^{page 17 note 1} Since writing this article I have had the privilege of reading Bloch, M. J.'s excellent book La formation de la langue maratheGoogle Scholar, which all students of Indian languages in particular and of comparative philology in general will welcome as one of the first scientific attempts to explain the phonetic history of a modern Indian language.

^{page 18 note 1} DrBarnett, 's statement in Antiquities of India, p. 34Google Scholar, that G. is descended from Āvanī seems without sufficient foundation. The Āvantī dialect of Ś. has left so few monuments that it is impossible to draw a safe conclusion.

^{page 18 note 2} A difficulty in the way of a close connexion is the different treatment of -m-. In A. every -m- tended to become -ṽ- e.g. nāṽa, caṽaru from nāma camaraḥ: G. nām cāmar (see 3. 3 ff.). The chief point of connexion is the close resemblance of verbal forms (cf. 5. 1).

^{page 21 note 1} Bloch, M., op. cit., p. 141Google Scholar, wrongly says that -m- changes to -ṽ- in all languages except Singhalese and North-West dialects.

^{page 21 note 2} The same writer, op. cit., pp. 50 ff., ignores any influence of stress accent in the development of the Mod. IA. languages. To do this in the case of G., at least, seems to me impossible. There is little or no stress now, but its effects are evident. It is worth noting that in many words it must have had a different position from any accent postulated to explain Marāṭhī changes.

^{page 24 note 1} M. Bloch gives the same form for M., but says that it is a matter only of writing, not of speech. The G. seems to me to be simply pāṇī.

^{page 25 note 1} My views have undergone some changes since my last note in the JRAS.

^{page 26 note 1} Only found in -aka-stems, i.e. where G. sing, is -ũ; -a- stems (e.g. ghar n. “house”) have pl. in -ō.

^{page 28 note 1} M. Bloch (§ 137) follows Grierson, and suggests a feeling of linguistic equivalence between m and v as the cause. This, however, would seem to be wrong in the light of the history of -m- in G.

^{page 29 note 1} There is a doublet āṇ: cf. M. āṇi “and”. M. Bloch (§ 135) suggests an Idg. difference of form.