Abstract
Saccades enable primates to sample their visual environments at high resolution. Yet, even before the saccade occurs, visual discrimination sensitivity increases at the saccade target location (Deubel, 1996). Does this benefit occur in all scenarios? To address this question, we designed a dual-task paradigm involving visual change detection. On each trial, participants (n=10) prepared and executed a saccade to one (cued) of four grating locations while concurrently localizing a change in grating orientation that immediately preceded the saccade. Surprisingly, we found no improvement in orientation change detection sensitivity at the saccade target (ST) location relative to the other locations: while the proportion of hits increased marginally, the proportion of false-alarms increased significantly. To investigate this result further, we performed another experiment in which participants (n=10), presaccadically, estimated the precise orientation of one of four gratings. Here, presaccadic precision of orientation estimates was indeed higher for ST gratings, as expected. Yet, on flashing a second set of gratings (oriented completely differently), immediately after the first set, the benefit in presaccadic precision shifted to the more recent (second) set and, surprisingly, vanished for the original (first) set. We modeled these behavioral findings with a Bayesian ideal observer model: the variable precision (VP) model (Keshvari et-al., 2012). Change detection behavior was best fit by a VP model that accounted not only for differences in precision across locations but, interestingly, also incorporated a different (lower) detection criterion at the ST location. In summary, presaccadic benefits on sensory precision occurred only for the most recent of multiple, sequential stimuli, suggesting that neural populations encode precision dynamically at the prioritized saccade target location. Moreover, saccade preparation lowered change detection criteria and increased false-alarms at the saccade target location. Our results show, for the first time, that presaccadic benefits do not extend to visual change detection scenarios.