Purchase this article with an account.
Nicholas Kleene, Melchi Michel; How Should Observers Allocate Limited Transsaccadic Memory in a Visual Search Task?. Journal of Vision 2016;16(12):1065. doi: 10.1167/16.12.1065.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Our primary means of gathering information about the world is through eye-movements. Each intervening fixation provides a sample of the visual world. Due to limited visual acuity in the periphery, integrating visual samples across fixations is a critical part of visual perception, and in particular visual search. Transsaccadic memory (TSM) is necessary to store samples from previous fixations to be integrated with the current one. Therefore, we sought to place a lower bound on transsaccadic memory capacity using two visual search experiments: a real saccade experiment and a simulated saccade experiment. The task in both was to localize a target signal (Gabor) embedded within a field of 1/f filtered noise. Participants were presented with one, two or four intervals of the stimulus, with the target either present in every interval (redundancy condition) or just one (uncertainty condition). In the real saccade experiment, participants were required to make a 5° eye-movement following each sample of the stimulus; in the simulated saccade task observer's maintained fixation while we simulated the dynamic transient caused by saccades. Performance was measured as target localization accuracy. Using an ideal observer model fit to each participant's sensitivity we found higher TSM capacity estimates in the real saccade experiment compared to the simulated saccade experiment. Previously, we assumed a simple encoding strategy wherein TSM capacity was divided evenly among the individual fixation intervals and target locations (Kleene & Michel, 2015). Removing this assumption, we found that a strategy that allocates TSM according to the posterior probability at each potential target location significantly improved performance for simulated observers with small TSM capacity relative to task demands. Additionally, simulated observers with smaller TSM capacity required greater weighting of previous information to achieve optimal performance. Therefore, as task demands on TSM increase, dynamic allocation of TSM capacity becomes more beneficial.
Meeting abstract presented at VSS 2016
This PDF is available to Subscribers Only