Purchase this article with an account.
Arash Fazl, Ennio Mingolla; Predicting eye movement trajectories in a multiple object tracking (MOT) task with free viewing. Journal of Vision 2008;8(6):103. doi: 10.1167/8.6.103.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
We previously reported that eye movements are similar across repetitions of the same trajectories for objects (targets and distractors) in MOT trials (Fazl, A., & Mingolla, E. (2007). [Abstract]. Journal of Vision, 7(9):902). Can we therefore predict eye movement trajectories from object paths in novel trials? In the present experiment subjects viewed two, three, or four targets among a maximum of 8 distractors in a MOT task, while their eye movements were recorded. Objects were 1° in diameter and were traveling at a constant speed of 9°/sec in a field of 30° X 30°. They bounced off the edges of the display and off each other. We found that the distance from the foveation spot to a particular target i (d_ETi) was highly correlated with a measure of that target's “clutter”, i.e. the sum of distances of all other objects to that target. The clutter scores for other targets, naturally, were negatively correlated with the magnitude of d_ETi; the more clutter around other targets, the more the foveation was attracted away from target i. We found that the eyes tended to move closer to a target that was surrounded by other targets, rather than by distractors. To quantify the relative contribution of clutter scores of targets and distractors on the d_ETi, we performed a multiple regression with d_ETi as the dependant variable. The coefficients of this regression showed that the clutter of a target by other targets is about 4 times more effective in attracting the eyes to that target as compared to clutter by distractors. Our model could predict about 70% of the variance in d_ETi for the scenes that it was trained on, and about 65% on novel scenes. Our prediction accuracy increased from two to three to four targets.
This PDF is available to Subscribers Only