Learning about the internal structure of categories through classification and feature inference

Abstract

Previous research on category learning has found that classification tasks produce representations that are skewed toward diagnostic feature dimensions, whereas feature inference tasks lead to richer representations of within-category structure. Yet, prior studies often measure category knowledge through tasks that involve identifying only the typical features of a category. This neglects an important aspect of a category's internal structure: how typical and atypical features are distributed within a category. The present experiments tested the hypothesis that inference learning results in richer knowledge of internal category structure than classification learning. We introduced several new measures to probe learners' representations of within-category structure. Experiment 1 found that participants in the inference condition learned and used a wider range of feature dimensions than classification learners. Classification learners, however, were more sensitive to the presence of atypical features within categories. Experiment 2 provided converging evidence that classification learners were more likely to incorporate atypical features into their representations. Inference learners were less likely to encode atypical category features, even in a “partial inference” condition that focused learners' attention on the feature dimensions relevant to classification. Overall, these results are contrary to the hypothesis that inference learning produces superior knowledge of within-category structure. Although inference learning promoted representations that included a broad range of category-typical features, classification learning promoted greater sensitivity to the distribution of typical and atypical features within categories.

Keywords:

• Category learning
• Classification
• Feature inference
• Representation
• Typicality

The authors thank Seth Chin-Parker for his comments on previous drafts of this paper, James Pellegrino, Susan Goldman, and Katie Witkiewitz for their input during the development of this project, and Megan Engleman and Erin Strand for their assistance with data collection.

This research was supported by a Natural Sciences and Engineering Research Council of Canada (NSERC) Postgraduate Scholarship to B.D.J. Portions of this work were completed by B.D.J. in partial fulfilment of the requirements for the degree of Doctor of Philosophy in Psychology at the University of Illinois at Chicago.

Notes

¹Following Markman and Ross (2003), a feature dimension is defined a variable (e.g., tail length) that can take one of several feature values (e.g., short vs. long). The term feature is used to refer to the value that an exemplar has along a particular dimension. The term category label refers to the symbol that corresponds to a particular group of exemplars.

²We credit an anonymous reviewer for raising this issue.

³Alternatively, the participant could learn that members of Category A have feature value 0 on dimension 1 except for Exemplar A₄. Thus, they could use a rule plus an exception in order to classify. It is also possible that participants memorize the category label of each individual exemplar, although this is arguably less efficient than the other two strategies.

⁴Two control variables—feature diagnosticity set, and feature ordering—were included in an alternative version of this analysis. Neither factor had an effect or interacted with another variable in this or any analysis for Experiment 1 or 2. The same was true of the control variables for the postlearning tasks—task ordering, induction task set, and block order. To streamline the Results sections, the control variables are not reported and are not mentioned further.

⁵See Giguère, Lacroix, and Larochelle (2007), and Lancaster, Shelhamer, and Homa (2012) for evidence that classification learners do learn the correlational structure of a category in some cases.