In many discrimination-learning tests, spatial separation between stimuli and response loci disrupts performance in rhesus macaques. However, monkeys are unaffected by such stimulus-response spatial discontiguity when responses occur through joystick-based computerized movement of a cursor. To examine this discrepancy, five monkeys were tested on a learning-set task that required them to touch computer-graphic "levers" (which differed in location across experimental phases) with a cursor in order to select an associated test stimulus. The task produced both first-order (joystick and lever) and second-order (lever and stimuli) spatial discontiguity between the stimuli to be discriminated and the discriminative response. Performance was significantly better than chance for all lever locations including locations in which selection of the correct lever required moving the cursor away from the positive stimulus. Thus, rhesus macaques do not attend simply to the region around the cursor in these computerized tests, but rather they attend to relevant stimulus loci even when these are discontiguous with response and reward areas. (Contains 4 figures.)