Abstract
Saccade latency - the time between target onset and the moment the saccade begins - varies greatly even in situations where other affecting factors are controlled for. This is an important issue when we consider ‘when’ and ‘where’ processes in saccade programming. In the case of moving targets, one can argue that ‘when’ processing (how long the latency will be on a given trial) will dictate the ‘where’ outcome - longer latencies result in greater target displacement, which requires a different saccade endpoint. Previously, we have shown that in situations where observers are unable to predict the onset of a moving target, they are still able to generate accurate saccades. Here, we investigate how positional information about moving targets is integrated in the run-up to a saccade being generated.
Participants were presented with two Gaussian patches moving at 18°/s. The contrast of one of the patches increased unpredictably, indicating the saccade target. In 50% of the trials, the patches were subject to a step-change to either a faster (30°/s) or slower (6°/s) speed at a variable interval after the contrast change. We find evidence of a surprisingly long dead time prior to each saccade of some 100–150 ms during which no new motion information is acquired. Preceding this there appears to be a period of motion integration of some 200 ms upon which the speed estimate determining saccade landing position is based. These results are inconsistent with the saccadic system taking a last-minute snapshot of target position and speed prior to the dead time, the latter presumably reflecting the operation of saccadic motor programming.