A generalization becomes suppressed over time in the context of exceptions

There has been a great deal of interest in how generalizations and exceptions are learned, with particular interest in how speakers learn to avoid overgeneralizations. Do overgeneralizations disappear only because exceptions become more strongly represented or does the generalization itself become suppressed? Novel labels were constructed by combining 56 syllables with one of two prefixes, and each label was assigned a unique image. Most labels with the first prefix were paired with images from a generalization category, whereas exceptional labels were paired with images from a different semantic category. All labels with the second prefix appeared with a third category (“baseline”). Participants used a computer mouse to choose one of two images for each label. Mouse-tracking results show that the generalization itself became suppressed over time in the context of exceptional labels. A post-test demonstrated that exceptions were learned with item-specific precision.