Four experiments examined the effect of Pavlovian conditioned inhibition on specific Pavlovian-instrumental transfer (PIT) in human participants. The task comprised an instrumental phase in which two responses (R 1 , R 2 ) were each paired with one of two outcomes (O 1 , O 2 : R 1 -->O 1 , R 2 -->O 2 ), and a Pavlovian phase, in which two CSs, CS 1 and CS 2 each signalled one of the two outcomes (CS 1 -->O 1 , CS 2 -->O 2 ). In Experiments 1-2 a conditioned inhibitor, X, predicted the omission of one of the outcomes (e.g. CS 1 -->O 1 , CS 1 X-->nothing). In a subsequent test, performance of R 1 and R 2 was examined in the presence of CS 1 and CS 2 . A specific PIT effect was observed: R 1 was performed more than R 2 during CS 1 , and R 2 more than R 1 during CS 2 . This PIT effect was significantly reduced by the presence of the inhibitor X in Experiment 1, in which the Pavlovian phase followed the instrumental phase, and in Experiment 2 in which it preceded it. No such effect was observed when X was presented in the absence of any expectation of the outcomes during the PIT test (Experiment 3a), or when X was trained as a signal for an alternative outcome (Experiment 3b). These results are consistent with the suggestion that the specific PIT effect occurs through a stimulus-outcome-response (S-O-R) mechanism, according to which the CS evokes a representation of the outcome which in turn elicits the response (e.g. CS 1 -->O 1 -->R 1 ). The conditioned inhibitor suppresses performance of the response by suppressing activation of the outcome representation.