This experiment tests the hypothesis that the amount of learning irrelevant to an e-induced set is inversely related to the strength of the incentive determining that set. Two groups of 37 Ss learned a sequence of 14 colored geometric forms to one perfect trial by the method of serial anticipation. Their instructions were to learn the names of the forms only. At the end of this training they were tested for incidental learning of the colors used in the forms. The high incentive group was offered a money bonus depending upon the speed of learning; the low incentive group was told not to try very hard. Results show that the learning of the form sequence was faster under the high incentive condition, but that the high incentive group scored significantly lower on the color recognition test. The difference was still significant (p=.O2) when the variable of rate of learning of the form sequence was controlled by analysis of covariance. This was interpreted to support the hypothesis that led to the expt.