Automatic and strategic aspects of knowledge retrieval

Abstract

An experiment was performed in which subjects retrieved members of natural categories. Subjects used a variety of retrieval cues in this task, making knowledge retrieval a two-tiered process: First, contexts were generated in which category members are likely to be found, and second, these were used as retrieval cues to produce the category members themselves. These retrieval cues were primarily episodic rather than abstract-semantic. Similarly, in a script-generation task, subjects tended to generate scripts from personal experiences rather than through the retrieval of an abstract scheme or script. A loosely interassociated memory network is suggested by these results. A model of memory retrieval developed to account for the data of list-learning experiments by Raaijmakers and Shiffrin was shown to be able to account for the automatic component of knowledge retrieval but requires a more complex control structure before it can successfully simulate the retrieval strategies used by the subjects in these experiments.