Statistical learning creates novel object associations via transitive relations

A remarkable ability of the cognitive system is the creation of new knowledge based on prior experiences. What cognitive mechanisms support such knowledge creation? We propose that statistical learning not only extracts existing relationships between objects, but also generates new associations between objects that have never been directly associated. Participants viewed a continuous color sequence consisting of base pairs (e.g., A-B, B-C), and learned these pairs. Importantly, they also successfully learned a novel pair (A-C) that could only be associated through transitive relations between the base pairs (Exp1). This learning, however, was not successful with three base pairs (e.g., learning A-D from A-B, B-C, C-D), revealing a limit in this transitive process (Exp2). Beyond temporal associations, novel transitive associations can also be formed across categorical hierarchies (Exp3), but with limits (Exp4&5). The current findings suggest that statistical learning provides an efficient scaffold through which new object associations are transitively created.