Abstract: Three experiments with participants of college students were done to investigate feature reasoning in uncertain categorization in the concentrative presentation of samples of category members. The results showed that it is the overall conditional probability of the prediction feature relative to the target feature rather than category uncertainty and representativeness of target features that affects participants’ feature reasoning, and that feature reasoning is direct reasoning based on the overall conditional probability of feature relevancy rather than indirect reasoning based on categories. The results supported the authors’ overall conditional probability model of prediction features.

Key words: categorization, feature reasoning, Bayesian analysis, single category explanation, the overall conditional probability model of prediction features