Abstract
The mechanism for controlling saccadic timing is not well understood. One possibility is that the eye stays fixated at a given location until the current target is analyzed, while another is that an internal timer is set on the basis of context, with minimal reference to the currently fixated target.
To distinguish these alternatives, subjects estimated the mean location of a cluster of steadily-appearing dots. Dot position was sampled from a Gaussian distribution (SD= 20′) whose mean was to the right or left of a reference line. Dots appeared at a rate of 1 every 53 ms while the reference line was fixated. Task difficulty was modulated by changing the spatial offset between the mean of the distribution and the reference line (offsets = 5′, 7′, 10′). Subjects were instructed to adjust viewing time to accumulate enough dots to achieve 95% accuracy. When task difficulty remained constant throughout an experimental session, viewing time per cluster increased with difficulty level (time/cluster ranged from ∼ 1-1.5 s). Performance approximated that of an ideal observer that was limited only by statistical fluctuations in sampling of dot location. However, performance in the most difficult case (offset = 5′) was only about 85% correct, showing a reluctance to prolong viewing time to improve accuracy. When task difficulties were mixed within a single experimental session, viewing time decreased for the harder difficulty levels and increased for the easier levels. Some subjects maintained saccade-free fixation on a cluster for the entire viewing time, while others made small saccades at a relatively constant rate.
These results show that subjects can adopt optimal viewing times (consistent with the ideal observer), but there are also modulations in viewing time reflecting both a preference to avoid prolonged fixation, and an influence of recent past history and task context.