Does the inspector have a memory?

Abstract

Serial and parallel search strategies are distinguished and illustrated. Investigators are urged to explore when and how these strategies are used and combined in normal search rather than to focus on which one is “right”. Although serial search would be more efficient if there were a system for discouraging wasteful reinspections, we should not be embarrassed by the possibility that serial search may be amnesic. A critical review and meta-analysis of studies exploring whether visual search is amnesic leads to the conclusion that it probably rarely is. In contrast, there is ample evidence for the existence of inhibitory tags (inhibition of return) that might discourage reinspections during search. Three strategies for linking these tags to increased search efficiency are described.

Notes

¹A real-world search may include ad hoc mental processes not represented in the idealized strategies illustrated in our idealized “urn” examples. For example, when the searcher fails to find the target after one pass through the candidate items, some of these items may be rechecked. Such a strategy would artificially inflate the estimate of inspection time derived from the slope on target-absent trials.

²Throughout this paper when we refer to the “efficiency” of search we are referring to the speed and accuracy of performance on the search task under discussion. The relative efficiency of different conditions (static and dynamic) or search strategies is assessed at each set size. Some investigators use the effect of set size upon performance (usually the slope of the RT/set size function) as the measure of search efficiency. We eschew this approach primarily because this measure overlooks the real-world consequences of relative slowness (or inaccuracy). That noted, it remains the case that if the slope is affected while the intercept remains the same, then there will be an efficiency difference in both senses.

³We refer the reader to the following paper (Thomas et al., in press) of which we just became aware. Participants were foraging for fruit while being on the lookout for flashing leaves in a virtual-reallty display. Methodologically, the study is a combination of Klein's (1988) manual detection probe and Klein & MacInnes's (1999) oculomotor search tasks. The results strongly support the foraging facilitator role attributed to IOR.

⁴This is not the only explanation. Such an improvement might have been accomplished if the passive instructions increased reliance upon the parallel strategy described earlier and if that strategy turned out to be more efficient than the serial strategy for the search displays used by Smilek et al. This differential strategy–emphasis explanation and the IOR explanation described in the text are not mutually exclusive.