EFFECTS OF PERCENT REINFORCEMENT ON PIGEON'S ACCURACY IN MIDSESSION REVERSAL TASKS [University of Kentucky]

Abstract

This experiment tests the effects of a reducing the value of one stimulus compared to another. We hypothesize that in a midsession reversal task, accuracy would improve by devaluing choice of S2 relative to S1. A midsession reversal task requires a simultaneous discrimination between 2 stimuli (S1, S2), during which, halfway through the session the reinforcing stimulus is reversed. A current theory, win-stay/lose-shift, is a strategy for optimal choices in this task. In this strategy, an animal would continue to repeat choice of a correct stimulus, but once the stimulus becomes incorrect, the animal should switch to the other stimulus. In this experiment, ten unsexed White Carneau pigeons were used as subjects. The experiment took place in an operant test chamber. Inside, three horizontally aligned circular response keys are attached to a response panel. The keys are lit by mounted projectors showing red and green hues. Reinforcement consisted of 1.5 seconds of access to mixed-grain from a feeder. In the experimental group, all correct responses to S1 were reinforced but only a random 20% of the correct responses to S2 were reinforced. Our findings suggest that paradoxically, the reduction in the probability of reinforcement for correct S2 responses from 100% to 20% had a net positive effect on correct responses. There was a significant reduction in errors prior to the feedback from the reversal as well as no increase in errors following the feedback from the reversal. Analysis of the trials on either side of the reversal made it clear that there was a net benefit in percentage correct due to the decrease in the probability of reinforcement associated with correct responses to S2. This study suggests the negative impacts of too much reinforcement on subjects performance in a task.