Abstract
Background and aims
Emotion dysregulation has been suggested to play a role in gambling-related harm, but past gambling research has typically assessed emotion dysregulation via self-report surveys rather than in a gambling context. Here, we sought to investigate how the severity of participants' hazardous gambling behavior was associated with their emotional reactivity and choice behavior within a simulated slot-machine task.
Methods
Participants (N = 100) recruited via Prolific completed a behavioral task involving repeated choices between two simulated slot-machines. When chosen, slot-machines could produce one of five outcome types (win/near-win/neutral/near-loss/loss). After each outcome, participants reported their subjective emotional valence. Emotion data were analysed using a beta-autoregressive computational model, allowing us to extract per-participant estimates of trial-by-trial emotional reactivity to different slot-machine outcomes.
Results
Correlation analyses revealed that people who engaged in more hazardous gambling behavior (higher PGSI scores) showed greater emotional reactivity to all slot-machine outcome types (all Spearman ρ > |0.31|, all p < 0.01, corrected for multiple comparisons). There were no significant associations between patterns of choice behavior and PGSI scores.
Discussion and conclusions
Within a simulated slot-machine task, individuals who engaged in more hazardous gambling behavior showed greater emotional reactivity in general (more positive emotional reactions to wins and more negative emotional reactions to unpleasant events such as losses and near-wins). These results are consistent with a model in which emotion dysregulation is a risk factor for gambling-related harm, and serve to validate this model in a more naturalistic setting.
Introduction
Emotion dysregulation—defined as a deficit in the capacity to modulate the time, intensity and valence of emotional experiences or expressions (Velotti, Rogier, BeomonteZobel, & Billieux, 2021)—is a key risk factor for gambling-related harm. Trait-level individual differences in emotion dysregulation are associated with individual differences in the development and maintenance of gambling-related harm (Rogier & Velotti, 2018; Velotti et al., 2021), and emotion dysregulation more broadly is considered a central component of clinical models of gambling disorder (Bonnaire et al., 2022; Buen & Flack, 2021; Rogier & Velotti, 2018). However, emotion dysregulation is not a unitary construct (Davidson, 1998), and it remains unclear which of its specific facets best explain individual differences in experiences of gambling-related harm.
An important distinction sometimes drawn within the affective science literature (see, e.g., Gross & Feldman Barrett, 2011) is between emotion dysregulation caused by a top-down failure of regulatory processes (e.g., Gross, 1998, 2002) and emotion dysregulation caused by an excessive bottom-up emotional reaction to stimuli or events in the environment (e.g., Gross et al., 1998; Becerra, Preece, Campitelli, & Scot-Pillow, 2019). The latter construct—termed emotional reactivity or affective reactivity—has been shown to be elevated among individuals with higher levels of gambling-related harm (Bonnaire et al., 2022). This suggests that emotional reactivity may explain part of the broader association between emotion dysregulation and gambling-related harm, potentially via excessive emotional reactions to the outcomes of gambling behaviour. However, previous work on emotional reactivity and gambling-related harm has not distinguished between emotional reactivity to gambling outcomes specifically versus emotional reactivity as a broader psychological phenotype. Moreover, like the broader emotion dysregulation literature, emotional reactivity has typically been measured using retrospective self-report questionnaires. Although valuable, questionnaire-based measures of emotion dysregulation rely heavily on individuals' autobiographical memory and insight, and may be subject to recall and social-desirability biases. There is relatively little empirical work investigating how emotion dysregulation broadly and emotional reactivity specifically manifest during gambling activity itself.
To address these open questions, the present study examined emotional reactivity to gambling outcomes produced by electronic gambling machine (EGM)-like stimuli. As well as being among the most harmful of commonly available gambling products (Binde, Romild, & Volberg, 2017; Russell et al., 2023), EGMs produce a myriad of different outcomes – not only wins and non-wins, but also more exotic outcome types such as near-wins (Clark, Crooks, Clarke, Aitken, & Dunn, 2012; Dixon et al., 2011) and losses disguised as wins (e.g., Dixon, Harrigan, Sandhu, Collins, & Fugelsang, 2010; Myles et al., 2023). In Australia, four out of five people who use EGMs at least weekly meet screening criteria for at-risk gambling, and almost half meet criteria for “problem gambling” (Australian Gambling Research Centre, 2023). As such, a topic of particular interest in gambling research is to identify what structural characteristics of EGMs (e.g., flashing lights, winning sounds, outcome displays; Livingstone, 2017; Yücel, Carter, Harrigan, Van Holst, & Livingstone, 2018) might be responsible for the strong associations between EGM use and gambling-related harm. The present study sought to contribute to this literature by identifying how individual differences in the severity of hazardous gambling behavior were associated with emotional reactivity following different outcome types in a simulated slot-machine task. Given that gambling-related harm is thought to be associated with increased emotional reactivity in general, one possibility is that stronger emotional reactivity to gambling outcomes might contribute to the harm associated with EGMs.
One particular topic of interest in the study of EGMs is the near-win: a kind of non-win outcome in which symbols are presented in a sequence that initially suggests a win might be about to occur (e.g., cherry, cherry, lemon). It has been suggested that near-wins have a reinforcing effect on EGM use, with participants showing increased behavioral persistence (Cote, Caron, Aubert, Desrochers, & Ladouceur, 2003; Ghezzi, Wilson, & Porter, 2006; Kassinove & Schare, 2001; MacLin, Dixon, Daugherty, & Small, 2007), prolonged post-reinforcement pauses (Belisle & Dixon, 2016; but see also Dixon, MacLaren, Jarick, Fugelsang, & Harrigan, 2013), and increased bet size and faster speed of play (Palmer, Ferrari, & Clark, 2024) following near-wins. However, further research is required to identify the cognitive mechanisms that best explain the behavioural effects of near-wins. Of note, however, near-win outcomes influence users' emotions as well as their behavior: near-wins are consistently rated as more unpleasant (Clark et al., 2012; Griffiths, 1991; Qi, Ding, Song, & Yang, 2011; Sharman & Clark, 2016; Sharman, Aitken, & Clark, 2015) and more arousing (Clark et al., 2012; Dixon et al., 2011, 2013) compared with other types of non-win outcomes. It remains unknown, however, whether the magnitude of people's emotional reactivity to near-win events differs as a function of their degree of gambling-related harm.
In this current study we used a simulated slot-machine task to investigate the effects of different gambling outcome types on both emotion and choice behavior. By embedding self-reports of trial-by-trial emotional valence within this task, we were able to capture a specific facet of emotion dysregulation—emotional reactivity—in a naturalistic gambling context. In this way, we were able to determine how both post-outcome choice behavior and emotional reactivity to simulated slot-machine outcomes were associated with individual differences in severity of hazardous gambling behavior (as measured by the Problem Gambling Severity Index [PGSI]).
Method
All data and analysis code are freely available in the project Open Science Framework (OSF) repository:
Participants
Participants were recruited in a two-stage process via the online research platform Prolific. We first invited 251 participants to complete a one-question screening survey regarding their gambling frequency (daily, weekly, monthly, yearly, or never). We then invited a subset of N = 100 participants to return to complete the behavioral task, oversampling those who gambled more frequently with the goal of increasing the proportion of participants in the final sample who engaged in some degree of hazardous gambling. Specifically, we sought to recruit a final sample in which one-third of participants gambled daily or weekly, one-third gambled monthly, and one-third gambled yearly or never. Full demographic information is provided in the Supplementary Material.
Participants were eligible to participate if they were aged between 18 and 65, fluent in English, and had unimpaired or adequately corrected eyesight. All participants provided informed consent via a web browser form. Participants were paid AUD $0.25 for completing the screening questionnaire, and those who were invited back to complete the behavioral task received a base payment of $6.67 plus a bonus payment up to $2 depending on performance in the task (mean total task payment = $7.87, SD = 0.07). Six participants were excluded from analysis for failing attention checks in the PGSI questionnaire or the behavioral task. The final sample comprised 94 participants (52 men, 39 women, 3 who did not report a binary gender) aged 20–64 (M = 38.26, SD = 11.65).
Measures and procedure
Within the task, trial-by-trial changes in emotional valence were measured using an emotion slider adapted from Betella and Verschure (2016), which ranged from extremely unhappy (0) to extremely happy (1). Individual differences in the severity of hazardous gambling were measured using the total score of the PGSI (Ferris et al., 2001). To identify inattentive survey respondents, the PGSI was modified to include an additional infrequency-item attention check (i.e., an item for which there was a clear correct response if participants were attending to the survey questions rather than responding at random). Identifying inattentive survey respondents is recommended when measuring individual differences in constructs that have a highly positively skewed distribution in the general population, as in the PGSI (Zorowitz, Solis, Niv, & Bennett, 2023).
Behavioral task
After completing a demographic survey and the PGSI, participants completed a 25-min behavioral slot-machine task with embedded high-resolution sampling of subjective emotional valence. The task consisted of 72 trials split into four blocks of 18 trials each. On each trial, participants chose to ‘play’ one of two simulated slot-machines (see Fig. 1A), each of which cost one ‘coin’ to play (out of an initial endowment of 150 coins that participants received at the start of the task). Participants were told that their goal was to collect as many coins as possible, and that coins could be cashed in at the end of the task for an additional monetary bonus. When chosen, each slot-machine could produce one of five outcome types: either a win (gain of 20 coins), a loss (loss of 20 coins), a near-win (no payout), a near-loss (no payout), or a neutral outcome (no payout). See Fig. 1B for a visual depiction of each outcome sequence, and the project OSF repository for corresponding videos. No matter the outcome, participants lost the initial coin that they paid to play the machine. Consequently, the net position after a win was +19 coins, the net position after a loss was −21 coins, and the net position after all other outcome types was −1 coin. Of these outcome types, wins, near-wins and neutral outcomes regularly occur within actual real-world EGMs, whereas losses and near-losses do not (Harrigan, MacLaren, Brown, Dixon, & Livingstone, 2014). However, we designed the simulated slot-machine task to include a wider range of possible outcomes than occur in real-word settings so as to quantify choice behavior and emotional reactivity in response to a broader spectrum of possible outcome types. Outcomes were revealed via the sequential display of symbols in the three ‘reels’ of the machine at a rate of three seconds per ‘reel’ (left to right). Participants rated their current emotional valence at the beginning of each block and after every trial (76 times in total).
Overview of the simulated slot-machine task. (A) Task schematic. On each trial, participants chose one of two simulated slot-machines, watched an animation that revealed its outcome (three cash and/or bomb symbols, sequentially revealed from left to right), and then reported their subjective emotional valence. (B) Outcome types. On each trial, a slot-machine produced one of five outcome types. Of the eight possible configurations of bomb and cash symbols, there were four configurations associated with a ‘neutral’ outcome and one configuration associated with each of the other four outcome types
Citation: Journal of Behavioral Addictions 2025; 10.1556/2006.2025.00003
The same two slot-machines were presented on each trial; unbeknownst to participants, the two machines had an identical and pre-defined likelihood of producing different outcome types on each trial. The trial sequence was pre-allocated to contain eight wins, eight losses, eight near-wins, eight near-losses, and 32 neutral outcomes (outcome order randomized separately for each participant). The pre-defined outcome for each trial was presented independent of which machine was actually chosen, though participants were not informed of this. These outcome frequencies correspond to an independent 50% probability of revealing a win symbol on each machine reel, and result in an expected value for each machine of −1 coins per spin. Finally, the task also included eight attention-check trials in which participants chose between one white slot-machine marked ‘Guaranteed Win’ and one black slot-machine marked ‘Guaranteed Loss’. Participants had been instructed that when presented with these choices they should choose the Guaranteed Win machine. Participants who chose the Guaranteed Loss slot-machine on any trial were deemed to have failed the attention check and were excluded from all further analysis.
Statistical analysis
Choice behavior was analysed using a mixed-effects logistic regression. For this analysis, the dependent variable was whether or not the participant chose the same slot-machine on trial t + 1 as on trial t (stay = 1, switch = 0, excluding trials followed by an attention check or the end of a block). Independent variables were the outcome type on the preceding trial (reference-coded), PGSI scores (z-scored for the logistic regression analysis only) and an outcome by PGSI interaction. Analyses included random intercepts for participants, but no random slopes because models including random slopes did not converge during estimation. Analyses were conducted using the lme4 package in R, with omnibus statistical significance assessed using Type-III Wald χ2 tests (from the car package). We chose choice repetition as a behavioral metric because choice persistence after an outcome (so-called “win-stay” behavior) is generally taken to be a signature of reinforcement in models of operant behavior: a reinforcing outcome for a choice is one that increases a participant's likelihood of making the same choice again at the next opportunity (Iyer, Kairiss, Liu, Otto, & Bagot, 2020).
In other words, emotional valence on a given trial was assumed to be a function of both a participant-specific baseline parameter
Crucially, to assess the associations between individual differences in gambling severity and emotional reactivity to various outcome types we calculated Spearman rank-order correlations between PGSI scores and participant-level estimates of emotional reactivity to each of the five outcome types (
We note that this model made several implicit assumptions about the data. First, we assumed that emotional valence could be adequately modelled by considering only the five most recent outcomes prior to a given self-report probe (as we discuss below, this assumption was validated by inspection of model parameter estimates). Second, we assumed independence of sequential effects, such that the effect of any given outcome on emotion was independent of the outcomes that occurred in the preceding trials. Finally, we assumed that emotion self-report data could be adequately modelled by a Beta distribution (in line with the results of a previous explicit comparison between a Beta-distributed model and a Gaussian model of emotion by Forbes & Bennett, 2024).
Ethics
Study procedures were carried out in accordance with the Declaration of Helsinki, and were approved by the Monash University Human Research Ethics Committee (#37651). All participants provided informed consent.
Results
Effects of outcome type on choice behavior
A mixed-effects logistic regression analysis revealed that there was a significant effect of outcome type on participants' likelihood of choosing the same machine again on the subsequent trial (χ2(4) = 48.15, p < 0.001; see Fig. 2A). Post-hoc contrasts (with Bonferroni-Holm correction for multiple comparisons) revealed that this was driven by a significantly increased likelihood of repeating a choice after wins (p = 0.02, odds ratio [OR] = 1.31) and near-wins (p = 0.003, OR = 1.43) compared with neutral outcomes, and a significantly decreased likelihood of repeating a choice after losses (p < 0.001, OR = 0.66) and near-losses (p = 0.02, OR = 0.79). There was no significant main effect of PGSI (χ2(1) = 0.22, p = 0.64) and no significant interaction between PGSI and outcome type (χ2(4) = 1.05, p = 0.90).
Effects of different slot-machine outcomes on choice behavior and emotion self-reports. (A) Probability of repeating a choice on trial t + 1 as a function of slot-machine outcome type on trial t. (B) Estimated change in emotional valence following different outcome types (positive: improvement in emotional valence; negative: deterioration in emotional valence) as estimated within the computational models. In both plots, bars and error bars represent group-level means and 95% confidence intervals, and grey points represent data from individual participants
Citation: Journal of Behavioral Addictions 2025; 10.1556/2006.2025.00003
Effects of outcome type on emotional valence
Emotional reactivity parameters of the computational model are visualised in Fig. 2B (raw measures of emotional reactivity are presented in the Supplementary Material). As would be expected, win outcomes were associated with large positive emotional reactions (wwin parameter = 0.64, 95% Bayesian highest density interval (HDI) [0.52, 0.75]) and loss outcomes were associated with large negative emotional reactions (wloss = −0.96, 95% HDI [−1.13, −0.79]). All other outcome types were associated with smaller negative emotional reactions (neutral outcomes: wneutral = −0.25 95% HDI [−0.33, −0.17]; near-win outcomes: wnearwin = −0.31, 95% HDI = [−0.40, −0.21]; near-loss outcomes: wnearloss = −0.19, 95% HDI [−0.28, −0.12]). The effects of near-wins and near-losses on emotion significantly differed from other kinds of non-win events: near-wins produced a significantly more negative emotional reaction than neutral outcomes (i.e., wnearwin - wneutral ; median difference = −0.06, 95% HDI [−0.11, −0.01]), whereas near-losses produced a significantly more positive emotional reaction than neutral outcomes (i.e., wnearloss - wneutral ; median difference = 0.05, 95% HDI [0.01, 0.10]). There were also substantial inter-individual differences in emotional reactivity parameters, and strong within-individual correlations both in emotional reactivity to different outcome types and between emotional reactivity parameters and the degree of recency-weighting in the effects of previous outcomes (see Table S3 in Supplementary Material for parameter correlation tables).
Inspection of other parameter estimates revealed a mean w0 parameter of 0.30 (Bayesian 95% HDI [0.14, 0.47]), indicating that participants reported significantly better-than-neutral emotional valence on average. The mean value of the appraisal decay parameter ɣ was 0.45 (95% HDI [0.36, 0.54]), indicating a relatively rapid decay of outcomes from previous trials of current emotional valence (i.e., only 45% of the emotional effects of the outcome of trial t – 1 were still evident in the emotion self-report following trial t, and only approximately 5% of a prior outcome's effect was still present outside the five-trial sliding window that we considered in our computational model).
Moderating effects of PGSI on emotional reactivity
Spearman correlations (with Bonferroni-Holm correction for multiple comparisons) were used to assess the association between individual differences in hazardous gambling severity (PGSI total score) and individual differences in emotional reactivity to different slot-machine outcome type. Notably, higher PGSI scores were associated with heightened emotional reactivity to all event types: significantly more positive emotional reactivity to wins (ρ(92) = 0.38, p = 0.001) but more negative emotional reactivity to losses (ρ(92) = −0.36, p = 0.002), near-wins (ρ(92) = −0.35, p = 0.002), near-losses (ρ(92) = −0.31, p = 0.01), and neutral outcomes (ρ(92) = −0.32, p = 0.01). These results indicate that a higher PGSI score predicted greater emotional reactivity to all slot-machine outcomes, regardless of type (see Fig. 3 below; equivalent plots for the raw emotional reactivity data are presented in Fig. S2 in the Supplementary Material). All results remained statistically significant when excluding participants who were outliers on the PGSI (scores >2 SD from the mean, or PGSI >9) and after controlling for age and gender as covariates (see Supplementary Section S4).
More severe hazardous gambling (as indicated by higher PGSI scores) was associated with greater emotional reactivity to all slot-machine outcome types as measured by the computational model. Positive values on the y-axes indicate more positive emotional reactions, negative values indicate more negative emotional reactions, and values of zero (horizontal dashed line) indicate no effect of a given outcome type on emotional valence. Shaded regions indicate the 95% confidence interval of the linear association between PGSI scores and emotional reactivity
Citation: Journal of Behavioral Addictions 2025; 10.1556/2006.2025.00003
Discussion and conclusions
In this study we investigated the emotional and behavioral effects of different gambling outcomes, with a specific focus on assessing how the severity of participants' hazardous gambling behaviour (measured using the PGSI scale) moderated post-outcome emotional reactivity and choice behavior. We developed a simulated slot-machine task with embedded self-reports of emotional valence, allowing us to directly assess participants' trial-by-trial emotional reactivity alongside their choice behavior. As would be expected, a computational model of subjective emotion revealed that win outcomes in this task tended to produce positive emotional reactions, whereas all other outcomes (each of which was associated with a net negative change in total winnings) produced negative emotional reactions. More notably, we found that participants with higher PGSI scores tended to have stronger emotional reactions to all outcome types, both positive and negative. By contrast, PGSI scores did not moderate the effects of outcome type on behavior, though there was an overall main effect of outcome type on choice persistence: participants on average showed increased choice persistence following win and near-win outcomes and decreased persistence following loss outcomes.
Our primary finding is that greater PGSI scores were associated with increased emotional reactivity to gambling outcomes, regardless of outcome type. In other words, participants who engaged in more hazardous gambling behaviors tended to experience more positive emotional reactions after win outcomes and more negative emotional reactions after near-wins, losses, near-losses, and neutral outcomes. This finding supports previous work positing emotion dysregulation as a risk factor for gambling disorder and gambling-related harm (Bonnaire et al., 2022; Buen & Flack, 2021; Rogier & Velotti, 2018), but builds on previous work by directly assessing emotional reactivity (a facet of emotion dysregulation related to the strength of bottom-up emotional reactions; Koole, 2009) during gambling behaviour. We speculate that excessive emotional reactions to various gambling outcomes may in part explain the broader association between emotion dysregulation and gambling related harm; under this perspective, our findings support theoretical models in which emotion dysregulation is a risk factor for gambling-related harm, and specifically suggest that the magnitude of immediate emotional reactions to gambling outcomes may be elevated among individuals most at risk for harm.
The direction of causality in the association between emotional reactivity and hazardous gambling remains unclear. One possibility is that gambling-related emotional reactivity may itself be a risk factor for gambling-related harm. That is, people who experience stronger emotional reactions to gambling outcomes may be more likely to experience gambling-related harm as a consequence. On the one hand, strong negative emotional reactions to aversive outcomes may motivate continued gambling as a way of escaping from negative emotions (Thomas, Allen, & Phillips, 2009). In line with this interpretation, Devos, Clark, Maurage, and Billieux (2018) showed that an experimental sad mood induction resulted in prolonged EGM gambling, suggesting a causal effect of negative emotion on gambling behavior. On the other hand, strong positive emotional reactions to win outcomes may also result in continued gambling (Cummins, Nadorff, & Kelly, 2009), either as a way of recapturing positive feelings (i.e., win-chasing; Zhang, Rights, Deng, Lesch, & Clark, 2024), because of a hyperactive neural reward response (Hewig et al., 2010), or because of a tendency to act impulsively in positive emotional states (i.e., positive urgency; Cyders et al., 2007). In line with this interpretation, both dysregulation of positive emotion and positive urgency are associated with the severity of harmful gambling among individuals with gambling disorder (Rogier, Colombi, & Velotti, 2022).
A second explanation for the association between hazardous gambling and emotional reactivity centers on the high prevalence of cognitive distortions in people experiencing gambling-related harm (Goodie & Fortune, 2013). For instance, Buen and Flack (2021) found that cognitive distortions partially mediated the relationship between emotion dysregulation and gambling-related harm. Extrapolating from this finding, one explanation of our results is that people with stronger emotional reactions to gambling outcomes may be more likely to develop cognitive distortions regarding gambling (e.g., illusions of control, beliefs in luck and superstition; Steenbergh, Meyers, May, & Whelan, 2002), which in turn increase risk of gambling-related harm. We note, however, that this explanation would also predict a moderating effect of PGSI on choice behaviour, which was not observed in the present study. Finally, a third alternative is that heightened emotional reactivity to gambling outcomes may be a consequence (rather than a cause) of gambling-related harm: individuals who gamble hazardously may perceive gambling outcomes to be more personally relevant, resulting in stronger emotional reactivity (Bayer, Ruthmann, & Schacht, 2017). Further research is required to disentangle these competing causal explanations of our study's correlational results.
Our behavioral finding that participants were more likely to persist in choosing the same slot-machine following near-win (and win) outcomes is conceptually consistent with previous studies showing that near-wins prolong gambling sessions (Cote et al., 2003; Ghezzi et al., 2006; Kassinove & Schare, 2001; MacLin et al., 2007) and produce longer post-reinforcement pauses (Belisle & Dixon, 2016; Daly et al., 2014; but see Dixon et al., 2013). Our findings are novel, however, in showing an effect of near-win outcomes on immediate preferences for re-choosing an individual machine. This ‘win-stay’ behavior (which has also previously been shown in response to ‘losses disguised as wins’; Myles, Carter, Yücel, & Bode, 2024) is a general signature of reward in reinforcement learning models (Iyer et al., 2020), and is in line with Thorndike's Law of Effect, which states that actions are more likely to be repeated if they are followed by satisfying consequences (Thorndike, 1911). Complicating this interpretation, however, we also found that near-wins also tended to produce negative emotional reactions. While this is in line with previous studies showing that near-wins are experienced as unpleasant (Clark et al., 2012; Griffiths, 1991; Qi et al., 2011; Sharman & Clark, 2016; Sharman et al., 2015), it is not in keeping with Thorndike's suggestion that only “satisfying” consequences reinforce choice. Indeed, under one perspective our results are a violation of Thorndike's Law of Effect, since they suggest that near-wins are an unpleasant outcome which nevertheless reinforce choice behavior. In addition to this near-win result, we found the opposite to be true for near-losses, where participants were significantly less likely to persist in choosing a machine following near-loss outcomes (as well as losses). This finding—which suggests that near-losses may punish rather than reinforce behaviour—is novel, as the literature has not previously assessed this aversive counterpart of the near-win.
Several limitations of our research approach should be acknowledged. Firstly, our behavioral task included some outcomes where participants could lose more than they initially bet. We included these outcomes because they afforded a broader perspective on emotional reactivity in response to different kinds of gambling outcomes, but it is important to acknowledge that near-losses and losses do not occur in real-world EGMs, and so this task feature reduced the ecological validity for studying EGMs specifically. Similarly, our simulated slot-machines included three reels, two symbol types, and a single payline; as such, these machines were markedly simpler than contemporary real-world EGMs (e.g., 5 reels, 20 symbols, as many as 100 different paylines; Schüll, 2012; Livingstone et al., 2017). This simplification was necessary to reliably isolate the effects of specific outcome types on emotion and behavior, but future research may benefit from utilising more realistic slot-machine stimuli to increase the generalisability of findings. In addition, although we assessed the severity of hazardous gambling for individual participants, we did not collect data on the types of gambling products that different participants typically used; as a consequence, we are unable to draw conclusions regarding whether our findings apply to all individuals with more hazardous gambling or just a subset of them (e.g., regular users of EGMs). In addition, the inclusion of emotion self-report probes after every trial may have disrupted participants' capacity to experience immersion or ‘flow’ within the task, which is notable given that these states have been associated with EGM play in those with high PGSI scores (Dixon et al., 2018; Kruger et al., 2022) and that dissociation within a gambling context has been identified as a feature linking PGSI scores and emotional introspection (Gori & Topino, 2024). However, our finding of amplified positive emotional reactivity to win outcomes in those with higher PGSI scores suggests that our task design did not disrupt the capacity of these individuals to experience reinforcement in response to win outcomes. Finally, our results leave open the question of whether increased emotional reactivity scores are specific to the outcomes of gambling behavior, or whether individuals with higher PGSI scores show a heightened emotional reactivity phenotype more generally, including in response to non-gambling stimuli. Future research should therefore test whether more hazardous gambling is also associated with greater emotional reactivity to stimuli unrelated to gambling (e.g., pleasant/unpleasant images, standardized laboratory mood inductions).
Conclusion
In summary, analyses of choice behavior and subjective emotion within a simulated slot-machine task revealed two novel findings. We found, first, that participants with more severe hazardous gambling behaviors showed stronger emotional reactivity to all outcome types, both positive and negative. This finding suggests that emotional reactivity in response to gambling outcomes may be a risk factor for gambling-related harm, and extends previous work on this topic by documenting heightened emotional reactivity specifically in response to gambling outcomes in individuals with high PGSI scores. Secondly, despite the fact that near-win outcomes tended to produce more negative subjective emotional states, we nonetheless found that near-wins increased the likelihood that participants would choose to immediately repeat their choice of a specific slot-machine. Given that this “win-stay” behavior is a general signature of reinforcement learning, our results support concerns that, despite representing a financial loss, near-win outcomes may directly reinforce choice behavior among EGM users.
Funding sources
This study was supported by internal research funds from the School of Psychological Sciences at Monash University.
Authors' contribution
AC contributed to study concept and design, analysis and interpretation of data, statistical analysis, and manuscript writing. DB contributed to study concept and design, analysis and interpretation of data, statistical analysis, study supervision, and manuscript writing. Both authors had full access to all data in the study and take responsibility for the integrity of the data and the accuracy of data analysis.
Conflict of interest
The authors declare no conflict of interest.
Acknowledgements
The authors gratefully acknowledge helpful comments by Laura Forbes and Dan Myles on an earlier draft of this manuscript.
Supplementary material
Supplementary data to this article can be found online at https://doi.org/10.1556/2006.2025.00003.
References
Australian Gambling Research Centre (2023). Gambling participation and experience of harm in Australia. https://aifs.gov.au/research/research-snapshots/gambling-participation-and-experience-harm-australia.
Bayer, M., Ruthmann, K., & Schacht, A. (2017). The impact of personal relevance on emotion processing: Evidence from event-related potentials and pupillary responses. Social Cognitive and Affective Neuroscience, 12(9), 1470–1479. https://doi.org/10.1093/scan/nsx075.
Becerra, R., Preece, D., Campitelli, G., & Scot-Pillow, G. (2019). The assessment of emotional reactivity across negative and positive emotions: development and validation of the Perth Emotional Reactivity Scale (PERS). Assessment, 26(5), 867–879. https://doi.org/10.1177/1073191117694455.
Belisle, J., & Dixon, M. R. (2016). Near misses in slot machine gambling developed through generalization of total wins. Journal of Gambling Studies, 32(2), 689–706. https://doi.org/10.1007/s10899-015-9554-x.
Betella, A., & Verschure, P. F. M. J. (2016). The affective slider: A digital self-assessment scale for the measurement of human emotions. Plos One, 11(2), e0148037. https://doi.org/10.1371/journal.pone.0148037.
Binde, P., Romild, U., & Volberg, R. A. (2017). Forms of gambling, gambling involvement and problem gambling: Evidence from a Swedish population survey. International Gambling Studies, 17(3), 490–507. https://doi.org/10.1080/14459795.2017.1360928.
Bonnaire, C., Devos, G., Barrault, S., Grall-Bronnec, M., Luminet, O., & Billieux, J. (2022). An empirical investigation of the Pathways Model of problem gambling through the conjoint use of self-reports and behavioural tasks. Journal of Behavioral Addictions, 11(3), 858–873. https://doi.org/10.1556/2006.2022.00055.
Buen, A., & Flack, M. (2021). Predicting problem gambling severity: Interplay between emotion dysregulation and gambling-related cognitions. Journal of Gambling Studies, 38(2), 483–498. https://doi.org/10.1007/s10899-021-10039-w.
Clark, L., Crooks, B., Clarke, R., Aitken, M. R. F., & Dunn, B. D. (2012). Physiological responses to near-miss outcomes and personal control during simulated gambling. Journal of Gambling Studies, 28(1), 123–137. https://doi.org/10.1007/s10899-011-9247-z.
Cote, D., Caron, A., Aubert, J., Desrochers, V., & Ladouceur, R. (2003). Near wins prolong gambling on a video lottery terminal. Journal Of Gambling Studies, 19(1), 433–438. https://doi.org/10.1023/A:1026384011003.
Cummins, L. F., Nadorff, M. R., & Kelly, A. E. (2009). Winning and positive affect can lead to reckless gambling. Psychology of Addictive Behaviors, 23(2), 287–294. https://doi.org/10.1037/a0014783.
Cyders, M. A., Smith, G. T., Spillane, N. S., Fischer, S., Annus, A. M., & Peterson, C. (2007). Integration of impulsivity and positive mood to predict risky behavior: Development and validation of a measure of positive urgency. Psychological Assessment, 19(1), 107–118. https://doi.org/10.1037/1040-3590.19.1.107.
Daly, T. E., Tan, G., Hely, L. S., Macaskill, A. C., Harper, D. N., & Hunt, M. J. (2014). Slot machine near wins: Effects on pause and sensitivity to win ratios. Analysis of Gambling Behavior, 8(2), 55–70.
Davidson, R. J. (1998). Affective style and affective disorders: Perspectives from affective neuroscience. Cognition & Emotion, 12(3), 307–330. https://doi.org/10.1080/026999398379628.
Devos, G., Clark, L., Maurage, P., & Billieux, J. (2018). Induced sadness increases persistence in a simulated slot machine task among recreational gamblers. Psychology of Addictive Behaviors, 32(3), 383–388. https://doi.org/10.1037/adb0000364.
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. International Gambling Studies, 11(3), 393–407. https://doi.org/10.1080/14459795.2011.603134.
Dixon, M. J., Harrigan, K. A., Sandhu, R., Collins, K., & Fugelsang, J. A. (2010). Losses disguised as wins in modern multi‐line video slot machines. Addiction, 105(10), 1819–1824. https://doi.org/10.1111/j.1360-0443.2010.03050.x.
Dixon, M. J., MacLaren, V., Jarick, M., Fugelsang, J. A., & Harrigan, K. A. (2013). The frustrating effects of just missing the jackpot: Slot machine near-misses trigger large skin conductance responses, but no post-reinforcement pauses. Journal of Gambling Studies, 29(4), 661–674. https://doi.org/10.1007/s10899-012-9333-x.
Dixon, M. J., Stange, M., Larche, C. J., Graydon, C., Fugelsang, J. A., & Harrigan, K. A. (2018). Dark flow, depression and multiline slot machine play. Journal of Gambling Studies, 34, 73–84. https://doi.org/10.1007/s10899-017-9695-1.
Ferris, J., & Wynne, H. J. (2001). The Canadian problem gambling Index final report. Ottawa, ON: Canadian Centre on Substance Abuse.
Forbes, L., & Bennett, D. (2024). The effect of reward prediction errors on subjective affect depends on outcome valence and decision context. Emotion, 24(3), 894–911. https://doi.org/10.1037/emo0001310.
Ghezzi, P. M., Wilson, G. R., & Porter, J. C. (2006). The near-miss effect in simulated slot machine play. In P. M. Ghezzi, C. A. Lyons, M. R. Dixon, & G. R. Wilson (Eds.), Gambling: Behavior, theory, research, and application (pp. 155–170). Reno, NV: Context Press.
Goodie, A. S., & Fortune, E. E. (2013). Measuring cognitive distortions in pathological gambling: Review and meta-analyses. Psychology of Addictive Behaviors, 27(3), 730. https://doi.org/10.1037/a0031892.
Gori, A., & Topino, E. (2024). Problematic gambling behavior in a sample of gamblers: The role of alexithymia, dissociation features, and external locus of control. Journal of Gambling Studies, 40(4), 2077–2091. https://doi.org/10.1007/s10899-024-10322-6.
Griffiths, M. (1991). Psychobiology of the near-miss in fruit machine gambling. The Journal of Psychology, 125(3), 347–357. https://doi.org/10.1080/00223980.1991.10543298.
Gross, J. J. (1998). The emerging field of emotion regulation: An integrative review. Review of General Psychology, 2(3), 271–299. https://doi.org/10.1037/1089-2680.2.3.271.
Gross, J. J. (2002). Emotion regulation: Affective, cognitive, and social consequences. Psychophysiology, 39(3), 281–291. https://doi.org/10.1017/S0048577201393198.
Gross, J. J., & Feldman Barrett, L. (2011). Emotion generation and emotion regulation: One or two depends on your point of view. Emotion Review, 3(1), 8–16. https://doi.org/10.1177/1754073910380974.
Harrigan, K., MacLaren, V., Brown, D., Dixon, M. J., & Livingstone, C. (2014). Games of chance or masters of illusion: Multiline slots design may promote cognitive distortions. International Gambling Studies, 14(2), 301–317. https://doi.org/10.1080/14459795.2014.918163.
Hewig, J., Kretschmer, N., Trippe, R. H., Hecht, H., Coles, M. G., Holroyd, C. B., & Miltner, W. H. (2010). Hypersensitivity to reward in problem gamblers. Biological Psychiatry, 67(8), 781–783. https://doi.org/10.1016/j.biopsych.2009.11.009.
Iyer, E. S., Kairiss, M. A., Liu, A., Otto, A. R., & Bagot, R. C. (2020). Probing relationships between reinforcement learning and simple behavioral strategies to understand probabilistic reward learning. Journal of Neuroscience Methods, 341, 108777. https://doi.org/10.1016/j.jneumeth.2020.108777.
Kassinove, J. I., & Schare, M. L. (2001). Effects of the “near miss” and the “big win” on persistence at slot machine gambling. Psychology of Addictive Behaviors, 15(2), 155–158. https://doi.org/10.1037/0893-164X.15.2.155.
Koole, S. L. (2009). The psychology of emotion regulation: An integrative review. Cognition & Emotion, 23(1), 4–41. https://doi.org/10.1080/02699930802619031.
Kruger, T. B., Dixon, M. J., Graydon, C., Larche, C. J., Stange, M., Smith, S. D., & Smilek, D. (2022). Contrasting mind-wandering,(dark) flow, and affect during multiline and single-line slot machine play. Journal of Gambling Studies, 38(1), 185–203. https://doi.org/10.1007/s10899-021-10027-0.
Livingstone, C. (2017). How electronic gambling machines work. Australian Gambling Research Centre, Australian Institute of Family Studies.
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. Behavior Research Methods, 39(2), 237–241. https://doi.org/10.3758/BF03193153.
Myles, D., Bennett, D., Carter, A., Yücel, M., Albertella, L., de Lacy-Vawdon, C., & Livingstone, C. (2023). “Losses disguised as wins” in electronic gambling machines contribute to win overestimation in a large online sample. Addictive Behaviors Reports, 18, 100500. https://doi.org/10.1016/j.abrep.2023.100500.
Myles, D., Carter, A., Yücel, M., & Bode, S. (2024). Losses disguised as wins evoke the reward positivity event‐related potential in a simulated machine gambling task. Psychophysiology, 61(6), e14541. https://doi.org/10.1111/psyp.14541.
Palmer, L., Ferrari, M. A., & Clark, L. (2024). The near-miss effect in online slot machine gambling: A series of conceptual replications. Psychology of Addictive Behaviors, 38(6), 716–727. https://doi.org/10.1037/adb0000999.
Qi, S., Ding, C., Song, Y., & Yang, D. (2011). Neural correlates of near-misses effect in gambling. Neuroscience Letters, 493(3), 80–85. https://doi.org/10.1016/j.neulet.2011.01.059.
Rogier, G., Colombi, F., & Velotti, P. (2022). A brief report on dysregulation of positive emotions and impulsivity: Their roles in gambling disorder. Current Psychology, 41(4), 1835–1841. https://doi.org/10.1007/s12144-020-00638-y.
Rogier, G., & Velotti, P. (2018). Conceptualizing gambling disorder with the process model of emotion regulation. Journal of Behavioral Addictions, 7(2), 239–251. https://doi.org/10.1556/2006.7.2018.52.
Russell, A. M. T., Browne, M., Hing, N., Rockloff, M., Newall, P., Dowling, N. A., … Woo, L. (2023). Electronic gaming machine accessibility and gambling problems: A natural policy experiment. Journal of Behavioral Addictions, 12(3), 721–732. https://doi.org/10.1556/2006.2023.00044.
Rutledge, R. B., Skandali, N., Dayan, P., & Dolan, R. J. (2014). A computational and neural model of momentary subjective well-being. Proceedings of the National Academy of Sciences, 111(33), 12252–12257. https://doi.org/10.1073/pnas.1407535111.
Schüll, N. (2012). Addiction by design: Machine Gambling in Las Vegas. Princeton University Press.
Sharman, S., Aitken, M. R., & Clark, L. (2015). Dual effects of ‘losses disguised as wins’ and near-misses in a slot machine game. International Gambling Studies, 15(2), 212–223. https://doi.org/10.1080/14459795.2015.1020959.
Sharman, S., & Clark, L. (2016). Mixed emotions to near-miss outcomes: A psychophysiological study with facial electromyography. Journal of Gambling Studies, 32(3), 823–834. https://doi.org/10.1007/s10899-015-9578-2.
Steenbergh, T. A., Meyers, A. W., May, R. K., & Whelan, J. P. (2002). Development and validation of the gamblers' beliefs questionnaire. Psychology of Addictive Behaviors, 16(2), 143. https://doi.org/10.1037/0893-164X.16.2.143.
Thomas, A. C., Allen, F. C., & Phillips, J. (2009). Electronic gaming machine gambling: Measuring motivation. Journal of Gambling Studies, 25, 343–355. https://doi.org/10.1007/s10899-009-9133-0.
Thorndike, E. (1911). Animal intelligence: Experimental studies. Macmillan.
Velotti, P., Rogier, G., Beomonte Zobel, S., & Billieux, J. (2021). Association between gambling disorder and emotion (dys)regulation: A systematic review and meta-analysis. Clinical Psychology Review, 87, 102037. https://doi.org/10.1016/j.cpr.2021.102037.
Yücel, M., Carter, A., Harrigan, K., Van Holst, R. J., & Livingstone, C. (2018). Hooked on gambling: A problem of human or machine design? The Lancet Psychiatry, 5(1), 20–21. https://doi.org/10.1016/S2215-0366(17)30467-4.
Zhang, K., Rights, J. D., Deng, X., Lesch, T., & Clark, L. (2024). Between-session chasing of losses and wins in an online eCasino. Journal of Behavioral Addictions, 13(2), 665–675. https://doi.org/10.1556/2006.2024.00022.
Zorowitz, S., Solis, J., Niv, Y., & Bennett, D. (2023). Inattentive responding can induce spurious associations between task behaviour and symptom measures. Nature Human Behaviour, 7(10), 1667–1681. https://doi.org/10.1038/s41562-023-01640-7.
