Aisea e sili atu ai ona fai ma vaisu nai lo isi: O Aafiaga o le Taimi ma Totogi i luga o le Tumau i se Taaloga Masini Slot (2016)

Richard J. E. James,¹ Claire O’Malley,^1,2 ma Richard J. Tunney^1,*

Faamatalaga tusitala ► Mataupu tusiga ► Faamatalaga o le Puletaofia ma Laisene ►

Manipulating different behavioral characteristics of gambling games can potentially affect the extent to which individuals persevere at gambling, and their transition to problematic behaviors. This has potential impact for mobile gambling technologies and responsible gambling interventions. Two laboratory models pertinent to this are the partial reinforcement extinction effect (PREE) and the trial spacing effect. Both of these might speed up or delay the acquisition and extinction of conditioned behavior. We report an experiment that manipulated the rate of reinforcement and inter trial interval (ITI) on a simulated slot machine where participants were given the choice between gambling and skipping on each trial, before perseverative gambling was measured in extinction, followed by measurements of the illusion of control, depression and impulsivity. We hypothesized that longer ITI’s in conjunction with the low rates of reinforcement observed in gambling would lead to greater perseverance. We further hypothesized, given that timing is known to be important in displaying illusory control and potentially in persevering in gambling, that prior exposure to longer intervals might affect illusions of control. An interaction between ITI and rate of reinforcement was observed, as low reinforced gamblers with a long ITI gambled for longer. Respondents also displayed extinction and a PREE. Gamblers exposed to a higher rate of reinforcement gambled for longer in acquisition. Impulsivity was associated with extended perseverance in extinction, and more depressed gamblers in the high reinforcement short ITI group persevered for longer. Performance in the contingency judgment failed to support the second hypothesis: the only significant contrast observed was that participants became better calibrated as the task progressed.

Delay, Trial Spacing, and ITIs

Increasing the interval between gambles might be instrumental in encouraging continued play and may be a component behind the popularity of certain games. Lottery games for example have extended delays between gambles and are often the most popular and frequently played games (Wardle et al., 2011). While this might be because lotteries are highly available (amongst numerous considerations), in some jurisdictions (e.g., the UK) other games are offered alongside lottery tickets (e.g., scratchcards), controlling for availability. Despite this, many more people play the lottery than similarly available games, and do so more frequently. However, the perceived risk of harm is very low, although it is unclear whether the ‘addictiveness’ of gambling lies in specific games (Afifi et al., 2014) or specific behavioral features (Griffiths ma Auer, 2013). Some mobile video games exploit similar effects by enforcing delays between plays of gambling-like games. In-play betting, which is associated with mobile (Hing et al., 2014) and problem gambling (Grey et al., 2012; LaPlante et al., 2014), combines continuous and discontinuous play. Understanding the role of timing and latency on gambling behavior has important consequences for newer forms of gambling, such as mobile gambling (where in-play betting is heavily promoted), as the manner in which people use smartphones is likely to increase latencies between gambles. In-play refers to bets made on an event (e.g., a soccer match) while the event itself is occurring whereas in traditional forms of betting the wager is made prior to the event. Griffiths ma Auer (2013) argue in-play betting might be more addictive because it is more continuous. However, considerable discontinuities persist in play as betting remains constrained within an event. Real data on in-play betting (LaBrie et al., 2007) reveals mixed findings: although there is a clear risk of problem gambling, the findings do not decisively conclude this is because of its continuous nature; in-play gamblers placed fewer bets and there was little difference in daily levels of betting. Although in-play bettors wagered more money overall, the median wagered was lower than traditional sports betting, and in-play bettors had a lower net loss. Gray et al. (2012) suggest the immediacy between wager and outcome may be instrumental in attracting risky or impulsive gamblers to in-play gambling.

The associative learning literature indicates that increased latencies between reinforcements facilitate acquisition of conditioned behaviors (Gallistel ma Gibbon, 2000). Gallistel and Gibbon’s (2000) timing model hypothesizes that a decrease in the ratio between reinforcements and ITI in classical and operant conditioning reduces the number of reinforcements to acquisition. This is claimed to be independent of partial reinforcement, which increases the number of tofotofoga ae leai tagata faʻamalosia. The literature on the ‘trial spacing’ effect, primarily studied in the context of classical conditioning (Barela, 1999; Stout et al., 2003; Sunsay et al., 2004; Moody et al., 2006; Sunsay and Bouton, 2008; Miguez et al., 2014), has found that dispersed trials facilitate conditioning.

It is less clear whether greater latencies in extinction affect performance. Gallistel and Gibbon (2000) claim that the interval without reinforcement rather than non-reinforcing events is key, and that omitted reinforcements in extinction are unaffected by partial reinforcement. Other research has identified ITI effects on extinction, with greater suppression of responding observed with shorter ITI’s (Mackintosh, 1974; Moody et al., 2006).

Timing is thought to be an important component of the illusion of control (Msetfi et al., 2005, 2007; Baker et al., 2010), a cognitive bias that is prevalent in problem gambling (Fortune ma Goodie, 2012). Illusions of control, operationalized as an overestimation of the relationship between a response and outcome, can be induced using a contingency judgment task in which these events are unrelated but the outcome occurs very frequently. Standard examples of this task include a button pushing task associated with the activation of a light (Alloy and Abramson, 1979), or a medical decision-making task judging the relationship between an experimental drug and patient improvement (Orgaz et al., 2013). The extent to which non-depressed individuals show illusions of control is affected by the latency between trials: longer ITI’s are associated with stronger illusory control in non-depressed individuals (Msetfi et al., 2005). Problem gamblers show stronger illusions of control in contingency judgment paradigms (Orgaz et al., 2013), although the causal direction of this relationship remains unclear: extensive exposure to certain schedules of reinforcement might increase illusions of control, or individuals susceptible to illusions of control may be more likely to develop gambling problems. We included a task derived from the same paradigm as Orgaz et al. (2013), which participants were asked to complete after the slot machine task. We also measured depression as depressed individuals appear to make more calibrated judgments in this paradigm (Alloy and Abramson, 1979) with a longer ITI (Msetfi et al., 2005). Disordered mood has also been identified as a potential pathway to problem gambling (Blaszczynski ma Nower, 2002).

Partial Reinforcement Extinction Effect and Impulsivity

The partial reinforcement extinction effect (PREE) is a behavioral paradox in which weakly reinforced behaviours persist for longer without reinforcement relative to more consistently occurring reinforcers (Mackintosh, 1974; Bouton et al., 2014), such as during an extended period of losses in gambling (Dickerson, 1979; Fantino et al., 2005; Horsley et al., 2012). Partial reinforcement deficits have been identified in high frequency gamblers¹, who take longer than recreational gamblers to extinguish these associations (Horsley et al., 2012), a change that might occur from chronic exposure to the schedules of reinforcement in gambling. Horsley et al. (2012) report that although partial reinforcement is hypothesized to be an important component in gambling, the evidence base is sparse. Failure to extinguish has been identified as a marker of problem gambling (Weatherly et al., 2004). Failure to extinguish also directly (e.g., unsuccessful efforts to stop gambling, gambling more than intended to) or indirectly (e.g., chasing losses) corresponds onto indicators for Gambling Disorder (American Psychiatric Association, 2013) or problem gambling (Lesieur ma Blume, 1987).

It is unsurprising that the PREE has been linked with gambling, and considerable attention has been devoted to studying this in slot machines. Slot machines tend to have a very low rate of reinforcement (although this varies on computerized machines), and gamblers persevere in play despite mounting sequences of losses. There is a literature that has used slot machine tasks to probe the effects of partial reinforcement on operant learning. Lewis and Duncan (1956, 1957, 1958a,b) conducted a series of experiments using simulated gambling to test theories of partial reinforcement, finding that lower reward probabilities were associated with greater perseverance. Poon and Halpern (1971) used a similar paradigm to test Capaldi’s (1966; Capaldi and Martins, 2010) partial reinforcement theories by manipulating trial order in a slot machine task with a small number of acquisition trials. Kassinove and Schare (2001) manipulated big wins and near-misses in perseverative behavior in extinction in a similar slot machine paradigm, finding that near-miss density affected the extent to which participants persisted gambling but not big wins.

Different schedules of reinforcement potentially affect how behaviors extinguish (Madden et al., 2007; Haw, 2008a) Gambling operates on a random ratio schedule of reinforcement, a subset of the variable ratio schedule. Less well understood than variable ratio schedules, it is informative to contrast how random ratio schedules differ from variable ratio schedules. The typical distribution the number of trials until a response is reinforced on a random ratio schedules follows an L-shaped pattern; the number of trials rapidly drops off after a small number of plays but continues indefinitely at very low probability. In contrast on a variable ratio schedule it is usually (but not necessarily) the case that the probability of the number of trials to reinforcement is evenly distributed, and there is an upper limit on the number of trials before a behavior is reinforced (Haw, 2008a). Studies comparing these schedules have not shown clear differences; Hurlburt et al. (1980) found no difference between variable and random ratio schedules in gambling, although weaknesses with this study have been identified (Haw, 2008b). Crossman et al. (1987) found no difference between three ratio reinforcement schedules (variable, fixed and random) in animals. Recent studies have suggested that random-ratio schedules demonstrate more perseverative behavior compared to fixed-ratio schedules, particularly when the number of trials to reinforcement is very large (Madden et al., 2007).

The slot machine task we outline in this report was designed so that participants were asked to risk money they had won during the experiment, but the amount of money won would gradually increase. The low-reinforcement conditions attempted to create a situation similar to real-money gambling. One criticism of many slot machine experiments was that these studies tended to utilize a high rate of reinforcement relative to real slot machines (Kassinove ma Schare, 2001; Harrigan, 2007). A mechanical three-reel slot machine has a win probability of 9%, but this varies on computerized machines (Harrigan ma Dixon, 2009). In gambling research (e.g., MacLin et al., 2007; Dixon et al., 2011) higher rates of reinforcement (20%) have been used in extinction paradigms. We decided to use a rate of reinforcement of 30%, operating on a random ratio schedule of reinforcement similar to real slot machine gambling.

Self-reported impulsivity was measured impulsivity predicts perseverative gambling in the face of mounting losses, and is a pathway to problem gambling. Breen and Zuckerman (1999) found that impulsive gamblers ‘chased’ losses for longer in a gambling game where the win probability decreased as the experiment continued. Impulsivity has been identified as risk factor for problem gambling, problem gamblers (MacLaren et al., 2011; Kräplin et al., 2014) show higher self-reported impulsivity.

To test whether these behavioral effects encourage perseverative gambling, we conducted a two-part experiment where ITI and rate of reinforcement were manipulated. Participants were assigned to one of four groups and exposed to a high or low rate of reinforcement, and a long or short ITI between gambles. Associations were extinguished after a certain amount of money had been won. Participants subsequently completed a contingency judgment task in which they judged the efficacy of an experimental drug. The literature on partial reinforcement predicts that individuals exposed to a lower rate of reinforcement will persevere longer. Trial based accounts of extinction predict that massed extinction trials should suppress responding faster, as opposed to a timing-based account where there ought to be no difference. Impulsive gamblers should persevere for longer in extinction as well, on the basis of previous experiments looking at perseverance in loss-chasing.

mamanu

The experiment was a 2 × 2 between-subjects factorial design, the rate of reinforcement and ITI were the factors manipulated. The rates of reinforcement were 0.7 and 0.3. ITIs were either long (10 s) or short (3 s).

On every trial the participants were given the choice either to gamble or not. The number of trials in which participants decided to gamble was the dependent variable. The outcome of the gamble and the amount of money participants had won was also recorded. The extinction phase was divided into blocks of 10 trials for analysis. Participants were also administered a contingency judgment task. In the contingency judgment task measures were of the proportion of trials in which the drug was administered, and the contingency judgment made by participants. Impulsivity and depression were measured using the Barratt Impulsiveness Scale (BIS-11; Patton et al., 1995) and Beck Depression Inventory (BDI; Beck et al., 1961) respectively. The BIS-11 is a 30-item measure that measures three higher order factors of attentional, non-planning, and motor impulsivity (Patton et al., 1995). The BDI is a 24-item measure that measures multiple levels of depression severity, discriminates depression from anxiety and has strong internal consistency (Beck et al., 1988). No further measurements of individual difference or behavior were taken apart from the ones reported herein.

Tagata e Auai

A total of 122 participants were recruited from the University of Nottingham community to take part in this study (Mean age = 22.63, SD = 3.96, gender – 69 females and 53 males)². This study was carried out in accordance of, and with ethical approval by the University of Nottingham School of Psychology Ethics Review Committee. All participants gave written consent prior to the beginning of the experiment.

There was no evidence of any trait differences between the groups. A one-way Analysis of Variance (ANOVA) was conducted on both questionnaires, and the ANOVAs for the BIS [F(4,166) = 1.543, p = 0.192] and the BDI [F(4,166) = 0.662, p = 0.619] were non-significant.

taualumaga

Participants were randomly assigned to one of four conditions. For the first part of the experiment, participants were asked to participate in a PREE paradigm in the context of a simulated slot machine (ata Figure11). Participants were told how the slot machine worked, and the magnitude of the payoff for each type of winning outcome. The simulated slot machine was a simple one-line slot machine with three reels. Participants won money if the icons on three reels matched. There were five different icons (lemon, cherry, pear, orange, and lucky seven), with winning values of 10, 15, 20, 25, and 30p. The likelihood of each winning outcome occurring was the same, so the mean winning outcome was 20p ($0.35).

 

Faʻaaliga 1

Screenshot of the slot machine display participants were given during the partial reinforcement task.

For each trial, participants were given the choice between gambling and skipping. The buttons were highlighted so that participants were aware of the two choices they had. Regardless of whether they chose to gamble or not, the images on the three reels presented on the screen refreshed every 500 ms to give the appearance of movement. At 1500, 3000, and 4500 ms, one of the reels (from left to right) stopped reeling. If the reels matched and the participant gambled, the participants was awarded money correspondent to value of the icons on the reel. If the reels did not match, they lost the wager they had made, which was fixed at 3p (£0.03, equivalent to around US$0.05). Wins and losses were accompanied by visual and auditory feedback which differed for each outcome. These noises were different if the participants skipped the gamble. Throughout the task participants were informed of their current balance. Between each trial, the buttons on the screen remained red, signifying that the participants were unable to make another wager. The ITI for the short ITI condition was 3000 ms, and 10000 ms for the long ITI condition.

Participants were presented with 10 practice trials before the game began crediting or deducting money from the player. Participants were informed when the practice trials had ended. Once the experimental trials began, participants played until they reached criterion, set as having won than £10.00 (US$15.40) in the bank. Once participants reached criterion, they were exposed to 50 trials of extinction, where it was not possible to win any money from the slot machine, and then the task ended automatically. Extinction was measured by the suppression of their gambling behavior; participants were not informed of the extinction phase at the end of the experiment. The practice trials had winning trials (which did not pay out), and the extinction phase had no wins or money. The practice and extinction phases were identical in each condition, bar the different ITI’s participants were exposed to.

After completing the PREE paradigm, participants were asked to make a series of contingency judgments about the effectiveness of a fictitious experimental drug related to patient recovery. The contingency judgment paradigm was adapted from a previously published study (Orgaz et al., 2013). In this paradigm participants were presented information about a fictional drug that was designed to cure a fictional infectious skin disease that had unpleasant consequences when an outbreak/crisis occurred. Participants were given the option of choosing between administering the drug and not administering the drug, and they were given feedback concerning the outcome immediately afterward (whether the patient’s situation had improved or not). The paradigm was designed to elicit illusions of control by having a high outcome density – the base rate of the desired outcome (patient recovered) was high (0.8), and was completely independent of the users decision. After making their decision, the participants were informed of the outcome of the choice, and there was a small pause (3500 ms) before being presented with the decision again.

After each set of 10 trials, participants were asked to judge the effectiveness of the drug. Participants were asked to judge the effectiveness of the drug on a scale from 0 to 100. This was represented by a shaded bar in the middle of the screen, on which they were given feedback about the number they chose, determined by how far along the bar they clicked. Participants could repeat clicking along the slider until they were happy with their choice, and were asked to confirm their choice using a separate button.

Analytic Approach

To assess the length of extinction for each group, the proportion of gambles made were averaged across five blocks of 10 trials. Data analysis proceeded in two stages. Firstly, factorial ANOVAs were conducted on the extinction and contingency judgment data, with a 5 (block) × 2 (ITI) × 2 (Rate of Reinforcement) mixed design ANOVA being conducted. A 10 × 2 × 2 mixed design ANOVA was carried out on the 10 contingency judgments participants made. To test the effects of individual differences on gambling behavior and perseverative gambling, a series of poisson regression models were estimated on the number of trials participants gambled on during acquisition and extinction. This was conducted in three steps. First, an initial model was constructed where no covariates were entered into the model. Then, a second regression model was constructed in which ITI, rate of reinforcement, BIS scores, BDI scores and an interaction term between ITI and rate of reinforcement were included. ITI and rate of reinforcement were dummy coded (high ROR = 1, low = 0; short ITI = 1, long = 0), and BIS/BDI scores were rescaled with a mean of 0. This was compared against a null model using a likelihood ratio test (LRT). LRT’s are typically used in latent variable modeling to compare between two nested models, for example in latent class analysis (Collins and Lanza, 2010), or between the fit of two regression models, as in this case. This was then compared against a full model in which interaction terms were modeled across each covariate.

At this point, the data was tested to examine whether the data fit a poisson distribution. Crucially, poisson regression assumes that the conditional mean and variance are equal. While deviations from this assumption have little effect on the overall regression coefficients, when overdispersion (the variance being larger than the mean) is substantial this tends to depress standard errors, increasing the risk of false positive findings. While robust standard errors can be used to adjust these (Cameron and Trivedi, 2009), an alternative is to estimate a negative binomial regression model, which includes an extra parameter to model overdispersion. For the acquisition data, this approach was taken. For the extinction data, while the data was overdispersed the level of dispersion was considerably less, and so robust standard errors were applied to the regression model.

A number of outliers were found in the low rate of reinforcement extinction data. An examination of the data indicated that a number of gamblers in the low reinforcement, long ITI condition stopped gambling less than two gambles into extinction occurring and that these were outlying data points. These participants (n = 3) reported in debrief they treated £10 as salient, either stopping immediately after they won £10 or stopped to remain above £10, independent of any change in contingency. These participants were excluded from further analysis.

Gambling Behavior

To study the effect of behavioral and trait variables on acquisition behavior, an offset negative binomial regression model was used to control for differential effects of exposure, where the same variables were used for the restricted and full factorial models as the extinction data. These revealed that the restricted model (laulau Table11) was a better fit than the null model (G² = 22.74, p < 0.001), but that a full factorial model was no better fit than the restricted model (G² = 6.359, p = 0.784). This revealed that participants exposed to a higher rate of reinforcement gambled more frequently in acquisition.

 

Laulau 1

Offset negative binomial regression model of acquisition data.

PREE Task

The ANOVA conducted on the extinction data revealed main effects of block, F(2.541,292.187) = 131.095, p <0.001, η2p = 0.533, where the linear contrast was significant, F(1,115) = 229.457, p <0.001, η2p = 0.666, and the rate of reinforcement, F(1,115) = 82.912, p <0.001, η2p = 0.419, but no main effect of ITI, F(1,115) = 1.455, p = 0.23. There was an interaction between block and rate of reinforcement, F(2.541,292.187) = 22.801, p <0.001, η2p = 0.165, and a further interaction between the rate of reinforcement and ITI, F(1,115) = 6.317, p = 0.0133, η2p = 0.052. There was no interaction between block and ITI, F(2.541,292.187) = 1.124, p = 0.334, or a three-way interaction, F(2.541,292.187) < 1. The main effect of block indicated that responses decreased as the block number increased (i.e., participants extinguished). This interacted with rate of reinforcement, as participants exposed to a higher rate of reinforcement extinguished more quickly, suggesting the presence of a PREE. The main effect of rate of reinforcement signified the same finding. The rate of reinforcement and ITI interaction indicated that when there was a low rate of reinforcement with a long ITI, participants gambled for longer in extinction (ata Figure22). The block and rate of reinforcement effects, and the interaction between block and rate of reinforcement were all large in size (η2p > 0.12), whereas the interaction between rate of reinforcement and ITI interaction was a small to medium effect.

 

Faʻaaliga 2

Plot of extinction data for all groups, in blocks of 10 trials.

Tasi Eseesega

To test the role of individual differences in perseverative gambling, a poisson regression procedure was used on the number of gambles in extinction The LRT indicated that the initial restricted model was a better fit of the data compared to the null model (G² = 581.15, p < 0.001). The restricted regression model (laulau Table22) indicated that lower rates of reinforcement and longer ITI’s predicted longer perseverative gambling. These terms interacted in the same manner as the factorial ANOVA. A further regression model including interaction terms between the different covariates was subsequently conducted (laulau Table33) with the same variables as the regression in laulau Table11. A LRT comparing the restricted and full factorial regression models indicated that the full factorial model was a better fit of the data (G² = 66.44, p < 0.001). This revealed the same significant effects as previously, but also that higher self-reported impulsivity predicted longer perseverative gambling. There was a trend suggesting that this interacted with rate of reinforcement, with less impulsive individuals appearing to persevere less in low reinforcement conditions. Scores on the two psychometric measures interacted, and there was a three way interaction between ITI, rate of reinforcement and BDI, with more depressed individuals in the high rate of reinforcement, short ITI group gambling for longer in extinction (ata Figure33).

 

Laulau 2

Restricted poisson regression model of extinction data with robust standard errors.

 

Laulau 3

Full poisson regression model of extinction data with robust standard errors.

 

Faʻaaliga 3

Boxplot of depression status and proportion of gambles in extinction for each of the four conditions.

Contingency Judgment Task

Analysis of the contingency judgment data revealed that a significant main effect of block, F(6.526,737.416) = 3.735, p = 0.001, η2p = 0.032. The main effect of block also included a significant linear contrast, F(1,113) = 10.312, p = 0.002, η2p = 0.084, indicating that participants became better calibrated as they subsequently made judgments about the efficacy of the drug (ata Figure44). Main effects of ITI, F(1,113) < 1, and rate of reinforcement, F(1,113) < 1, were not observed. Interactions between block and ITI, F(6.526, 737.415) < 1, block and rate of reinforcement, F(6.526,737.415) < 1, and ITI and rate of reinforcement, F(1,113) = 1.109, p = 0.295, were not significant. A three way interaction between block, rate of reinforcement, F(6.526,737.416) = 1.048, p = 0.399, was not significant either.

 

Faʻaaliga 4

Plot of mean contingency judgments across the 10 judgments participants made.

Faatupeina. This research contained in this report was funded by the Economic and Social Research Council (ES/J500100/1) and Engineering and Physical Sciences Research Council (EP/GO37574/1).

¹This study reports that their sample of high-frequency gamblers (n = 19) contained only three pathological gamblers, and the mean number of DSM-IV Pathological Gambling criteria endorsed was 2.3, indicating this is a difference found in low to moderate levels of problematic gambling.

²A number of participants across conditions dropped out (n = 18). Participants who withdrew were resampled. All the participants who dropped out completed measures of depression and impulsivity. The majority of these dropouts (82%) were in the low rate of reinforcement, high ITI condition. Non-parametric tests were carried out to test whether the participants who dropped out differed from other participants from the same condition in any regard. No significant differences were observed in impulsivity or depression scores, nor the rate they were gambling prior to dropping out (Wilcoxon’s signed rank test, p > 0.05). All participants were debriefed upon withdrawal from the experiment. Participants who dropped out reported that they withdrew from the experiment because the length of the study conflicted with other engagements (e.g., lectures).

• Abramson L. Y., Garber J., Edwards N. B., Seligman M. E. (1978). Expectancy changes in depression and schizophrenia. J. Abnorm. Toma. 87 102. [PubMed]
• Afifi T. O., LaPlante D. A., Taillieu T. L., Dowd D., Shaffer H. J. (2014). Gambling involvement: considering frequency of play and the moderating effects of gender and age. Fa. J. Mental Health Addict. 12 283–294. 10.1007/s11469-013-9452-3 [Cross Ref]
• Alloy L. B., Abramson L. Y. (1979). Judgment of contingency in depressed and nondepressed students: Sadder but wiser? J. Exp. Psychol. Gen. 108 441. [PubMed]
• American Psychiatric Association (2013). Diagnostic and Statistical Manual of Mental Disorders, (DSM-5^®). Uosigitone, DC: American Psychiatric Publishing.
• Baker A., Msetfi R. M., Hanley N., Murphy R. (2010). “Depressive realism? sadly not wiser,” in Clinical Applications of Learning Theory, eds Haselgrove M., Hogarth L., editors. (Hove: Psychology Press; ), 153–179.
• Barela P. B. (1999). Theoretical mechanisms underlying the trial-spacing effect in Pavlovian fear conditioning. J. Exp. Toma. Manu. Faʻatasi. Faʻamasinoga. 25 177. [PubMed]
• Beck A. T., Steer R. A., Carbin M. G. (1988). Psychometric properties of the Beck Depression Inventory: twenty-five years of evaluation. Falemaʻi. Toma. Faaa. 8 77-100.
• Beck A. T., Ward C. H., Mendelson M. M., Mock J. J., Erbaugh J. J. (1961). An inventory for measuring depression. Faʻau. Gen. Psychiatry 4 561–571. 10.1001/archpsyc.1961.01710120031004 [PubMed] [Cross Ref]
• Blaszczynski A., Cowley E., Anthony C., Hinsley K. (2015). Breaks in play: do they achieve intended aims? J. Gambl. Stud. [Epub ahead of print]10.1007/s10899-015-9565-7 [PubMed] [Cross Ref]
• Blaszczynski A., Nower L. (2002). A pathways model of problem and pathological gambling. onāga i vailāʻau 97 487-499. [PubMed]
• Bouton M. E., Woods A. M., Todd T. P. (2014). Separation of time-based and trial-based accounts of the partial reinforcement extinction effect. Behav. Process. 101 23-31. [PMC free article] [PubMed]
• Breen R. B., Zuckerman M. (1999). Chasing’in gambling behavior: personality and cognitive determinants. Pers. Ta'itoatasi. Eseesega. 27 1097-1111.
• Cameron A. C., Trivedi P. K. (2009). Microeconometrics Using Stata. College Station, TX: Stata Press.
• Campbell-Meiklejohn D. K., Woolrich M. W., Passingham R. E., Rogers R. D. (2008). Knowing when to stop: the brain mechanisms of chasing losses. Biol. Fomaʻi 63 293-300. [PubMed]
• Capaldi E. J. (1966). Partial reinforcement: a hypothesis of sequential effects. Toma. Faaa. 73 459. [PubMed]
• Capaldi E. J., Martins A. P. (2010). Applying memories of reinforcement outcomes mainly to Pavlovian conditioning. Aoao. Motugaafa. 41 187-201.
• Coates E., Blaszczynski A. (2014). Predictors of return rate discrimination in slot machine play. J. Gambl. Suesuega. 30 669–683. 10.1007/s10899-013-9375-8 [PubMed] [Cross Ref]
• Collins L. M., Lanza S. T. (2010). Latent Class and Latent Transition Analysis: With Applications in the Social, Behavioral and Health Sciences. Hoboken, NJ: John Wiley and Sons.
• Crossman E. K., Bonem E. J., Phelps B. J. (1987). A comparison of response patterns on fixed-, variable-, and random-ratio schedules. J. Exp. Anal. Faʻatasi. 48 395-406. 10.1901 / jeab.1987.48-395 [PMC free article] [PubMed] [Cross Ref]
• Dickerson M. G. (1979). FI schedules and persistence at gambling in the UK betting office. J. Appl. Behav. Anal. 12 315-323. [PMC free article] [PubMed]
• Dickerson M. G. (1984). Compulsive Gamblers. London: Addison-Wesley Longman Ltd.
• Dixon M. J., Fugelsang J. A., MacLaren V. V., Harrigan K. A. (2013). Gamblers can discriminate ‘tight’ from ‘loose’ electronic gambling machines. Int. Gambl. Stud. 13 98-111. 10.1080 / 14459795.2012.712151 [Cross Ref]
• Dixon M. J., Harrigan K. A., Jarick M., MacLaren V., Fugelsang J. A., Sheepy E. (2011). Psychophysiological arousal signatures of near-misses in slot machine play. Int. Gambl. Stud. 11 393-407.
• Dymond S., McCann K., Griffiths J., Cox A., Crocker V. (2012). Emergent response allocation and outcome ratings in slot machine gambling. Toma. Faʻaaoga. Faʻatasi. 26 99-111. 10.1037 / a0023630 [PubMed] [Cross Ref]
• Fantino E., Navarro A., O’Daly M. (2005). The science of decision-making: behaviours related to gambling. Int. Gambl. Stud. 5 169-186.
• Fortune E. E., Goodie A. S. (2012). Cognitive distortions as a component and treatment focus of pathological gambling: a review. Toma. Faʻaaoga. Faʻatasi. 26 298. [PubMed]
• Gallistel C. R., Gibbon J. (2000). Time, rate, and conditioning. Toma. Faaa. 107 289. [PubMed]
• Gray H. M., LaPlante D. A., Shaffer H. J. (2012). Behavioral characteristics of Internet gamblers who trigger corporate responsible gambling interventions. Toma. Faʻaaoga. Faʻatasi. 26 527-535. 10.1037 / a0028545 [PubMed] [Cross Ref]
• Griffiths M. D., Auer M. (2013). The irrelevancy of game-type in the acquisition, development and maintenance of problem and pathological gambling. I luma. Toma. 3: 621 10.3389 / fpsyg.2012.00621 [PMC free article] [PubMed] [Cross Ref]
• Harrigan K. A. (2007). Slot machine structural characteristics: distorted player views of payback percentages. J. Gambl. Issues 215–234. 10.4309/jgi.2007.20.7 [Cross Ref]
• Harrigan K. A., Dixon M. J. (2009). PAR Sheets, probabilities, and slot machine play: implications for problem and non-problem gambling. J. Gambl. Issues 81–110. 10.4309/jgi.2009.23.5 [Cross Ref]
• Haw J. (2008a). Random-ratio schedules of reinforcement: the role of early wins and unreinforced trials. J. Gambl. Issues 56–67. 10.4309/jgi.2008.21.6 [Cross Ref]
• Haw J. (2008b). The relationship between reinforcement and gaming machine choice. J. Gambl. Suesuega. 24 55–61. 10.1007/s10899-007-9073-5 [PubMed] [Cross Ref]
• Hayden B. Y., Platt M. L. (2007). Temporal discounting predicts risk sensitivity in rhesus macaques. Curr. Biol. 17 49-53. 10.1016 / j.cub.2006.10.055 [PMC free article] [PubMed] [Cross Ref]
• Hing N., Gainsbury S., Blaszczynski A., Wood R. T., Lubman D., Russell A. (2014). Interactive Gambling. Report for Gambling Research Australia by the Office of Liquor Melbourne, VC: Gaming and Racing Department of Justice; Available at: http://www.gamblingresearch.org.au/resources/6482d5fa-f068-41e5-921f-facd4f10365e/interactive+gambling.pdf [Accessed August 1 2014].
• Horsley R. R., Osborne M., Norman C., Wells T. (2012). High-frequency gamblers show increased resistance to extinction following partial reinforcement. Faʻaaloalo. Brain Res. 229 438-442. [PubMed]
• Humphreys L. G. (1939). The effect of random alternation of reinforcement on the acquisition and extinction of conditioned eyelid reactions. J. Exp. Psychol. 25 141.
• Hurlburt R. T., Knapp T. J., Knowles S. H. (1980). Simulated slot-machine play with concurrent variable ratio and random ratio schedules of reinforcement. Toma. Sui. 47 635–639. 10.2466/pr0.1980.47.2.635 [Cross Ref]
• Kassinove J. I., Schare M. L. (2001). Effects of the “near miss” and the “big win” on persistence at slot machine gambling. Toma. Faʻaaoga. Faʻatasi. 15 155. [PubMed]
• Kräplin A., Dshemuchadse M., Behrendt S., Scherbaum S., Goschke T., Bühringer G. (2014). Filifiliga le lelei i le faia o taaloga faʻapitoa: faʻapitoa faʻamaʻi ma le matafaioi o le le mautonu. Maualuga Resitala. 215 675-682. [PubMed]
• LaBrie R. A., LaPlante D. A., Nelson S. E., Schumann A., Shaffer H. J. (2007). Assessing the playing field: a prospective longitudinal study of internet sports gambling behavior. J. Gambl. Suesuega. 23 347–362. 10.1007/s10899-007-9067-3 [PubMed] [Cross Ref]
• LaPlante D. A., Nelson S. E., Gray H. M. (2014). Breadth and depth involvement: understanding Internet gambling involvement and its relationship to gambling problems. Toma. Faʻaaoga. Faʻatasi. 28 396-403. 10.1037 / a0033810 [PubMed] [Cross Ref]
• Lesieur H., Blume S. (1987). The South Oaks Gambling Screen (SOGS): a new instrument for the identification of pathological gamblers. Am. J. Fomaʻi 144 1184-1188. [PubMed]
• Lewis D. J., Duncan C. P. (1956). Effect of different percentages of money reward on extinction of a lever-pulling response. J. Exp. Psychol. 52 23. [PubMed]
• Lewis D. J., Duncan C. P. (1957). Expectation and resistance to extinction of a lever-pulling response as functions of percentage of reinforcement and amount of reward. J. Exp. Psychol. 54 115. [PubMed]
• Lewis D. J., Duncan C. P. (1958a). Expectation and resistance to extinction of a lever-pulling response as a function of percentage of reinforcement and number of acquisition trials. J. Exp. Psychol. 55 121. [PubMed]
• Lewis D. J., Duncan C. P. (1958b). Vicarious experience and partial reinforcement. J. Abnorm. Soc. Toma. 57 321. [PubMed]
• Linnet J., Rømer Thomsen K., Møller A., Callesen M. B. (2010). Event frequency, excitement and desire to gamble, among pathological gamblers. Int. Gambl. Stud. 10 177-188. 10.1080 / 14459795.2010.502181 [Cross Ref]
• Mackintosh N. J. (1974). The Psychology of Animal Learning. Oxford: Academic Press.
• MacLaren V. V., Fugelsang J. A., Harrigan K. A., Dixon M. J. (2011). The personality of pathological gamblers: a meta-analysis. Falemaʻi. Toma. Faaa. 31 1057-1067. [PubMed]
• MacLin O. H., Dixon M. R., Daugherty D., Small S. L. (2007). Using a computer simulation of three slot machines to investigate a gambler’s preference among varying densities of near-miss alternatives. Amio. Re. Metotia 39 237-241. [PubMed]
• Madden G. J., Ewan E. E., Lagorio C. H. (2007). Toward an animal model of gambling: delay discounting and the allure of unpredictable outcomes. J. Gambl. Suesuega. 23 63–83. 10.1007/s10899-006-9041-5 [PubMed] [Cross Ref]
• Miguez G., Witnauer J. E., Laborda M. A., Miller R. R. (2014). Trial spacing during extinction: the role of context–us associations. J. Exp. Psychol. Anim. Learn. Cogn. 40 81. [PubMed]
• Miller N. V., Currie S. R., Hodgins D. C., Casey D. (2013). Validation of the problem gambling severity index using confirmatory factor analysis and rasch modelling. Int. J. Methods Psychiatr. Res. 22 245-255. [PubMed]
• Moody E. W., Sunsay C., Bouton M. E. (2006). Priming and trial spacing in extinction: effects on extinction performance, spontaneous recovery, and reinstatement in appetitive conditioning. Q. J. Exp. Psychol. 59 809-829. [PubMed]
• Msetfi R. M., Murphy R. A., Simpson J. (2007). Depressive realism and the effect of intertrial interval on judgements of zero, positive, and negative contingencies. Q. J. Exp. Psychol. 60 461-481. [PubMed]
• Msetfi R. M., Murphy R. A., Simpson J., Kornbrot D. E. (2005). Depressive realism and outcome density bias in contingency judgments: the effect of the context and intertrial interval. J. Exp. Toma. Gen. 134 10. [PubMed]
• Orford J., Wardle H., Griffiths M., Sproston K., Erens B. (2010). PGSI and DSM-IV in the 2007 British Gambling Prevalence Survey: reliability, item response, factor structure and inter-scale agreement. Int. Gambl. Stud. 10 31-44.
• Orgaz C., Estévez A., Matute H. (2013). Pathological gamblers are more vulnerable to the illusion of control in a standard associative learning task. I luma. Toma. 4: 306 10.3389 / fpsyg.2013.00306 [PMC free article] [PubMed] [Cross Ref]
• Patton J. H., Stanford M. S., Barratt E. S. (1995). Factor structure of the Barratt impulsiveness scale. J. Clin. Toma. 51 768-774. [PubMed]
• Poon L., Halpern J. (1971). A small-trials PREE with adult humans: resistance to extinction as a function of number of NR transitions. J. Exp. Psychol. 91 124.
• Sharpe L. (2002). A reformulated cognitive–behavioral model of problem gambling: a biopsychosocial perspective. Falemaʻi. Toma. Faaa. 22 1–25. 10.1016/S0272-7358(00)00087-8 [PubMed] [Cross Ref]
• Stout S. C., Chang R., Miller R. R. (2003). Trial spacing is a determinant of cue interaction. J. Exp. Toma. Manu. Faʻatasi. Faʻamasinoga. 29 23. [PubMed]
• Sunsay C., Bouton M. E. (2008). Analysis of a trial-spacing effect with relatively long intertrial intervals. Learn. Behav. 36 104-115. [PubMed]
• Sunsay C., Stetson L., Bouton M. E. (2004). Memory priming and trial spacing effects in Pavlovian learning. Learn. Behav. 32 220-229. [PubMed]
• Wardle H., Moody A., Spence S., Orford J., Volberg R., Jotangia D., et al. (2011). British Gambling Prevalence Survey 2010. London: The Stationery Office.
• Weatherly J. N., Sauter J. M., King B. M. (2004). The“ big win” and resistance to extinction when gambling. J. Psychol. 138 495-504. [PubMed]