Although there has been extensive research in both humans and rodents regarding the influence of excitatory predictions on action selection, the influence of inhibitory reward predictions is less well understood. We used a feature-negative conditioned inhibition procedure to generate Pavlovian excitors and inhibitors, predicting the presence or absence of specific outcomes, and assessed their influence on action selection using a Pavlovian–instrumental transfer test. Inhibitors predicting the absence of a specific outcome reversed the bias in action selection elicited by outcome-specific excitors; whereas excitors promoted responding on the action associated with the same outcome as the cue, inhibitors shifted responding away from such actions and toward other actions. Furthermore, the influence of the inhibitors on choice reflected the nature of the inhibitory associations learned by participants; those encoding outcome-specific inhibitory associations showed a strong reversal in the bias elicited by the excitors, selectively biasing performance away from the action associated with the to-be-omitted outcome and toward other actions. In contrast, those encoding only general inhibitory associations did not show any bias during the transfer test and instead reduced their performance of both actions.