Reversal (R) and extradimensional (ED) shifts were studied in a large Jordanian sample which included six age groups (4, 5, 6, 9, 10, and 11 years). No sex differences were found. Consistent with other findings based on American samples, a highly significant Age by Shift interaction effect was found, reflecting the ease of executing a R relative to an ED shift at a very young age and the reversal of this effect at about 9 years of age. The results provide further support for the Kendlers' two-stage mediational theory.The literature is replete with studies which have shown that, while rats and preschoolers solve an extradimensional (ED) shift faster than a reversal (R) shift, the opposite is true of older children and young adults (Kendler & Kendler, 1975). Moreover, children of approximately 5 years of age solve both shifts at comparable rates. In explaining these ontogenetic changes in problem-solving behavior, Kendler and Kendler (1962) proposed a dual-stage mediational model, according to which the behavior of rats and very young children is assumed to be under the control of specific external stimuli, whereas the behavior of older children and young adults is not governed by such a singleunit system. Instead, these subjects behave according to a mediational system in that they solve a discrimination task by utilizing conceptual mediational links between the overt response and the external stimulus. The lack of difference in executing either shift at about age 5 is viewed as marking a transition period that reflects a progression from stimulus-bound responding to mediational responding. As Nehrke (1973) has recently suggested, conceptual shift behavior appears to develop from a stage which is dominated by mediationally deficient behavior to one in which mediationally proficient functioning becomes more prevalent.Initially, the mandatory shift paradigm was used to investigate problem-solving behavior as reflected in discrimination learning in both humans and animals. According to this design, subjects are required to learn This research was carried out by the second author in partial fulfillment of his master's degree under the direction of the frrst author, who was on sabbatical leave at the University of Jordan, Department of Psychology, Amman, Jordan, during the 1975-1976
109to discriminate first between two stimuli that differ on two dimensions, with one dimension being arbitrarily designated as correct and the other as incorrect; then a shift in conditions is introduced (see Figure 1). In a R shift, the previously correct dimension continues to be correct but the previously correct cue in that dimension becomes incorrect. In an ED shift, the previously correct dimension becomes irrelevant and the previously Figure 1. mustration of the R-ED design.