Evidence for a learning bias against saltatory phonological alternations

This study provides new experimental evidence that people learn phonological alternations in a biased way. Adult participants were exposed to alternations between phonetically dissimilar sounds (i.e., those differing in both voicing and manner, such as [p] and [v]). After learning these alternations, participants assumed, without evidence in the input, that more similar sounds (e.g., [b] and [v]) also alternated (Exp. 1). Even when provided with explicit evidence that dissimilar sounds (e.g., [p] and [v]) alternated but similar sounds ([b] and [v]) did not, participants tended to make errors in assuming that the similar sounds also alternated (Exp. 2). By comparison, a control group of participants found it easier to learn the opposite pattern, where similar sounds alternated but dissimilar sounds did not. The results are taken as evidence that learners have a soft bias, considering alternations between perceptually similar sounds to be more likely.