Previously, we have demonstrated that a subpopulation of visual cortical neurons exhibit oscillatory responses to their preferred stimuli at a frequency near 50 Hz (Gray & Singer, 1989). These responses can selectively synchronize over large distances of cortex in a stimulus-specific manner (Gray et al., 1989; Engel et al., 1990 alpha). Here we report the results of a new analysis which reveals the fine temporal structure inherent in these interactions. We utilized pairs of recordings of the local field potential (LFP) activity from area 17 in the anesthetized cat which met two criteria. The LFP was correlated with the underlying unit activity at each site and the recording sites were at least 5 mm apart in cortex. A moving-window technique was applied to compute cross correlograms on 100-ms epochs of data repeated at intervals of 30 ms for a period of 3 s during each direction of stimulus movement. A statistical test was devised to determine the significance of detected correlations. In this way we were able to determine the magnitude, phase difference, frequency, and duration of correlated oscillations as a function of time. The results demonstrate that (1) the duration of synchrony is variable and lasts from 100-900 ms; (2) the phase differences between and the frequencies of synchronized responses are also variable within and between events and range from +3 to -3 ms and 40-60 Hz, respectively; and (3) multiple correlation events often occur within a single stimulus period. These results demonstrate a high degree of dynamic variability and a rapid onset and offset of synchrony among interacting populations of neurons which is consistent with the requirements of a mechanism for feature integration.