Between self-reinforcing behaviors and substances, there may exist a “drug interaction” which exacerbates addictive patterns. Addictive substances stimulate the brain’s natural reward pathways [1], and behavioral addictions involve the endogenous cannabinoid and related brain systems [2]. Dopamine plays a role in the self-reinforcing nature of substance use [3–5], and elevated dopamine plays a “gain-amplifying role” [6] in responses to “rewarding” Pavlovian stimuli [7]. As would be expected based on these shared biological mechanisms, survey studies find relationships between substance use problems and self-reinforcing behaviors like eating, gambling, internet use, sex, exercise, "workaholism," shopping, television viewing, and video games [8–13].

Such a “drug interaction” between self-reinforcing behaviors and addictive substances has been discovered in survey and laboratory findings that alcohol exacerbates problem gambling [14–18]. Part of the explanation offered for this is that alcohol contributes to disinhibition and risky decision making [19]. If this “drug interaction” effect is not, however, unique to the specific context of the alcohol/gambling combination but part of a larger pattern in which behavioral and substance addictions contribute to each other, then it should be evident in other substance/behavior combinations. It should also hold that engaging in self-reinforcing behavior while using or feeling the effects of a substance is uniquely associated with not only behavioral addiction, but substance use problems as well. In the present study, we focus on video game play. Video gaming is both similar to and distinct from gambling [20] in its self-reinforcing features [21] and degrees of prevalence and social acceptability, which all contribute to its unique characteristics as an addictive behavior [13].

The premise that video game playing is a potentially addictive behavior has empirical support [22]. Although the idea of video game addiction is not universally accepted [23], particularly as a discrete diagnostic category [24], and is not apparently likely to be included in the forthcoming DSM-V [25], problem video gaming behavior remains an area of scholarly and clinical [26,27] concern. A report by the American Medical Association’s Council on Science and Public Health called for more research on it [28]. A growing body of research on problem video gaming behaviors has emerged from within work on problem gambling [29] and adapted its measures [27], which originally came from Diagnostic and Statistical Manual (DSM) substance abuse/dependence criteria [30,31]. Prevalence estimates of significant problem video gaming behavior hover between 4.9% and 9% among video gamers internationally [11,32–35], with smaller numbers fitting stricter criteria for dependence [22,36].

Video game addiction has identifiable biological dimensions, which are similar to those for other behavioral addictions as well as substance dependence. An electroencephalography study found differences in processing of game-related cues between “casual” and “excessive” video game players suggestive of an addiction-like process among the “excessive” players [37]. Video game play stimulates dopamine [38] through the “reward” structures programmed into video games [39,40]. Dopaminergic medications indicated for ADHD and substance dependence help remediate problem video gaming behavior [41,42], and video games are associated with development of attention problems in children [43,44].

In addition to shared biological mechanisms, problem video gaming behavior and substance use have several risk correlates in common. Examples include mood disorders [33,45,46], impulsivity [33,47], attention deficit hyperactivity disorder symptoms [42,48–52], low social competence [53,54], low academic performance [55–58], and (in the case of violent video games) violence [59,60]. Although many of these studies on negative effects of problem video gaming behavior focus on children and youth [24], there are similar findings on adults [22], which implies that it would not be accurate to frame video game addiction as a youth problem behavior. Additional studies on adults would help distinguish problem video gaming behavior as an issue with addiction, rather than something parsimoniously explained by youth problem behavior theory [61].

Given shared biological mechanisms and risk correlates of problem video gaming behavior and substance use problems, as well as the general trend for behavioral and substance addictions to co-occur [13], this study investigated unique potential of playing video games while using or feeling the effects of a substance—referred to herein as “concurrent use”—to contribute to problem video gaming behavior and substance use problems. Its hypothesis was that concurrent use would explain variance in problem video gaming behavior and substance use problems that would not be explained by substance use frequency, video gaming as an enthusiastic hobby (video game play frequency, enjoyment, and consumer involvement), or demographics, including gender [30,34,62,63], race [64], age, and socioeconomic indicators. Analyses tested this hypothesis with respect to caffeine, tobacco, alcohol, and marijuana.

Table 1 provides descriptive statistics on video game and substance use variables, as well as differences by categorical demographic factors.

Concurrent use with all substances was prevalent. Males reported higher consumer involvement, but females report higher enjoyment, and males (within this video gamers only sample) exhibited only marginally higher PVP. Females had more frequent use of and problems with tobacco. Males were, however, more frequent concurrent users with caffeine and alcohol than females. Although Blacks, Asians, and Native Americans had higher PVP than whites, Blacks had higher enjoyment and consumer involvement but not frequency, Asians had higher frequency but not enjoyment or consumer involvement, and Native Americans had higher consumer involvement but not frequency or enjoyment. Whites, however, were the only group to exhibit clearly higher rates of concurrent use issues, and then only with caffeine. The only demographic factor consistently associated with risk for problem use patterns was non-working status, and this only held for video games, tobacco, and alcohol.

Table 2 describes bivariate correlations among continuous study variables. Frequency of game playing, enjoyment of average game, consumer involvement, and PVP were all correlated with each other. All substance use problems variables were correlated with each other except for marijuana with tobacco and PVP. The positive correlation between days played and age was not, according to follow-up analyses (not shown), because of meaningful curvilinearity—the slope of the positive relationship between age and days played was steeper for younger participants and still positive, although more shallow, for older participants. Age was, however, negatively correlated with other game playing variables, including PVP. Consistent with findings about employment in Table 1, income and education were negatively correlated with video game playing frequency, consumer involvement, PVP, tobacco use frequency, tobacco use problems, and alcohol use problems.

Figures 1–4 describe results of path analysis models for caffeine, tobacco, alcohol, and marijuana users (respectively). For all models, the path from concurrent use to substance use problems was significant, and the correlation between PVP and substance use problems also was significant. Concurrent use was not directly associated with PVP. Video gaming as an enthusiastic hobby was indirectly associated with substance use problems through two paths. The first path was via PVP, which was significant for all four substances. The second was via concurrent use, which was only significant for caffeine, tobacco, and alcohol. An effect via concurrent use with marijuana may have been hard to distinguish because of the low sample size and the high rate of concurrent use among marijuana users (see Table 1). The pattern of significant paths described here also held if a dosage variable (e.g., number of caffeinated drinks, cigarettes, or alcoholic drinks per day of alcohol use) was used instead of the substance use frequency variable.

Results confirmed that, in models accounting for shared variance between substance use problems and PVP [8,10] and controlling for frequency of substance use, video gaming as an enthusiastic hobby, and demographics, concurrent use (playing video games while using or feeling the effects of substances) was uniquely associated with substance use problems. They did not, however, confirm the hypothesized direct association between concurrent use and PVP. The same pattern of results occurred for all four substances studied. The lack of a significant direct association between concurrent use and PVP makes these results not wholly congruent with previous research findings that alcohol exacerbates problem gambling [14–19]. Rather, video gaming as an enthusiastic hobby emerged as a possible “third variable” associated with both concurrent use and PVP. Our models also distinguish the enthusiastic hobby factor as indirectly associated with substance use problems, through PVP (for all for substances) and concurrent use (for caffeine, alcohol, and tobacco). Although demographics are background variables in our models, our bivariate results for demographic variables echo earlier findings that socioeconomic stressors are just as relevant to problem video gaming behavior [64,81] as they are to substance addiction [82].

Factors contributing to this study’s validity are its nationally representative sample collected by a provider whose data are frequently used in research [83] and its use of sampling weights calculated by the provider to correct for biases and authentically represent the population under study [76]. The online nature of the sample and 59% response rate to the screening instrument probably cannot be argued to be limitations in and of themselves: Online data collection has demonstrated validity in this area [84] and, while new communications technology has diminished response rates to all survey methods (phone and postal mail included), point estimates remain stable across methods and at much lower response rates than ours [85]. Another strength of this study was its sample of adults, filling a need noted in previous work for more studies on adults [24] and ensuring that our findings cannot be parsimoniously explained by youth problem behavior theory [61].

In order to meet the 10-minute median length requirement for this survey, established measures for PVP [31] and substance use [65] had to be abridged. We ensured validity of our measures, in part, through use of structural equation modeling – had measures really been invalid, our models would not have had adequate overall fit. Further confidence in our substance use problems and problem video gaming behavior measures can be drawn from their significant bivariate correlations with each other, which are consistent with other research using these concepts [8,10,11]. Another limitation of our measures was that operationalization of our key construct of concurrent use depended on a single question: “During the past 30 days, have you played video games while using [substance in question] or feeling its effects?” Some participants may have interpreted it either too strictly (as an exclusive rather than inclusive “or”) or loosely (reporting on usual behavior rather than thinking specifically about the past 30 days). Although this variation in interpretation may have increased random error, it did not necessarily introduce bias. Further confidence in the validity of this measure can be drawn from the same pattern of associations involving it holding for all four substances under study. Other limitations of this study are those that it shares with every survey study, e.g., dependence on self-report data and participants’ recall of events 30 days or even a year ago. Finally, the scope of the implications of these findings is limited by their basis in cross-sectional data. Although our graphics include arrows indicating directions of effects, structural equation modeling in the context of a study like this mainly offers a heuristic for understanding associations among variables. Our results are not meant to support firm conclusions about causality or how these factors influence each other over time.

Even with these limitations, these findings contribute to a growing understanding that behavioral and substance addictions co-occur [8–13] and may contribute to each other through influencing or complementing each other’s actual use practices [14–18] and activating biologically mediated addictive processes [2–7]. Understanding concurrent engagement in self-reinforcing behavior and use of addictive substances may be an important consideration in addiction specificity [86], i.e., variability in the development of co-occurring addictive patterns among those engaged in the prerequisite behaviors. Clinical implications of these findings, to the extent that they can be drawn from a non-clinical population, are that clients who present with behavioral addictions or substance use problems should be screened for both and also assessed for concurrent use.