LingLang Lunch (11/16/2016): Uriel Cohen Priva (Brown University)

An interplay between information, duration, and lenition

What are the causal mechanisms that lead to lenition, and why do languages tend to lenite particular segments? Predicting the actuation of lenition should explain why lenition tends to be exceptionless, why it may result in more effortful articulatory production, why it seems to ignore certain merger-avoidance properties that exist elsewhere in language (Wedel, Kaplan, and Jackson 2013), and why different language varieties may be similar in leniting the same segment, but differ in the output (e.g. /t/ to /ʔ/ vs. /t/ to /ɾ/). I propose that all these properties can be explained if we assume that lenition is caused by reduction in duration, and that one of the leading factors in systematic reduction in duration is low segment informativity. I show that low informativity is correlated with shorter duration in American English, above and beyond contextual predictability. I show that cross-linguistically, relative lower informativity of a particular segment matches the languages in which the selective lenition of particular segments occurs. I further show that lenition processes match duration-reduction better than they match undershoot alone, or effort-reduction alone, and that durational reduction occurs in leniting environments even when lenition itself does not happen.