¹ ‘The songs of the Todas’, Proceedings of the American Philosophical Society 77.543–60 (1937); ‘Oral poets of South India—the Todas’, Journal of American folklore 71.312–24 (1958) (also Traditional India: structure and change [ed. Milton Singer] 106–18 [Philadelphia, 1959]; reprinted in Language in culture and society [ed. Dell Hymes] 330–40 [New York, 1964]). One other relevant article, on the language, is ‘Toda, a Dravidian language’, Transactions of the Philological Society 1957.15–66. References to the projected publication of the songs (verbal texts only, since the music was not recorded in any way) are by the numbers of the songs and the couplets, or verses, within the songs (e.g. 78.30, i.e. song 78, couplet or verse 30). The chief authority on the Todas remains W. H. R. Rivers, The Todas (London, 1906). The field work which is the basis of my analysis was done at intervals in a three-year period from 1935 to 1938.

² They only partly duplicate the statements of the 1937 article.

³ A complete description of all Toda utterances would identify at least one more special type of voice production, that used by dairy-priests in ritual utterances; I am assuming that there is only one such used by them. The shamans, when performing, also have a special type of voice-production, but with hardly an exception they purport to be speaking another language, Malayalam, which is that of the various gods who possess them. Rivers (251) reported a shaman who was possessed by a Toda divinity and spoke in Toda. Presumably his voice production would have been similar to that of the shamans who speak Malayalam during possession. In my time in the field there were no shamans who spoke Toda while possessed.

⁴ Many of the four-syllable units end with the vocative suffix -a· or -ya· as the 4th syllable; e.g. ï nesofya. 'O buffalo Nesof!', aryin guxya. 'I do not know, O wife!' Others need not be analyzed here.

⁵ An em space within a sentence, or in a literal translation, indicates the end of a song-unit.

⁶ /n/ does not have an allophone Ŋ before /g/.

⁷ As in Kota; cf. Kota texts, part one 21 (UCPL 2; 1944).

⁸ If numerous examples have been given, it is because this type of construction is so unusual (or perhaps rarely reported) in inflected languages. Moreover, if one should look for archaic traits in Toda song language, this has as good a chance as any of being one, since it is apparently paralleled in Old Tamil, but not in Modern Tamil.

⁹ The exceptions include :

(1) unpaired sentences :

(a) those whose key units have no paired units; there are occasional sets of three such sentences, which might well be interpreted as triplets rather than sets of three unpaired sentences.

(b) those whose key units have paired units which are exceptionally omitted in these passages; the passages are usually sequences (running at times to 10 or 12 or more sentences) in which secondary key units seem to dictate the failure of the pairs to occur; the peahen's song (196) has a sequence of this kind, from which several examples are given at the end of section 3 below.

(2) sentences which within the one syntactic unit contain paired song-units. In section 2b below, one of the examples of redundancy is constructed in this way, the sentence beginning: kenofïn (230.27). Similarly, in section 3, the sentence beginning: kïr nasθad (241.16) ; in section 4, the sentence beginning kas wïda·r (233.61) ; in the section on figures of speech based on meaning, the sentence beginning pi·s o kwï·s (242.16).

Many of the examples given in the text of the article are single sentences, which in the texts occur as members of pairs; this needs no further remark. The parallel sentences are indicated by the punctuation symbols / and //, beginning in section 1 below.

¹⁰ ‘Toda dream songs’, JAOS 85.39–44 (1965).

¹¹ Treated in Burrow and Emeneau, A Dravidian etymological dictionary, as from two distinct groups of etyma, 2655 and 2815. The ‘translate’ meaning is a borrowing from Tamil tiri (S² tiritt-).

¹² For a discussion of this matter in less linguistic terms, see my 1958 paper.

¹³ The 1958 paper 321 f.

¹⁴ See JAOS 83.188–93, esp. 192–3 (1963).