关键词: 类别使用, 特征诊断效应, 类别判断, 特征预测

Abstract: Introduction
Category use mainly includes category judgment and feature prediction. The important issue is how people retrieve diagnosticity information from features that are distinct among different categories. Previous studies mostly investigated how the number of overlapped diagnosticity features influenced category learning. In order to reduce the variables, researchers commonly differentiate feature diagnosticity into 2 states, namely “none” and “all.” However, in natural situations, feature diagnosticity is often multilevel, the effect of which depends on 2 aspects: single feature’s diagnosticity and the number of diagnosticity features. In the present study, these 2 aspects of category use were investigated along with category judgment and feature prediction in order to determine the real effect of feature diagnosticity.
Method
We conducted 2 experiments, each focused on category judgment and feature prediction. The experimental design adopted was 3 (single feature’s diagnosticity level: high, medium, and low) × 2 (number of diagnosticity features: 1 and 2) within-subjects design.
Forty-eight 21-year-old juniors participated in the study. Drawings of fish contours were assessed in the preliminary experiment. There were 2 fish categories, one of which comprised 5 category members. The fish contour was designed to comprise 4 feature dimensions according to the 4 different levels of feature diagnosticity, included high, medium, low and non-diagnosticity level. The experimental drawings were presented using the DMDX software. In the category learning phase, participants attempted to classify the fish drawings into 2 target categories until their correct score reached 90%. Subsequently, the category use phase started. In experiment 1, the participants were asked to judge the category labels of novel testing items, while in experiment 2, the novel testing items were presented along with its category label, and the participants were asked to predict the missing features. Finally, a repeated-measures MANOVA was used to analyze the data.
Results
In experiment 1, statistically significant results were obtained for the main effects associated with the 2 independent variables but not for the interaction effect. The effect of high-level diagnosticity feature was significantly better than that of medium level; however, there was no significant difference between medium- and low-level diagnosticity feature. The effect of 2 diagnosticity features was also significantly better than that of 1 diagnosticity feature.
In experiment 2, the main effect was obtained for single feature’s diagnosticity but not for number of diagnosticity features. The effect of high-level diagnosticity feature was significantly better than that of medium- and low-level diagnosticity feature; however, no significant difference was observed between medium- and low-level diagnosticity feature. No interaction effect was found.
Conclusion
The experimental results suggest that both the single feature’s diagnosticity and the number of diagnosticity features could facilitate category judgment, but only the former could facilitate feature prediction

Key words: category use, effect of feature diagnosticity, category judgment, feature prediction