The effects of instructional variables and individual differences upon structural learning tasks were studied. Eighty college students were given pretasks to test mathematical problem solving and paired-associate (PA) learning skills. Each subject was then given two PA learning tasks presented in a study-test format. Kinship terms served as stimuli in one task and line-drawn faces in the other. Following learning to criterion, a PA test was presented to each subject consisting of old instructional stimuli to test retention, related stimuli to test acquisition of structure, and new stimuli to test learning generalization. Manipulated instrucational variables were: (1) type of information received (GIG group received a complete description of each stimulus dimension and attribute while the RIG group were told to try to discover the relationship in the material); and (2) stimulus underlying structure (for ER group, stimulus material represented each of the three possible stimulus dimensions and corresponding attributes equally and the UR group used unequal dimensions). Data indicated that GIG subjects out-performed RIG subjects. All groups gave more acquisition and retention responses to the kinship terms than to the faces. There were no differences between ER and UR groups. Pretask scores were more related to performance for RIG than for GIG subjects. (Author/JP)