Attention in the visual guidance of saccades

Pro Saccades

The original definition of attention refers to the sensitivity of perception of a sensory stimulus. When perception of a stimulus is improved reaction times of motor responses to that stimulus are often shortened.

Because one usually looks at what catches our attention (overt shift of attention), the question of how the eye finds the goal for the next saccade has often been answered by assuming that attention allocation precedes the shift in gaze direction. This leaves the question of how covert attention shifts are guided from one location to the next.

Anti Saccades

By using precues in conjunction with pro- and antisaccades it was possible to dissociate the direction of voluntary and automatic attention allocation from the direction of reflexive and voluntary saccades: a precue could indicate the direction of the next saccade either by its spatial location or by its meaning (= the saccade has to be made to the opposite side).

With prosaccades and procues one indeed obtains more express saccades and a shortening of reaction times.

However, with procues and antisaccades the reaction times become longer and the error rate increases. The subject behaves as if the precue was also treated as indicating the opposite side, eventhough it indicated the correct side. In other words: the subject is unable to use the instruction anti for the stimulus and the instruction pro for the cue.


By systematically variing the cue lead time in each case of the possible combinations of cue and saccade direction we showed that attention is not a unique mechanism of selection of saccade targets, but rather should be considered as having a voluntary and an automatic component, which have different effects on saccade generation, as listed below.

Attention Orienting
voluntary involuntary, automatic
spatially selective direction selective
sustained time course transient time course
top down bottom up
endogeneous exogeneous


Surprisingly, if the subjects were asked to indicate trial by trial whether they believed they made an error or not in an antisaccade task (gap condition) it turns out that about half of the errors and the corresponding corrections escaped the subjects' conscious perception. The subjects tell what they wanted to do, not what they really did.

Unrecognized errors are corrected faster: On trials with unrecognized errors the corrections occurred after an average of 95ms as compared to 145ms on trials with recognized errors. The reaction times of the errors were the same.

Even more surprising, when the subjects were asked to produce a sequence of saccades equivalent to an error and a corrective saccade on purpose almost all of them were unable to return the eyes from the first saccade target as quickly as they did when this first saccade happened as an unvoluntary error: a saccade made on purpose implies that the fovea spends a minimum of time (about 200 ms on average) at its goal and there is nothing the subject can do.

When the antisaccade task was combined with a dicrimination task it turned out that on trials with unrecognized errors the subjects had allocated their attention to the "wanted" side although they made saccades to the "unwanted" side. This was the first time a dissociation of saccade direction and attention direction could be demonstrated.

Author: Prof.Dr.B.Fischer
back to Express Saccade Laboratory