In visual areas of the cerebral cortex, most neurons exhibit preferences for particular features of visual stimuli, but in general, the tuning is broad. Thus, even simple stimuli evoke responses in numerous neurons with differing but overlapping feature preferences, and it is commonly held that a particular feature is encoded in the pattern of graded responses of the activated population rather than in the optimal responses of individual cells. To decipher this population code, responses evoked by a particular stimulus need to be identified and bound together for further joint processing and must not be confounded with responses to other, nearby stimuli. Such selection of related responses could be achieved by synchronizing the respective discharges at a time scale of milliseconds, as this would selectively and jointly enhance their saliency. This hypothesis predicts that a given set of neurons should exhibit synchronized discharges more often when responding to a single stimulus than when activated by different but simultaneously presented stimuli. To test this prediction, recordings were performed with two electrodes from spatially segregated cells in the middle temporal area (MT) of the awake behaving macaque monkey. It was found that cells with overlapping receptive fields, but different preferences for directions of motion, can engage in synchronous activity if they are stimulated with a single moving bar. In contrast, if the same cells are activated with two different bars, each moving in the direction preferred by the cells at the two respective sites, responses show no or much fewer synchronous epochs. Control experiments exclude that this effect is attributable to changes in response amplitude, the mere presence of two stimuli, or the specific orientation of the bars. The critical variable determining the strength of correlation is the extent to which both sites are activated by a common stimulus or by two different stimuli with different directions of motion.