Divided stimulus control depends on global, but not local, reinforcement contingencies

Research suggests that divided stimulus control is graded according to the relative reinforcer probabilities associated with each stimulus, much like choice depends on reinforcer ratios. This experiment investigated whether parallels between divided stimulus control and choice extend to local levels of analysis. Six pigeons responded in a matching-to-sample procedure in which sample stimuli were composed of a color (blue or white) and frequency of color alternation (fast or slow) and subjects reported the identity of one dimension in each trial. In some conditions, the dimension to report in a trial signaled the dimension likely to be reinforced or to be reported in the following trial. Unlike choice-which favors the likely next-reinforcer location as signaled by last-reinforcer location-local probabilities had little effect on divided stimulus control. Instead, divided control depended on overall reinforcer probabilities associated with each stimulus dimension. Such global reinforcer effects were weaker when one dimension was less discriminable and were well described by a model asserting that divided control depends on differential and nondifferential reinforcement. These findings highlight the importance of the discriminated reinforcer differential in divided stimulus control and suggest that parallels between divided control and choice may not extend to local levels.