The Perception of Involuntary Prosaccades in an Antisaccade Task

In an antisaccade task subjects are required to generate a voluntary saccade to the side opposite to a small visual stimulus. With fixation-point offset preceding stimulus onset (gap) subjects produce some involuntary saccades to the stimulus and correct them by a second saccade. We wanted to know whether the subjects recognised their errors and whether a recognised sequence (error followed by correction) is different from an unrecognised sequence. To test the access to the correction mechanism, subjects were asked in subsequent experiments to produce the error-correction sequence voluntarily (voluntary sequence).

We used the gap = 200 ms condition. A valid cue was presented 100 ms before stimulus onset. This manipulation increased the error rate (Fischer and Weber, 1996 Experimental Brain Research 109 507 – 512). Subjects indicated errors by key-press. The rate of recognised and unrecognised errors, saccadic size, reaction times (SRT), and correction times (CRT) were determined.

Altogether 93 data sets (400 trials each) from 38 subjects were analysed. The mean error rate was 20%, of which 62% went unrecognised. In sessions with high error rates the fraction of unrecognised errors was high. The SRT of the errors ranged from 80 to 170 ms with a strong mode of express saccades at 100 ms. Both types of errors had the same mean SRT of 117 – 119 ms. The unrecognised errors were 0.4 deg smaller. They were corrected after a mean CRT of 95 ms. The recognised errors were corrected after 127 ms; in the voluntary sequence the correction occurred after 217 ms. The CRT distributions differ from each other with the unrecognised errors having an extra peak around 45 ms, suggesting different modes of correction, to which perception has different access. These results raise the question why the large and long-lasting changes of the retinal image escape the conscious perception so often.