A key was illuminated on the average of every 30 sec for a duration of 6 sec and this was followed by food presentations. When key pecks in the presence of the light produced immediate access to grain (autoshaping procedure) pigeons were likely to peck. When pecks terminated the keylight but prevented access to grain (automaintenance procedure) pigeons were much less likely to peck. Seven of 12 pigeons failed to develop responding during the automaintenance procedure. Four of the five pigeons that responded during the automaintenance procedure were exposed to a procedure in which responses could not immediately terminate the light. Three of the four ceased to respond during optimal automaintenance conditions, suggesting that the response-dependent offset of the keylight had been reinforcing their pecking. Responding during the automaintenance procedure was eliminated for a fifth pigeon by eliminating the contiguity of light-offset and food-onset on those trials in which the pigeon did not peck. These results suggest that: (1) automaintenance (unlike autoshaping) is not an effective procedure for reliably generating responding; (2) responding that does occur during the automaintenance procedure is reinforced by the response-dependent offset of the keylight.