This paper presents production data and acoustic modelling from 16 speakers of Telugu, showing that what has typically been described as a ternary distinction between alveolar, palatal, and retroflex sibilants is rather more accurately an emerging binary system, with the majority of speakers showing no distinction between the latter two postalveolar categories. The relative acoustic discriminability of the system (and ultimate support for merger patterns) was not, however, uniform and varied by position, vowel context, and lexical characteristics such as frequency, neighbourhood density, and the presence and number of sibilant competitors in the lexicon.
Charles Redmon (Wed,) studied this question.