ASSESSING THE EXISTENCE OF MENTAL ROTATION IN AN OBJECT-BASED SACCADE TASK.

P. N. Sabes^1,2, B. Breznen¹, and R. A. Andersen¹

¹Biology Division, 216-76 Caltech, Pasadena, CA 91125, U.S.A.,
²Sloan Center for Theoretical Neurobiology, Salk Institute, La Jolla, CA 92037, U.S.A.

A macaque monkey was trained to saccade to prespecified locations on a 2D object which rotated in space, as described in the accompanying abstract. One strategy for computing the movement vector would be to match the redisplayed object with the original object by mentally rotating the image. Thus we tested whether mental rotation was used to saccade to a remembered location on a rotated object. In one experiment, the object reappeared at any one of eight equally spaced orientations. Errors were not dependent on the degree of rotation. And while the reaction time covaried significantly with absolute angle of rotation, the values differed by only 30msec over the range of rotations, with a slope of 0.15 msec/deg. Neither of these results is consistent with a mental rotation strategy for this task. In a second experiment, the object was always rotated +/-135 degrees. Here, we recorded single neurons in the posterior parietal cortex of the monkey. Responses were analysed using a population vector approach. If the coded movement plan swept between the cue and saccade angles (rotation), then cells with preferred directions between those locations should become preferentially active at some point during the reaction time. We saw no such increase in activity. On the other hand, if the code for the cue is phased out while the movement direction code is generated (replacement), the coherence of the population vector estimate should drop during the reaction period, as observed. We conclude that the monkey did not use a mental rotation strategy for this particular Task. This work was supported by the NEI and the Sloan Foundation.