Association for Behavior Analysis International

Gambling is a popular social behavior that occurs worldwide. An estimated 16.5 million Americans experience problems similar to those found in substance abuse due to pathological or problem gambling. Despite widespread interest in determining mechanisms involved in problem gambling, gambling can be difficult to study in humans for ethical and practical reasons. Pigeons have been proposed as an animal model of problem gambling because they sometimes select probabilistic schedules with variable-magnitude reinforcers over fixed schedules even when the expected amount of food on the probabilistic schedule is much lower. This symposium focuses on recent experimental contributions to the development of pigeon models of gambling. Two talks describe how near wins, stimuli that approximate winning stimuli but deliver no reinforcers, impact behavior in a pigeon slot matching analog. Ward and Kyonka describe interactions between near-win effects and alcohol. Rice and Kyonka present results of experiments showing how near wins function in extinction and in higher-order schedules. Laude, Zentall, Case, Stagner and Sticklen present results on near misses and choice behavior. Tan and Hackenberg present results of a study of free-operant gambling involving the use of a token economy. The utility and implications of using pigeon analogs for gambling will be discussed.

Four pigeons responded in a slot machine analog in which keys were lighted either red or green for a three-peck sequence. A sequence with three red lights in a row signaled a win and three green lights in a row signaled a loss. Two red lights followed by one green light signaled a near win. Following the three-peck sequence, a white key was illuminated for five seconds to signal a collect phase. Food was delivered on winning trials if at least one peck occurred on the collect key during the five-second collect phase. For all other trial types, no food was delivered for pecks on the collect key. Ethanol was administered in three increasing doses semi-weekly after responding stabilized. Ethanol administration decreased discrimination between trial types and increased response rates on all losing trials. Sensitivity to near wins increased as a function of dose, a further indication of impaired discrimination between wins and losses.

In slot machine play, near wins are stimuli that approximate winning stimuli but deliver no reinforcers (e.g., two winning symbols and a losing symbol). In two experiments, keys were illuminated red or green following pecks in a 3-peck sequence. Three successive red lights signaled a win and three green lights signaled a loss. Two red lights followed by a green light was a near win. Following the 3-peck sequence, a collect key was illuminated white and a peck was required to deliver reinforcement on winning trials. In Experiment 1, pigeons were trained on the slot-machine analog and then switched to extinction. Near wins were either present or absent in extinction. In Experiment 2, the slot-machine analog was embedded in a concurrent-chains procedure. Preference was investigated as the probability of a win, probability of a near-win trial, and magnitude of reinforcement was systematically altered. Across both experiments, collect-key response rates were proportional to the amount of reds shown in a trial. The prediction of behavioral momentum that conditioned reinforcers (e.g., near wins) cause responding to persist for longer in extinction and cause shifts in preference was tested across experiments.

Problem gambling in humans has garnered significant research interest due to its maladaptive consequences. Recently, animal models of human gambling behavior have uncovered some of the underlying mechanisms that contribute to maladaptive choice that may also have implications for human gambling behavior. Perhaps one of the most important and reliable findings in animal models is the critical role of conditioned reinforcement. One paradoxical effect is the near miss phenomenon which appears to lead to persistence in gambling behavior despite it having no informative value. A near miss is characterized by the successive presentation of stimuli that closely resemble a winning trial with the exception that the last stimulus in the sequence signals a loosing outcome. Although the near miss should serve to devalue the win, it appears to have the opposite effect. Recently, we have attempted to model this human phenomenon with pigeons. The results of our work and the importance of capturing certain aspects of human gambling that are oftentimes not integrated in animal analogs of near miss will be discussed.

Pigeons were trained in a free-operant gambling task in which they were able to earn tokens, and then then exchange earned tokens for food or gamble them, by completing separate FR requirements on three simultaneously available keys presented on a touchscreen. The response cost for earning tokens, as well as the size and probability of gambling wins was manipulated over a series of conditions. The cost of earning tokens had little effect on gambling behavior. Subjects were more sensitive to the probability than the size of wins and would continue to gamble even when the costs outweighed the gains. This procedure shows promise as an animal model for the investigation of gambling behavior.