Introduction

Gambling can be an enjoyable pastime for many. However, a substantial minority of those that gamble regularly experience at least some harm1. These harms can include relationship problems, reductions in money available for leisure activities, absenteeism, psychological distress, insomnia, and suicide2,3,4,5,6 Gambling harm stems from engaging with gambling in such a way that causes adverse consequences. How people gamble, i.e., the decisions they make within a gambling session, can increase the likelihood of these adverse consequences through the spending of more money or time gambling than they intended1,2.

Late-night opening hours of licensed venues and late-night play online are common features of gambling. The extended periods of wakefulness associated with this late-night activity may influence the decisions that people make when they gamble. General cognitive performance is estimated to decline after 15.8 ± 0.7 h of continuous wakefulness7 and this decline is marked. A seminal study by Dawson and Reid8 demonstrated that after 17 h of extended wakefulness, decrements in task performance were significant and in the same order of magnitude as subjects who had a blood alcohol concentration of 0.05%. Specific areas of cognition have been shown to be negatively affected by extended wakefulness. These include attention, mood, reaction time, working memory, spatial reasoning, and cognitive processing speed7,9 These impairments manifest in decision making deficits10,11,12, thus, potentially affecting how people gamble.

Despite the gambling environment facilitating the placement of wagers at almost any time of the day or night13 there is limited research on the relationship between lack of sleep and gambling intensity. Studies examining decision-making and risky choices, more generally, are more common. In a review of the literature, Harrison and Horne14 concluded that sleep deprivation impairs a wide range of decision-making tasks. The most common findings were large decrements in repetitive, familiar tasks. However, the authors noted that sleep deprivation also negatively affects other tasks that involve higher-order cognition.

In examining the research on gambling and sleep, a variety of ‘gambling’ tasks appear in the literature. These tasks range from the very simple, for example, a simplified blackjack task15 to the more complex, for example, the Iowa Gambling Task16,17,18. It is important to note that these tasks are proxy gambling tasks that are not representative of the most common real-world gambling products, such as electronic gaming machines (EGMs). For instance, the Iowa Gambling Task (IGT)19 is a card game created for use in patients with frontal lobe damage. It is a learning and decision-making task in which participants choose cards from two decks. One deck delivers small wins but even smaller losses, resulting in long-term gains. The other deck delivers large wins but even larger losses, resulting in long-term losses. Healthy participants learn to repeatedly choose the deck associated with long-term gains. However, subjects that had 23 hours18, 45 hours18, 49.5 hours16, 51 hours17, and 75 hours17 of total sleep deprivation exhibited a preference for the long-term losses deck of cards. Pointedly, this effect was seen regardless of those subjects having previously participated in the IGT whilst rested and had learned to choose the long-terms gains deck of cards. These findings show a significant effect of extended wakefulness on decision-making in proxy gambling tasks, with participants appearing to focus on large gains and ignoring losses, despite having experience with the task.

A similar study utilised a simplified blackjack card game to assess gambling decisions15 Army recruits were subject to either no sleep restriction (controls) or five hours or fewer of interrupted sleep per night over two or four nights (no difference was found between two and four nights) then tested on the proxy gambling task. Both controls and sleep restricted participants were more likely to rate objectively more risky blackjack hands as riskier15 However, when rating the attractiveness of blackjack hands, those in the control and sleep restricted conditions differed. Controls’ ratings of the attractiveness of a gamble aligned with the probability that hand would win. Whereas sleep restricted participants did not significantly alter their rating of the hands’ attractiveness depending on how risky that hand was. Sleep restricted participants also gambled more money, overall, due to not adapting their bet size to suit the level of risk. This study shows that restricting individuals’ sleep may lead to a greater attraction to risk or to attending less to risk.

The results from these studies that specifically examined gambling-like tasks suggest that gambling whilst tired may contribute to the experience of gambling-related harm. However, there are concerns around the lack of applicability of these studies to real-world situations. This is due to the use of unrealistic protocols—particularly in the extreme hours of wakefulness. For example, Killgore et al.’s17,18 protocol was 75 h of continuous wakefulness. These levels of extended wake are far in excess of what we would expect to see occurring at a typical gambling venue for most patrons. For example, it is unlikely that people would stay awake for 40–70 h and then subsequently choose to gamble. A more typical scenario is one that involves staying up past one’s usual bedtime time while at a venue and gambling late into the night. In this way, gambling is often something that occurs after work or on a ‘night out’20 Thus, a more conservative protocol, for example, 18–20 h of extended wake (the equivalent to waking up at 0700 h and going to bed at between 0100 and 0300 h), is more ecologically valid.

While Frings15 used a slightly more realistic protocol of sleep restriction compared to Killgore et al.17,18 (5 h or less of interrupted sleep per night for 2–4 nights, rather than total sleep deprivation), the sample was atypical: a group of army cadets (86% male) undergoing an intensive physical training course. In a further challenge to ecological validity, the recruits for this study were not regular gamblers. Consequently, it may be that changes in behaviour can be attributed to the extreme conditions experienced and the type of participants, rather than sleep restriction alone. Lastly, the use of proxy gambling tasks prevents the application of results to real-world gambling products and regular gamblers21,22,23,24. In sum, recruitment of experienced gamblers drawn from the general population, manipulation of wakefulness to reflect moderate levels of fatigue, and assigning individuals to gamble on realistic gambling products using ‘real money’ is likely to result in more ecologically relevant findings.

The aim of this study was to understand how moderately extended wake time affects decision-making on gambling tasks, among regular gamblers, using ecologically valid gambling tasks. We hypothesised that after 18.75 h of extended wake, people would gamble more intensely (place more bets per minute, select longer odds, and/or wager higher amounts) than at baseline and perform worse on tasks of cognitive functioning following extended wake compared to baseline.

Material and methods

Participants

Healthy adults (n = 19; 9 male, 10 female) were recruited from Adelaide, Australia, via online advertisements and physical advertisements placed at university campuses and student accommodation. Participant characteristics are reported in Table 1. Participation in the study was voluntary. Participants provided written informed consent and were remunerated at least $100 (AUD) for their time (see Procedure for detail on compensation structure). The informed consent provided details on the types of activities they would undertake but not the study hypotheses.

Table 1 Participant Demographics (n = 19).
Full size table

Participants were screened using a general health questionnaire to determine their eligibility to take part in the study. Inclusion criteria were aged between 18 and 35 years; experienced gamblers, defined as those who gambled for money on continuous forms of gambling at least bimonthly; non-smoker; caffeinated beverage consumption of < 3 per day; habitual bedtimes between 2200-0000 h; rise times between 0600-0800 h; no transmeridian travel or shift work in the last month; free from any drugs or medications that may interfere with sleep. Participants were excluded from the study if they: (1) had a diagnosed sleep condition or reported snoring or choking in their sleep; (2) had an Epworth Sleepiness Scale score > 1025; (3) were not fluent in English; or (4) were pregnant or breastfeeding. Of the 426 individuals aged between 18 and 35 years who applied for inclusion in the study, 392 did not meet at least one of the screening criteria. For the week prior to the study, participants were instructed to keep to their regular sleep and wake schedule. To ensure fidelity of participants’ habitual sleep schedule, participants completed a sleep diary to record sleep timing, quantity and quality and were excluded from the study if their sleep habits deviated from their habitual sleep.

Experimental design

The laboratory-based study had a repeated measures design. The test battery was administered at 1800 h (baseline) and again at 0300 h during a period of extended wake. Participants stayed voluntarily confined in the laboratory for one night. While real-world gambling sessions can involve prolonged continuous play, the current testing protocol (baseline and during extended wake) was intentionally selected to isolate the acute impacts of extended wakefulness on specific gambling tasks. The median EGM session length in Australia is about 2 h, with only 20% of sessions lasting more than 5 hours26. Our two-session protocol is consistent with common experience where people may take breaks for meals or other entertainment, although to our knowledge this observation has not be quantified in past research. Although our approach somewhat limits ecological validity, it provides greater experimental control and clearer insights into performance changes directly attributable extended wakefulness. An overview of the experimental protocol is shown in Fig. 1.

Fig. 1
figure 1

Experimental protocol.

Full size image

Procedure

On the day of the study, participants were instructed to wake at 8:15am and they were called to ensure compliance. On arrival, participants were shown to their private rooms and briefed on the sleep laboratory safety procedures and schedule. All test battery measures, aside from the gambling tasks, were demonstrated and participants undertook training on each task a set number of times to mitigate learning effects. Gambling games were demonstrated to participants using pre-recorded videos. This vicarious exposure was employed to ensure participants were familiar with the games but that prior wins or losses resulting from game practice were not able to influence subsequent play. To enhance external validity, participants were advised that they would be gambling with a portion of their study compensation (AUD$5 per game) and that any losses would be subtracted from their total compensation and any winnings added. In reality, participants were paid out any winnings, but no net losses were subtracted. Compensation was in line with time commitments, with participants receiving between $100 and $125.

The first test battery (T1) began at 1800 h, concluded at 1930 h and followed the order specified in Fig. 1. At 0300 h, the second test battery (T2) began and followed the same order as the first test battery, concluding at 0430 h. Social interaction was minimised by having participants spend the majority of their time, including during testing sessions, alone in their bedrooms, specifically to control for external social influences.. Daily energy intake was calculated using a modified Harris-Benedict Eq. 27, incorporating an activity factor of 1.5, consistent with previous methodologies28. A standardised dinner meal was provided at 2000 h. Snacks were administered at 1600 h and 0100 h. The macronutrient composition of all meals reflected a typical Western dietary profile, consisting of 26% protein, 20% fat, and 54% carbohydrate. Participants consumed the entirety of each meal and snack as provided. Water was ad libitum. Participants were directed to avoid any foods or beverages containing caffeine on the day of the study to avoid potential confounding effects on performance tasks. During the supervised evening mealtime, participants were prohibited from discussing gambling to avoid any competition between participants that may have impacted on betting behaviour.

Materials—the environment

Participants stayed overnight at the Appleton Institute Sleep Laboratory. The laboratory has six single bedrooms and is sound-attenuated and temperature controlled (21 ± 2ºC). To simulate the experience of being in a gambling venue where attention to clocks is limited, the laboratory was time isolated. Participants were not allowed access to clocks, watches, or their cellular phones throughout the study period; television viewing was permitted during non-testing periods, but no other electronic devices were allowed. Light levels in the laboratory were set to 75 lx. This level was based on covert measurements taken at the Adelaide Casino in South Australia in April 2018 and was consistent with the mandated minimum lighting levels for casinos in Australia29.

Measures

Cognitive measures

Neurobehavioural performance was measured by the Psychomotor Vigilance Task (PVT)30 a 10-min task in which participants respond to visual stimuli with a button press. This task has been shown to be a valid measure of neurobehavioural performance and is robust to repeated administration within subjects31,32 Outcome measures of interest were the number of lapses (response time > 500 ms) and mean reciprocal response times (RRT; 1/ms × 1000).

Subjective measures

Perceived ability to perform on the PVT was determined using a 100 mm visual analogue scale anchored with statements ‘extremely poorly’ (far left, 0) and ‘extremely well’ (far right, 100) (VAS-Pre)33 Participants marked the scale in response to the question ‘How well do you think you will perform?” with higher scores indicating better predicted performance. Subjective sleepiness was determined using the Karolinska Sleepiness Scale (KSS),34 a subjective 9-point scale rating one’s sleepiness at that present moment anchored with statements ‘extremely alert’ (top, 1) and ‘very sleepy, great efforts to keep awake, fighting sleep’ (bottom, 9). Participants marked the scale in response to the question ‘How do you feel right now?” with higher scores indicating a greater feeling of sleepiness. Mood was measured using the Profile of Mood States (POMS),35 a 65-item questionnaire that uses a 5-point Likert-type scale to measure five negative mood domains and one positive mood domain. Higher scores indicate greater negative mood.

Gambling measures

Problem gambling status was measured by the Problem Gambling Severity Index (PGSI)36 The PGSI has been widely recognised as a reliable and standardised instrument for assessing at-risk gambling behaviours36,37. Comprising nine items, the PGSI is designed to evaluate varying levels of problem gambling involvement. Each item is scored on a four-point Likert scale, with response options weighted as follows: ‘never’ = 0, ‘sometimes’ = 1, 'most of the time’ = 2, and ‘almost always’ = 3. The total PGSI score is obtained by summing the responses across all items, yielding a possible range from 0 to 27. Based on the total score, individuals are categorised into risk levels: scores of 1–2 indicate low-risk gambling, 3–7 reflect moderate-risk gambling, and scores of 8 or above signify problem gambling.

Gambling decisions were measured by four online gambling paradigms developed by Central Queensland University for research: electronic gaming machine (EGM);38,39 roulette;40 sports betting;41 and bingo42. Each game was played on a laptop computer and resembled popular applications available in the current marketplace. The games were accompanied by sounds/music typical of each gambling product. Sound was accessed through headphones so that participants were not affected by sounds from other games, since exposure to other peoples’ wins has been shown to have an influence on gambling behaviour43. At the onset of each gambling game, participants were told to “keep playing until I tell you to stop.” After 10 min, the researcher stopped each gambling task, creating a fixed time period for each task. Participants were not told of the length of each task, nor were they given means to measure the length. Tasks were programmed pseudo-randomly with participants designed to break even approximately every two minutes, with a combination of large and small wins and losses. However, bet sizes—and betting speed in the EGM task—were chosen by players and, therefore, break-even performance was not guaranteed. This protocol gave a more natural gambling experience, as opposed to the common experimental gambling paradigm in which small wins are followed by an indefinite losing sequence to test persistence44.

Statistical analysis

Statistical analyses were carried out in the statistical software packages of SPSS (Version 25) and R (Version 4.0.2). For each outcome measure, paired sample t-tests were used to investigate the differences in responses between T1 (1800 h) and T2 (0300 h). Gambling data were also analysed with a MANOVA model, with BingoBetSize (mean size of bets made), SportBetRisk (long vs. short odds) (odds of winning bet), RouletteBetSize (mean size of bets made), and SlotBetSize (mean size of bets made), as repeated measures dependent variables; each with two levels labelled as “time;” being measured during the evening and early next morning. PGSI gambling problems was a between-subjects independent variable in the model. Finally, we calculated the 95% confidence intervals for each experimental effect as a proportion, i.e., the likely true bounds of the percentage increase or decrease in gambling behaviour outcomes as a result of sleep deprivation.

Ethics

The study procedures adhered to the Declaration of Helsinki. Ethical approval was obtained from the Human Research Ethics Committee of Central Queensland University (H16/06–166). Participants were briefed regarding the study and provided written informed consent.

Results

Cognitive measures

The number of lapses on the psychomotor vigilance task (PVT) at T2 (M = 2.84, SD = 4.11) was significantly more than at T1 (M = 1.10, SD = 1.52), t(18) = 2.47, p = 0.024. There was a significant increase in the mean reciprocal reaction time from T1 (M = 3.98, SD = 0.58) to T2 (M = 4.25, SD = 0.53), t(18) = -3.87, p = 0.001 (higher scores indicating slower reaction time).

Subjective measures

There was a significant difference between T1 and T2 for the VAS-Pre, in that objective ratings of expected performance on the PVT were higher at T1 (M = 58.37, SD = 12.53) compared to T2 (M = 27.53, SD = 18.69), t(18) = 8.43, p < 0.001. The KSS also showed significant increases in ratings of subjective sleepiness between T1 (M = 4.21, S = 1.51) and T2 (M = 7.84, SD = 1.50), t(18) = -9.28, p < 0.001. There was also a significant increase in negative affect as measured by the POMS from T1 (M = 37.53, SD = 11.90) to T2 (M = 48.00, SD = 21.49), t(18) = -2.91, p = 0.009.

Gambling intensity

Table 2 shows the differences in gambling intensity as measured by bet size, bet risk (long vs. short odds) and bet count for the four modes of gambling between T1 and T2. As shown in the table, there were no significant differences on these measures for any of the outcomes.

Table 2 Summary of differences between T1 (1800 h) and T2 (0300 h) on indicators of gambling intensity.
Full size table

Multivariate within-subjects effects revealed no significant overall effect for “time,” Wilk’s Λ = 0.483, partial η2 = 0.23, revealing no overall identifiable changes amongst the set of gambling behavioural outcomes of BingoBetSize, SportBetRisk, RouletteBetSize and SlotBetSize between evening and early next morning. Similarly, there was no multivariate between-subjects effect for PGSI on any outcome, Wilk’s Λ = 0.456, partial η2 = 0.23. As shown in Table 2 below, there were also no univariate effects for each of the outcomes considered separately, p > 0.05, ns.

Given the non-significant effects, we considered it useful to calculate the (95% CI) upper and lower bound for each of our estimates, expressed in terms of the proportion change in gambling behaviour after the manipulation, as presented in Table 2. This provides some indication of the bounds of the maximum effect that sleep deprivation might have on gambling behaviour, given the power limitations of the current limited-sample experimental study. The mean estimated effect size for all modes was 4%.

Discussion

This study examined the effect of 18.75 h of extended wake on cognitive performance, subjective experience, and gambling intensity (measured variously using total bet size, bet risk, and bet speed) on four different online gambling modes. As expected, ratings of subjective alertness were lower, and ratings of subjective sleepiness were higher following extended wake. Also as expected, performance on the PVT was impaired (great number of lapses and slower reaction time) following extended wake and participants also displayed more negative affect at 0300 h than at 1800 h. Gambling intensity across all modes was not significantly different at 0300 h compared to at 1800 h and, contrary to predictions, problem gambling (PGSI) status, subjective sleepiness, subjective alertness, and PVT performance did not predict differences in gambling intensity between the two time points.

When investigating gambling intensity as a composite measure, the current findings again show that extended wake did not yield a significant change in gambling intensity. Furthermore, we can calculate that the mean likely nonsignificant increase in betting intensity, measured in various ways, at the 95% Confidence Interval was 4% when averaged across gambling modes. This shows that using more realistic wakefulness protocols and gambling using both realistic games and ‘real’ money results in no changes to gambling intensity or, at least, far smaller effect sizes than previous studies have reported15,16,17,18. When examining the broader literature on decision-making and cognitive performance, extended wakefulness typically results in performance decrements8, particularly in simple, repetitive tasks such as the PVT45 However, although the gambling tasks employed in the current study were repetitive and familiar to participants, they inherently involved more complex cognitive processes, including decision-making and contextual evaluation. Thus, these tasks differ substantially from simpler reaction time tasks like the PVT. Given this complexity, our observed null findings i.e., no measurable performance decrements following moderate extended wakefulness, may be less surprising. However, considering the existing literature demonstrating clear cognitive impairment following periods of extended wakefulness7,8, further research exploring how varying task complexities interact with extended wake is needed.

The current study methodology diverged from the current literature examining gambling and sleep. Importantly, we used measures that were both familiar (realistic) gambling tasks and a shorter period of extended wakefulness than previous studies. It may be that gambling on realistic gambling products is not as vulnerable to decrements in performance as when gambling on proxy or unrealistic gambling tasks. In this way, previous studies using proxy gambling tasks may have yielded significant results due to the novelty of the task. Moreover, the novelty of the gambling tasks was likely increased due to the participants in these previous studies not being experienced (regular) gamblers. For instance, it is likely that a lot more attention is needed from a novice gambler to engage with a game of blackjack, even when simplified, when compared to an experienced gambler playing an online EGM. It is well established in the gambling literature that continuous gambling products, especially EGMs, contribute to players entering ‘the zone’, a state of dissociation, immersion, or flow46,47. Hence, decisions made when gambling on ecologically valid gambling tasks may not be vulnerable to the same degree of decision-making decrements as when gambling on more novel proxy gambling tasks.

Previous studies examining gambling utilised designs that deprived participants of sleep for between 23 and 72 h,16,17,18 or restricted them to five or fewer hours of disturbed sleep for two to four nights15. It is therefore unsurprising that significant effects on gambling performance were observed given the severity of these sleep manipulations.. At the far end of these manipulations, it is akin to an individual waking up at 7am, staying awake for three days straight and then deciding to go out gambling. More realistic levels of extended wake, such as those used in the current experiment, would be akin to that same person who awoke at 7am gambling at 1:45am that next morning. Therefore, the results that we found in the current study indicate that extended wakefulness of a moderate magnitude is unlikely to play a large role in increasing gambling intensity on real-world gambling products. To give context, annually on average, problem gamblers spend over 700% (7 ×) more than non-problem gamblers48. When we compare the results of the current study to that of other studies examining environmental factors that influence gambling behaviour, we can see that the effect sizes are much larger for other factors. For example, in comparison to gambling alone, the presence of other gamblers caused people to place a significantly greater number of bets, with an effect size of approximately d = 0.4543 Kyngdon and Dickerson49 demonstrated that, compared to placebo, those that consumed three standard drinks prior to gambling gambled significantly more intensively, with an effect size of d = 1.39. Harm minimisation policies based on concern about people gambling after extended periods of wakefulness should be weighed up against other less costly measures that focus on environmental factors that show large effect sizes on behaviour. However, future research may wish to explore a combination of factors, such as alcohol consumption and extended wake experienced concurrently, as previous research has identified that people who get less sleep overnight are more likely to also consume greater amounts of alcohol than those that get more sleep overnight50.

Limitations and future directions

Although one of the strengths of the study was the within-subjects design which limits variability between persons, a limitation is that participants may have been subject to learning effects. Experienced gamblers were recruited to minimise learning effects for the gambling tasks and practice trials were implemented to attempt to mitigate learning effects on other measures. However, we cannot conclude with certainty that learning effects did not take place. Hence, future research should employ a control group tested twice with the same length of time but without being exposed to extended wake. Further, the controlled laboratory environment, which restricted electronic device use, caffeine restriction, social interactions, outdoor exposure, and included structured daily activities, differs substantially from typical real-world conditions, thus limiting the overall ecological validity of the study design.

Some differences in gambling intensity between 1800 and 0300 h might have reached significance with a larger sample (see Table 2). The lack of significant findings for our primary hypothesis may therefore reflect small effect sizes and limited statistical power due to the modest sample size, typical of sleep studies due to their expense (e.g.,11,51,52,53). However, these results are still of interest given that previous studies reported larger effects even with similarly small samples.

One of the exclusion criteria for this study is that people should not be currently experiencing sleep problems. This was an appropriate criterion, because we were most interested in exploring the mechanism of extended wake on gambling choices. However, we found many instances where people who gamble regularly also had sleep disturbances and mental health problems that are associated with sleep problems, and therefore were excluded from participation. Of the 426 people who applied, 392 were excluded based on the exclusion criteria. It may be that our exclusions, while reasonable with respect to internal validity, have restricted our sample to a range of persons who handle extended wakefulness relatively well.

Finally, the literature shows that some individuals cope with periods of extended wake well and any deficits in functioning are small, whereas some individuals handle it poorly resulting in comparatively large deficits to cognitive functioning7. As the results from the current study are averaged across participants, assessment of individual variability was unable to be characterised. To reduce variability in sleep timing and circadian influences, participants with consistent, ‘normal’ sleep/wake behaviours were recruited; however, this may limit the generalisability of findings to gamblers who engage in late-night or irregular gambling behaviours. Future studies should also consider assessing diurnal preference, as morning types (‘larks’) have been shown to experience greater adverse effects from late-night wakefulness compared to evening types (‘owls’)54.

Conclusion

The current study improved on past literature that has demonstrated that extended wake increases gambling intensity through the novel use of a more highly controlled environment, more ecologically valid measures of gambling, a more realistic period of extended wakefulness and experienced gamblers as participants. These methodological improvements yielded results that differed from those reported in previous literature in that gambling after 18.75 h of continuous wakefulness did not result in a reliable increase in gambling intensity. Harm minimisation policies should focus on more influential factors in the current gambling environment that contribute to increased gambling intensity rather than extended wakefulness.