Proceedings of the Annual Meeting of the Cognitive Science Society

Two learning mechanisms have been suggested to underlie statistical learning: computation of transitional probabilities and chunking. It remains an open question though what determines which mechanism is used. In this study, we examined whether learning mechanisms are exploited differentially depending on the structure of the input to be learned. More specifically, we investigated whether the strength of the relationships between elements in the input structure and the presence of higher-order relationships influence the employment of the mechanisms. Participants were presented with three different input structures. We measured reaction times in a self-paced statistical learning task and created Bayesian models that formalised different learning mechanisms. The results show that the employment of the learning mechanisms indeed depends on the input structure. Further studies will need to examine a more specific mapping between the input structures and the learning mechanisms.