Introduction

Nearly all young people in the developed world now play video games, and this popularity has driven concerns about the possible negative effects of this recreational activity. Studies polling members of the general public as well as scientists suggest views concerning the effects of gaming on young people vary widely as a function of demographics and personal experiences with games. Some researchers conclude that gaming has social and cognitive benefits, whereas others argue the medium contributes to mass-shooting events, and for consistent and strong effects on aggressive Royal Online behaviour, more broadly . Recently, a series of open letters published by scholars have cautioned the public and policy-makers that both the positive and negative effects of time spent gaming, their addictive potential , cognitive benefits and aggressive effects may have been overstated.

Like individuals, policy and professional organizations have expressed varied positions regarding video game effects. In general, most organizations' initial guidance was framed by the precautionary principle—an approach to mitigating societal harm that puts protections in place when there is a plausible risk. Policy-makers guided by this mindset have discretion to take measures in cases where scientific knowledge about something new is lacking. In line with this principle, some organizations like the American Psychological Association err on the side of caution and warn to limit youngsters’ time spent playing video games. Such steps are far from universal as other organizations conducting their reviews of the science, such as the Australian and Swedish  government reports, and the APA's own Media Psychology and Technology Division , have concluded there is no actionable evidence that aggressive behaviour results from youth gaming. As more nuanced empirical understanding of media effects has emerged, other policy positions, once stridently aligned against gaming and screen time, such as the American Academy of Pediatrics, have softened their prescriptions concerning digital media and psychosocial development . These changes have been reflected in the statutory arena: in 2011, the United States Supreme Court  judged that there is insufficient evidence that games cause harm to uphold laws restricting the sale of violent games to minors. These changes in law and policy follow closely from a shifting empirical landscape.

There is a good reason to believe that violent video game engagement might be associated with human aggression, though this idea is a controversial one . To date, the main theoretical framework used to study the links between violent game engagement and aggression has been the general aggression model . Briefly, the GAM is an appetitive social learning theory that proposes that repeated exposure to violent media increases the accessibility of aggressive thoughts, which in turn increases the probability of aggressive cognitive schema, emotions and behaviour . Some reviews  and recent studies  informed by the GAM framework report consistent, though modest, support for the idea that violent gaming is linked to human aggression. This interpretation is not uniform; other analyses of the literature conducted by Sherry  and Ferguson  provide evidence the GAM framing, and the idea that games cause aggression more broadly, is incomplete, not evidenced or flawed. Indeed, motivation research indicates many factors key to understanding games are often overlooked by GAM researchers, such as the observation that aggressive individuals gravitate towards violent games , and that violent games might foment player aggression in experimental studies not because they prime aggressive cognitive schema, but rather that they frustrate the basic psychological need for competence .

One noteworthy attempt to bring a measure of harmony to the existing literature is that by Hilgard et al. , who re-analysed widely cited metanalytic data, drawing together results derived from GAM research  that form the basis of a number of past and existing policies regarding violent game effects . Their analysis detected the presence of publication bias not uncovered in the original reporting of the data. Upon adjusting for publication bias, the observed aggregate effect sizes relating gaming to aggression were smaller than those originally presented. Worryingly, this analysis also suggested that studies originally deemed to be following best practices showed particularly strong evidence of publication bias. That understood, the naive and corrected estimated effect sizes relating violent video game play to self-reported aggression extracted from this meta-analysis across a total of 37 studies (n = 29 113) were relatively consistent and small to medium (r = 0.21; 95% CI = 0.20–0.22) in size.

With this in mind, there is reason to think that outstanding methodological challenges might be inflating this metanalytic estimate. First, there is a noticeable degree of flexibility in how violent game play is operationalized in survey studies. For example, in multiple published studies of gaming effects drawn from the Effects of Digital Gaming on Children and Teenagers in Singapore project (EDGCTS;), violent video game engagement has been computed differently across presentations of findings from the dataset. In one case, the researchers  measured violent gaming by combining responses to three questions into a single variable, one about non-violent gaming (reverse scored): ‘How often do other players help each other in this game?’, and two questions that assessed violent gaming ‘How often do you shoot or kill other players in this game?’ and ‘How often do characters try to hurt each other's feelings in this game?’. Working with the same data, researchers [35] later selected four items to reflect game content, of which two were not part of the original analysis. These were: ‘How often do you shoot or kill creatures in this game?’ and ‘How often do you help others in this game?’. In place of computing one violent game content variable, the researchers created two separate variables for their analysis, one reflecting non-violent content and a second reflecting violent game content. Subsequent work by the same group [36], again using the same data, relied on a single violent gaming construct but used four items without identifying which of the available items reflecting game content were included. This flexibility, described as part of the ‘garden of forking paths problem’ increases the chance of false-positive results and serves to reduce our confidence in the inferences linking gaming to aggression [37,38].

Adding to problem of survey measurement flexibility is the fact that violent gaming effects research relies on self-reported data entirely provided by young people. Said differently, this work depends on children and young people accurately reporting on their video game play, the level of violent content present in this play and their own trait- or state-level aggression (for an exception, see ). This is problematic because studies of young people , health  and gaming  may be susceptible to the so-called mischievous responding—a phenomenon in which research participants exaggerate their responses by selecting extreme, and sometimes implausible, response options when providing self-report data. Mischievous responding can have the effect of introducing measurement noise that inflates relations that are logically incoherent or absurd to take at face value. Indeed, it is possible that some might respond to surveys in such a way that both their video game play and their experiences of intimate sexual behaviours are exaggerated. Such a pattern could lead researchers to make the spurious claim that playing the 2004 Xbox game Spider Man 2 is a significant catalyst for adolescent promiscuity.

Measurement flexibility also extends to a number of outcome assessments employed in the gaming literature. In experiments evaluating gaming aggression, methods for computing self-reported measures of aggressive emotions alternate between approaches that use all of the available scale items and those that use a subset of items thought by some researchers to be ‘most sensitive to an experimental manipulation of video game play’ . A similar tractability is present in behavioural measurements of aggression. The most widely used laboratory-based method for measuring aggression, the competitive reaction time task, has been used in more than 125 published papers, and surprisingly, task scores have been quantified in more than 155 different ways in this literature. In many cases, more than one computational approach is used to operationalize behavioural aggression in the same paper. For both self-reported and behavioural aggression measures, this flexibility affords otherwise well-meaning researchers the ability to select between different operationalizations of predictors and outcomes until they find a combination in line with their pre-existing biases or theories. Because this work is almost entirely exploratory in nature (i.e. not preregistered), it is difficult to know what to make of studies that report positive findings under these conditions of routinized methodological flexibility .

A handful of preregistered studies have rigorously tested the links between violent game play and human aggression , and do not detect an effect of brief exposure to violent gaming on aggression in the laboratory. Given these findings and the wider importance of conclusions drawn from this work, it is important to use preregistered study methodology to evaluate whether the existing literature may be under- or over-estimating the extent to which violent video game play relates to aggression. With this in mind, the present study examined the fundamental dynamic of concern in this subfield through a purely confirmatory lens following a registered reports protocol.

Our aim was to rigorously test the hypothesis that time spent playing violent video games is positively associated with adolescents' everyday behavioural aggression. The study examined the extent to which there are detectable positive linear and parabolic, ‘U’-shaped, relationships linking these factors. To this end, we analysed data collected from a large and representative cohort of British young people and their carers. Of interest was the significance, direction and effects sizes observed between video game engagement, operationalized as time spent playing violent video games, and aggressive behaviour, operationalized using carer reports of adolescents' aggression.