Methods
The following framework was employed to conduct the systematic review: (1) formulated the research questions; (2) defined the inclusion and exclusion criteria for study selection; (3) conducted the systematic literature search; (4) screened search results through title, abstract and full text; (5) extracted relevant evidence from the included studies; (6) synthesized and summarized evidence. Our review team consisted of knowledge experts in the fields of gambling research, systematic reviews, psychology, consumer behaviour, as well as two information specialists.
The systematic review investigated two research questions: what are the effects of the near misses/LDWs on the player? and Does the gambling status of a player (i.e., normal, at-risk, or problem gambler) alter the effect of near misses/LDWs on the player? Effects were defined as changes in psychological or cognitive state, behaviour, or psychobiology. Finally, data were examined for the effects of specific aspects of how the EGMs provide feedback to identify key characteristics that may be driving the response to near misses and LDWs.
Literature Search
A systematic literature search for relevant studies was conducted in November, 2015 using Scopus, PubMed, PsycINFO(PsycNet), and Proquest Sociology databases as well as the Gambling Research Exchange Ontario Knowledge Repository—Synopsis collection which contains concise summaries of peer-reviewed gambling research articles. Comprehensive search methods were developed by the information specialists in consultation with the research team. Search strategies consisted of author keywords (those appearing in the title or abstract of the paper) and subject headings (controlled vocabulary specific to each database) focusing on two sets of search terms: (1) the structural characteristics of EGMs of interest: near miss, just missed, almost winning, just missing, narrow win, virtual reel, losses disguised as wins, multiline, multiple paylines, and small wins; and (2) gambling-related terms intended to focus the review on the effects of EGMs alone, with culturally-appropriate terms for EGMs from a variety of nations: gamble, lottery, gaming, pokies, poker, slot, and fruit machines. The primary and secondary search terms were combined using the logical AND, OR, truncation and wildcard, and WITHIN operators to identify all pertinent studies within each research database.
Full search strategies are available as electronic supplementary material.
Study Selection
Studies selected for inclusion in the review were required to be published, peer-reviewed, written in English, include experiments on human participants, and using an EGM or simulation of an EGM. Studies were included if they specifically addressed the effects of near misses or LDWs in gambling on single line slot machines, multiline slot machines, or computer generated simulations of either single or multiline slot machines. In addition, studies were required to quantify the effect on the player either psychologically (i.e., self-report), behaviourally (e.g., reaction times, amount gambled, etc.), or psychophysiologically (e.g., skin conductance, heart rate, etc.). Studies that were either randomized controlled trials or observational in nature were included in the review. The relevance of each study identified by the literature search was assessed in two stages, by title and abstract screening and by full text screening. A screening tool was developed by the research team to guide reviewers in both stages of the relevance assessment process. To ensure reliability of the screening tool, the screening tool was piloted separately on ten randomly selected publications from the literature search for both the title and abstract screening and full text screening steps. Each title and abstract, and later full text, were reviewed by two independent readers using the screening tool. At both stages of screening, the decision to include or exclude a study was made by reaching consensus. In cases where consensus could not be reached, the appropriate knowledge experts on the research team were consulted to make the final decision.
Data Extraction and Review
Data and evidence were extracted from all included studies through the use of a data extraction tool. The data extraction tool was developed by the research team to guide each reviewer throughout the data extraction stage. The data extraction tool used the following categories: primary author, jurisdiction, study design, sample size, population employed, type of gambling task studies, topic of research (either near miss or LDW), a description of psychological, physiological, or behavioural outcome or effect, concluding remarks, and recommendations for policy makers and practitioners. To ensure reliability of the data extraction tool, the tool was piloted on ten randomly selected studies.
The evidence collected by the data extraction was summarized using thematic analysis, identifying all consistent and unique effects of near misses and LDWs on the player. Due to the sizable diversity in the methods, types of collected data, and populations studied within each included study, a meta-analysis of the results was not possible.
Discussion
This systematic review examined the psychological, behavioural, and psychobiological responses of individual players in response to near misses and LDWs across 51 studies published from 1991 to 2015.
For near misses, a number of consistent findings were observed. A number of studies suggest that near misses increase the frequency with which a player will estimate that they are winning and motivate continued play (nine studies); encourage longer play(three studies); lead to overestimation of the frequency of winning (four studies). Near misses also appear to result in an increase in SCL in a large number of studies (10 out of 11 studies). Near misses were also found to be viewed as negative or aversive events in 6 out of 9 studies. The present review also found no evidence for these effects to be the product of game feedback, suggesting that the response to near misses is a product of the phenomenology of seeing matching symbols alone and not some byproduct of how the EGM produces an exciting or engaging gameplay experience to generalize near misses as wins, such as through the use of visuals, animations, and sounds.
Notably, however, findings were not completely uniform across all the sampled studies, with considerable spread in the reported effects of near misses in a variety of different measured outcomes. For example, near misses were found to be associated with increasing one’s bet, decreasing one’s bet, or having no effect, each in a different study, making it difficult to determine whether near misses are capable of influencing per-play betting behaviour. The precise reasons for this inconsistency is presently unclear. However, existing work on gambling behaviour has found that the choice of behaviour is likely influenced by a number of individual (such as traits, motives, and gambling status) and situational (options to play, amount of money available, etc.) factors (Smith et al.
2007), many of which can vary greatly from player-to-player. Another inconsistent result was observed in how players respond emotionally to a near miss event, both in terms of self-reported measures and in the neuroimaging data. One likely possibility for the inconsistency in these studies is that the response to a loss event, or an event which is similar to a loss, like a near miss, is driven by more complex cognitive constructs (such as the degree of counterfactual thinking, see: Henderson and Norris
2013), leading to more varied responses when this and other variables are not accounted for. Other work has also shown that individual differences with response to gambling losses is strongly associated with the expectation of success and the degree to which the game is enjoyable or reinforcing (Campbell-Meiklejohn et al.
2008). Taken as a whole, betting and gambling behaviour appears to be the result of a potentially large number of different factors which vary between players, each of which must be better identified and measured in further studies so that the precise reason for the effect of near misses can more completely and accurately be understood. This is considered particularly important, in light of the consistent finding that EGMs featuring near misses appear to encourage or extend play—the underlying reason for the motivation to play remains unclear, at least at present.
Another area that requires further investigation is the ability for near miss events to produce a response in the player at the physiological level, such as is the case in heart rate and HRV outcomes. Despite finding that 10 out of 11 studies showed a significant elevation in SCL in response to a near miss, no such consensus was observed for heart rate or HRV and near misses. One reason for this may be that skin conductance responses are physical responses brought on by fundamentally different types of processing, but both occurring, at least in part, through activity in the autonomic nervous system. For example, when recording SCL during rest versus the performance of eight different tasks, tasks which captured both internalized processing (such as, solving complex arithmetic problems) and the processing of external stimuli (i.e., distinguishing between different levels of white noise), SCL was found to show a response in all tasks (Lacey et al.
1963). In contrast, heart rate was shown to decelerate in response to tasks requiring attention to external stimuli and accelerate in response to more internal processing. Other work has also suggested that heart rate may be less sensitive to certain kinds of emotional processing, such as that of sadness, than skin conductance measures (Kreibig et al.
2007). A study using simultaneously EEG, heart rate, and SCL measures in tasks requiring vigilance or sustained attention has shown that heart rate may be more sensitive to changes in overall vigilance in performing a task, whereas skin conductance was found to be associated with effort or time-on-task (Olbrich et al.
2011). Thus, heart rate and heart rate variability, rather than SCL, appear more strongly influenced by the type of processing being engaged in by the player in an average gaming session, something that can vary considerably across individual players. Future studies including these measures would profit from a more rigorous account of the types of processing (e.g., internal versus external processing) and degree to which the participant is attending to the gambling task, in addition to the existing practice of measuring self-reported emotional state and overall gambling severity level. Without more stringent control and investigation of gambling phenomenology, it is difficult to conclude how exactly near miss events are affecting the player, at a cognitive or neural level. This also suggests that conclusions raised from HR data, without further specification, should be interpreted with caution.
Diffuse activity in the brain reported across five studies of the effects of near misses, but a number of common areas were found to be active when processing near miss events, including the insular cortex (three studies), ventral striatum (two studies), and inferior parietal tissue (two studies), though activity was also reported in other areas in the brain, ranging from prefrontal tissue to occipital tissue. While each reported area is likely meaningful, in some way, to the context of gambling and the response to near misses, the less consistent activity may be the product of idiosyncrasies in task design or analysis regime. In terms of the most consistently significant activity in the insular cortex, inferior parietal, and striatum, these areas have often been implicated in the processing of uncertainty and in the assessment of reward or punishment status. For example, in a set of non-gambling tasks, the activity in the ventral striatum has been correlated with the magnitude of reward or punishment (Hsu et al.
2005) and is immediately separable from activity related to uncertainty or risk assessment. In contrast, activity inferior parietal, but particularly the inferior parietal lobule, appears related to the processing of uncertainty (Vickery and Jiang
2009). It is reasonable to conclude that near misses would recruit either of these modes of processing, so the observed activity is consistent with these other fields of research. The extent to which either uncertainty processing or reward/punishment assessment is engaged in when experiencing near miss events is a topic for further work. But the strongest and most convincing evidence for the direct role of this type of tissue or processing can be found in studies using patients with lesions, such as in Clark et al. (
2014). In patients with lesions to the insula, the presence of a lesion significantly reduced the response to near miss events and their reinforcement on behaviour. Despite the relatively few studies identified by our systematic review addressing near miss events and the brain, it is clear that future work should further investigate the specific role of the insula and the complex interplay between tissue in the insular cortex, striatal, and inferior parietal lobule in processing near miss events.
In terms of real-time processing in the brain, consistent EEG signals (i.e., FRN, P300, and P3b) were observed in relation to near miss events in a number of studies. In one study, P300 and P3b were shown to be associated with false spatial feedback, while FRN activity was associated with unexpected negative feedback alone (Balconi and Crivelli
2010). This is consistent other studies that show that FRN activity related to fairness precedes the P300 signal, which has been associated with a state of uncertainty or dissonance (Yu et al.
2015). Together, this suggests that near miss events may quickly be identified by the player’s brain as a negative event, but the false or uncertain nature of the near miss may take longer to be processed or revealed. How the timing and magnitude of these neural events relates to the selection and initiation of the next gambling act (be it continued play, pausing for a distinct period of time, or ceasing of a play session) is something that will require further investigation. One potential avenue for this future work would be to explore the clarity of the near miss event as aversive, as it is expected that a stronger differentiation between FRN and P300 signals would be possible when the outcome of the game was more directly manipulated.
With regard to LDWs, a strong and clear picture of their effect on the gambler was found, despite being a topic of less concerted study. LDWs were consistently found to inflate win estimates and be a component of players’ preferred games. There was also some evidence, in the form of SCL response and self-report, to indicate that the number of credits awarded by the LDW is what induces the reinforcement of play, despite ongoing losses. Three studies also specifically identified that game sounds were important in this reinforcement process. However, only one study was explicitly assessed how LDWs are perceived by the player or are effecting the player, making it difficult to determine if the arousal present in the SCL and HRV responses are the product of excitement or are instead the product of a change in the level of effort or vigilance expended while gambling, as was introduced previously. Through further work differentiating the various modes of cognitive processing from the excitement of an LDW event, much stronger models could be established to describe and assess the effect of LDWs on healthy and problem gamblers alike, and could highlight new methods for producing a safe and healthy gambling environment. Studies on the neural substrates of the response to LDWs could also place the behavioural findings in a more general context, allowing a more direct comparison between the cognitive processing underlying both types of simultaneously reinforcing loss events—the near miss and the LDW—to be accounted for.
In both the study of near misses and LDWs, whether the effects differ in problem or at-risk populations remains to be seen. Only a few studies explored research questions related to non-healthy players. Of the studies that were identified, the research approach and question of interest varied widely, making it difficult to draw any strong conclusions on whether near misses or LDWs affect problem gamblers in any consistent way. However, some evidence was found to indicate that problem gamblers may have a reduced or suppressed emotional response to near misses. This was indicated by diffuse patterns of elevated activity in prefrontal cortex, amygdala, and striatum across a number of studies (Goldin et al.
2008). Many of the brain areas, such as the midbrain and striatum, have also been implicated in altered dopamine transmission and reception throughout the brain, something that appears relevant in gamblers as a whole (Bergh et al.
1997; Meyer et al.
2004). In two of the sampled studies, one on near misses and one on LDWs, some evidence was reported for players to be hyposensitive to wins and losses, neurologically (Lole et al.
2015) or in terms of SCL level (Lole et al.
2014). Problem gamblers were also found to prefer multiline slot machines more than non-problem gamblers. Taken as a whole, these works suggest that the complex emotional or reinforcing responses that problem gamblers have to near misses and LDWs may be blunted, potentially encouraging continued play when it is otherwise inadvisable. One reason for this may be that near misses and LDWs produce brief levels of excitement, causing spikes in the otherwise reduced response, sufficient to produce engagement and continued play, but insufficient to cause a change in behaviour, though this possibility requires further study to confirm.
The present review was limited by the inclusion of studies using a diverse number of techniques and approaches to studying the effects of near misses and LDWs. Given the number of studies identified investigating the topic of LDWs, this choice represented a practical necessity. However, within the topic of near misses, it remains possible that a more stringent review could establish with greater certainty the particular effect of near misses on the player through a statistically rigorous meta-analysis. As no review of this scope exists in the field of gambling studies at this time, we instead chose to describe and relate the effects of near misses and LDWs on the player to provide a current account of the effect of each mechanism and the current state of knowledge, however, rather than precisely target one specific research question. The current review was also limited by its inclusion of only peer-reviewed work. It remains possible that books, dissertations, or grey literature could provide more detail on the understanding of the effects of near misses and LDWs, given the multi-disciplinary nature of the field. However, these forms of media were excluded to ensure a consistent level of quality throughout the review.
In spite of these limitations, the current review provides an important foundation for future work in this area. The present systematic review on near misses and LDWs establishes how each of these two systematic characteristics of slot machines that appear to mislead the player into gambling through a variety of different mechanisms or outcome. Where clear evidence was found for near misses to be perceived as “almost winning”, reinforcing continued play through their surprising or exciting nature, LDWs appear to be viewed as a type of win more directly (despite actually being a loss). This systematic review provides a comprehensive description of the effects of near misses and LDWs, highlighting both consistent and inconsistent findings. Clear directions for future research were also provided, addressing topics of theoretical or conceptual importance to the understanding of EGM play and gambling behaviour as a whole and providing a framework for future work to build upon.