Original Article
Illusionary pattern detection in habitual gamblers

https://doi.org/10.1016/j.evolhumbehav.2014.02.010Get rights and content

Abstract

Does problem gambling arise from an illusion that patterns exist where there are none? Our prior research suggested that “hot hand,” a tendency to perceive illusory streaks in sequences, may be a human universal, tied to an evolutionary history of foraging for clumpy resources. Like other evolved propensities, this tendency might be expressed more stongly in some people than others, leading them to see luck where others see only chance. If the desire to gamble is enhanced by illusory pattern detection, such individual differences could be predictive of gambling risk. While previous research has suggested a potential link between cognitive strategies and propensity to gamble, no prior study has directly measured gamblers' cognitive strategies using behavioral choice tasks, and linked them to risk taking or gambling propensities. Using a computerized sequential decision-making paradigm that directly measured subjects' predictions of sequences, we found evidence that subjects who have a greater tendency to gamble also have a higher tendency to perceive illusionary patterns, as measured by their preferences for a random slot machine over a negatively autocorrelated one. Casino gamblers played the random slot machine significantly more often even though a training phase and a history of outcomes were provided. Additionally, we found a marginally significant group difference between gamblers and matched community members in their slot-machine choice proportions. Performance on our behavioral choice task correlated with subjects' risk attitudes toward gambling and their frequency of play, as well as the selection of choice strategies in gambling activities.

Introduction

Why do we gamble? The simple answer, of course, is to win. But when games of chance are truly random and entirely unaffected by human skill, as many are, the rationale for engaging in them is far from obvious. If the statistics of the game mean that the best one can expect in the long run is to break even—and usually not even that—why play? Why do so many people around the world spend substantial portions of their income on games of chance, such as lotteries, that will only make them poorer on average?

One possibility is that gamblers do not fully grasp the random nature of the games they are playing. There is a large psychological literature documenting what is sometimes called “apophenia:” a human tendency to perceive patterns in random data that simply do not exist (e.g., Falk & Konold, 1997). In particular, people seem to have difficulties when perceiving independent events, or series of events whose outcome has no influence on the outcome of future events (Nickerson, 2002). One of the best known of these biases is the “hot-hand” phenomenon, first identified in a study of observers' predictions about basketball shots (Gilovich, Vallone, & Tversky, 1985). Both players and fans tended to judge a player's chance of hitting a shot to be greater following a successful shot than a miss, despite the fact that hit rate was statistically the same in both cases. Perhaps not surprisingly, illusory pattern perception of this kind has also been found among gamblers. For example, roulette players often bet on more numbers after winning than after losing (Wagenaar, 1988). Lottery players tend to redeem winning tickets for more tickets rather than for cash, reflecting a belief that they are more likely to win again (Clotfelter & Cook, 1989). Many lottery players believe in “hot” and “cold” numbers, returning to previously “hot” numbers once they've been given time to cool off (Rogers, 1998). And lottery tickets are sold more often at stores that have just issued a winning ticket, reflecting a hot hand or “lightning strikes twice” mentality (Guryan & Kearney, 2008). Another fallacy, known as the “gambler's fallacy,” is in some ways the flip side of hot hand, reflecting a belief that a streak is coming to an end—leading roulette gamblers, for example, to bet on black after several reds in a row (e.g., Ayton and Fischer, 2004, Croson and Sundali, 2005). Both hot hand and the gambler's fallacy, then, seem to reflect illusory perception of clumps or streaks in data that do not contain them.

Where do these beliefs come from? Why might people systematically perceive streaks where there are none? In previous work, we have proposed that hot hand may have its roots not in basketball or in financial markets, but in a much more ancient mode of human cognition: foraging (see Scheibehenne et al., 2011, Wilke and Barrett, 2009, Wilke and Mata, 2012, Wilke and Todd, 2012; see also Reifman, 2011). On this view, a tendency to look for clumps is not just a quirk of modernity, but a deep-seated part of our psychology that evolved because there are many contexts in which the world is not random, and looking for clumps is therefore adaptive. In particular, clumped distributions of resources such as plants, animals, water sources, and human settlements are common in natural environments, and animal and human foragers appear to adapt their search strategies to these observable statistical regularities in their foraging landscape (Bell, 1991, Hills, 2006, Krause and Ruxton, 2002, Taylor, 1961, Taylor et al., 1978).

Consistent with this, we found that hot hand occurs in both Western cultures and a traditional foraging culture, and seems to be a kind of psychological default which is only partly erased by experience with true randomizing mechanisms like coin tosses (Wilke & Barrett, 2009). Importantly, hot hand is not necessarily irrational when clumps actually do exist, and in cases where they do not—for example, when trying to predict random sequences of independent and equiprobable events (such as when playing roulette)—hot hand does not decrease accuracy, because all strategies produce chance-level performance (c.f. Scheibehenne et al., 2011). Therefore, the tendency to assume or look for patterns or regularities in a given sequence may be a reasonable default strategy: If there is in fact a pattern, expecting that particular pattern can be advantageous by providing an edge in predicting future events, and if there is no pattern, expecting one will do no worse than any other strategy.

Could hot hand play a role in human gambling behavior? The tendency to search for patterns in random data could explain part of the pleasure humans experience in gambling—for example, the experience of winning several times in a row could be highly compelling, leading one to believe that one is on a hot streak. But in addition to this universal propensity, there could be differences between individuals in just how “hot-handed” they are—just how prone they are to perceive streaks, even when they do not exist—that could lead to differences in how much gambling on random outcomes is enjoyed. As for many evolved traits, individual differences in pattern perception could arise from a variety of factors, including genetic differences, environmental differences, or differences in individual experience. If such individual differences are predictive of propensity to gamble, this could have implications both for developing assessment tools to detect risk of developing a gambling problem and for interventions that might be effective, such as targeting gamblers' perceptions of randomness (c.f. Petry & Armentano, 1999). Moreover, if the hypothesis that hot hand is a universal human cognitive adaptation is right, then the risk of developing a gambling problem might also be a human universal, not restricted to those with a cultural history of gaming, or individual experience with it.

Several prior studies have suggested that beliefs about illusory patterns, such as hot hand and its opposite effect the gambler's fallacy, may play a role in preferences for gambling (Ayton and Fischer, 2004, Croson and Sundali, 2005, Gaboury and Ladouceur, 1989, Joukhador et al., 2004, Oskarsson et al., 2009, Toneatto, 1999). However, these studies were based on questionnaires about subjects' gambling beliefs, not on direct measurements of subjects' predictions of streaks. While questionnaires can be useful, asking subjects to report their beliefs about their own behavior is not the same as measuring what subjects actually do (c.f. Nisbett & Wilson, 1977). For example, many college students report understanding that coin tosses are perfectly random. However, their judgments of actual sequences of coin tosses reveal that they expect them to contain fewer streaks than they actually do, revealing a bias that they might not realize they have (Falk & Konold, 1997). To truly measure hot-handedness then, observation of actual behavior is necessary. In the case of gambling, one might expect subjects to prefer to bet on sources that they perceive as containing clumps, compared to sources they perceive as less clumpy—even when the clumps are illusory. Here we gave subjects a choice between paired sequence generators that varied in how hot-handed they actually were, and measured which the subjects preferred to bet on (c.f. Shaffer, Peller, LaPlante, Nelson, & LaBrie, 2010).

In order to assess the possible role of the hot-hand phenomenon in the propensity to gamble, we adopted a mixed-method approach that looked for within-subject correlations between hot-handedness, as measured with a behavioral task, and separate measures of proneness to gamble. Our behavioral task, adapted from Scheibehenne et al. (2011), presented subjects with a choice between two simulated slot machines (see Fig. 1). One was slightly anti-clumpy, or negatively autocorrelated, while the other was entirely random, with no clumps. In a prior study, Scheibehenne et al. (2011) found that subjects preferred to play the truly random machine, consistent with the perception that it generated more streaks. Thus, degree of preference for the random over the negatively autocorrelated machine is a direct behavioral measure of preference for an illusorily hot-handed machine.

Our research design looked for correlations between performance on the gambling task, and separate, independent measures of gambling propensity. The latter involved both a natural between-group component and a variety of individual difference measures. For the between-group component, we tested two groups of people: habitual gamblers and a control population. For the individual differences component, we examined several factors potentially related to gambling: measures of cognitive capacity as well as standardized screens of gambling history and psychometric measures of risk-taking propensity (a version of the DOSPERT scale, described below). In addition, because our task involved a long sequence of individual gambling decisions, we used quartile analysis to look for changes in strategy, including possible learning effects, within the task.

Our study examined two main hypotheses. First, we predicted that as a group, habitual gamblers would be more prone to see illusionary patterns in random data sets than a matched sample of non-gamblers (Hypothesis 1). Second, we predicted that individual differences in hot-handedness across groups would correlate with gambling-related risk attitudes and individual differences in gambling experience (Hypothesis 2).

Section snippets

Participants

We collected data from two target populations. In close proximity to Clarkson University is the territory of the St. Regis Mohawk Tribe, or Akwesasne, who are presently situated on more than 30,000 acres of tribal land extending from New York into Quebec and Ontario. With the permission of the Akwesasne Mohawk Casino, a gaming enterprise under the supervision of the St. Regis Mohawk Tribe, we recruited 92 experienced adult gamblers [58 females (63%), 34 males (37%)]. The Akwesasne Mohawk Casino

Clinical sample recruitment and subjects gambling history

The two left-hand subplots of Fig. 2 show histograms of subjects' clinical score on the South Oaks Gambling Screen (SOGS). In line with published prevalence rates of gambling disorders, only very few subjects in our non-gambling community sample had SOGS scores high enough to indicate a serious gambling disorder (about 3%; see Kessler et al., 2008, Shaffer et al., 1997). The SOGS score distribution in our gambling group, however, showed that our recruiting strategy was successful in recruiting

Discussion

Pathological gambling is a psychological and medical disorder identified in both the DSM-V and ICD-10 (American Psychiatric Association, 2013, World Health Organisation, 1992; see also National Center for Responsible Gaming, 2013). The lifetime prevalence varies worldwide, but has been estimated at 0.6% to 1.1% of adults in the United States, with an additional 2.3% defined as problem gamblers (Kessler et al., 2008, Shaffer et al., 1997). Problem gambling is on the rise in older populations

Conclusion

The present study provides the first evidence of a direct link between illusory pattern perception and gambling risk. Because of the magnitude of gambling and gambling addiction as a global socioeconomic phenomenon, replication and further investigation are required before our findings can be translated into actual policy recommendations or interventions. However, the results are both consistent with the commonsense intuition that people who are more prone to perceive “luck” where there is none

Supplementary Materials

The following are the supplementary data to this article.

gambling_0224_data.sav

Acknowledgments

We thank Amanda Sherman, Aisha Wood and the members of the Evolution and Cognition Lab for their help in data collection, and Todd Papineau, the executive director of the St. Regis Mohawk Tribal Gaming Commission, for his support. Daniel Nettle and an anonymous reviewer gave very helpful advice and criticism.

References (50)

  • W.J. Bell

    Searching behaviour: The behavioural ecology of finding resources

    (1991)
  • C.T. Clotfelter et al.

    Selling hope: State lotteries in America

    (1989)
  • R. Croson et al.

    The gambler's fallacy and the hot hand: Empirical data from casinos

    Journal of Risk and Uncertainty

    (2005)
  • R.M. Cunningham-Williams et al.

    Computerized gambling assessment module (C-GAM)

    (2003)
  • J. Derevensky

    Teen gambling: Understanding a growing epidemic

    (2012)
  • R. Falk et al.

    Making sense of randomness: Implicit encoding as a basis for judgment

    Psychological Review

    (1997)
  • A. Gaboury et al.

    Erroneous perceptions and gambling

    Journal of Social Behavior and Personality

    (1989)
  • J. Guryan et al.

    Gambling at lucky stores: Empirical evidence from state lottery sales

    American Economic Review

    (2008)
  • Y. Hanoch et al.

    Domain-specificity in experimental measures and participant recruitment: An application to risk-taking behavior

    Psychological Science

    (2006)
  • T.T. Hills

    Animal foraging and the evolution of goal-directed cognition

    Cognitive Science

    (2006)
  • E.E. Johnson et al.

    The lie/bet questionnaire for screening pathological gamblers

    Psychological Reports

    (1997)
  • J.G. Johnson et al.

    Beyond a trait view of risk-taking: A domain-specific scale measuring risk perceptions, expected benefits, and perceived-risk attitude in German-speaking populations

    Polish Psychological Bulletin

    (2004)
  • J. Joukhador et al.

    Superstitious beliefs in gambling among problem and non-problem gamblers: Preliminary data

    Journal of Gambling Studies

    (2004)
  • R.C. Kessler et al.

    DSM-IV pathological gambling in the National Comorbidity Survey Replication

    Psychological Medicine

    (2008)
  • J. Krause et al.

    Living in groups: Oxford series in ecology and evolution

    (2002)
  • Cited by (26)

    • An empirical investigation of wagering behavior in a large sample of slot machine gamblers

      2020, Journal of Economic Behavior and Organization
      Citation Excerpt :

      Concrete policy implications also arise from our results, specifically in connection with behavioral “nudges” and their effects on problem gamblers. Turner et al. (2011) and Wilke et al. (2014) have shown that people with gambling problems are more likely to perceive patterns in random sequences and less likely to identify random sequences than people without gambling problems. Similarly, Orford et al. (2003), summarizing several studies, note that problem gamblers have a greater tendency to misunderstand randomness (e.g., believe random patterns should look random, believe in hot and cold numbers, exhibit the gambler's fallacy) and have a greater belief in their ability to affect an outcome (e.g., belief in luck, illusion of control, and experienced a big win early in their gambling history) than non-problem gamblers.

    • Decision-making in gambling disorder: Understanding behavioral addictions

      2017, Decision Neuroscience: An Integrative Perspective
    • Cognitive bias

      2016, The Curated Reference Collection in Neuroscience and Biobehavioral Psychology
    View all citing articles on Scopus

    The research was supported by grants from the National Center for Responsible Gaming and the T. Urling and Mabel Walker Research Fellowship Program of Northern New York, which were both awarded to the first author. The National Science Foundation (DBI-0926568) provided financial support to the fourth author.

    View full text