The impact of reinforcement density on response differentiation in configural discrimination problems

Abstract

Two human Pavlovian conditioning experiments investigated the impact of reinforcement density (the number of reinforced trials divided by the total number of trials) on discrimination learning. Experiment 1 used a negative patterning problem (A+, B+, AB−) and Experiment 2 used a positive patterning problem (A−, B−, AB+). In both experiments, reinforcement density varied across four levels. Response differentiation between reinforced and nonreinforced stimuli was linearly related to the decrease in reinforcement density. Responses to nonreinforced stimuli did not differ between the four groups in either experiment. In contrast to this, responses to reinforced stimuli were considerably more pronounced in conditions with lower reinforcement density. For negative patterning, this replicates and extends similar observations from other species. For positive patterning, this is a finding that has not yet been reported in other experiments. The results are in agreement with the comparator hypothesis (Miller & Matzel, 1988) and with Wagner's (1981) “standard operating procedures” (SOP) model.