Interaction does not lead to spontaneous category-based conditioning in an artificial language

Abstract Variation is present in every language at every structural level. Though extremely complex, linguistic variation is not fully unpredictable. Previous research suggests that cognitive biases in learning favour conditioned variation: learners often make languages more predictable by eliminating variation or by conditioning it on context, pointing to the presence of biases against random variation. Learning biases favour lexical conditioning over more general category-based conditioning, though both occur in natural languages. Interaction may also contribute to shaping conditioned variation by providing a mechanism for interlocutors to develop a shared system through the coordination of individual preferences. In the present study, we investigated the role of dyadic interaction in the emergence of conditioned variation. We trained participants on an artificial language with unpredictable variation in plural marking and objects representing one or two semantic categories and had them play a communication game using the newly learned language. We hypothesised that interaction would introduce category-based conditioning, this being the simplest conditioned system in the language. Contrary to our expectations, we found no evidence of spontaneous category-based conditioning: participants either removed variation or conditioned marker use on lexical items. Further experiments are needed to explain the emergence of this common linguistic pattern.