[PS-2.2] Illusory correlations and the acquisition of probabilistic linguistic variation

Saldana, C. , Loy, J. & Smith, K.

Centre for Language Evolution, The University of Edinburgh

Subjective correlations that exaggerate objectively presented contingencies are known as illusory correlations. While they are a well-documented phenomenon in social-cognition, little is known about their role in language learning, particularly in the context of probabilistic variation. In this paper we explore how the frequency of categories within a cue affects the emergence or accentuation of correlations between linguistic variants and contextual cues. We test this across three artificial language learning experiments, manipulating the frequency distribution of two categories within a cue (50%-50% or 25%-75%). Participants are trained and tested on two plural marker variants. In Experiment 1, the two variants are equally frequent across cue categories (no majority variant). In Experiment 2, we skew the distribution of variants equally across cue categories (same majority variant) . In Experiment 3, we probabilistically condition variants on cue category (the majority variant within a category is the minority within the other and vice versa). We conduct these experiments across different cue types: semantic (animate vs. inanimate referents) and social (male vs. female speakers). Results suggest that learners are sensitive to the distribution of cue categories in the learning and formation of conditioned variation, and more so with semantic than with social cues.