736Research on behavioral choice has studied how nonhumans choose between rewarding outcomes that differ in terms of the immediacy or amount of food reinforcement. This research originated from Herrnstein's (1961) matching law and has led to models for choice that have had substantial impact for understanding human decision making, such as the hyperbolic discounting model for intertemporal choice (see Green & Myerson, 2004, for review). The question posed by the present study is whether choice between two delayed food rewards is affected by providing additional food deliveries independently of responding.Our experiment uses a popular paradigm known as concurrent chains, which measures choice between stimuli that signal different rewarding outcomes. In this procedure, subjects (usually pigeons or rats) respond on two alternatives during a choice phase (initial links) to produce access to one of two mutually exclusive outcome schedules (terminal links). The typical result is that subjects respond more to the initial link that leads to the terminal link that provides the more immediate, larger, or more probable reinforcement. Various models for choice in concurrent chains have been proposed, including the contextual choice model (CCM; Grace, 1994), the hyperbolic value added model (HVA; Mazur, 2001), and delay reduction theory (DRT;Fantino, 1969). Although differing in specific details, the equations for these models assume that choice is determined solely by the temporal relations between stimuli and reinforcement. By contrast, there is no explicit role for the relationship between responding and reinforcement, particularly the contingency and temporal contiguity between choice responses and food.An experiment by Mazur (2003) in which the effect of free food deliveries on choice was investigated is relevant to this issue. In his experiment, pigeons responded on a concurrent-chains schedule in which the initial link was a single variable-interval (VI) 30-sec schedule that arranged equal access to the terminal links, which were associated with 3-sec and 12-sec delays to food. In some conditions of his experiment, the contingency between responding and reinforcement was degraded by adding a variabletime (VT) 30-sec schedule during the initial links, which provided food independently of responding. Although overall initial-link responding would be expected to decrease during sessions with added VT food because of the degraded contingency (Rachlin & Baum, 1972), Mazur (2003) showed that the HVA model predicted that choice for the 3-sec alternative should increase. The reason for this prediction was that the average time to reinforcement from the onset of the initial links, which is an important variable in the HVA model, decreased when the VT schedule provided free food. By contrast, according to CCM, there should be no effect of the free food on choice, because its predictions are affected by the time spent in the initial links, which did not change when the VT schedule operated. Mazur (2003) found that results were c...