We present an extension of the search of associative memory (SAM) model that simulates the effects of both prior semantic knowledge and prior episodic experience on episodic free recall. The model incorporates a memory store for preexisting semantic associations, a contextual drift mechanism, a memory search mechanism that uses both episodic and semantic associations, and a large lexicon including both words from prior lists and unpresented words. These features enabled the model to successfully account for the effects of prior semantic knowledge and prior episodic learning on the pattern of correct recalls and intrusions observed in free recall experiments.
In this article, we present a model that simulates the effects of prior experience on episodic free recall, including the effects of previously learned semantic relations and prior episodic learning. The model presented here, which we term eSAM, is an extension of the search of associative memory (SAM) model (Gillund & Shiffrin, 1984; Raaijmakers & Shiffrin, 1981; Shiffrin & Raaijmakers, 1992). SAM is an associative model of memory in which it is posited that, during study, list items become associated with each other and with the study context in proportion to the amount of time the items spend in a limited-capacity rehearsal buffer. SAM further assumes that retrieval is cue dependent, with the list context and previously recalled items serving as retrieval cues for other items, and the probability of retrieving an item being determined by strength-dependent competition among all items associated to a given set of cues. SAM has been applied to a broad range of free recall phenomena, including the effects of presentation rate and list length (Raaijmakers & Shiffrin, 1980), part-set cuing (Raaijmakers & Shiffrin, 1981), word frequency (Gillund & Shiffrin, 1984), interference and forgetting (Mensink & Raaijmakers, 1988), list strength (Shiffrin, Ratcliff, & Clark, 1990), generation (Clark, 1995), and temporal contiguity (Kahana, 1996).
Notwithstanding SAM's far-ranging ability to simulate recall phenomena, instantiations of SAM to date have made certain simplifying assumptions that impair a test of whether the model can simulate the effects of prior experience on free recall. A vast body of empirical evidence suggests that episodic recall of a list learned during an experiment is affected both by preexperimental semantic relations involving list items (see, e.g., Deese, 1959; Glanzer, Koppenaal, & Nelson, 1972; Howard & Kahana, 2002; Kahana & Wingfield, 2000; Pollio, Richards, & Lucas, 1969; Roediger & McDermott, 1995; Romney, Brewer, & Batchelder, 1993; Tulving, 1968) and by the prior study of other lists during the experiment (e.g., Anderson & Bower, 1972; Kahana, Howard, Zaromb, & Wingfield, 2002; Postman & Underwood, 1973; Tulving, 1966; Zaromb et al., 2005).
Although the general theory underlying SAM allows for the storage in memory of associations formed prior to the study of a list, previous applications of the model to recall have represented only the items appearing in, at most, two lists presented experimentally for recall, and they generally have assumed that the only relations involving such list items are those that arise during study and recall of the list.
The eSAM model implements the SAM theory more generally by avoiding many of the simplifying assumptions of earlier work. The eSAM model extends the SAM framework to address the effects of prior experience by incorporating four major features. First, eSAM explicitly represents preexperimental, pairwise semantic associations among words. Although eSAM is a priori neutral as to the best measure of semantic association strength for this purpose, we use two particular measures of semantic rciatedncss, because they have previously been applied to large corpora of words: latent semantic analysis (henceforth, LSA; Landauer & Dumais, 1997) and word association space (henceforth, WAS; Steyvers, Shiffrin, & Nelson, 2005). …