In a series of three experiments the effects of variation in grain duration on automaintenance were evaluated. In the first experiment, key illumination was followed by grain only when pigeons did not peck the key. Each subject was exposed to 2‐, 4‐, and 8‐second feeder durations in blocks of 10 sessions. Subjects pecked on a high percentage of trials at all feeder durations. The mean peck latency was shorter in the 8‐second condition than in the two other conditions in five of six subjects. The conditional probability of pecking given successive keylight‐grain pairings did not increase as the number of pairings increased. The second experiment was identical to the first, except that key pecking had no scheduled consequence. Under these conditions, all three subjects showed substantial responding. The recorded measures showed no systematic relationship to feeder duration in this study. In the third experiment, two different stimuli were followed by feeder presentations of either identical (2‐ or 8‐second) or different (2‐ and 8‐second) durations within each session. Subjects tended to respond sooner and with a higher overall rate in the presence of the stimulus associated with the longer feeder duration only when different feeder durations were presented within the same session. This result was confirmed by direct observation of the pigeons. The results of these experiments suggest that the effects of varying grain duration may be small, compared to the effects of varying other variables. The results also suggest that the location as well as the frequency of pecking may be an important measure in the analysis of factors controlling the