Temporal summation and reciprocity were studied in the retinula and eccentric cells of the excised Limulus lateral eye as a function of variation in response measure: Using the latency instead of the peak of the receptor potential as a response measure produced considerably shorter critical durations. Using the area under the receptor potential as a response measure produced no critical duration up to a stimulus duration of 640 msec; instead, supersummation occurred at long durations. Similar effects were observed in the optic nerve spikes, where the response measures were first spike latency and maximum spike number sampled in time windows that ranged from 40 to 640 msec. The critical durations clearly depended on the response measure used and, when a 640-msec window was used, no critical duration occurred; supersummation again occurred. Increasing the sampling period within which maximum spike number was measured increased the critical duration and changed the formal properties from those characteristic of the receptor potential's peak to those characteristic of the receptor potential's area. The implications of the more central portions of the nervous system using different summation times for different perceptual tasks are discussed; it is suggested that the choice of response measures is crucial in studies of temporal summation.