Joint control and generalized nonidentity matching: Saying when something is not.

This study investigated how the absence of a specified stimulus can control behavior. Four children were trained in nonidentity matching, and as a control, four were trained in identity matching. Both performances were produced by training over mediating responses, so that in identity matching, the selection of a particular comparison was evoked by the repetition of a sample tact to the comparison, and in nonidentity, by the inability to repeat the sample tact to the comparison. Successful generalization of the performances indicated that they were indeed controlled by these general features rather than by stimulus-specific features. Comparison selection thus served as an autoclitic report about other verbal behavior. In particular, generalized nonidentity matching indicated that sensitivity to discrepancies between what a sample specifies, and what is actually presented, can be accounted for behaviorally, without recourse to hypothesized cognitive mediators.