Besides traditional supervised classification learning, people can learn categories by inferring the missing features of category members. It has been proposed that feature inference learning promotes learning a category's internal structure (e.g., its typical features and interfeature correlations) whereas classification promotes the learning of diagnostic information. We tracked learners' eye movements and found in Experiment 1 that inference learners indeed fixated features that were unnecessary for inferring the missing feature, behavior consistent with acquiring the categories' internal structure. However, Experiments 3 and 4 showed that fixations were generally limited to features that needed to be predicted on "future" trials. We conclude that inference learning induces both supervised and unsupervised learning of category-to-feature associations rather than a general motivation to learn the internal structure of categories. (Contains 16 figures and 4 tables.)