Purchase this article with an account.
Christopher Myers, Nicole Jardine, Joseph Houpt, Andrew Howes, Richard Lewis; A Boundedly Optimal State Estimation & Control Model of Detecting Targets Among Salient Distractors. Journal of Vision 2014;14(10):88. doi: https://doi.org/10.1167/14.10.88.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
"Salient" distractors can capture attention and affect saccades, accuracy, and response speed in target-detection tasks. Salience may be based on some combination of the physical conspicuity of information in a scene and deliberate control. Recently, a model of boundedly optimal state estimation and control demonstrated saccadic bias toward the minority distractor in a distractor-ratio task without appealing to physical conspicuity/salience (Myers, Lewis, & Howes, 2013), but salience could also explain the result (Theeuwes, 1993). The present experiment was designed to determine whether saccade and manual response behavior in a singleton-distractor task could be accounted for with the same boundedly optimal model without a salience calculation. Perceptual errors in the model are caused by spatial uncertainty of features at greater eccentricities, leading to a greater likelihood of feature migration. The model optimally estimates the likelihood of each of the displays given the noisy perceptual samples (which may be corrupted by feature migration) to produce manual and oculomotor responses for each trial. Participants detected the presence of a target (red X) in a speeded search task. All displays contained two pairs at 8°, 12°, 16°, or 20° visual angle on each side of the center. Displays contained target-colored items (red O) and a conspicuous color singleton that may share the target shape (green X or O). Participants' manual responses were slower and more accurate for target-absent trials, and saccades were initiated toward the color singleton on 25%—40% of trials, depending on target presence and whether the color singleton was the target shape. The optimal model completed the same task by estimating the content of the display. The model produced RT distributions and accuracy performance similar to humans, suggesting that an explicit conspicuity calculation may not be necessary to explain behavior in simple search tasks with a "salient" singleton.
Meeting abstract presented at VSS 2014
This PDF is available to Subscribers Only