Search

Animals choose suboptimally when provided with cues that signal whether reinforcement is coming or not. For example, pigeons do not prefer an alternative that always provides them with a signal for reinforcement over an alternative that provides them with a signal for reinforcement only half of the time and a signal for the absence of reinforcement the rest of the time. In the present research, we tested the hypothesis that if the results of the choice are delayed, pigeons will choose less suboptimally. We tested this hypothesis by forcing pigeons to wait following their choice, requiring them to complete a fixed-interval 20-s schedule prior to receiving the signals for reinforcement. In Experiment 1, we gave the pigeons a choice between (a) a 50% chance of receiving a signal for reinforcement or a 50% chance of receiving a signal for the absence of reinforcement and (b) a 100% chance of receiving a signal for reinforcement. When the signal for reinforcement was delayed, most of the pigeons chose optimally. When it was not delayed, most of the pigeons chose suboptimally. In Experiment 2, we gave the pigeons a choice between (a) a 25% chance of receiving a signal for reinforcement or a 75% chance of receiving a signal for nonreinforcement and (b) a 100% chance of receiving an unreliable signal for reinforcement (predicting reinforcement 75% of the time). When the signal was not delayed, the pigeons showed a strong tendency to choose suboptimally but they chose suboptimally much less when the signal was delayed.

Pigeons show relatively poor memory when trained on 0-s delayed matching and tested with longer delays. We hypothesized that one reason for the effect of delay may be a loss of association between sample responding and reinforcement. To test this hypothesis, we compared the effect of standard delays inserted between the offset of the sample and the onset of the comparison stimuli with similar delays inserted between the comparison response and reinforcement. We also manipulated whether the delay was dark or lit because there is some evidence that lit delays are disruptive but filled intervals may also help bridge delays. In Experiment 1, pigeons were trained with a 0-s delay and were tested with longer delays. When both delays and delay illumination were novel, we found no effect of delay of reinforcement and only a small effect of delay illumination. In Experiment 2, to eliminate the effect of the novelty of the delay and delay illumination, pigeons were trained with delays from the start. Results indicated once again that delay of reinforcement had little effect on matching accuracy. However, in Experiment 2, delay illumination produced a general decline in matching accuracy and when it preceded comparison choice its effect increased with increasing delay. The results indicate that the rate of forgetting seen in the typical delayed matching experiment is not due to delay of reinforcement for attention to the sample. The results also confirm that lighting the delay after the sample is detrimental to delayed matching. Surprisingly perhaps, lighting the delay after the comparison response is also detrimental to delayed matching. Finally, it was concluded that a more accurate assessment of the rate of forgetting in delayed matching can be obtained by training with variable delays from the outset.