Limits of behavioral control by temporally extended response -reinforcer relations

Three experiments were performed to determine the extent to which the behavior of rats can be controlled by response-reinforcer relations that are extended in time. In Experiment 1, bonus pellets delivered at the end of the session were contingent upon a shift in choice responding within the session. Experiment 2 examined control of aggregated responses by a delayed consequence over a much shorter time period than an entire session. The reinforcing efficacy of bonus pellets was assessed using a chained-schedule procedure. The relation between aggregated responses and a delayed reinforcing consequence was assessed several times per session and with shorter delays than in Experiment 1. Experiment 3 used an adjusting-delay procedure to assess whether differential reinforcer magnitudes have a differential effect on choice behavior when the delay between choice and subsequent reinforcement is equal for the two alternatives. The experiment was designed to determine the longest delay at which differential reinforcement is effective. Taken together, these three experiments were designed to determine the extent to which aggregated responses may be controlled by aggregated reinforcers or a single reinforcing event, and the extent to which a single response may be reinforced by its delayed consequence. Experiment 1 failed to produce reliable control of choice responding by the post-session consequence. Experiment 2 established control of responding by the delayed reinforcer, but such control was reliable for all rats only at delays of 40 s and less. Experiment 3 was unsuccessful in establishing discriminated choice performance by the large reinforcer, even at short delays, preventing the determination of the temporal limit of control by differential reinforcer magnitude. Overall, the results of this series of experiments suggest that the operant behavior of rats can be controlled by delayed consequences, but a finite limit to such control exists. It seems that reinforcers delayed on the order of several minutes or more are unlikely to control the behavior that produces them. Thus, response-reinforcer contiguity determines whether response-reinforcer correlations can control behavior