Association for Behavior Analysis International

Investigating slot machine gambling from a behavioural perspective offers a unique insight into problems involving pathological gambling. We add to a growing body of research with a set of experiments that examined the possible reinforcing qualities of a near win (i.e. four out of five winning symbols on a machine) and its effects on subsequent responding. Previous research (e.g. Dixon & Schreiber, 2004) suggests that near wins may act similar to reinforcers despite remaining functionally a loss trial. The simulated gambling procedure consisted of two concurrent slot machine reels consisting of five symbols each. Winning trials (all five symbols) across reels were arranged according to the following ratios: 2:10, 4:10, 10:4 and 10:2. The effects of near win trials were examined as a function of presence vs. absence (study 1) or the relative percentage of near win trials (study 2). Analysis using the generalised matching law and examination of post-reinforcement pauses confirmed the near win as possessing reinforcing qualities distinct from those of wins.

Dynamic choice behaviour of 15 human participants was examined in a Rock Paper Scissors (RPS) game. The RPS game provides a valid procedure for investigating choice amongst multiple alternatives, and potential for applications of game and foraging theory in behavioural research. In the current study, participants played against a computer opponent; on a single trial, either the participant or the computer could earn points by beating the other in a round of RPS. Sessions lasted for 360 trials. Reinforcement probabilities varied across sessions; there were three conditions in which the ratios of the probabilities of the computer playing Rock, Paper or Scissors varied across blocks of 6, 12 or 36 trials within a session. A total of nine different reinforcement ratios were used: 5-10-85, 20-50-30, 10-20-70, 33-33-33, with locations counterbalanced. Response variability and sensitivity were examined; models of choice behavior to three discrete options were fitted to the data, and changes in responding over time were analysed as a function of reinforcement variability.

Over the past decade, behavior-analytic research has yielded several extensions to the generalized matching law. These extensions typically increase variance accounted for relative to the already-high variance accounted for by generalized matching. When goodness of fit is similar across candidates, residual analyses for systematic trends provide an alternate means of distinguishing between models. Natapoff relations (Schneider & Davison, 2005) were computed in a reanalysis of Kangas and colleagues (2009) Concurrent performance in a three-alternative choice situation: Response allocation in a Rock/Paper/Scissors game. Kangas reported results of generalized matching analyses for ratios of responding and reinforcement for a target option versus both other options, for example, rock/(paper+scissors). Goodness of fit was adequate for both ratios and Natapoff relations. However, residual analyses revealed polynomial trends were sometimes present for ratios but not for Natapoff relations, identifying Natapoff relations as a quantitatively superior account of choice between the three alternatives rock, paper and scissors. Residual analyses of 3-alternative choice may continue to be a fruitful means of comparing extensions of generalized matching.