PURPOSE. When an observer searches for a known target, can the oculomotor system use shape differences between the target and distractors to guide search saccades?

METHODS. Four observers searched for a target letter (D, U, or X) among distractors (O in the discrimination task, nothing in the detection task), with the target and distractors shown at several contrast levels spanning the observers’ psychometric functions. We measured the accuracy of observers’ first saccadic decisions on each trial. We also measured the accuracy of observers’ perceptual (i.e., keypress) responses in separate runs, with the stimulus duration approximately matched to the time required to program a saccade (~150 ms), thereby providing the same amount of task-relevant information while precluding any actual saccades. We quantified saccadic decision accuracy by calculating the relative efficiency of saccadic and perceptual decisions, defined as (d'_saccade/d'_perception)^2.

RESULTS. The relative efficiency of saccadic decisions was high in the letter detection task (mean ± SD across observers: 60% ± 20%), and much lower but nevertheless greater than zero in the letter discrimination task (16% ± 11%). Furthermore, when we randomized the contrast of the target and distractors, thereby rendering local contrast cues uninformative, the relative efficiency of saccades in the letter discrimination task did not decline (44% ± 8%). This control experiment demonstrates that in the letter discrimination task, saccades actually were guided by shape differences between the target and distractor letters.

CONCLUSION. The saccadic targetting system can use information about shape differences between objects, but less efficiently than the perceptual system can.