Evaluative Conditioning (EC) is the change in liking of a conditioned stimulus (CS) due to its pairing with a positive/negative unconditioned stimulus (US). EC research has so far focused exclusively on pairings with a single US within a situation. However, natural environments are arguably more complex, leading to pairings with more than one affective stimulus at a time. Here, we investigate EC in situations with multiple USs. In five experiments (total N = 389, all preregistered) with different materials and pairing procedures, we paired CSs with one/two positive/negative USs simultaneously. We find that two USs do not influence people’s conditioned attitudes in a mere additive fashion. Instead, people average the valence of multiple USs with a stronger weight of negative USs. This weighted averaging leads to two non-additive patterns: First, pairings with multiple US of the same valence exert the same conditioning effect as pairings with only one US of that valence. Second, the EC effects of one US is strongest if no second US is present and weakest if another negative US is present. We show that these non-additive patterns cannot be explained by weakened memory for stimulus pairings or by US revaluation effects. Further findings indicate that people most likely integrate multiple affective experiences in a situation already at learning. We discuss the implications of our research for theories on and practical applications of evaluative conditioning.