Human sensitivity to reinforcement in operant choice: How much do consequences matter?

Abstract

The results of many human operant conditioning experiments appear to show that humans are less sensitive than nonhumans to operant consequences, suggesting species discontinuities in basic behavioral processes. A reanalysis of 311 data sets from 25 studies employing variable-interval schedules of reinforcement designed to assess sensitivity to reinforcement corroborates the claim that human behavioral allocation among alternatives often deviates from predictions based on rates of experimentally programmed consequences. Close inspection of the studies in question, however, suggests that methodological issues contribute heavily to the differences noted so far between humans and nonhumans and that an explanation based upon species discontinuities is not tenable.Consequences clearly influence the behavior of nonhuman organisms. The rich operant conditioning literature shows that reinforcing consequences alter the strength of behavior that produces them and forge relations between behavior and antecedent stimuli. These effects hold across a wide range of species, settings, response classes, and types of reinforcers, placing them among the most widely replicated outcomes in the biological and behavioral sciences. As a result, the threeterm operant contingency, which encompasses the relations among antecedent conditions, operant behavior, and reinforcers, has been proposed as the foundation of a broad range of complex capabilities both in nonhuman species (e.g., Donahoe, Burgos, & Palmer, 1993) and, perhaps more speculatively, in humans (e.g., Skinner, 1953Skinner, , 1957.Successful applications to human affairs, based on the core notion that human behavior is sensitive to its consequences, provide reason for optimism about the generality of operant principles. As can be expected ofa successful science, applications have flowed from the laboratory since the earliest days of operant psychology. They include, but are not limited to, animal models of substance abuse and treatment (e

Human choice among five alternatives when reinforcers decay

“…Both of these aspects -slopes and variability -were closer to those in animals studies under the decay condition than under hold. Thus, consistent with conclusions by Bradshaw and Szabadi (1988), and by Kollins et al (1997), procedural details may contribute importantly to how choice distributions relate to obtained reinforcements.…”

“…Most research on concurrent schedules has been with animal models, such as pigeons, rats, mice, or monkeys, but more than 25 human studies have been described in excellent reviews by Bradshaw and Szabadi (1988) and by Kollins et al (1997). These reviews indicate that people manifest lower power-law slopes than nonhuman animals, indicating lower sensitivity to changes in reinforcements, and greater variability in those slopes.…”

“…These reviews indicate that people manifest lower power-law slopes than nonhuman animals, indicating lower sensitivity to changes in reinforcements, and greater variability in those slopes. Kollins et al (1997) offer many hypotheses to account for such differences, but ultimately implicate procedural differences. The present method attempted to increase both attention to task and motivation for immediate reinforcement, two factors thought to be especially important.…”

“…Choices by humans often show a wide range of sensitivity to reinforcement frequencies, and many explanations have been offered (Bradshaw and Szabadi, 1988;Kollins, Newland & Critchfield, 1997). We devised a procedure that differed in a number of ways from those commonly employed and tested the influence of differentially reinforcing rapid reinforcer acquisition (explained in the next paragraph).…”

Risk Aversion in Human Subjects under Conditions of Probabilistic Reward

“…The parameters sa and Sf designate individual sensitivities to amount and probability (or its equivalent in this context, rate 1 ). We set the exponents sa and Sf to 0.50 and 1.50, respectively, based on extensive previous data suggesting that subjects are undersensitive to amounts and oversensitive to probabilities/rates (e.g., de Villiers, 1977;Goodie & Fantino, 1995;Herrnstein, 1997;Kollins, Newland, & Critchfield, 1997;Rachlin et aI., 1986;Siovic & Lichtenstein, 1968). This equation was employed not to test its utility as a model of risk taking, but rather as a descriptive tool, providing a standard upon which to judge the obtained data with respect to subjective versus expected values of the response options.…”