Under second-order schedules of morphine injection, high rates of responding by squirrel and rhesus monkeys were maintained when morphine was injected intravenously only at the end of each session. Every 30th key-pressing response during a 60-min interval produced a 2-s light; the first 30-response component completed after 60 min produced both the light and intravenous injection of morphine. A mean rate of approximately one response per second was maintained by doses of morphine ranging from 0.75-1.5 mg/kg. A pause in responding after each light presentation was followed by rapid responding until the light was produced again; pauses became shorter as the 60-min interval progressed. When brief light presentations were omitted, but morphine was still injected, response rates decreased and patterns of responding were altered. When saline injections were substituted for morphine injections, but the brief light was still presented, responding decreased markedly within three to five sessions and patterns of responding were altered.