The use of recurrent signals about adaptation for subsequent saccade programming depends on object structure.

Executing sequences of accurate saccadic eye movements supposes the use of signals carrying information about the first saccade for updating the predetermined motor plan of the subsequent saccades. The present study examines the signals used in planning a second saccade when subjects made two successive saccades towards one long or two short peripheral objects displayed before the first saccade execution. Different first eye movement signals could be used: desired eye movement signals, representing the movement necessary for attaining the intended target, or actual eye movement signals, representing the movement actually executed. Experimental dissociation of desired and actual eye movement signals is made possible by adaptive modifications of the first saccade, obtained by transfer of single saccade adaptation, during which the motor vector was progressively modified in response to the systematic intra-saccadic step of a single target. Whether the second saccade used the actual eye movement signal to compensate or not for the adaptive changes in the first saccade depended on which object properties were relevant for saccade planning. Compensation was observed for saccades that aimed for a new object (between-object saccades) because adaptation modifies relative object location. No compensation was observed for saccades that explored an extended object (within-object saccades). Implications for the on-line control of subsequent eye movements are discussed.