Where do grammars come from? The emergence of epenthesis
Many languages seem to avoid sequences of vowels within their words. This can be observed via morphological alternations which arise in just those cases where two vowels come together, effectively ‘repairing’ the output. This ‘repair’ can take the form of deletion of one of the two vowels, coalescence of the two vowels into one, or insertion of a novel segment between the two. This last process, known as epenthesis, is the focus of this talk. Within standard generative theory, all possible phonological repairs have implicitly equal status. New typological evidence I will present suggests, however, that epenthesis in vowel hiatus is much rarer than would be expected under this account. I will argue that this outcome can be understood via a new modeling approach that traces phonological paradigms from phonetic inception to phonological realization. This model is developed by integrating the results of three distinct types of research. Firstly, the revised typology prompts a division of epenthesis patterns into two main types, the first of which arises directly from phonetic epenthesis (glides and other vowel-like segments), and the second of which emerges indirectly through historic consonant loss (any segments likely to be found at the ends of morphemes). A further subdivision of the two types characterizes the degree of regularity of the resulting phonological process. The implicit hypotheses about how localized sound changes become language-wide grammar changes are refined using the second methodology: experiments on artificial language learning under impoverished input. Thirdly, the problem is reverse-engineered via analysis of the logically necessary conditions required for the emergence of a fully productive epenthesis process. The rarity of this end result thus emerges as the product of a set of special historical circumstances. Minus any of these, synchronic patterns of deletion, prosodically conditioned allomorphy, or suppletion will obtain instead. The results suggest a fundamental difference between epenthesis and other types of phonological processes that apply in vowel hiatus. New insight is gained into this classical phonological phenomenon by analyzing the language processing system as a whole. Furthermore, this methodology – which tests the interaction of perception, language learning, usage, and cognitive organization – is one that has broad promise for revealing universal principles in any phonological domain.