Johan Hulleman¹ (), Katerina Sergi¹, Jeremy M. Wolfe^2,3; ¹University of Manchester, ²Brigham and Women's Hospital, ³Harvard Medical School

Visual searchers miss targets. Every search task produces stochastic errors, where searchers miss a target due to random bad luck. Some search tasks also produce deterministic errors, where properties of specific scenes or specific stimuli make the target hard to find. If a search display is presented twice, a deterministic error would be likely to occur both times. The chance of a stochastic error occurring twice would simply be the square of the probability of missing once. Do properties of individual targets or of the search display as a whole cause deterministic errors? We investigated by having 22 participants search for Ts amongst Ls, with zero, one, or two targets. Critically, each 24-item display was presented twice (576 trials). Ts in one-target displays were either high or low contrast. In two-target displays, one T was always high contrast; the other T was either high (high-high) or low contrast (high-low). Eye movements were tracked. For high-low trials, 17.8% of search displays yielded a miss error on both presentations. For high-high trials, this only occurred for 4.8% of displays. The double miss rate for high-high is consistent with purely stochastic misses, but the elevated high-low rate indicates a substantial deterministic component. If participants miss a target on first presentation of a high-low trial, which target do they miss -if they miss- on second presentation? Cases where the missed target changed from high to low or vice versa were in line with the stochastic prediction (4.9% vs. 4%). In contrast, double high and, especially, double low misses occurred almost twice as often as predicted stochastically (12.8% vs. 6.8%). So, the deterministic component of miss errors in high-low trials arises from double misses of the same target. In these tasks, it is the camouflaged item, rather than the camouflaging scene that causes miss errors.

Acknowledgements: JH and KS were supported by UKRI grant ES/X000443/1. JMW was supported by NIH-NEI: EY017001, NSF: 2146617, and NIH-NCI: CA207490.