Event frequency, excitement and desire to gamble, among pathological gamblers

Abstract

In this study we compared gambling behaviour of 15 pathological gamblers (PG) and 15 non-problem gamblers (NPG) on two conditions of a commercially available slot machine. One condition used a commercially available two-second event frequency (games per minute), while the other condition used an experimental three-second event frequency. The payback percentage (wins relative to losses) and reward frequency (wins over number of games played) varied randomly across conditions. The results showed that PG had significantly higher measures than NPG on time spent gambling, excitement level and desire to play again in the two-second condition. In the three-second condition there were no differences in excitement level and desire to play again. The number of PG playing the maximum time (60 minutes) was reduced in the three-second version, and reward frequency contributed to reduction in time spent gambling. The results may have implications for understanding behavioural mechanisms of pathological gambling among slot machine players.

Keywords:

• pathological gambling
• clinical psychology
• neuroimaging
• addiction
• addictive behaviour
• gambling

Pathological gambling is an impulse control disorder characterised by repeated maladaptive gambling behaviours that persist despite negative consequences or impaired social functioning of the individual (American Psychiatric Association [DSM-IV], 1994). Slot machine players represent the largest group of treatment seeking pathological gamblers, accounting for 35–93% of the treatment seeking population (Linnet, 2009; Meyer, Hayer & Griffiths, 2009; Petry, 2003).

The structural characteristics of slot machines may contribute to the high prevalence of pathological gambling among slot machine players (Griffiths, 1999; Parke & Griffiths, 2006a, 2007; Wood, Griffiths & Parke, 2007), although some studies raise questions about the addictive nature of slot machines (Dowling, Smith & Thomas, 2005). Structural characteristics are game properties, which “facilitate the acquisition, development, and/or maintenance of gambling behaviour irrespective of the individual's psychological, physiological, or socioeconomic status” (Parke & Griffiths, 2007, p. 212). Griffiths and colleagues (Griffiths, 1999; Parke & Griffiths, 2006a, 2007; Wood et al., 2007) have suggested a taxonomy dividing structural characteristics into factors that: (1) make gambling fun, interactive, and/or engaging; (2) relate to how one pays to gamble (‘payment’); (3) relate to how one receives financial rewards or winnings (‘reward’); (4) relate to the frequency, duration, and expediency of the game or reward; (5) educate or provide information to players (‘educational’); (6) may influence the immediate situation of the game or may contribute to other factors already mentioned (e.g. the use of colour and sound) (Parke & Griffiths, 2007, p. 215).

Research on structural characteristics in slot machines has focused on different aspects of gambling behaviour, for example frequent small gains and ‘near misses’ (Dixon & Schreiber, 2004; Kassinove & Schare, 2001; Maclin, Dixon, Daugherty & Small, 2007), infrequent big wins (Kassinove & Schare, 2001; Weatherly, Sauter & King, 2004; Wohl & Enzle, 2003), manipulation of symbol sequence and frequency of reward contingencies (Haw, 2008; Ladouceur & Sevigny, 2002, 2005; Loba, Stewart, Klein & Blackburn, 2001; Young, Wohl, Matheson, Baumann & Anisman, 2008; Zlomke & Dixon, 2006), and changes in bet size (Blaszczynski, Sharpe, Walker & Coughlan, 2005; Sharpe, Walker, Coughlan, Enersen & Blaszczynski, 2005).

Structural characteristics that reinforce gambling behaviour are particularly interesting, as they may strongly affect pathological gambling behaviour. They often have properties that resemble classical and operant conditioning. Learning theory of classical and operant conditioning (see, e.g., Schultz, 2006) suggests that optimal reinforcement of behaviour is achieved when conditioned stimuli: (1) precede the reward in close proximity of a few seconds (‘contiguity’); (2) signal a high probability or magnitude of reward (‘contingency’); and (3) are presented at random (‘random’ or ‘variable reward schedules’).

In slot machines reward contiguity is represented by ‘event frequency’, i.e. the time period between games. Slot machines have a very high event frequency of approximately two seconds per game. Incidentally, this is the optimal reward contiguity of reinforcement (Schultz, 2006). Most animal studies of behavioural conditioning use frequency of reward or punishment as reward contingency. Slot machines use two types of reward contingency: ‘reward frequency’ and ‘reward magnitude’. Reward frequency refers to how frequently rewards occur, i.e. the average number of games between reward, while reward magnitude refers to the size of reward or payback percentage, i.e. how much the person wins or loses. Reward frequency and payback percentage are variable reward schedules, i.e. both the frequency and magnitude of reward are random in slot machines. It is therefore never possible for the individual to predict when a reward is going to occur, prompting the individual to continue gambling. Finally, the payback percentage, i.e., the payout amounts relative to losses, is very high in slot machines, often around 70–95% (Parke & Griffiths, 2007), which is also believed to reinforce gambling behaviour.

In this study we compared pathological gamblers (PG) and non-problem gamblers (NPG) on two conditions of a commercially available slot machine: one condition used a commercially available two-second event frequency, while the other condition used an experimental three-second event frequency. Reward frequency and payback percentage varied randomly in the study. To determine behavioural reinforcement of the structural characteristics, we compared individuals on the time they spent gambling. We also investigated motivational factors of ‘excitement’ and ‘desire to play again’ in relation to the structural characteristics.

We hypothesized that pathological gamblers would have significantly higher excitement levels and desire to play again, and would spend significantly more time gambling compared with non-problem gamblers. We further hypothesised that lower event frequency, reward frequency and payback percentage would reduce excitement, desire to play again and time spent gambling.

Methods and materials

Participants

Fifteen pathological gamblers (five women and ten men) were recruited through the treatment Center for Ludomani [Centre for Pathological Gambling], which is the largest treatment facility for pathological gamblers in Denmark. Pathological gamblers were included in the study if they: (1) had a score of five or more on the South Oaks Gambling Screen (SOGS) (Lesieur & Blume, 1987, 1993); (2) were actively gambling; and (3) were primarily addicted to slot machines. We defined ‘active gambling’ as having gambled within the past month. Most pathological gamblers were in treatment at the time of participation. For these individuals controlled gambling behaviour was the stated or achieved treatment goal. Controlled gambling behaviour is an important therapeutic goal (Blaszczynski, McConaghy & Frankova, 1991; Dowling & Smith, 2007; Dowling, Smith & Thomas, 2009; Ladouceur, 2005), which is emphasised at Centre for Pathological Gambling. Due to time restrictions in the intake procedure we only used SOGS to access gambling pathology. The SOGS has been criticised for having too many false positive classifications of pathological gambling. However, Stinchfield (2002) found no false positives in SOGS scores of 10 and above. The lowest SOGS score of pathological gamblers in our sample was 7, and only three out of 15 pathological gamblers had a SOGS score below 10. Potential confounds as a result of false positives were therefore minimal in our population.

Fifteen non-problem gamblers (seven women and eight men) were recruited through local newspaper advertisement. They were included if they scored below two on the SOGS. Non-problem gamblers gambled infrequently and used small amounts of money; the gambling frequency could vary from weeks to months and even longer. The two groups were matched on age and gender. The average age of pathological gamblers was 43.6 years (SD = 10.20) and 38.5 years for non-problem gamblers (SD = 13.96), t(28) = 1.13, ns. The average SOGS score of pathological gamblers was 12.07 (SD = 2.52) and 0 (SD = 0.00) among non-problem gamblers, t(28) = 18.54, p < 0.000001. There were no significant differences in gender distribution, χ²(2, N = 30) = 0.56, ns. We used a ‘rolling intake’ of pathological gamblers and non-problem gamblers, which made it possible to match non-problem gamblers to pathological gamblers as the study progressed. Therefore, none of the participants were excluded based on matching criteria, although some were excluded for other reasons during screening. For instance, one person was too young to participate, but re-enrolled after he turned 18.

The study was approved by De Videnskabetiske Komiteer For Region Midtjylland [Midtjyllands Regional Science Ethics Committee], Denmark. All participants gave written informed consent prior to participating in the study and received financial compensation for their time and travel expenses. As described in the procedure section, participants who won could keep part of their gains, but they could not lose money. All participants had access to the treatment facilities at Centre for Pathological Gambling, in case they experienced an increase in gambling symptoms following the experiment, such as loss of control over gambling or a strong desire to gamble more or spend more time gambling. Participants were debriefed after the study and had access to on-call staff from the project. None of the participants showed increased gambling symptoms following the experiment.

Since slot machine gambling could potentially lead to problem gambling behaviour among non-problem gamblers or increase symptom severity in pathological gamblers, the study went through a rigorous review process, which involved several steps of approval by the local science ethics committee. This included: (1) gambling time was limited to a maximum of two hours (a literature search revealed no prior reports of exposure of this time frame leading to pathological gambling); (2) participants had no contact with money during the experiment; (3) the credits were ‘pre-loaded’ and provided at no cost to the participants; (4) potential gains were limited and it was not possible to lose money; (5) players could not activate the payout button; (6) the money participants won was added to their compensation, which was paid several days later, thereby weakening the association between winning and payout; (7) participants were carefully debriefed by experimenters with substantial psychological training, who were supervised by a clinical psychologist; (8) follow-up sessions were always offered after the experiment was completed; (9) researchers were on call in case participants experienced increased gambling urges or hazardous gambling episodes; and (10) the experimenters collaborated closely with the treatment centre, and could enrol participants in a treatment program if necessary. Neither pathological gamblers nor non-problem gamblers experienced an increase in gambling symptoms after participating in the experiment.

Procedure

Participants gambled on ‘Orientexpressen’ [the ‘Oriental Express’], one of the most popular Danish slot machines. Dansk Automatspil A/S, the slot machine subsidiary of Danske Spil A/S [Danish Games, Inc.], provided two versions of the slot machine in a collaboration that promoted responsible gambling. One machine, the ‘two-second condition’, used the commercially available settings, which included a two-second event frequency, i.e. two-second reel spin between games, and an average of 83% payback percentage, i.e. for every unit wagered the long-term return was 0.83 units or an average loss of 0.17 units per wager. In Danish Kroner, the value of one credit was DKK0.5 (around US$0.1). It was not possible to change the bet size on the slot machine, as Danish slot machines are mandated by law to have a maximum bet size of DKK0.5. The other machine, the ‘three-second condition’, used a three-second event frequency and an average of 50% payback percentage. The study originally had a 2 × 2 × 2 design (group × event frequency × payback percentage). However, it was not possible to detect differences in payback percentage between the 50% and the 83% condition over the study period. The study was therefore revised to have a 2 × 2 design (group × event frequency), comparing gambling behaviour of pathological gamblers and non-problem gamblers across the two-second and three-second conditions; payback percentage and reward frequency were used as covariates of time spent gambling.

Before the participants arrived, each machine was loaded with 2000–2400 credits (around US$200), which yielded a total gambling time of at least 60 minutes (one participant managed to lose all credits after 50 minutes and this was recorded as a way of ‘self-stopping’). To make the experiment resemble a real life gambling situation participants could keep 10% of the gains exceeding the pre-loaded amount, up to a limit of 400 credits (around US$35). Four out of 30 (13.3%) individuals won money in the two-second condition, three of those won the maximum amount; three out of 30 (10%) individuals won money in the three-second condition, two of those won the maximum amount.

Upon arrival, we presented the participants with the two slot machines placed next to each other. Participants were told that the slot machines differed in the way they responded, and that the objective of the study was to investigate slot machine gambling behaviour under different conditions. Participants were given the following instructions prior to gambling: “You can gamble as long as you want, and when you don't feel like gambling on this machine anymore, just let me know”, and “As a rule, you decide how long you want to gamble, but at some point I might stop you”. Participants were stopped after 60 minutes, but this information was not disclosed to them.

Participants unfamiliar with the slot machines had a training session before the experiment. Here, the rules were explained, and they had the opportunity to briefly play to secure game familiarity. All of the pathological gamblers knew the slot machine in advance, so in the end training was only administered to some of the non-problem gamblers.

Participants gambled on each slot machine in a randomised order to avoid a preference bias in the data, i.e. a consistently lower or higher rating of one machine over another. After the first gambling session participants were given a brief interview about their gambling experience, and immediately continued to the next session, or took a quick break, if needed. The slot machines had an ‘autoplay’ button, which is a common feature, that automatically spins the reels without players pressing the play button (Parke & Griffiths, 2006b, 2007). The reels only stop when players need to make decisions (e.g. hold or nudge). Gamblers often use this function to play on several machines at the same time. In the study participants were not allowed to use the autoplay function, to ensure maximum attention on the slot machine.

After each session participants were interviewed about their gambling experience. In the interview participants were asked to rate different aspects of their gambling experience using a ten point scale, ranging from 1 (least) to 10 (most). Participants were asked to rate how exciting it was for them to gamble (“How exciting was it to play on this machine?”), and how strong their desire was to play on the machine again (“How much would you like to play on this machine again?”). Other questions (not reported here) included questions about experienced differences in sound, light, speed, winning percentage and autoplay on the machines. After participants completed the experiment, data were retrieved from each of the slot machines using a hyper-terminal. Sessions were video recorded, so the reward frequency could be counted.

Measures

The South Oaks Gambling Screen (SOGS)

The SOGS is a widely used 20-item self-report questionnaire that assesses lifetime and past year gambling-related difficulties (Lesieur & Blume, 1987, 1993). A score of five or more suggests a diagnosis of pathological gambling.

Event frequency, reward frequency, and payback percentage

The event frequency was set to either two seconds per game (30 games per minute) or three seconds per game (20 games per minute). In reality, these frequencies were lower because they only referred to reel spins and did not include reward payouts or response latency.

Reward frequency (wins over number of games, wins/games)

Pathological gamblers were faster at pressing the ‘play’ button, both in the two-second condition and the three-second condition. This is consistent with previous findings (Fisher, 1993; Griffiths, 1994). We therefore measured reward frequency as the number of wins per game rather than per minute. A reward frequency of 5%, for instance, meant that, on average, one win occurred every twenty games. The reward frequency was on average 3–5% across the two slot machine conditions. The number of wins were counted using videotapes of participants' gambling sessions and divided by the number of games recorded by the slot machine.

Payback percentage (gains relative to losses)

This was calculated as the percentage of the total number of credits won divided by the total number of credits lost. Payback percentage was calculated by the machines and transferred to a computer via a hyper-terminal.

Statistics

We used independent sample t-tests to measure group differences between pathological gamblers and non-problem gamblers on age and gambling variables, and a χ² test to determine gender differences. Fisher's exact test was used to determine if pathological gamblers and non-problem gamblers differed in the number of individuals playing until stopped (i.e. 60 minutes). We used paired sample t-tests to test differences within pathological gamblers and non-problem gamblers between the two-second and three-second conditions. We used repeated measures analysis of covariance with excitement level, desire to play again and time spent gambling as within subject variables, pathological gamblers and non-problem gamblers as the grouping variable, and reward frequency and payback percentage as covariates. Finally, we used a Mantel–Cox χ² analysis to determine differences in stop-time between pathological gamblers and non-problem gamblers across conditions.

Results

As shown in Table 1 pathological gamblers had a significantly higher excitement level than non-problem gamblers in the two-second condition, but not in the three-second condition. The distribution of scores are shown in Figure 1A. Pathological gamblers also had a significantly higher desire to play again than non-problem gamblers in the two-second condition, but not in the three-second condition, see Table 1 and Figure 1B.

Table 1. Differences between pathological gamblers and non-problem gamblers on gambling measures.

	Pathological gamblers (n = 15)		Non-problem gamblers (n = 15)
	Mean	SD	Mean	SD	r	t
Two-second condition
Reward frequency (%)	3.43	1.19	4.55	3.20	− 0.23	− 1.27
Payback percentage (%)	52.47	44.55	77.13	103.42	− 0.15	− 0.85
Desire to play again	6.47	3.40	2.67	1.91	0.57	3.77****
Excitement from gambling	6.33	2.66	4.33	2.09	0.39	2.29*
Time spent gambling	51.40	12.78	37.00	15.05	0.46	2.83**
Three-second condition
Reward frequency (%)	3.31	1.79	3.51	2.10	− 0.05	− 0.27
Payback percentage (%)	69.27	86.13	44.47	34.41	0.19	1.04
Desire to play again	3.93	3.35	2.20	1.90	0.30	1.74
Excitement from gambling	4.47	2.70	3.20	1.66	0.27	1.55
Time spent gambling	47.67	16.50	33.47	17.68	0.38	2.27*
Notes: *p ≤ 0.05. **p ≤ 0.01. ***p ≤ 0.005. ****p ≤ 0.001.

Figure 1 Excitement level and desire to play again from gambling across two-second and three-second conditions in pathological gamblers and non-problem gamblers. (A) Excitement level. In the two-second condition (2s) pathological gamblers (PG, solid circles) have a significantly higher excitement level than non-problem gamblers (NPG, solid triangles). In the three-second condition (3s) pathological gamblers (PG, open circles) do not differ in excitement level from non-problem gamblers (NPG, open triangles). (B) Desire to play again. In the two-second condition (2s) pathological gamblers (PG, solid circles) have a significantly higher desire to play again than non-problem gamblers (NPG, solid triangles). In the three-second condition (3s) pathological gamblers (PG, open circles) do not differ in desire to play again from non-problem gamblers (NPG, open triangles).

Pathological gamblers had a significantly decreased desire to play again from the two-second to three-second condition, t(14) = 2.79, p < 0.05, while the decrease in excitement failed to reach significance level. Non-problem gamblers did not differ in desire to play again or excitement level between conditions.

Pathological gamblers on average spent more time gambling than non-problem gamblers both in the two-second and three-second condition, see Table 1. The results were confirmed by a repeated measures analysis of co-variance using excitement level, desire to play again and time spent gambling as within-subject variables, group as the grouping variable, and reward frequency and payback percentage as covariates. Pathological gamblers and non-problem gamblers differed significantly on excitement level, F(1, 28) = 7.51, p < 0.05, desire to play again, F(1, 28) = 11.41, p < .01, and time spent gambling, F(1, 28) = 8.64, p < 0.01. Furthermore, reward frequency in the three-second condition was a significant covariate of time spent gambling, suggesting that lower reward frequency in the three-second condition reduced the time spent gambling.

With regard to stop behaviour Fisher's exact χ² test showed that a significantly higher number of pathological gamblers (60%) than non-problem gamblers (6.7%) continued gambling until they were stopped in the two-second condition, χ²(2, N = 30) = 9.60, p < 0.01. In the three-second condition pathological gamblers (40%) and non-problem gamblers (20%) did not differ in the number of players who played until stopped, χ²(2, N = 30) = 1.43, ns. The findings were confirmed by a Mantel–Cox χ² survival analysis of stop behaviour. Figure 2A shows that in the two-second condition non-problem gamblers stopped significantly faster than pathological gamblers, χ²(2, N = 30) = 9.54, p < 0.01. In contrast, pathological gamblers and non-problem gamblers did not differ in stop time in the three-second condition, χ²(2, N = 30) = 3.36, ns, see Figure 2B.

Figure 2 Mantel–Cox survival analysis of differences in time spent gambling between pathological gamblers and non-problem gamblers. (A) In the two-second (2s) condition significantly more pathological gamblers (PG, solid circles) than non-problem gamblers (NPG, solid triangles) gamble until stopped. (B) In the three-second (3s) condition PG (solidk circles) and NPG (solid triangles) do not differ in stop behaviour.

Finally, pathological gamblers had significantly higher gambling urge at arrival, Urge 0: F(1, 28) = 5.24, p < 0.05; after first gambling session, Urge 1: F(1, 28) = 8.31, p < 0.01; and after second gambling session, Urge 2: F(1, 28) = 13.62, p < 0.005. The results are summarized in Figure 3. It can be seen from the figure that pathological gamblers maintained a high gambling urge over time, while non-problem gamblers reduced their gambling urge over time. Gambling urge was measured over time (0, 1, 2), not over conditions (two-second vs three-second condition). Repeated measures analysis showed that pathological gamblers had significantly higher gambling urge over the three time observations compared with non-problem gamblers, F(1, 28) = 13.19, p < 0.005.

Figure 3 Gambling urge among pathological gamblers and non-problem gamblers. Pathological gamblers (PG, solid circles) have significantly higher gambling urge than non-problem gamblers (NPG, solid triangles) at arrival (Urge 0), after first gambling round (Urge 1), and after second gambling round (Urge 2).

Discussion

In this study we found that pathological gamblers had significantly higher excitement level, desire to play again, and spent significantly more time gambling compared with non-problem gamblers. Lower event frequency reduced the desire to play again and the number of pathological gamblers who played until stopped. Lower reward frequency was associated with reduced gambling time in the three-second gambling condition. This suggests that lowering event frequency and reward frequency reduces the time spent gambling in pathological gamblers.

Pathological gamblers had significantly higher excitement level and desire to play again in the two-second condition compared with non-problem gamblers. The largest difference was found in the desire to play again, while a smaller difference was found in excitement level. This is consistent with studies of physiological measures and self-reported accounts of excitement levels, which show that pathological gamblers do not necessarily differ in excitement level or arousal from non-pathological gamblers (Brown et al, Rodda & Phillips, 2004; Coventry & Hudson, 2001). Emotional experiences such as ‘excitement’ contain dimensions of both physical arousal and emotional valence, which appear to overlap in the gambling experience (Brown et al., 2004). Self-reported measures of excitement may therefore be less specific than the desire to play again in relation to pathological gambling, as individuals may report on different aspects of excitement. Since we did not focus on physiological measures in our study, we were unable to test the physiological aspects of excitement. This reduces the validity of our results to the vantage point of excitement level as a self-reported measure.

Pathological gamblers significantly decreased their desire to play again in the three-second condition, while the decrease in excitement level failed to reach significance level. Again, this suggests that the desire to play again was more sensitive to changes in structural characteristics of event frequency. Non-problem gamblers showed no differences in level of excitement or desire to play again across conditions. The results suggest that pathological gamblers have a clear preference for the commercially available two-second condition. The differences in desire to play again may reflect a high degree of conditioning in pathological gamblers toward the ‘known’ version of the slot machine, and this differentiation was not present in non-problem gamblers.

Significantly more pathological gamblers than non-problem gamblers continued gambling until stopped in the two-second condition, but these differences did not reach significance level in the three-second condition. The Mantel–Cox survival analysis confirmed the relationship between time spent gambling and stop behaviour. Lower reward frequency significantly contributed to reductions in time spent gambling. These results are consistent with studies showing that gamblers prefer slot machines with more frequent wins, rather than lower reward frequency (Dixon, MacLin & Daugherty, 2006). From a learning theory perspective it is possible that reward frequency is a stronger reinforcer than payback percentage. While payback percentage represents the ‘running total’ of gambling, reward frequency represents the reward schedule of gambling, and pathological gamblers might therefore be more focused on anticipating the next reward rather than the overall outcome.

Pathological gamblers maintained a significantly higher level of gambling urge throughout the study than non-problem gamblers. The results are consistent with finding of increased gambling behaviour in pathological gambling. However, it is interesting to note that the gambling urge of pathological gamblers remain high over time, while it relatively quickly is reduced among non-problem gamblers.

The study faced several limitations. Firstly, it was not possible to control the measures of payback percentage and reward frequency. While payback percentage of the three-second experimental condition was programmed to 50% (vs 83% in the commercially available version), it was not possible to detect actual differences in payback percentage within the study period. Secondly, the maximum time limit for gambling was 60 minutes on each condition, due to ethical considerations of the risk of increased in gambling symptoms following the experiment. While these restrictions were necessary, they also resulted in a ceiling effect among pathological gamblers in time spent gambling, where 40–60% of pathological gamblers continued gambling until stopped. Further studies of structural characteristics in slot machines should address these design limitations. Thirdly, for ethical and practical reasons, slot machine gambling was limited and our laboratory experiment therefore clearly differed from a naturalistic setting. However, we note that while gambling with own money in a naturalistic setting is clearly more arousing than not gambling (Meyer et al., 2000, 2004), similar effects can be produced in laboratory environments (Yucha, Bernhard & Prato, 2007).

In conclusion, pathological gamblers have increased levels of excitement, desire to play again, and spend more time gambling than non-problem gamblers. Lower event frequency reduces gambling motivation and some aspects of gambling behaviour in pathological gambling, and lower reward frequency contributes to the reduction of time spent gambling. The results may have implications for understanding behavioural mechanisms of pathological gambling among slot machine players. Our data suggest that pathological slot machine gambling behaviour (i.e. time spent gambling) is more persistent than the gambling experience. This might suggest that substantial changes in structural characteristics of slot machines are necessary for harm reduction initiatives and responsible gambling policies to be successful.

Acknowledgements and disclosure of interests

This study was supported by the Danish Agency for Science, Technology and Innovation grant number 2049-03-0002 and 2102-05-0011. The authors declare that they have no competing financial interests.