Coherent category training enhances generalization in prototype-based categories

Abstract

A major question for the study of learning and memory is how to tailor learning experiences to promote knowledge that generalizes to new situations. In two experiments, we used category learning as a representative domain to test two factors thought to influence the acquisition of conceptual knowledge: the number of training examples (set size) and the similarity of training examples to the category average (set coherence). Across participants, size and coherence of category training sets were varied in a fully crossed design. After training, participants demonstrated the breadth of their category knowledge by categorizing novel examples varying in their distance from the category center. Results showed better generalization following more coherent training sets, even when categorizing items furthest from the category center. Training set size had limited effects on performance. We also tested the types of representations underlying categorization decisions by fitting formal prototype and exemplar models. Prototype models posit abstract category representations based on the category's central tendency, whereas exemplar models posit that categories are represented by individual category members. In Experiment 1, low coherence training led to fewer participants relying on prototype representations, except when training length was extended. In Experiment 2, low coherence training led to chance performance and no clear representational strategy for nearly half of the participants. The results indicate that highlighting commonalities among exemplars during training facilitates learning and generalization and may also affect the types of concept representations that individuals form. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

Figure 1.

Category-learning task. Category representations under the assumptions of the A. prototype model and B. exemplar model. Prototype: categories are represented by their central tendencies (prototypes). New items are classified into the category with the most similar prototype. Exemplar: categories are represented as individual exemplars. New items are classified into the category with the most similar exemplars. C. Example stimuli. The leftmost stimulus is the prototype of category A and the rightmost stimulus is the prototype of category B, which shares no features with prototype A. Members of category A share more features with prototype A than prototype B and vice versa for members of category B. D. Participants underwent feedback-based training with one of six possible training sets that varied the size of the training set and the coherence of the examples. E. In recognition, participants were shown training (old) items and never seen category members and made old/new judgments (see Supplemental Materials for detailed methods and results). F. In categorization, participants were shown training (old) items and never seen category members and made categorization judgments without feedback.

Figure 2.

A. Training accuracy by block for the small set size conditions (dotted lines) and large set size conditions (solid lines) that matched the small set size in terms of the number of repetitions of each item. High-coherence conditions in black, low-coherence conditions in gray. B. Same as A, but depicting the large set size conditions (solid lines) that matched the small set size in terms of the total number of trials during training. For small training sets, the same data are depicted in A and B, but collapsed from 8 blocks of 16 trials each in A into 4 block-pairs of 32 trials in B to match 32 trials per block in the large set size (trial matched) conditions. Error bars depict the standard error of the mean across subjects.

Figure 3.

Categorization accuracy for each item type separated by training set. Small = sets with 4 items/category. Large = sets with 8 items/category. High = high coherence: 6 prototypical features for all training examples. Low = low coherence: 5 prototypical features for all training examples. Accuracies on training (old) items re-presented during the categorization test are depicted with striped bars. Accuracies for new items varying in their similarity to category prototypes are depicted with solid bars. Accuracies for prototypes are depicted in the darkest bars with increasingly lighter bars for new items sharing fewer features with the prototypes. Error bars depict standard error of the mean across subjects.

Figure 4.

Training set coherence x test item typicality interaction effect in generalization. Test item typicality is depicted on the x-axis with the prototypes on the far left and the items closest to the category boundary (5 prototypical features) on the far right. Dark bars depict generalization accuracy for those trained with high coherence sets (6 prototypical features for all training examples), and light bars depict generalization accuracy for those trained on low coherence sets (5 prototypical features for all training examples). Both coherence groups were collapsed across set size conditions. Stars indicate differences between groups of p < .05 corrected, and tildes indicate differences between groups of p < .05 uncorrected. Error bars depict the standard error of the mean across subjects.

Figure 5.

Prototype and exemplar model fits to categorization data. Small = sets with 4 items/category. Large = sets with 8 items/category. High = high coherence: 6 prototypical features for all training examples. Low = low coherence: 5 prototypical features for all training examples. Subjects better fit by the prototype than exemplar model are indicated in blue. Those better fit by the exemplar than the prototype model are indicated in red. Those will comparable fits across models are in purple (‘similar fits’). Those whose fits did not outperform the random model are in grey. Model fits are separated by training set typicality, set size, and whether large sets were matched to small sets in terms of the number of repetitions of each item (repetition-matched) or the number of total training trials (trial-matched).

Figure 6.

A. Training accuracy by block for the small set size conditions (dotted lines) and large set size conditions (solid lines) that matched the small set size in terms of the number of repetitions of each item. High-coherence conditions in black, low-coherence conditions in gray. B. Same as A, but depicting the large set size conditions (solid lines) that matched the small set size in terms of the total number of trials during training. For small training sets, the same data are depicted in A and B, but collapsed from 8 blocks of 20 trials each in A into 4 block-pairs of 40 trials in B to match 40 trials per block in the large set size (trial matched) conditions. Error bars depict the standard error of the mean across subjects.

Figure 7.

Categorization accuracy for each item type separated by training set. Small = sets with 5 items/category. Large = sets with 10 items/category. High = high coherence: 8 prototypical features for all training examples. Low = low coherence: 6 prototypical features for all training examples. Accuracies on training (old) items re-presented during the categorization test are depicted with striped bars. Accuracies for new items varying in their similarity to category prototypes are depicted with solid bars. Accuracies for prototypes are depicted in the darkest bars with increasingly lighter bars for new items sharing fewer features with the prototypes. Error bars depict standard error of the mean across subjects.

Figure 8.

Training set coherence x test item typicality interaction effect in generalization. Test item typicality is depicted on the x-axis with the prototypes on the far left and the items closest to the category boundary (6 prototypical features) on the far right. Dark bars depict generalization accuracy for those trained with high coherence sets (8 prototypical features for all training examples), and light bars depict generalization accuracy for those trained on low coherence sets (6 prototypical features for all training examples). Both coherence groups were collapsed across set size conditions. Stars indicate differences between groups of p < .05 corrected. Error bars depict the standard error of the mean across subjects.

Figure 9.

Prototype and exemplar model fits to categorization data. Small = sets with 5 items/category. Large = sets with 10 items/category. High = high coherence: 8 prototypical features for all training examples. Low = low coherence: 6 prototypical features for all training examples. Subjects better fit by the prototype than exemplar model are indicated in blue. Those better fit by the exemplar than the prototype model are indicated in red. Those will comparable fits across models are in purple (‘similar fits’). Those whose fits did not outperform the random model are in grey. Model fits are separated by training set typicality, set size, and whether large sets were matched to small sets in terms of the number of repetitions of each item (repetition-matched) or the number of total training trials (trial-matched).