Proceedings of the Annual Meeting of the Cognitive Science Society

People exhibit a tendency to generalize a novel noun to thebasic-level of a hierarchical taxonomy – a cognitively salientcategory such as “dog” – with the degree of generalization de-pending on the number and type of exemplars. Recently, achange in the presentation timing of exemplars has also beenshown to have an effect, surprisingly reversing the prior ob-served pattern of basic-level generalization. We explore theprecise mechanisms that could lead to such behavior by ex-tending a computational model of word learning and word gen-eralization to integrate cognitive processes of memory and at-tention. Our results show that the interaction of forgetting andattention to novelty, as well as sensitivity to both type and to-ken frequencies of exemplars, enables the model to replicatethe empirical results from different presentation timings. Ourresults reinforce the need to incorporate general cognitive pro-cesses within word learning models to better understand therange of observed behaviors in vocabulary acquisition.