Context, integration, and retrieval.

Three experiments examined the effectiveness of external retrieval cues when encoding context varied with respect to the integration of the representation. In all three experiments, it was found that nonencoded cues led to greater improvement if the initial representation was not well integrated. Strong-associate cues led to more improvement when encoding context consisted of weak-associate pairs than when the pairs were embedded in sentences (Experiment 1). The cues were more effective when subjects studied a list of words without instructions than when they were instructed to form images integrating the list members (Experiment 2). The third experiment demonstrated that well integrated material takes longer to access, and a control experiment argued against an encoding interpretation of the data. The results demonstrated both a flexibility of retrieval and a restriction from context, such that the better the representation, the harder it is to retrieve using external retrieval cues.