Abstract: Recent debates over the role of memory in visual search (Shore & Klein, 2000; Horowitz & Wolfe, 1998, 2001, 2003; McCarley et al., 2003) pose a dilemma: The emerging view of vision as ‘just-in-time’ processing (Hayhoe & Ballard, 2005) suggests that visual search is largely amnesic and routinely relies on external information. Simultaneously, it is clear that people build, and benefit from, internal representations of their visual environment.
Despite notable exceptions (e.g., Thornton & Horowitz, 2004; Chun & Jiang, 1998) laboratory investigations of visual search typically involve brief and single stimulus presentations and hence do not address the gradual and incidental construction and subsequent use of internal representations. We investigate trade-offs between external and internal information acquisition using a serial search paradigm (Neth, Gray & Myers, 2005), which requires participants to locate a series of up to 20 targets on a display of 10 stable singletons. Thus, singletons assume the functional roles of both targets and distractors throughout a trial.
Eye-tracking data was analyzed and compared to computational model predictions (ACT-R, Anderson & Lebiere, 1998). Strong evidence for memory processes supplementing visual search was found. Both repeated and novel targets were found with fewer fixations as a function of frequency, recency, and total dwell duration. As predicted by rational analysis (Anderson, 1990), the effects of memory are more pronounced as the cost for external information access increases. We conclude that memory processes operate routinely during visual search. The use of internal representations varies as a function of memory activation for both target and distractor information.