ABSTRACT
This talk will focus on two biases that influence (morpho-)phonological learning: neutralization avoidance and paradigm uniformity. The first part of the talk will focus on neutralization avoidance. Phonological neutralization increases ambiguity in a language, particularly when the neutralization leads to significant amounts of homophony. Because too much ambiguity would result in a functionally ineffective language, we can hypothesize that some pressure works against neutralization during the course of language change. Previous work has tended to model this pressure as being external to the learning process, stemming from a production/perception loop (e.g. Wedel 2004, 2006, 2012) or communicative pressures (e.g. Kirby 2015). In this talk, I will argue on the basis of two artificial language experiments that a neutralisation avoidance bias is (additionally) active during phonological learning itself, driven primarily by a desire to avoid homophony. I will also outline a real case of language change in the history of Campidanian Sardinian where evidence suggests that neutralization avoidance caused learners to shift to a drastically different phonological system, even though the new system was itself highly marked from a phonological perspective.
The second part of the talk will focus on paradigm uniformity (i.e. the avoidance of alternations altogether). Previous work has shown that learners, and especially children, prefer uniform paradigms (e.g. Do 2018). We investigated how the preference for paradigm uniformity influences the way that adult learners acquire a variable pattern in an artificial language. Using the mixture-shift experimental paradigm (Culbertson et al. 2012), we exposed learners to a language in which plurals were variably marked by one of two prefix forms, one of which was frequent and one infrequent. One of the prefix forms (either the frequent or infrequent variant, depending on condition) additionally caused stem alternations. Learners generally matched the frequency of the two variants in their input, with one exception: when the infrequent variant also caused stem alternations, learners used it even less often than it occurred in their input (i.e. they regularized the frequent variant that did not cause alternations). The results suggest that the preference for paradigm uniformity could hinder the adoption and proliferation of forms if they cause alternations.
