Abstract
Transsaccadic integration is the process whereby we combine information across visual fixations and, as such, plays an important role in many visual tasks including visual search. Because information from previous fixations must be preserved to allow integration with the current fixation, a necessary component of transsaccadic integration is transsaccadic memory (TSM). We developed an ideal observer model of transsaccadic integration that takes into account the critical role of TSM and used this model to estimate a lower bound on TSM capacity in a simple visual search task. Prior to the main experiment, subjects completed a forced-choice detection experiment, allowing us to measure their psychometric functions at each potential target location. Subjects then performed a visual search task that required localizing a target signal in noise. In this initial study, we used a single-fixation paradigm to control for variations in visual sensitivity across the visual field. We simulated successive visual fixations using multiple intervals, and saccades using visual 'sweeps'. In each block subjects were presented with one, two, or four intervals of 1/f noise. A Gabor target was present in every fixation interval (redundancy condition), or only one (uncertainty condition). Performance was measured as the proportion of trials in which subjects correctly localized the target. The ideal observer model allowed us to place a lower bound on each subject's TSM capacity, quantified in terms of the minimum amount of memory (in bits) required to explain the search performance obtained for that subject. Across subjects, this estimated lower bound on TSM capacity was about 5.5 bits, with bounds for all subjects increasing as a function of the number of fixation intervals. We discuss the relationship of our obtained bounds to previous estimates of transsaccadic memory and visual short term memory and to existing accounts of the role of visual memory in search.
Meeting abstract presented at VSS 2014