Sixteen kindergarten and 28 first-grade children were tested on two-choice discrimination problems. A prompt light indicated the positive (rewarded) object P on all training trials, and these were followed by a single nonprompted test trial during which a new object (X) replaced either P (X-N problems) or N (P-X problems) or neither (P-N control problems). Two additional control problems assessed verbal responses to the P and N objects alone. All Ss followed the prompt (i.e., displaced only P) and therefore never directly observed the nonreward value of N on prompted trials. However, performance was significantly above chance on nonprompted X-N trials. Control conditions and verbal reports permitted the conclusion that the negative (nonrewarded) value of N had been inferred while responding to P on prompted trials. Replicating previous findings, the present results further suggest that stimulus novelty is not an important factor in cue-substitution procedures. (Author)