The purpose of the present study was to investigate the effect of mediation-generative instructions upon the performance and learning modes in reversal shift with verbal materials. Prior-instructions were given to Ss as follows: (a) Instructions to generate a visual image of the stimulus pair member (Verbal), (b) presentation of the picture of the stimulus pair member (Pictorial), (c) the presentation and the instructions were given together (Pictorial-Verbal), or (d) neither Pictorial nor Verbal instructions were given (Control). In the Pictorial-Verbal, Verbal, Pictorial groups (in order), the mediational mode responses (conceptual, response-switching learning mode) were found more numerous than the associative learning mode, while in the Control group the inverse was true. In the group that efficient mediational mode responses were found more numerous, the shift problem was easy. These results were interpreted in terms of the depth of information-processing of Ss.
!!!All Science Journal Classification (ASJC) codes