Thursday, 18 November 2010

Gambling on the Near-Miss

Recent studies from the growing literature on the neuroscience of gambling have focused on the "near miss" effect in slot machine gambling. Briefly, "near miss" is a term used to describe an effect resembling 'almost winning' on a slot machine when, for instance, two of three matching symbols appear on the payout line. This outcome is still a loss though (a win would be three matching symbols).

It has been speculated that near misses serve to maintain persistent gambling. There are several strands of data that support this assertion. For instance, gamblers and non-gamblers initiate trials following a trial with a near miss faster than trials with either wins or losses (Dixon & Schrieber, 2004). Animals also show similar effects: Weatherly and Derenne (2007) and Peters et al. (2010) developed an animal model of slot machine gambling and found that the pause durations of rats was shortest after losses and longest after large "wins".

Gamblers also rate their chances of winning on the next trial far higher than on trials following other outcomes. These effects serve to suggest that gamblers ignore the "independence of turns" rule and act in way that implies that outcomes of the next trial will, in some way, be related to the outcome of the previous trial. They are not. Slot machines operate according to complex random ratio schedules, where each response is independent of the last. This is what makes gambling, generally, and the near miss effect, in particular, so interesting to study.

The proportion and physical form of the near misses that a gambler experiences has also been shown to effect persistence. A well-cited experiment by Kassinove and Schare (2001) found that a proportion of 30% of near miss trials resulted in greatest persistence (as measured by the number of trials played in the absence of feedback), while Ghezzi and colleagues (2006) showed that near misses on three-reel slot machines of the form X-X-Y were preferred over near misses of the form X-Y-X or Y-Y-X. The findings on the effects of proportion of near misses suggest that they do not resemble direct losses to the gambler. They still lose on these trials. The fact that the form of the near misses exerts an influence highlights how seemingly irrelevant, structural features of the stimuli involved in slot machine gambling may come to exert an influence over the behaviour of seemingly "free", language-able humans.

B.F. Skinner argued that near misses may function as conditioned reinforcers:
Gambling devices make an effective use of conditioned reinforcers which are set up by pairing certain stimuli with the economic reinforcers which occasionally appear. For example, the standard slot machine reinforces the player when certain arrangements of three pictures appear in a window on the front of the machine. By paying off very generously—with the jack pot—for "three bars," the device
eventually makes two bars plus any other figure strongly reinforcing. Almost hitting the jack pot" increases the probability that the individual will play the machine, although this reinforcer costs the owner of the device nothing.
(1953, p. 397).

Recent neuroscience evidence supports this view (see the excellent Skepticon posting, for instance). fMRI studies by Chase and Clark (2010) and Habib and Dixon (2010) have each shown that near-misses activated the same brain regions as wins for gamblers and regions related to losses in non-gamblers. This suggests that near misses have conditioned reinforcing functions for gamblers (i.e., they behave "as if" they are wins). Interestingly, the extent of brain activation in mid-brain regions declines as a function of gambling severity. The more severe the gambling problem, the less activation. This implies that a degree of tolerance has occurred: repeated experience with near-misses leads to habituation-like effects on activation, and suggests that gambling persistence must be, at least in part, maintained in other ways. There is an awful lot more work to be on the behavioural and neural correlates of gambling, but the handful of studies just described have already made quite an impact on our understanding of this fascinating behaviour.

Chase, H., & Clark, L. (2010). Gambling severity predicts midbrain response to near-miss outcomes Journal of Neuroscience, 30 (18), 6180-6187 DOI: 10.1523/JNEUROSCI.5758-09.2010

Dixon, M. R., & Schreiber, J. (2004). Near-miss effects on response latencies and win estimations of slot machine players. The Psychological Record, 54, 335–348.

Ghezzi, P. M., Wilson, G. R., & Porter, J. C. K. (2006). The near-miss effect in simulated slot machine play. In P. M. Ghezzi, C. A. Lyons, M. R. Dixon, & G. R. Wilson (Eds.), Gambling: Behavior theory, research and application (pp. 155-170). Reno, NV: Context Press.

Habib, R. & Dixon, M.R. (2010). Neurobehavioral evidence for the “near-miss” effect in pathological gamblers. Journal of the Experimental Analysis of Behavior, 93 (3), 313-328. DOI: 10.1901/jeab.2010.93-313

Kassinove, J. I., & Schare, M. L. (2001). Effects of the ‘near miss’ and the ‘big win’ on persistence at slot machine gambling. Psychology of Addictive Behaviors, 15, 155–158.

Peters, H., Hunt, M., & Harper, D. (2010). An animal model of slot machine gambling: The effect of structural characteristics on response latency and persistence. Journal of Gambling Studies, 26, 521-531.

Weatherly, J.N., & Derenne, A. (2007). Rats playing a slot machine: A preliminary attempt at an animal gambling model. Analysis of Gambling Behavior, 2, 79-89.

Friday, 12 November 2010

Children's "Erroneous Beliefs" About Gambling

The empirical literature on gambling behaviour is small but growing, and has tended to be dominated by qualitative/survey, cognitive, and behavioral approaches. Qualitative approaches to gambling focus on the prevalence, nature, and extent of different forms of gambling. Largely demographic or survey-based, these studies provide important details about changing social trends in gambling and gambling related problems. When writing an empirical article on gambling one has to cite at least one gambling prevalence survey in the opening paragraphs; you know, to set the scene, to introduce 'the problem'. And, let's face it, that's all these studies allow us to do: to get a feel for what's happening "out there", in the "real world".

Cognitive approaches emphasise the role played by a gambler's thoughts, feelings, attitudes or beliefs in initiating and maintaining gambling. Gamblers may be deemed to suffer from "erroneous beliefs" or to engage in "cognitive switching" whenever the outcomes of games of chance go against their "expectations". This is all well and good as a description of behaviour, but it is not a complete explanation of the causes of the behaviour observed. Woaahh, did I lose you? Let me back up.

In behaviour analysis, saying that a gambler persists in betting his/her final dollar on "red 16" because of faulty cognitive processes or beliefs etc. is to offer a mentalistic explanation, which is a mediational-based account that says, "the behavior you observe (in this case, betting on "red 16") is caused or initiated by a mysterious, nonphysical realm called 'beliefs'". We can't directly manipulate "beliefs". All we can manipulate are environmental events, like peoples' histories with "red 16" and money, and with instructions to behave in certain ways, etc. Mentalism seduces us into thinking that a complete explanation has been provided, when in fact it hasn't. If we can't manipulate it (i.e., can't predict and influence it's occurrence), then it's of no use to a behaviour analyst. We are arch-pragmatists after all.

Faced with this, and arriving under-dressed, grumpy and late to the party on empirically analyzing gambling behaviour, behaviour analysts have sought to "behaviouralize" common cognitive findings, while seeking to preserve their philosophical integrity. For instance, an article from my lab published in the Journal of the Experimental Analysis of Behavior set about studying 'erroneous beliefs' from a behavioural perspective in a gambling-like task with young children. The article kicked off with the now-customary critique of all things non-behavioural, which, if you don't mind, I think is worth repeating here:

"A limitation of mediational explanations is that they tend to treat mediational  "responses" (e.g., self-stated rules, beliefs, etc.) as primary causes. Individual differences with regard to mediational responses is purported to explain observed differences in gambling situations. Unfortunately, such an approach places the causes  of gambling in the thoughts and other behaviors of the participants, and we are left unable to determine the conditions that cause those thoughts. Information about the mediational variables is important in that they allow us to predict gambling. However, they do not allow us to control the occurrence of gambling." (Dymond, Bateman, & Dixon, 2010, pp. 353-354).

Key here are "the conditions that cause those thoughts". What kind of thoughts might young kids have about a gambling-like task? Well, a previous study showed that 7-10 year olds could be taught to believe that one of two dice was "luckier" or "hotter" than another. The procedures were as follows. Kids chose between a red die and a blue die that 'rolled' a random number between 1 and 6, and a preselected game piece moved the predetermined number of spaces along a racetrack. Kids "raced" against the computer.

They love stuff like this, so the task was very reinforcing and they played it until it was clear that they did not have a preference for either the red or blue die. After all, why should they? Each die randomly rolled between 1 and 6, so kids distributed their choices relatively easily between them, as expected. Next, the kids were taught to associate one color (red) with "greater than" and another color (blue) with "less than". Then, they were re-exposed to the dice task and, guess what, they showed an increased preference for the red die, despite the identical outcomes of the two die! Red still rolled as it did (as indeed blue did) during the pretest phase, but now, after a relational intervention, kids behaved "as if" red was more likely to roll higher numbers than blue. Cool.

We wanted to extend these promising findings by examining whether or not we could first attach fixed, high and low roll outcomes to two die labelled with meaningless nonsense words (let's call them Bill and Ben) and then testing selections of indirectly related die (let's call them Dave and Nick). Right, I may have lost you again, which is bad, particularly as you have read this far. So, let me summarise the design of the study as follows:

1. Die labelled "Bill" always roll high numbers (4, 5, 6) and die labelled "Ben" always rolls low numbers (1, 2, 3).

2. Imagine if Mary was related to Bill, and Mary was related Dave; also, that Anne was related to Ben, and that Anne was related to Nick.

3. Now, if I asked you, would you prefer Bill or Ben, you'd probably opt for Bill. After all, it's loaded; it always rolls high numbers. Okay, but what if I asked you to choose between Dave and Nick, and you had no prior experience of these labelled die? Which one would you choose?

We predicted that young kids, given this history, would choose Dave over Nick. Why? Because Dave is indirectly related to the directly-experienced Bill die. And boy do I like that Bill die! So, despite never have "played" the Dave or Nick dice before, and the fact that, in reality, each rolled the same, I'm going to choose that there Dave more than Nick. And that's what the kids did, well, all but one (there's always one!). The kids also rated the Dave die more as more likable than the Nick die.

What does this all mean? What's it all got to do with knocking cognitive-based explanations of seemingly irrational gambling behavior? Well, it suggests a behavioural process whereby novel, indirectly experienced stimuli may come to control behavior. Direct experience with all stimuli and the good/bad things attached to them (such as high and low rolls) may not be necessary for complex gambling behavior to develop. In this way, "beliefs" about outcomes, and the behavior based upon these beliefs, can be said to be mediated by words. This allows us to remove ourselves from the spatio-temporal present and 'imagine' how good a nice Dave would be right now...

Gaining an empirical handle on these word-webs, and the behavioural processes responsible for them, is what my students, colleagues and I spend a lot of time doing. I'll blog about this on an ongoing basis.

Here's a link to a PDF of the paper described above.

Wednesday, 3 November 2010

Prompting Therapists to Deliver More Accurate & Varied Discrete Trials

A student from my lab, Richard J. May, has just published a paper in Research on Autism Spectrum Disorders on a novel method of training therapists to deliver discrete trial teaching (DTT). DTT is probably the most common instructional strategy used in early intervention work with children with autism. It involves breaking complex skills down into component parts and arranging for clearly specifiable antecedents to precede the behaviour (such as request or instruction) and consequences to follow the desired behaviour (such as praise). Progress is often slow on DTT programs, for a host of reasons, and applied behaviour analysts have recently made great headway in developing new and exciting ways of boosting outcomes and of designing easy-to-use programs and instructional materials for use with therapists working on home intervention programs with children with autism.

Richard's clinical experience lead him to design a simple prompting system to facilitate delivery of DTT across entire teaching sessions. Teaching therapists to implement DTT effectively and efficiently has obvious implications for educational intervention with children with autism. Thus, "VB-TiPS" was born: the Verbal Behaviour Teaching integrity Prompting System (although it isn't actually called this in the paper).

Richard evaluated VB-TiPS with and without existing behavioural skills training (BST) procedures and found that all three therapists studied delivered a higher rate of accurate teaching trials and a greater variety of trials when VB-TiPS was present. The prompting system helped therapists to provide greater numbers of different verbal operant trials: tacts, intraverbals, and listener relations were all presented more often compared to traditional BST procedures.

These encouraging preliminary findings highlight the need for continued research on developing facilitative DTT procedures that are applicable to a host of different participant populations (not just children with autism) and are sensitive to the varied qualifications and background of therapists working on early intervention programs.

As the paper concludes: "Behavior analytic consultants are often faced with the challenge of training a large number of therapists in a relatively short amount of time, particularly when starting a program for a new client or when therapist turnover is high. The continued identification of training strategies that are both efficient and effective is imperative not only for therapists, but for the children they endeavor to teach." (May, Austin, & Dymond, 2011, p. 315).

You can get a (PDF) copy of the paper here.