The role of working memory and long-term memory in visual search

Abstract

Models of attentional deployment in visual search commonly specify that the short-term, or working memory, system plays a central role in biasing attention mechanisms to select task relevant information. In contrast, the role of long-term memory in guiding search is rarely articulated. Our review of recent studies calls for the need to revisit how existing models explain the role of working memory and long-term memory in search. First, the role of working memory in guiding attentional selection and search is much more complex than many current theories propose. Second, both explicit and implicit long-term memory representations have such clear influences on visual search performance that they deserve more prominent treatment in theoretical models. These new findings in the literature should stir the conception of new models of visual search.

Acknowledgments

We thank Steve Luck, Gordon Logan, Andrew Rossi, and Andrew Hollingworth for valuable discussions regarding the issues addressed in this paper. GFW is supported by an individual NRSA from the National Institute of Health (F32 EY015043) and MMC is supported by a grant from the National Eye Institute (R01 EY014193).

Notes

¹Because visual search tasks are typically comprised of discrete trials, it is useful to distinguish different roles of memory according to how memory influences performance within or across trials. We assume that within-trial effects are best subserved by working memory that has limited-capacity and requires active maintenance to perform a task at hand. Because target and distractor locations typically change unpredictably from trial to trial, the system should reset itself on each new trial to minimize debilitating proactive interference. Such resetting is naturally performed by the working memory system, according to theories of memory and models of visual search. Across-trial influences appear to be best explained by LTM processes that have larger capacity and less susceptibility to interference and erasure, serving to extract useful regularities that may occur over time. Although we will discuss this distinction in more detail later, we acknowledge that the distinction between working memory and LTM is a simplification, and in fact, we will conclude that visual search benefits from both working memory and long-term memory systems.

²The majority of evidence supports a distinction between spatial and object working memory stores, but this does not mean that object and spatial working memory representations cannot be linked (e.g., Jiang, Olson, & Chun, 2000; Rao, Rainer, & Miller, 1997). In fact, it is likely that one role of attention is to bind information across such separate working memory stores (e.g., Wheeler & Treisman, 2002).