Abstract
Visual information presented during saccadic preparatory time does not influence the saccade destination. Less is known about how visual information is integrated prior to the saccade preparatory time. Here, we use a search task for a target added to a time sequence of independent samples of spatially uncorrelated noise (i.e., spatiotemporal noise) to examine how the information used to choose where to make saccades accumulates over time.
We recorded eye movements of two observers in a 5 alternatives force choice (AFC) search task. On each trial a sequence of frames was presented at 40 Hz for 500 ms. The observer's task was to saccade to a bright Gaussian target among 4 dim Gaussian distractors (SNR of a single frame was 2.8). The target and the distractors were placed within boxes equidistant along an imaginary circle with diameter of 6.4 deg. We considered the saccadic decision to be correct if its endpoint was closest to the target location. Performance (d′) of the 1st search saccade was measured as a function of the number of frames presented before the onset of the saccade (number of frames = saccadic latency / duration of a frame).
For short latencies (< 200 ms.) observers made few saccades and accuracy was low (d′ < 1.0). For medium latencies (200–325 ms) observers made many saccades and d′ increased linearly as a function of the square root of the number of noise samples. For long latencies (> 325 ms), observers made few saccades and d′ was approximately constant (d′ = 2.3).
An ideal observer integrates information from all available frames predicting that performance (d′) should increase as the square root of the number of frames presented prior to the saccade. In contrast, human observers appear to only accumulate information from frames within a brief temporal window prior to the saccadic preparatory time.