Experimental Studies
Griffiths (
1994) conducted an ecologically valid gambling experiment in the UK using slot machines in a gambling arcade to assess both the cognitive biases by regular non-problem gamblers (N = 30, 29 males) and non-regular non-problem gamblers (N = 30, 15 males), as well as their overt gambling behaviour (mean age = 23.4 years). Cognitive biases were assessed using the ‘thinking aloud’ method where gamblers’ verbalisations are recorded and categorised (Ericcsson and Simon
1980). Overt gambling behaviour variables included total plays, time spent gambling, and speed of play. Results relating to speed of play demonstrated that on average, regular gamblers played significantly faster (eight gambles per minute) compared to non-regular gamblers (six gambles per minute). The mean speed of play rate was reduced in the thinking out loud condition for non-regular gamblers from 6.5 to 5.3 gambles per minute, and increased in the thinking out loud condition for regular gamblers from 7.5 to 8.4 gambles per minute, though both of these differences were not statistically significant.
Because cognitive biases were the main focus of this experiment and not speed of play, and the fact that speed of play was used as one of several dependent variables, knowledge gained in terms of the impact of speed on the gambler is limited. However, the study did provide empirical evidence that regular gamblers play on slot machines significantly quicker than non-regular gamblers (p < .01). Reasons for this may simply be due to the fact that regular gamblers are more familiar with the gambling product and consequently, the game mechanics, allowing them to operate the games at a faster pace through familiarity and competence. This was supported by the verbalisations from both regular and non-regular gamblers in the ‘thinking aloud’ condition. Compared to regular gamblers, non-regular gamblers made significantly more verbalisations that were classed as ‘confused questions’ (p < .001) and ‘confused statements’ (p < .001), suggesting that the lower level of competence may slow down the speed of gambling for non-regular gamblers.
Loba et al. (
2001) conducted a laboratory-based experiment in Canada using commercially available video lottery terminals (VLTs) to examine the effects of structural characteristic manipulations on subjective game experiences. Participants comprised 60 regular VLT players (38 males), with 29 being classed as a ‘pathological gambler’ and 31 as ‘non-pathological’ gamblers, as determined by the SOGS (Lesieur and Blume
1987). Participants were on average 34.7 years of age (SD = 11.6). Game manipulations included increasing and decreasing the speed of play for a video poker and ‘reel spin’ game, as well as other sensory manipulations such as sound/no sound, stop button/no stop button, and display counter/no display counter. Results indicated that when compared to non-pathological gamblers, pathological gamblers’ ratings of enjoyment, excitement, and tension-reduction was significantly reduced when speeds of play were reduced, as well as when sound was turned off during the game. Of note, pathological gamblers reported significantly more difficulty in stopping gambling than non-pathological gamblers when speed of play was increased accompanied by sound.
However, it is not made clear to what extent the game speeds were increased or decreased relative to a control condition, as no information on VLT event frequency was provided. This is an important omission, as it is not known if the pathological gamblers were sensitive to small changes in event frequency, or if in fact the speed manipulations were large. In addition, the use of dichotomous participant groupings, non-pathological vs pathological gamblers, overlooked the fact that pathological gambling behaviour is viewed along a continuum of problematic behaviours and intensities, where several intermediate levels of risk between non-pathological and pathological gambling exist (Currie and Casey
2007). In terms of the impact of speed of play on self-reported gambling experiences, it is important to acknowledge that speed of play was manipulated concurrently to other multiple structural game changes. This makes it difficult to ascertain the proportional impact of each manipulation on reported gambling experiences, and therefore does not shed light on the impact of speed of play on gambling experiences in isolation. However, it is understandable why speed was not isolated in Loba et al.’s experimental procedure given the already lengthy experiment duration (i.e., two hours).
Sharpe et al. (
2005) conducted a naturalistic experiment, in which various structural manipulations to eight gaming machines in gambling venues and hotels in the New South Wales region of Australia were made. Participants comprised 779 gamblers, from which 634 participants provided SOGS scores. Participant mean age was 46.1 years (SD = 17.9), and the mean SOGS score was 2.43 (SD = 3.43) out of 20. One-fifth (20%) of the participants were classed as problem gamblers having scored five or more on the SOGS. All other participants were grouped as non-problem gamblers due to sub-categories of ‘at-risk’ gamblers being too small for reliable statistical analysis. Speed of play was one of the independent variables, being manipulated at two levels: 3.5-second, and 5-second event frequencies, with maximum bet size and maximum size note acceptors as the two other structural characteristics being experimentally manipulated.
The speed manipulations had little effect on gambling behaviour. There was no statistical significance in terms of the difference in time spent on the gaming machines, number of bets placed, amount lost, number of lines or credits played, and alcohol/cigarette consumption, as a result of manipulations in speed of play. However, it is not possible to tell from this study whether reductions in speed of play would be differentially effective for problem gamblers as compared to non-problem gamblers, because there were insufficient numbers of problem gamblers included in the study. In addition, that fact that gambling behaviour was being observed by the researchers may in turn have produced demand characteristics, possibly resulting in gamblers behaving in a more controlled and moderate manner, gambling more slowly and deliberately as a result. Of the three proposed modifications, only a reduction in maximum bet size to $1 demonstrated evidence for a potential reduction in harm associated with gambling, because those gambling on $1 maximum machines played for less time, made fewer bets, lost less money, and consumed less alcohol and cigarettes during play.
Blaszczynski et al. (
2005) similarly demonstrated that a reduction in speed of play on EGMs from a three-second to five-second event frequency had no impact on a gambler’s intentions to continue playing. They conducted a live experiment in hotels and clubs in the Sydney region of Australia, comprising more than 400 participants of various non-problem and problem gambling statuses who played on modified experimental and non-modified gaming machines. As well as manipulating speed of play, experimental machines were modified to limit the maximum bet size and reduce the high denomination note acceptors compared to control machines. Limiting the maximum bet size and note acceptor modifications had a non-significant impact on self-reported satisfaction and enjoyment levels for both social and problem gamblers. However, satisfaction ratings were reduced significantly when both social and problem gamblers played the machines modified to a five-second event frequency, when compared to the unmodified machines with three-second event frequencies. There was a non-significant impact of slowing the event frequencies on self-reported enjoyment levels, although Blaszczynski et al. (
2005) report this as a trend towards reduced enjoyment levels given the
p value of .065. There was no interaction effect between levels of enjoyment of three- and five-second event frequencies and problem gambling status, although overall, problem gamblers rated all EGMs as less enjoyable than social gamblers. While satisfaction ratings reached statistical significance, the largest difference in satisfaction and enjoyment scores between the modified and control machines was just 8.75%, suggesting a small effect size.
Despite the seemingly negative impact of reducing speed of play on satisfaction and enjoyment levels, this did not impact gamblers’ intentions to continue gambling on EGMs, as respectively, 54% and 53% reported intentions to continue play on the control and experimental machines. Speed of play was the only modification to the machines that gamblers were able to identify, although detection rates were low, with only 14% of gamblers able to identify the modifications. This suggests that reasons for the reduced satisfaction and enjoyment ratings were subconscious, at least for the majority of the gamblers in this experiment. An alternate explanation could be that the overall effect of reduced satisfaction and enjoyment was driven only by those gamblers that were able to detect the reduced speed modification. Further post hoc statistical analysis would be required to provide evidence for such claims.
Delfabbro, Falzon and Ingram (
2005) conducted three laboratory-based experiments in South Australia assessing the impact of parameter variation on simulated EGMs in terms of their impact on subjective gambling experience and observable gambling behaviour. The EGM manipulations included reinforcement magnitude and frequency (Experiment 1), sound and screen illumination (Experiment 2), and outcome display and speed manipulation (Experiment 3). The speed of play in Experiment 3 was manipulated at two levels to provide machines with both a 3.5- and five-second event frequency.
Participants exposed to the speed of play manipulations were 24 gamblers (15 males) with various gambling experiences, participation rates, and problem gambling statuses. The mean age of participants in Experiment 3 was 47.92 years (SD = 15.6), with 10 of the gamblers being classed as a problem gambler using the SOGS. Participants were asked to play for three minutes on each of the four machines programmed with the varying parameter settings (credit display/fast speed, credit display slow speed, dollar display fast, and dollar display slow). After this mandatory exposure, participants were given a free choice to continue gambling on one of the four machines.
Speed of play was shown to significantly influence excitement ratings, with faster speeds yielding a significantly higher rating than slower speed games. Preference ratings were again, significantly higher for faster speed machines. Display type (dollars vs. credits) did not significantly impact excitement or preference ratings. There was no significant impact of speed of play in terms of the amount spent gambling on the machines overall, but the total amount of games played was significantly higher in the fast speed condition. Control measures indicated that these differences in subjective experience ratings and gambling behaviour could not simply be attributed to specific machines yielding a higher return to player or win rate, indicating the effects were driven by the speed manipulations alone. Neither gender, nor problem gambling status, interacted with the manipulations to produce significant effects, though these small sub-sample comparisons may not be reliable given the low number of participants in each category (e.g., Experiment 3 comprised just 10 problem gamblers).
Ladouceur and Sevigny (
2006) investigated the impact of VLT game speed on gamblers’ levels of concentration, motivation, self-control, and number of games played. Participants comprised 43 gamblers (22 females) from the Quebec City region of Canada. Gambling participation rates ranged from 0-24 times over the past six months, with an approximate overall mean average of three times in the past six months. A majority of the sample (n = 32) scored zero on the SOGS, six had a score of one, and five had a score of two, indicating the sample did not contain any problem or at-risk gamblers.
Speed of play was manipulated at two levels, with one group being exposed to a VLT game with a five-second event frequency, the other group a 15-second event frequency. Gamblers in the five-second condition played more games and underestimated the number of games they had played compared to participants in the slow speed condition. However, speed of play did not have a statistically significant impact on participant levels of concentration during gambling, motivation to continue gambling, or self-control in terms of time and money spent gambling. The authors concluded that the slower speed VLT game did not appear to have any positive impact in terms of facilitating more controlled gambling behaviour among the participants studied.
The use of both a five-second event frequency for the ‘fast’ condition and 15-second event frequency for the ‘slow’ condition is questionable, particularly given that event frequencies on electronic gaming machines can reach three seconds for offline EGMs, and even higher ones in their online form. Consequently, a five-second event frequency would arguably be considered slow for specific forms of EGM gambling. Motivation to continue playing was extremely low in both speed conditions, with mean motivation scores of 2.6 and 2.5 out of 10 being recorded in the fast and slow conditions respectively. Enjoyment ratings of both games were also arguably very modest, with mean enjoyment ratings 2.7 and 2.5 out of 4 for the fast and slow condition respectively. Of note, was that 67% of participants in the slow condition reported that they would like the game to go faster (compared to just 33% in the fast condition). Taken together, it could be argued that the gambling in this experimental study failed to replicate the exciting and arousing nature of real-world gambling, although it is acknowledged that this is often a trade-off for high-levels of experimental control. In addition, mean participation rates in gambling were very low for this sample, with mean participation rates equating to just once every couple of months, meaning that the gamblers were already participating at highly controlled levels, potentially masking the effects of the speed modification, and failing to be representative of gambling behaviour typically exhibited by more regular gamblers.
Linnet et al. (
2010) conducted a laboratory-based experiment in Denmark to investigate the effects of event frequency on the behaviour and experiences of problem and non-problem gamblers. The study comprised 15 pathological gamblers (10 males) and 15 non-problem gamblers (eight males). Event frequency on a popular and commercially available slot machine was manipulated at two levels to produce a two-second and three-second event frequency slot machine. The dependent variables included self-reported excitement levels, desire to play again, and time spent gambling.
Pathological gamblers reported significantly higher levels of excitement in the two-second condition compared to non-problem gamblers. This significant effect was not maintained in the three-second condition. Pathological gamblers also reported significantly higher desire to play again than non-problem gamblers in the two-second condition, but again, this effect was not maintained in the three-second condition. Pathological gamblers spent more time gambling than non-pathological gamblers in both the two-second and three-second condition. In addition, significantly more pathological gamblers (60%) continued gambling until they were told to stop in the two-second condition compared to non-pathological gamblers (6.7%). In the three-second condition, twice as many pathological gamblers (40%) continued gambling until stopped compared to non-pathological gamblers (20%), although this effect did not remain statistically significant.
Overall, the results supported the notion that the behaviour and gambling experiences of pathological gamblers differs significantly from non-pathological gamblers at the faster two-second event frequency, but that their behaviour and experience was more similar at the slower three-second event frequency. However, upon close examination of the statistics, pathological gamblers report approximately 40-60% higher ratings of excitement and desire to continue gambling compared to non-pathological gamblers in the three-second condition. While these figures did not differ at a statistically significant level, this lack of statistical significance is likely due to the small sample size of just 15 for each problem gambling status, and represents a significant limitation of the study. An additional limitation of the experimental procedure was that the experimenters were not able to control payback and win percentages across the two slot machines. As a result, the researchers were not able to control for extraneous variables such as emotion as a result of wins and losses, which has been demonstrated to be an important determinant in a range of gambling behaviours (Harris and Parke
2015; Harris et al.
2016).
Choliz (
2010) manipulated reward delay, and thus, event frequency in a repeated-measures laboratory experiment conducted in Spain. The sample comprised 10 problem gamblers recruited from gambling treatment services, and they took part in a simulated slot machine study. Whilst controlling for reel speed, the experimenter manipulated the reward delay at two levels: a two-second, and 10-second delay. While the reward delay was the main variable of interest, it is important to note that as a result of this experimental manipulation, event frequency duration coincided with the reward delay, to produce a condition with a two-second and 10-second event frequency.
Key results indicated that more games were played in the two-second event frequency condition (n = 56) when compared to the 10-second condition (n = 39). Choliz (
2010) reported that this difference could not be attributed to volume or frequency of winning outcomes because there were no significant differences in gambling outcome across the two conditions. However, it is questionable whether results were driven by the reward delay or the event frequency. It may have been the case that fewer games were played in the ten-second condition due to each game cycle taking longer to complete, and participants may simply be constrained for time resulting in fewer games being played. Caution must also be taken relating to the reliability of results given the small sample of just 10 participants.
In a Norwegian laboratory-based experiment using a computer-simulated slot machine, Mentzoni et al. (
2012) investigated the impact of various bet-to-outcome-intervals (BOIs) on subjective gambling experience, illusions of control, and observable gambling behaviour. The authors define BOI as the time delay between the initiation of the bet and receiving the outcome of that bet. However, there was an important distinction overlooked by Mentzoni et al. (
2012) between event frequency and BOI. It is possible to have a short BOI structure within a relatively slower event frequency if the outcome of the event does not signify the point at which a new game cycle or bet can begin. For example, a slot machine may spin for two-seconds and reveal the outcome of the wager immediately following the reel spin, but there may be a further one-second delay before a new wager can be made. In this hypothetical example, the machine would have a two-second BOI, yet a three-second event frequency. This distinction is not made by the authors, so it has to be assumed that BOI and event frequency are of the same length of time in this study.
Sixty-two undergraduate students (31 males) with a mean age of 20.8 years (SD = 3.26) participated in the between-participants experiment. Three participants were probable pathological gamblers, 27 had some problems with gambling, and 32 had no problems with gambling, as indicated by the SOGS. Of note, the three participants scoring five or more on the SOGS were excluded from further analysis. Participants were allocated to one-of-three BOI condition: 400 ms; 1700 ms; and 3000 ms respectively. The results showed no overall main effect of BOI on average bet size, illusion of control, or subjective enjoyment ratings, and therefore, little evidence to support the notion that speed of play leads to more intensive and risky gambling, and no evidence that the faster game was preferred by the participants. However, results did indicate an interaction effect in that at-risk pathological gamblers made significantly higher average bet sizes than non-problem gamblers in the short BOI condition. The differences in bet sizes between these two sub-groups did not reach statistical significance in the moderate or long BOI condition. This may indicate that at-risk gamblers may be particularly susceptible to elevated risk-taking in games with high event frequencies and short BOIs.
To reiterate, one of the limitations of this study was the lack of distinction between BOI and event frequency, so it is not possible to ascertain whether the short BOI or high event frequency resulted in at-risk gamblers escalating their average bet sizes. Further research would be required to control for this distinction. In addition, the absence of a meaningful sample size of pathological gamblers means results cannot be extended to account for the behaviour of those at the extreme end of the problem gambling continuum.