Across six experiments, we tested the hypothesis that acquired attitudes are moderated by memory retrieval of relevant evaluative information previously experienced with the attitude object. In Experiment 1, an evaluative conditioning effect was moderated by presenting the unconditioned stimulus (US) as a retrieval cue, using a design that controlled for valence priming effects. Experiment 2 replicated this moderation effect in an impression formation paradigm, where attitudes were based on behavioral descriptions. Experiments 3a and 3b used a similar design as Experiment 1, but with intermixed positive and negative information. Again, retrieval cues influenced attitudes, although the effects were smaller in magnitude and not significant in one of the studies. In Experiments 4 and 5, we examined memory retrieval in a setting with higher external validity: We assessed people’s liking of everyday objects and showed that liking was related to the valence of a related memory after the memory’s retrieval. We found this effect on valence ratings and on an adapted Affect Misattribution Procedure, but not in an evaluative priming task. Together, these studies demonstrate across multiple paradigms that acquired attitudes depend not only on the encoding and storage of evaluative information, but also on its retrieval. These findings suggest that, within the paradigms and timescales we investigated, acquired attitudes do not become fully semantically integrated into the representation of attitude objects but remain contingent on the retrieval of the original evaluative experiences.