Abstract
Background: Kirchner and Thorpe (2006 Vision Research) showed that when two scenes are simultaneously flashed left and right of fixation human observers can reliably make saccades to the side containing the target (e.g., an animal) in as little as 120 msec. It is suggested that rapid object categorization can be performed without conscious recognition using highly automatic and data driven routines. This question was addressed in two patients with visual agnosia. Method: two agnostic patients and age-matched control participants were presented with photographs of natural scenes or isolated objects. Two scenes, one containing the target (an animal) were simultaneously displayed for 3 sec. In separate sessions participants were instructed either to locate the target with a keypress or to make a saccade toward the target. Results: Patients, who identified very few objects with unlimited exposure time were able to detect the target with 70% accuracy and saccade latencies of 250–300 ms in the categorization task. Performance was better with saccades than with keypress. For all observers accuracy was equivalent for objects in scenes and isolated objects but RTs were shorter for scenes . Conclusion: The results support the hypothesis that rapid categorization can be accomplished without conscious recognition.