The determinants of random choice

The purpose of the present experiment was to describe a pattern of reinforcement sufficient to produce an unpredictable pattern of choice response by rats. On each trial two levers were inserted into an experimental chamber. If the reinforcement was always contingent upon single alternation (a simple pattern), Ss learned to alternate at significantly better than chance level; if reinforcement was contingent upon alternation on 50% of the trials (an insoluble pattern), Ss developed a position preference. To produce apparently random responding, the less preferred response (an alternation) was differentially reinforced on 75% of the trials. A simple stochastic model adequately described the results.