US-revaluation refers to the observation that subsequent changes in the valence of an unconditioned stimulus (US) after pairing it with a neutral, conditioned stimulus (CS) also changes the valence of the associated CS. Experiment 1 found evidence for the US-revaluation effect using an unobtrusive measure of evaluation. However, US-revaluation effects were more pronounced for positive-to-negative compared to negative-to-positive revaluations. Experiment 2 replicated this finding for self-reported evaluations, further showing that US-revaluation effects are stable over time and independent of explicit memory for the revaluating information. Using a modified paradigm, Experiment 3 ruled out method-related explanations for these findings and showed that changes in CS evaluations are correlated with parallel changes in US evaluations. These findings encourage the view of evaluative conditioning as an instance of stimulus–stimulus (S–S) rather than stimulus-response (S–R) learning. Implications for basic and applied research are discussed.