The insistence of vision: Why do people look at a salient stimulus when it signals target absence?

Abstract

Researchers and practitioners across many fields would benefit from the ability to predict human search time in complex visual displays. However, a missing element in our ability to predict search time is our ability to quantify the exogenous attraction of visual objects in terms of their impact on search time. The current work represents an initial step in this direction. We present two experiments using a quadrant search task to investigate how exogenous and endogenous factors influence human visual search. In Experiment 1, we measure the oculomotor capture—or the tendency of a stimulus to elicit a saccade—of a salient quadrant under conditions in which the salient quadrant does not predict target location. Despite the irrelevance of quadrant salience, we find that subjects persist in making saccades towards the salient quadrant at above-chance levels. We then present a Bayesian-based ideal performer model that predicts search time and oculomotor capture when the salient quadrant never contains the search target. Experiment 2 tested the predictions of the ideal performer model and revealed human performance to be in close correspondence with the model. We conclude that, in our speeded search task, the influence of an exogenous attractor on saccades can be quantified in terms of search time costs and, when these costs are considered, both search time and search behaviour reflect a boundedly optimal adaptation to the cost structure of the environment.

Keywords:

• Visual search
• Exogenous
• Endogenous
• Ideal performer model
• Saccadic reaction time

Acknowledgements

The authors would like to thank Brett R. Fajen, Michael J. Schoelles, and members of the CogWorks Laboratory for their comments on earlier versions of this report. Support for this work was provided in part by grants to Wayne D. Gray from the Air Force Office of Scientific Research (No. F49620-03-1-0143) as well as the Office of Naval Research (No. N000140310046). Portions of this research were reported at the 26th annual meeting of the Cognitive Science Society, Chicago, IL, 2004.

Notes

¹We thank an anonymous reviewer for pointing this out and suggesting this comparison.

²Note there are four competing actions, one pro-SQ strategy, and three instances of non-SQ, corresponding to each of the three nonsalient quadrants.

³Note our switch in terminology from non-SQ to anti-SQ. When initial saccades are under the control of exogenous forces we refer to saccades to nonsalient quadrants as non-SQ. In contrast, when an adopted strategy or statistical adaptation dictates the salient quadrants be avoided, we refer to this as an anti-SQ effect, similar to the antisaccade effect.