Negative patterning tasks are a key tool to unveil the mechanisms by which stimulus representations are acquired—a central concern in Robert Rescorla’s research. In these tasks, target stimuli are reinforced when presented individually (A+/B+) but not when presented in compound (AB−). The discrimination of single stimuli from their compound presentation is a challenge for theories of associative learning, because it cannot be explained by the simple accrual of associative strength. The present study examined the conditions under which mice learn this part–whole discrimination in olfactory stimuli using a novel instrumental methodology. In two experiments, reinforcement was contingent on distinct responses depending on whether a set of odor mixtures were presented in isolation or as a compound. Using C57BL/6 mice, Experiment 1 showed a mutual interference between learning a response to individual odors and learning a different response to those odors presented in compound. Using wild-type APP/PS1 mice (a control strain for a murine model of Alzheimer’s disease), Experiment 2 replicated this interference and showed that it is stimulus-specific. These experiments show that the instrumental patterning task may not only complement Pavlovian negative patterning tasks but may also motivate its own questions on the representation of complex stimuli.