Journal of Gambling Studies

, Volume 29, Issue 4, pp 661–674

The Frustrating Effects of Just Missing the Jackpot: Slot Machine Near-Misses Trigger Large Skin Conductance Responses, But No Post-reinforcement Pauses

  • Mike J. Dixon
  • Vance MacLaren
  • Michelle Jarick
  • Jonathan A. Fugelsang
  • Kevin A. Harrigan
Original Paper

DOI: 10.1007/s10899-012-9333-x

Cite this article as:
Dixon, M.J., MacLaren, V., Jarick, M. et al. J Gambl Stud (2013) 29: 661. doi:10.1007/s10899-012-9333-x


Near-misses in slot machines resemble jackpot wins but fall just short. Previous research has demonstrated that near-misses are behaviorally reinforcing despite the absence of monetary reward. We assessed the hedonic properties of near-misses by measuring the time between outcome delivery and the initiation of the next spin—the post-reinforcement pause (PRP) and skin conductance responses (SCRs) for losses, near-misses, and a range of wins (5, 15, 25, 50 or 250 credits) while participants (N = 122) played a slot machine simulator. PRPs and SCRs were compared for 40 low frequency and 22 high frequency slots players who were non-problem gamblers, 37 at risk players, and 23 problem gamblers. For winning outcomes, PRPs and SCRs tracked monotonically with win size such that progressively larger wins were associated with progressively larger PRPs and SCRs. Near-misses with jackpot symbols landing on the first two reels had significantly larger SCRs than regular losses, and other types of near misses. Crucially, PRPs for this kind of near-miss were significantly smaller than all wins, and when non-parametric statistics were used, significantly smaller than regular losses. This pattern of large SCRs and small PRPs suggest that these are highly frustrating outcomes that stimulate appetitive components of the reward system to promote continued gambling.


Near-missesPost-reinforcement pausesSkin conductance responsesSlot machinesGambling


In mechanical reel slot machines, near-misses are outcomes that resemble jackpot wins, but fall just short (Reid 1986). If the jackpot involves three red sevens falling on the payline, a classic near-miss would be two red 7s on the payline and the third red 7 just above or just below the payline. For mechanical reel slot machines, slots designers use virtual reel mapping to ensure that the high paying symbols frequently land on reel locations that lead to near-misses (Harrigan 2008).

Near-misses are important because they influence the psychology, and psychophysiology of the player as well as their overt behaviour. Near-misses evoke high levels of subjective arousal (Griffiths 1990a, b, c, 1991; Parke and Griffiths 2004) and have been shown to foster the urge to continue gambling (Clark et al. 2009, 2011. Importantly, near-misses have been shown to extend play in simulated (Kassinove and Schare 2001; MacLin et al. 2007) and actual slot machines (Cote et al. 2003).

The effect of near-misses has been characterized as a cognitive misrepresentation. Dixon and Schreiber (2004) had participants rate how similar different slot machine outcomes were to wins. All twelve of their participants rated near-misses as being closer to wins than regular losses (see also Habib and Dixon 2010 for similar findings). As such they concluded that “the near-miss, or in Skinner’s words “almost a win” appears to be evaluated by participants as approximating a winning trial more so than a total loss trial type” (p. 344). Griffiths (1991) conjectured that “if these excessive gamblers become physiologically aroused when they win or nearly win, then in their minds they are not constantly losing but constantly nearly winning” (p. 356). We refer to such mentations as cognitive misrepresentations since (as noted by Dixon and Schreiber 2004) the randomly selected outcomes in slot machines ensure that no losing outcomes are any closer to wins than any other losing outcomes. According to the cognitive misrepresentation theory, it is the surface similarity of near-misses to wins that causes players to treat these outcomes as somehow akin to a win.

The cognitive misrepresentation account was rejected by Clark et al. (2009). They hypothesized that the rewarding property of the near-miss is related to the illusion of control. In their view, players appear to equate near-misses in slots to near-misses in other skill games like basketball. Thus, just as a three point shot that hits the back of the rim is seen as more skillful than an ‘air ball’ that completely misses the rim, to certain gamblers, a near-miss in slots is seen as a reflection of their skill that enables them to get very close to the jackpot.

Clark et al. had participants play a simplified 2-reel slots game simulator. In one condition, the computer selected a symbol that landed on the payline on reel 1. This was the to-be-matched symbol. The second reel then spun and stopped. If the symbol on the payline on reel 2 matched the symbol on the payline on reel 1, the player won, if not they lost. Near-misses were mismatches (losses), in which the matching symbol appeared just above or below the payline on reel two. In a second condition the participants (rather than the computer) selected the to-be-matched symbol. The software controlled only the second reel. The outcomes in both conditions were identical, but participants in the illusory control condition rated their chances of winning to be significantly higher before spinning than those who did not have control over the first reel. The participants also rated wins as ‘pleasant’, and near misses as even more ‘unpleasant’ than regular losses. Despite their unpleasantness, near-misses in the illusory control condition were also associated with greater urges to continue playing. The fact that players rated near-misses as being less pleasant than both wins, and regular losses is at odds with the cognitive view that near-misses are somehow miscategorized as being somehow like regular wins.

Near-Misses Activate Brain Reward Systems

Clark et al. (2009) had participants undergo functional Magnetic Resonance Imaging (fMRI) while playing their 2-reel simulator. Regular wins were associated with greater activation in the ventral putamen, anterior insula, midbrain, and anterior cingulate cortex than losses. Near-misses were also associated with greater activation of the ventral putamen and right anterior insula. Habib and Dixon (2010) found similar results, with wins activating the left midbrain near the substantia nigra and ventral tegmental area. Finally, Chase and Clark (2010) found that near-misses activated areas of the ventral striatum similarly to regular wins, and that midbrain activation correlated positively with the severity of symptoms of problem gambling among regular gamblers. Similar results were shown by Habib and Dixon (2010)—for non-problem gamblers, the areas activated by wins were distinct from the areas activated by near misses. For problem gamblers wins and near misses activated overlapping areas. Such activation of the dopaminergic centers is compelling evidence that near-misses activate structures that are widely known to be part of a subcortical system that mediates behavioral reinforcement.

Although near-misses are unpleasant outcomes, neuroimaging research clearly shows that they activate some of the brain areas associated with reward. This puzzling result may be resolved by the fact that the dopaminergic mesolimbic system is not a homogenous entity; rather, it contains at least two distinct functional components that may be difficult to discern with the temporal resolution of fMRI. The consummatory component (Alcaro et al. 2007) is responsible for the subjective ‘liking’ of hedonic enjoyment (Robinson and Berridge 2000). The appetitive component generates the anticipatory ‘wanting’ of opportunities to activate behaviors that lead ultimately to the satisfaction of needs. These two subsystems of the reward circuitry are closely interconnected, but serve distinct motivational and affective functions and have dissociable neurophysiological substrates (Berridge 2007). The pathophysiology of addiction largely consists of sensitization of the appetitive system, and habituation of the consummatory system with chronic exposure to powerful sources of hedonic reward such as drugs of abuse (Koob and Le Moal 2008).

If near-misses promote appetitive ‘wanting’ but not consummatory ‘liking’, they could reinforce persistent gambling behavior. Furthermore, when goal pursuit is frustrated (Amsel 1958), as clearly happens in the case of near-misses, the appetitive system of some individuals may respond by activating increased effort and resistance to extinction of the reward-seeking behavior (Carver and Harmon-Jones 2009).

Frustration elicits marked psychophysical changes that can be documented using skin conductance responses (SCRs) (Lobbestael et al. 2008), and by phasic changes in heart rate (Osumi and Ohira 2009). Clark et al. (2011) used their 2-reel slot simulator to investigate the psychophysiological responses elicited by near-misses. Near-misses led to higher SCRs than full losses but only when the gambler chose the to-be-matched symbol. They also showed that after an initial heart rate deceleration, near-misses triggered larger heart rate acceleration than full misses. Both findings are hallmarks of frustration.

Dixon et al. (2010) recorded SCRs and heart rate decelerations (HRDs) of slots players as they played a realistic three-reel slots simulator. They recorded SCRs and HRD for each loss, win, and near-miss during play. Average SCRs and HRD were both significantly larger for near-misses than either wins or losses. Based on the complementary SCR and HRD findings, Dixon et al. proposed that these arousal patterns are due to the frustration of just missing a big win.

In the Dixon et al., study, only “classic” near misses were analyzed (where the first and second reels stopped with a jackpot symbol on the payline, but the rightmost reel stopped with the jackpot symbol just off the payline). These classic near misses were thought to be particularly frustrating, because of the build-up of anticipation of a jackpot as the first reel stops first with a jackpot symbol on the payline, followed by the second reel stopping with a second jackpot symbol on the payline. This build up of the anticipation of a jackpot win, followed by the deflation of just missing the jackpot might make these classic near-misses especially frustrating—more frustrating than other near misses where the first reel stops with a non-jackpot symbol on the payline, and the jackpot just off the payline. Even if the last two reels stop with jackpot symbols on the payline, anticipation would not build up, as the player knows that they cannot win the jackpot after the first reel stops spinning. In the current study, classic near-misses will be compared to other near misses to document their preferentially frustrating properties.

In the Dixon et al. study, the largest win size was a relatively small win of 25 credits. Concerning SCRs, a current goal therefore was to replicate and extend this earlier work by including a fuller range of win sizes to see if the size of the SCRs would scale with the sizes of wins, and to see if classic near-misses would show larger SCRs than other near-misses as well as regular losses.

Post-reinforcement Pauses Indicate Enjoyment of Hedonic Reward

Behavioral studies and fMRI indicates that near-misses support persistent gambling and also activate brain areas that are associated with behavioral reinforcement. If near-misses are misconstrued wins, they should be inherently pleasurable. If near-misses are highly frustrating losses they should be unpleasant, yet may stimulate further play via the appetitive reward system. One potential way of adjudicating between these competing alternatives is to look at post-reinforcement pauses (PRPs). If one measures the time between outcome delivery and initiation of the next spin, the typical finding is that this duration is longer for winning versus losing outcomes, and that the duration of the PRP increases with the magnitude of the win (Delfabbro and Winefield 1999; Peters et al. 2010). One explanation of these pauses is that the consummatory enjoyment of reward exerts an inhibitory influence on continued appetitive reward seeking (see Leslie 1996). If so, then one would predict large PRPs for hedonically pleasurable outcomes, and no PRPs for frustrating outcomes.

Comparative and human studies have used PRP methods to understand the different reinforcing roles of wins, losses, and near-misses. Scarf et al., (2011) showed that pigeons playing a slots-like pecking game have different dedicated neurons that respond specifically to wins, losses, and near-misses. Behaviourally, they noted larger PRPs for wins compared to losses. Similarly in rats, wins led to longer PRPs than losses (Peters et al. 2010). When PRPs for near-misses were compared to standard losses, no effects were found for the pigeons, but longer PRPs were noted for near-misses than standard losses for rats. Peters et al. concluded that because of the perceptual similarity of near-misses to wins, the near-misses acquired reinforcing properties due to Pavlovian generalization. In human participants, Dixon and Schreiber (2004) showed that 8 of 12 participants playing a real slot machine had longer PRPs for wins compared to losses, but 4 participants showed no effect. For these latter participants, PRPs for near-misses were actually larger than either wins or losses. It should be noted that in Dixon and Schreiber, Scarf et al., and Peters et al., participant numbers were quite small (5–12). To rectify these shortcomings, we measured the PRP following losses, near-misses, and wins of various sizes (5, 15, 25, 50, and 250 credits) and collected data on a large number of participants (n = 122).

In the following study, we sought to demonstrate that PRPs following wins of different sizes would track with the magnitude of the win (Dixon and Schreiber 2004; Delfabbro and Winefield 1999). Next, we sought to demonstrate that the amplitude of the SCR would also scale with win size (Lole et al. 2011; Wilkes et al. 2010). Such findings would allow us to adjudicate between two competing predictions concerning near-misses: if near-misses are misclassified as wins, then they should stimulate the consummatory reward system, and consequently have longer PRPs than regular losses. Alternatively, if near-misses are interpreted as frustrating losses (Dixon et al. 2010), then the consummatory reward system should not be activated and the PRP duration following near-misses should be no greater than that which follows losses. Indeed, near-misses might even lead to PRPs that are actually shorter than for regular losses, since players might wish to quickly escape that frustrating state by initiating the next spin as quickly as possible. Thus, PRPs combined with SCRs provide a powerful means to adjudicate between the miscategorization and frustration hypotheses of near-misses. Here, a pattern of small PRPs but high SCRs would favour the frustration hypothesis.



We tested 129 people in total. Technical problems precluded recording PRPs and SCRs for 7 participants. For one other participant, PRP only was recorded due to an SCR-electrode breakage. This left a usable sample of 122 individuals (56 males) for PRPs and 121 for SCRs. Gamblers ranged in age from 19 to 65 (mean = 37). Gambling status was assessed using the Problem Gambling Severity Index (PGSI). Non-problem gamblers (0–2 on the PGSI) were divided into frequent gamblers (twice per month or more) and low-frequency gamblers. There were 40 low-frequency non-problem gamblers (Lo-Freq NPGs), 22 high-frequency non-problem gamblers (Hi-Freq NPGs), 37 at-risk gamblers (3–7 on the PGSI) and 23 problem gamblers (8 or more on the PGSI). Gamblers were recruited from the community and the University, using newspaper and web based (e.g., Kijiji) advertisements. Demographic and gambling related means of the participants are shown in Table 1.
Table 1

Means (standard deviations) of demographic and gambling related variables of 122 participants


Mean age

Slots playing frequency per year


Lo-freq NPGs

33.68 (15.50)

4.11 (5.01)

1.05 (3.15)

Hi-freq NPGs

41.23 (13.70)

45.09 (43.02)

.95 (.84)


37.38 (14.81)

28.93 (21.05)

4.68 (1.29)


38.48 (13.32)

51.87 (55.34)

12.48 (4.29)

NPGs non problem gamblers, PGs problem gamblers, PGSI Problem Gambling Severity Index


All testing was conducted using a slot machine simulator created by Game Planit Interactive Corp. This realistic slots simulator contained three animated reels, pay table, counters indicating bet size, spin outcome displays, and running totals (see Fig. 1). For wins, the winning symbols flashed on and off repeatedly, and a 1-s celebratory song was played. This winning song was identical and equivalent in length for all win sizes. Participants did not have to wait for the winning song to finish before initiating the next spin. The winning songs (and flashing symbols) were absent on losing and all types of near-miss outcomes. The timing of the reels, and celebratory songs is shown in Fig. 2.
Fig. 1

Slot machine simulator used in all experiments
Fig. 2

Timing of the reels, outcome delivery, winning songs, and PRPs


The current study extracted data from two experiments designed to investigate a different research question (the subjective reactions to winning and losing streaks). Both experimental protocols involved playing a single practice block of 40 spins followed by 12 experimental blocks of 40 spins each. For this study, the first experimental trial, and last trial in each block were eliminated. Near-misses were either of the classic JJX type (where Js stand for jackpot symbols, X for non-jackpot symbols) or non-classic types. On JXJ near-misses, the player knew they had lost by the stopping of the second reel. On XJJ near misses, the player knew they had lost by the stopping of the first reel were also employed. After eliminating the first and last spins, protocol 1 had 284 losses, 27 five-credit wins, 25 fifteen-credit wins, 35 twenty-five-credit wins, 9 fifty-credit wins, 8 two-hundred and fifty credit wins, 44 classic (JJX) near misses, and 24 other (JXJ or XJJ) near misses. Protocol 2 had 275 losses, 28 five-credit wins, 26 fifteen-credit wins, 36 twenty-five-credit wins, 12 fifty-credit wins and 12 two-hundred and fifty credit wins and 46 anticipatory build-up near misses and 21 other near misses.

In each protocol, 6 blocks contained near-misses. In these blocks 12 of the 40 spins were near misses (30 %). In protocol 1, blocks comprised spins with more wins than normal, an average amount of wins, or fewer wins than normal. In protocol 2, blocks either had either more wins than normal (winning streaks) or fewer wins than normal (losing streaks). For each game, the 12 blocks were presented using one of two randomized orders counterbalanced across participants.


Participants signed a consent form and were administered the PGSI. Skin conductance electrodes were attached to the first and third fingers of the left hand. Clamp-on EKG electrodes were also attached to the upper arms and right wrist (but this data was not analyzed). Participants were told they would be playing a slot machine game divided into brief sessions (i.e., blocks). They were to verbally indicate whether the session was a winning or losing streak. They bet 3 credits per spin with each credit being worth $1.00 CDN of simulated money. Participants were shown the pay table and told how to look up the yield associated with each winning combination. Participants were informed that three red 7s on the payline was a ‘jackpot’ that would yield 2,500 credits. Each block began with the participant cashing out the balance from the previous block and then loading the machine with $120.00 simulated dollars via a mouse-click interface.

Participants received $10 for participating, but could also win up to an additional $10 based on the slot machine outcomes. Portions of the $10 were paid after each block (sessions with total credits less than 100 paid .50¢, credits between 100 and 500 paid .75¢, and credits more than 500 paid $1.25).

At the end of each block, participants were asked to evaluate the strength of the winning or losing streak on a scale of −100 to +100. Streak data were not analyzed in this study.

Data reduction

PRPs for each outcome were calculated by taking the time between the current outcome and the initiation of the next spin (see Fig. 2). For each participant, 8 average PRPs were calculated based on the average PRP length for that particular condition (e.g., average PRP for the 275 losses). Averages were based on as few as 8 observations (wins of 250) and as many as 275 observations (losses). Prior to calculating averages, outliers were eliminated using the Van Selst and Jolicouer (1994) procedure that adjusts for samples with different numbers of observations per cell (on average 2.5 % of the data were removed).

SCRs for each outcome were collected by defining a two second window beginning one second after outcome delivery (Dawson et al. 2000). The skin conductance level at the beginning of the window was subtracted from the maximum amplitude within the window to calculate the SCR. To minimize the positive skew associated with SCRs, a square root transformation was applied to all SCRs. After removing outliers (on average 3.6 % of the data) for each participant, 8, outlier-free, average SCRs were calculated based on the SCRs for each outcome type.


Post-reinforcement Pauses

The average PRPs for losses, wins of 5, 15, 25, 50 and 250 credits, classic near-misses, and other near-misses were analyzed using a mixed model analysis of variance (ANOVA) with outcome (losses, 5, 15, 25, 50 and 250 credit wins, classic near-misses and other near-misses) as the repeated factor, and gambling group (Lo-freq NPG, hi-freq NPG, At-Risk, and PG) as the between factor (see Fig. 3). There was no main effect of group F(3, 118) = .659, n.s., or condition by group interaction F(18, 708) = .530, n.s. There was a main effect of outcome F(7, 826) = 249.007, p < .001, η2 = .678. For winning outcomes, as the credit value increased, so too did participants’ PRPs (Least Significant Difference post-hocs indicated all winning means were significantly different from one another (all ps < .001 except for the wins of 25 and 50; p = .039). The different types of losses, (regular losses, classic near misses, and other near misses) all differed from all wins (all ps < .001), but did not differ from one another (smallest p value = .136). These relations are depicted in Fig. 3.
Fig. 3

Mean PRPs for losses, non-classic, and classic near-misses and wins of different magnitudes (error bars represent standard errors of the mean)

Of the 122 participants, 81 (66.4 %) showed shorter PRPs for classic near-misses compared to losses. A related-samples Wilcoxon signed rank test indicated that PRPs for classic near-misses were significantly shorter than for regular losses. By contrast only 61 participants (50 %) showed smaller PRPs for the non-classic near-misses compared to losses—the value one would expect if non-classic near misses were equivalent to losses.

Skin Conductance Responses

Figure 4 shows the timing of the measurement of all SCRs for the different outcomes, along with when participants initiated the next spin. As can be seen in Fig. 4, there are two key windows: the one-second delay (during which eccrine gland activity begins to change for significant outcomes), and the 2-s window during which the SCRs are actually manifested. At the bottom of Fig. 4, one can see that, on average, key presses to initiate the next spin following all the losing outcomes (classic, other near-misses, and regular losses) were all made during the one second delay period, whereas for all winning outcomes key presses were made during the SCR measurement window. Given this categorical difference between winning and losing outcomes in terms of when key presses were made to initiate the next spin, we chose to analyze SCRs for losing outcomes in one set of analyses, and winning outcomes in another set of analyses.
Fig. 4

Timing of the PRPs for wins and losses in relation to the SCR recording window

Average SCRs for wins of 5, 15, 25, 50, and 250 credits were analyzed using a mixed model ANOVA with outcome (5, 15, 25, 50, and 250 credit wins) as the repeated factor, and gambling group (Lo-freq NPG, Hi-freq NPG, At-Risk, and PG) as the between factor. There was no main effect of group F(3, 117) = .791, n.s., or condition by group interaction F(18, 702) = .190 n.s. There was a main effect of outcome F(4, 68) = 11.82, p < .001, η2 = .082. For winning outcomes, as the credit value increased, so too did participants’ SCRs (Least Significant Difference post-hocs indicated all winning means were significantly different from one another (all ps < .021) except for the wins of 5 and 15, and 25 and 50 which did not differ). Additionally, trend analysis showed a strong, significant linear trend in this data F(1, 117) = 18.701, p < .001. The linear relations between win size and SCRs are shown in Fig. 5.
Fig. 5

Mean skin conductance response amplitudes for wins of different magnitudes (error bars represent standard errors of the mean)

Losing outcomes were analyzed using a mixed model ANOVA with outcome (regular losses, classic near misses, and other near misses) as the repeated factor, and gambling group (Lo-freq NPG, Hi-freq NPG, At-Risk, and PG) as the between factor. There was no main effect of group F(3, 117) = 2.314, n.s., or condition by group interaction F(6, 234) = .290 n.s. There was a main effect of losing outcome F(2, 234) = 19.109, p < .001, η2 = .140. As shown in Fig. 6, classic near misses triggered significantly larger SCRs than both the regular losses, and the non-classic near-misses (Least Significant Difference post-hocs both p values <.001). Regular losses and non-classic SCRs were not significantly different from one another (p = .630). These relations are shown in Fig. 6.
Fig. 6

Mean skin conductance response amplitudes for losses, non-classic and classic near-misses (error bars represent standard errors of the mean)


Classic near-misses which afforded a build up of the anticipation of winning the jackpot over the course of the spin were associated with short PRPs but large SCRs—larger than all other outcomes except the largest win presented in the game. This combination of large SCRs but no discernable PRP supports the hypothesis that near-misses are interpreted as frustrating losses rather than misconstrued wins.

The current study indicates that not all near-misses are created equal. Near-misses are typically defined as any two jackpot symbols on the payline with a third just off the payline. Here we showed that the order in which the near misses land on the payline is crucial. If the first or second reel stops on a non-jackpot symbol, these outcomes trigger the same PRPs and psychophysiological arousal signatures as regular losses. By contrast, during classic near-misses when the first reel stops with a jackpot on the payline, and then second reel stops with another jackpot symbol on the payline, we propose that hopes of a big win accrue—hopes that are dashed by the third reel stopping with the jackpot symbol just off the payline. This hypothesis is supported by the fact that the majority of participants initiated the next spin more quickly for these classic near misses than for regular losses, and the SCRs were far greater for this type of near miss than for other near-misses.

In this study, there was a large discrepancy between the PRPs associated with losses and the PRPs associated with wins. Even for the smallest win, PRPs were well over half a second longer than for any type of loss. A number of features could serve to enhance the PRPs associated with wins. First, because the simulator was designed to emulate actual slot machines, on winning outcomes the symbols on the winning line repeatedly flashed on and off in attention-capturing fashion. Second, as in actual slot machines, winning outcomes were accompanied by a winning song—which stood in stark contrast to the silence that followed any type of loss. As such one might argue that near-misses are still hedonically rewarding, but in terms of PRPs, the reason they look so much more like losses than like wins is due to the marked difference between wins and losses in terms of the auditory and visual feedback that is present in wins, and completely absent in losses and near misses. If, however, near misses were even minimally hedonically pleasurable, then they should have had longer PRPs than regular losses. The non-parametric analysis shows that the classic near misses actually have significantly shorter PRPs than those associated with regular losses.

The current PRP analysis informs us not only about near-misses and losses, but also about wins. For wins, PRPs increased in a relatively linear fashion as a function of win size. Just like in the casino, winning spins were accompanied by reinforcing sights (flashing symbols) and sounds (celebratory songs). Since all winning songs were of exactly the same length, the linear relation between PRP length and win size indicates that players were not merely waiting until the song’s end prior to initiating the next spin (otherwise all PRP lengths would have been the same). Rather the length of the pause depended on the size of the win. Although participants could (and some fast responders did) initiate spins during the winning song, on average (as can be seen in Fig. 3), this was not the case. Even following the smallest win, the next spin was initiated after the song ended, and the pauses extended for longer and longer periods into this period of silence with the size of the pause depending on the size of the win.

These results extend the findings of Delfabbro and Winefield (1999) who analyzed PRPs in three bins (wins of 1–25, 26–50, and 50 plus) and showed increasing PRPs with win size. Here we show a more fine-grained titration between win size and PRP—players appear to be sensitive to the difference between win sizes of 5 versus 15 versus 25 credits (outcomes that were combined in that study).

It is of interest that in video slot machines found in casinos, the length of the winning song is actually yoked to the size of the win—the bigger the win, the longer (and more complex) the song. As such, modern slot machines may fill the PRP with celebratory stimuli—a situation that could heighten the reinforcing properties of the win.

The current data also extends the findings of Dixon and Schreiber (2004). In their study the majority of participants (8 out of 12) showed PRPs that were greater for winning versus all types of losing trials. In our study with 10 times as many participants we replicated this pattern of results, but showed that when one takes into account the size of the win these effects are even more robust. In our data set when wins of 5 credits were compared to regular losses, and classic or non-classic near-misses, 82 % of participants showed larger PRPs for the small win, than any type of loss. For wins of only 15 credits this value rose to 95 %, for wins of 25 credits, 98 % of participants showed a larger PRP for the win than any type of loss.

Dixon and Schreiber (2004) have conjectured that losing outcomes constitute a type of aversive event that participants wish to escape. They escape by initiating the next trial quickly. If classic misses are even more aversive than regular losses (as suggested by their elevated SCRs), then it is possible that participants would attempt to liberate themselves from the aversive state caused by the classic near miss even more quickly than for losses. Our data suggest that for the majority of participants (64 %) this is the case. When non-parametric statistics are used to assess the PRPs of classic near misses compared to regular losses, we found that the PRPs following near-misses are shorter than following regular losses. Such a finding supports the hypothesis that these types of near-misses are a particularly frustrating form of loss, and contradicts the supposition that they are a miscategorized win. Specifically, following these types of near-misses, participants may be driven to spin again as quickly as possible to remove themselves from a particularly frustrating state.

Classic near-misses generated larger SCRs than other losses, indicating that they were psychologically different than regular losses, or non-classic near misses. The fact that near-misses generated SCRs that were larger than all outcomes except wins of 250 credits, is also an indication of their frustrating properties. Recall that in the Clark et al. study, fMRI indicated that near-misses recruited the same neural reward circuitry as monetary wins, yet were rated as being unpleasant events. In the current study, credit wins were positively rewarding and indeed led to substantial PRPs—in general, the bigger the win the longer the pause. Credit wins also led to increased SCRs—in general, the bigger the win, the larger the SCR. Importantly, the same players who showed a titration between win size, SCRs and PRPs, showed shorter PRPs for classic near-misses than regular losses. This finding contradicts the notion that the visual similarities between Jackpot wins and near-misses cause near-misses to acquire win-like status via Pavlovian Generalization (Peters et al. 2010). If visual similarity led to the acquisition of win-like status, they should have had longer PRPs than losses.

Although this experiment fails to support the notion that near-misses are hedonically pleasurable, this does not suggest that near-misses are not reinforcing. Indeed many studies have shown that near-misses cause players to persist in slots play (Kassinove and Schare 2001; MacLin et al. 2007; Cote et al. 2003). If near-misses are frustrating losses, they may increase the propensity to keep playing by activating the appetitive component of the mesolimbic reward system and by perhaps creating a hopeful subjective impression that the next win is imminent. This may ultimately contribute to the sensitization of the appetitive system, which plays a key role in addictive behavior (Koob 2009), and our results are therefore consistent with an addiction model of compulsive gambling (Zack and Poulos 2009).

Copyright information

© Springer Science+Business Media, LLC 2012

Authors and Affiliations

  • Mike J. Dixon
    • 1
  • Vance MacLaren
    • 1
  • Michelle Jarick
    • 2
  • Jonathan A. Fugelsang
    • 1
  • Kevin A. Harrigan
    • 1
  1. 1.University of WaterlooWaterlooCanada
  2. 2.MacEwan UniversityEdmontonCanada