Proceedings of the Annual Meeting of the Cognitive Science Society

Unpredictable variation is widely used to investigate how cognitive and communicative biases impact on language evo-lution and change. Learning, interactive and cultural biases all contribute to universal linguistic patterns. We exploredthe effects of social cues using a miniature artificial language exhibiting unpredictable lexical variation distributed eitherwithin or between multiple speakers. We compared the effects of testing modality (spoken vs. forced-choice), experimen-tal population (students vs. online workers) and setting (laboratory vs. online). Learners were sensitive to social cues,but reliable differences only emerged in the laboratory. In an online setting, students were much more likely to regulariseacross conditions. In addition, task difficulty increased rates of regularisation but only online. Online workers showedhigh levels of regularisation throughout. Our experiments suggest that the conditions in which learning and recall takeplace have a large impact on the biases which shape language and our ability to measure them.