The mystery of word-initial consonant loss in Kera'a

Abstract

Kera’a, a Trans-Himalayan language, shows a theoretically and typologically unexpected loss of word-initial consonants in CV syllables (e.g. /kalondrõ/ > /alondrõ/ ‘mosquito’, /tahru/ > /ahru/ ‘to ask’, /miku/ > /iku/ ‘dog’). Word-initial consonant loss challenges certain widespread assumptions: that the CV syllable is the preferred syllable structure, and that consonants in word-initial position are strengthened rather than reduced. In addition, we will see that the Kera’a data also challenges the notion that sound change is necessarily phonetically grounded. So far, this kind of unexpected loss has otherwise mostly been found in numerous Australian languages, where it is described as a typological and theoretical oddity (Blevins 2001, 2007, Verstraete 2022). Attempts to explain word-initial consonant loss in the Australian context are diverse (Blevins 2001), and in several instances, there is no satisfactory explanation. In this paper, we argue that Kera’a poses a perhaps even greater challenge to phonological theory, as none of the Australianist explanations — stress shift, phonetically “weak” consonants, or contextual predictability — holds for Kera’a. In contrast to several Australian languages, Kera’a also shows little evidence for a phonetic foundation of the loss in lenition. This paper explores this unexpected loss through a discussion of Kera’a synchronic and diachronic phonology, and through a detailed acoustic analysis of stress. Based on our findings, we argue for a qualification of some of the most basic tenets of phonological theory.