Evaluative conditioning (EC) is the valence change of a stimulus (conditioned stimulus, CS) that is due to the previous pairing with another stimulus (unconditioned stimulus, US). We investigated whether EC can occur also when the CS-US pairings are not experienced directly by the participant but are implied by other events that the participant encounters. In two experiments, positive USs were presented in some trials and negative USs in other trials. Afterwards, participants were given information from which it was possible to conclude that CSs were covertly present during these trials. Finally, the valence of these CSs was registered using both implicit (Implicit Association Test, affective priming) and explicit measures (valence ratings). In line with the assumption that EC effects can be based on CS-US pairings that are not directly experienced, the valence of the CSs changed in the direction of the US with which they were covertly paired. This effect was observed both on explicit and on implicit measures. We argue that several aspects of our results are in line with propositional models of EC and fit less well with association formation models.