The W t-R2 verbal discrimination transfer paradigm was found to produce pronounced positive transfer, relative to a nonspecific control, when subjects were explicitly informed of the interlist relationship. The instructional variable was suggested as important in reconciling discrepancies in previous literature.The present study investigated the influence of instructions concerning the interlist relationship between wrong-(W) and right-(R) item alternatives in the W 1 -Rl , W l -R2 verbal discrimination transfer paradigm. In this paradigm the W items of List 2 are identical to the W items of List I, while the R items of List 2 are new, unrelated items. Performance is typically compared to a nonspecific transfer control, W 1 -Rl' WrR2' in which List 2 involves both new Wand R items.According to the frequency theory of verbal discrimination learning (Ekstrand, Wallace, & Underwood, 1966), the W 1 -R2 paradigm should be governed by Rule 2 (select the subjectively least frequent item) early in List 2 practice. As List 2 practice continues, however, the initially "less frequent" R item should accumulate frequency units at a more rapid rate than the W item, based on the counting postulates of frequency theory, and the discrimination based on differential frequency should break down followed by a switch to a Rule I mode of responding (select the subjectively most frequent item). Thus, relative to the W2-R2 control, frequency theory predicts early positive transfer followed later by negative transfer, since Rule I is the appropriate mode of responding for the control group throughout List 2 practice.A number of studies have shown support or at least partial support for the predictions. Underwood, Jesse, and Ekstrand (1964), in a study important to the later developmen t of frequency theory, supported the prediction with a relatively fast anticipation interval presentation rate (1.5: 2 seconds), but not with a slower rate (3: 2). King and Levin (1971), in a study varying degree of List I learning, generally supported frequency theory, but the across-trial effects relative to the control were not conSistently present within each degree of List I learning. Unlike the Underwood et al. (1964) study, subjects were not informed of the interlist relationship. Schneider and Goulet (I973), varying the type of Requests for reprints should be addressed to N.