The unit of selection: what do reinforcers reinforce?

We begin by stipulating three central points upon which we and the broad consensus of commentators agree. First, the effects of reinforcers on behavior can be readily demonstrated under conditions in which the antecedents of behavior are not identified. That is, orderly functional relations emerge between operants and reinforcers when the experimental analysis of the effects of antecedents is impossible or impractical. Second, response-contingent reinforcers most commonly alter the control of responses by their antecedents. That is, discriminative control of responding is ‘‘practically inevitable’’ (Skinner, 1937, p. 273). Third, the manipulated antecedents of behavior are typically environmental events in cases that are amenable to experimental analysis, but may include covert or private (intraorganismic) events as well. Covert events (characterized behaviorally or neurally) invariably accompany environment–behavior relations and indispensably contribute to scientific interpretation. Where we differ from some commentators (but fewer than we might have supposed) is with respect to whether control by antecedents is inescapable, including on those occasions in which the antecedents cannot be manipulated (i.e., those in which functional relations between antecedents and behavior