Subjects were given the attribute-identification task of unidimensional and conjunctive concepts under a condition in which a list of attributes instead of an array of stimulus patterns was available. Subjects tested their strategies by selecting one attribute (unidimensional concepts) or two attributes (conjunctive concepts) at a time. Since a correct strategy was chosen by the experimenter in advance, an appropriate feedback could be given depending on subjects' selecting a correct, a wrong, oran irrelevant strategy as specified by the strategy-selection theory. In comparison with a control condition in which an array of stimulus patterns was shown, subjects took a greater number of trials to reachthe criterion. It was concluded that subjects seemed to process information quite efficiently by comparing the stimuli selected on the successive trials to arrive at the correct strategy in the usual condition of attribute identification. This means that the selection probabilities of a correct, a wrong, or an irrelevant strategy may change suddenly in the course of the task.