Roger Ratcliff Gail McKoon Northwestern University
An important concept championed by Tulving ( 1974, 1983) is the notion that retrieval conditions are critical in assessing memory: Memory cannot be assessed independently of retrieval, and a theoretical description of memory cannot be formulated without specification of the retrieval environment. In this chapter, we illustrate the profound influence that this view has had on the development and testing of memory models, and we show how the cue dependent view has begun to have an influence on research concerned with text processing.
There is a new generation of memory models that are more ambitious than models that were developed in the 1970s (with the notable exceptions of HAM [ Anderson & Bower, 1973] and ACT [ Anderson, 1976]). The new models attempt to deal with a range of phenomena across experimental paradigms at a level of detail that in the past has been found only in extremely limited models designed for a single task. In this chapter, five models that vary in their commitment to cue-dependent retrieval are considered and evaluated on the dimension of encoding/retrieval interaction. In models dealing with recall and recognition, the treatment of results from the recognition failure procedure is described because these results provide serious problems for superficial accounts of recognition and recall. In the models that do not deal with recall and recognition, the treatment of cue-target interactions is discussed.
In the recognition failure procedure, pairs of words are studied and then the second member of a pair is tested; in one test, the second member is presented by