It has been observed (Emeneau (1957), 63–4) that Toda and Kota show loss of many vowels in the non-initial syllables of reconstructed proto-Dravidian (PDr.) or proto-South Dravidian (PSDr.) forms. An attempt was made to state a rule as follows: ‘every non-initial syllable of PSDr. (or perhaps PDr.) that has a short vowel loses its vocalic peak in Toda; vowels of all qualities are lost except that *i > y'. Exceptions were noted, and an attempt was made, with some misgivings, to save the rule by several manoeuvres. One of these was to insist that the rule applied only to pre-Toda short vowels and to note that some at least of the short vowels that appear in non-initial syllables in Toda words ‘originally’ were long vowels, i.e. some Toda short vowels in non-initial syllables were derived from pre-Toda by a rule that stated that PSDr. (or PDr.) long vowels in non-initial syllables became short vowels in Toda and remained as such. This latter rule can be demonstrated to hold for a greater number of examples than the few that were mentioned in the 1957 paper. It should be noted at once that the rule applies only to Toda, and that when cognates are to be found, Kota retains the length of these vowels.