Similarity-Based Reasoning is Shaped by Recent Learning Experience

Popular approaches to modeling analogical reasoning have captured a wide range of developmental and cognitive phenomena, but the use of structured symbolic representations makes it difficult to account for the dynamic and context sensitive nature of similarity judgments. Here, the results of a novel behavioral task are offered as an additional challenge for these approaches. Participants were presented with a familiar analogy problem (A:B::C:?), but with a twist. Each of the possible completions (D1, D2, D3), could be considered valid: There was no unambiguously “correct” answer, but an array of equally good candidates. We find that participants’ recent experience categorizing objects (i.e., manipulating the salience of the features), systematically affected performance in the ambiguous analogy task. The results are consistent with a dynamic, context sensitive approach to modeling analogy that continuously updates feature weights over the course of experience.