Crowding refers to the deleterious visual interaction among nearby objects. Does maximal crowding occur when objects are closest to one another in space and time? We examined how crowding depends on the spatial and temporal proximity, retinally and perceptually, between a target and flankers. Our target was a briefly flashed T-stimulus presented at 10° right of fixation (3-o’clock position). It appeared at different target-onset-to-flanker asynchronies with respect to the instant when a pair of flanking Ts, revolving around the fixation target, reached the 3-o’clock position. Observers judged the orientation of the target-T (the crowding task), or its position relative to the revolving flankers (the flash-lag task). Performance was also measured in the absence of flanker motion: target and flankers were either presented simultaneously (closest retinal temporal proximity) with different angular spatial offsets, or were presented collinearly (closest retinal spatial proximity) with different temporal onset asynchronies. We found that neither retinal nor perceptual spatial or temporal proximity could account for when maximal crowding occurred. Simulations using a model based on feed-forward interactions between sustained and transient channels in static and motion pathways, taking into account the differential response latencies, can explain the crowding functions observed under various spatio-temporal conditions between the target and flankers.