Abstract
In visual search tasks, observers look for a target among some number of distractors. Civilization has created demanding and important search tasks like airport baggage screening and routine mammography. Can we use performance on laboratory search tasks to identify people who would be successful if employed in these critical real world tasks? In the current study, 20 observers (14 women) performed a battery of search tasks. The battery included one conjunction task (red vertical target among red horizontal and green vertical and horizontal distractors), two spatial configuration tasks (T among Ls and 2 among 5s), one search for arbitrary objects in photographs of indoor scenes, and a simulated x-ray baggage-screening task (two conditions: targets appear frequently or targets appear rarely). We used corrected reaction time (RT/d', Townsend & Ashby, 1983) as an index of performance. For a given search task, reliability was high. We conducted test-retest reliability for two of the 5 tasks: r = .76 on the T among Ls task, .55 on the rare target baggage-screening task, and .72 on the frequent target baggage task. Split-half reliability ranged from .83 on rare target baggage search to .94 on 2 among 5s. Rather surprisingly, however, correlations between performance on one task and performance on another were generally quite low. Only the conjunction task correlated with the frequent target version of the baggage-screening task (r=.63, p[[lt]]0.01), while only the 2 among 5s task correlated significantly with the more ecologically valid rare target version (r=.67, p[[lt]]0.01). The high reliability scores suggest that we had sufficient power to detect correlations if they were present. However, in general, performance on one task fails to predict performance on another for this set of tasks.
Research funded by NIH and DHS