Spatial representations for programming saccadic eye movements

We often initiate movements to objects which are not currently visible, such as reaching for a plate which is stored behind an opaque cabinet door. What is the nature of spatial representation which enables these movements to remembered locations? We asked human subjects to make eye movements, after a delay of a few seconds, to the remembered locations of targets presented briefly, alone in darkness. In the absence of landmarks such a target can only be localized in an egocentric reference frame (relative to some part of the body). We tested whether this egocentric reference frame is retina- or head-centered by studying the precision of the targeting saccades after a varying number of intervening saccades. A retina-centered frame would require updating to account for intervening eye movements during the memory period which would cause decreased targeting precision. The data support a head-centered representation which uses an eye position signal with a standard deviation of less than 1.4$\sp\circ$. The results also replicate the recent finding that relationships between visible objects (exocentric representations) aid movement precision.

Might such representations play a role in guiding movements even to visible targets? We tested this in a more natural visual-motor task which involved copying random patterns of colored blocks on a computer display. Eye movements revealed fixation near a previously-placed neighboring block during each additional block placement. This natural, consistent fixation allowed the previously-placed neighboring block to serve as an implicit target, without instructions to the subject. Monitoring eye movements and updating the display quickly, we removed target blocks during the saccade preceding the targeting saccade. Disappearance and reappearance of target blocks made them invisible during the fixation preceding the targeting saccade. Subjects were usually unaware of display changes. Precision degraded slightly when saccades were launched without a visible target suggesting that we use current visual information to plan saccades. However, minimal disruption when forced to direct the eyes to a location without a visible target indicates that other representations must be a natural part of saccadic programming.

Both sets of experiments support a model of movement planning that relies on combining multiple representations.