. Is it parallel to the -sio- which forms so many derivates, Plana-sia, contrā-riu-s, aquā-riu-s, etc.?

^{page 298 note 1} See also C. R. iii. p. 469.

^{page 299 note 1} Brugmann Grundr. ii. § 140, 4, Whitney, Skr. Gramm. §§ 965, 966, 1217.

^{page 299 note 2} dispennite from dispendo may be thought to prove that -nd- did become -nn- in some dialect, since pando certainly contains an I. -Eu. -nd-. But it is probably only an incorrect dialecticism, on the pattern of distennite. Would more exactness be expected in representing dialects in a modern farce?

^{page 300 note 1} φο⋯υιοϛseems a confusion of φουιοϛ and φοιυ⋯ϛ (the latter is only quoted from Nicander).

^{page 300 note 2} See Dr. Fennell's pamphlet on ‘The Indo-European Vowel System.’ Its main contention, the importance of syllabism in phonetic change, and its connexion with accent, appears to me most important, though I cannot accept the wider applications of ‘Fennell's Law’ which its author would give to it.