Daily rhythms in a complex operant: targeted percentile shaping of run lengths in rats

Daily rhythms in response output and accuracy were examined when reinforcement for a complex operant was uncoupled from accuracy of performance. Rats housed in operant conditioning chambers earned their daily ration of food under a targeted percentile procedure for responding on two levers. The targeted pattern was a series of consecutive responses on the left lever (a "run"), followed by a single response on the right lever. The targeted run length was either "O" (i.e., undefined, under the nondifferential baseline), 6, 12 or 24. Under baseline, a random third of all trials ended in pellet delivery; under the percentile conditions, trials with runs closer to the target than two-thirds of the runs from the most recent 24 trials ended in pellet delivery. This contingency shaped run lengths while ensuring that approximately one-third of all trials produced pellets. Responding tracked the target value well, with mean obtained run lengths equal to 90% of the target or better. Daily rhythms were clearly evident in measures of overall output, with subjects responding primarily 3-7 h into the dark period. The only substantial light-period responding observed in all subjects occurred during the 2 h after noon, when the chambers were serviced. No systematic variation in this pattern was observed as a function of target. Run length was much less variable across the daily cycle than was response output, with only a suggestion under the longest target that response accuracy was lower during periods removed from the period of peak activity.