Visual cortical areas are richly but selectively connected by "patchy" projections. We characterized these connections physiologically with cross-correlograms (CCHs), calculated for neuron pairs or small groups located one each in visual areas 17 and 18 of the cat. The CCHs were then compared to the visuotopic and orientation match of the neurons' receptive fields (RFs). For both spontaneous and visually driven activity, most non-flat correlograms were centered; i.e. the most likely temporal relationship between spikes in the two areas is a synchronous one. Although spikes are most likely to occur simultaneously, area 17 spikes may occur before area 18 or vice versa, giving the cross-correlogram peak a finite width (temporal dispersion). Cross-correlograms fell into one of three groups according to their full-width at half peak height: 1-8 ms (modal width, 3 ms), 15-65 ms (modal width 30 ms), or 100-1000 ms (modal width 400 ms). These classificatory groups are nonoverlapping; the three types of coupling appeared singly and in combination. Neurons whose receptive fields (RFs) are nonoverlapping or cross-oriented may yet be coupled, but the coupling is more likely to be the broadest type of coupling than the medium-dispersed type. The sharpest type of coupling is found exclusively between neurons with at least partially overlapping RFs and mostly between neurons whose stimulus orientation preferences matched to within 22.5 deg. The maximum spatial dispersion observed in the RFs of coupled neurons compares well with the maximum divergence seen anatomically in the A18/A17 projection system. We suggest three different mechanisms to produce each of the three different degrees of observed spatial and temporal coherence. All mechanisms use common input of cortical origin. For medium and broad coupling, this common input arises from cell assemblies split between both sides of the 17/18 projection system, but acting synchronously. Such distributed common-input cell assemblies are a means of overcoming sparse connectivity and achieving synaptic transmission in the pyramidal network.