Working Memory Capacity Predicts Selection and Identification Errors in Visual Search. (January 2017)

• Record Type: Journal Article
• Authors: Peltier, Chad; Becker, Mark W.
• Is Part Of: Perception. Volume 46:Number 1(2017)
• Page Start: 109
• Page End: 115
• Publication Date: 2017-01
• Subjects: visual search -- low prevalence -- working memory -- attention; Perception -- Periodicals; Perception -- Periodicals; Perception; Periodicals; 153.705
• DOI: 10.1177/0301006616678421
• Languages: English
• ISSNs: 0301-0066

Abstract

As public safety relies on the ability of professionals, such as radiologists and baggage screeners, to detect rare targets, it could be useful to identify predictors of visual search performance. Schwark, Sandry, and Dolgov found that working memory capacity (WMC) predicts hit rate and reaction time in low prevalence searches. This link was attributed to higher WMC individuals exhibiting a higher quitting threshold and increasing the probability of finding the target before terminating search in low prevalence search. These conclusions were limited based on the methods; without eye tracking, the researchers could not differentiate between an increase in accuracy due to fewer identification errors (failing to identify a fixated target), selection errors (failing to fixate a target), or a combination of both. Here, we measure WMC and correlate it with reaction time and accuracy in a visual search task. We replicate the finding that WMC predicts reaction time and hit rate. However, our analysis shows that it does so through both a reduction in selection and identification errors. The correlation between WMC and selection errors is attributable to increased quitting thresholds in those with high WMC. The correlation between WMC and identification errors is less clear, though potentially attributable to increased item inspection times in those with higher WMC. In addition, unlike Schwark and coworkers, we find that these WMC effects are fairly consistent across prevalence rates rather than being specific to low-prevalence searches.