Proactive interference and the dynamics of free recall.

Abstract

Proactive interference (PI) has long been recognized as a major cause of forgetting. We conducted two experiments that offer another look at the subject by providing a detailed analysis of recall latency distributions during the buildup of and release from PI. These functions were accurately characterized by the convolution of the normal and exponential distributions (viz., the ex-Gaussian), which previously has been shown to describe recognition latency distributions. Further, the fits revealed that the increase in recall latency associated with the buildup of PI results from a slowing of the exponential retrieval stage only. The same result was found even when a short retention interval was used (and recall probability remained constant). These findings suggest that free-recall latency may be a sensitive index of the increased search set size that has often been assumed to accompany the buildup of PI. A central insight emerging from the memory literature of the 1950s and 1960s was that previously learned information can result in the rapid forgetting of more recently learned information. Underwood (1957) argued that this phenomenon, termed proactive interference (PI), was by far the major cause of forgetting in everyday life. Indeed, even in laboratory experiments, the degree of retroactive interference encountered over the course of hours or days was assumed to pale in comparison with the degree of proactive interference resulting from years of prior learning. Although its preeminent (and still unexplained) role in the process of forgetting continues to be recognized, interest in the subject of PI has waned in recent years. The present article contributes a new empirical analysis of this important subject and pursues a detailed theoretical exploration into its underlying nature. In a typical PI experiment, subjects receive blocks of Brown-Peterson trials involving words from a single category (Wickens, 1972). Within a block, free-recall performance declines with each successive trial (the buildup of PI) but recovers each time a new category is introduced (release from PI). In most cases, the dependent variable used in these experiments was the percentage of correct free-recall responses. However, in the research to be presented here, we focus on latency to free recall. Research on free-recall latency in any context is very limited, and in the study of PI it is almost nonexistent. Why might free-recall latency be an interesting variable to investigate? Because such a measure provides important information about the process of retrieval that is likely to be missed by static measures, such as probability of recall. Before addressing the question of exactly what that information might be, we review the scant literature pertaining to the more general and purely empirical question of whether these