The encoding specificity principle predicts that a change in context between input and test will adversely affect recognition memory. Experiment I tested this with sentences from a prose passage and no context effects were obtained. Experiments II, III, and IV compared context effects for words in random sentences versus connected discourse. In most cases, the usual encoding specificity effects were obtained for a list of unrelated sentences. But context effects were attenuated or wiped out when the test items had been embedded in a meaningful passage. The theme of a passage serves as a powerful retrieval cue that overrides the context effects that have typically been obtained. (Author)