Few entertainment franchises have attracted as much cultural attention and public anxiety as Grand Theft Auto. With the November 2026 launch of Grand Theft Auto VI ending a 13-year wait, the video game series is once again drawing massive audiences and renewed scrutiny.
The franchise began in 1997 as a top-down, two-dimensional crime game, and its 2001 entry, Grand Theft Auto III, moved the action into a fully three-dimensional city that players could roam freely. Across the series, players steal cars, complete criminal missions, and attack bystanders, while later installments added mundane pastimes like bowling and getting haircuts.
When Grand Theft Auto V launched in 2013, it grossed $1 billion faster than any entertainment product in history. That cultural reach has long invited questions from lawmakers, journalists, and parents about whether stepping into the shoes of a virtual criminal changes how people think, feel, and behave once they put down the controller.
Over the past two decades, psychologists, neuroscientists, and criminologists have used the franchise to put that question to the test. They have measured players’ anger and blood pressure, scanned their brains for signs of dulled empathy, tracked teenagers for years, and compared game release dates against national crime records.
Don’t miss the next big psychology finding
Every morning, PsyPost sends a short briefing on the most interesting new peer-reviewed research in psychology and neuroscience.
The evidence has rarely pointed in a single direction. Some effects show up within minutes of play and others fail to appear after months, and a few studies have set off public disputes among the researchers themselves.
Short-term anger versus long-term aggression
Much of the early scientific debate surrounding Grand Theft Auto centered on the General Aggression Model, a psychological framework proposing that repeated exposure to violent media reinforces aggressive thoughts, hostile expectations, and angry feelings. According to this theory, short-term spikes in aggression during or right after play can gradually accumulate into lasting personality changes.
Short-term laboratory experiments have provided some evidence for immediate emotional shifts, particularly when gameplay feels highly immersive. In a 2016 study published in Psychology of Popular Media Culture, Robert B. Lull and Brad J. Bushman assigned 194 college students to play Grand Theft Auto IV for 15 minutes either violently by killing bystanders or nonviolently by going bowling inside the game.
Participants played on a small 2D monitor, a large 2D projector screen, or a large 3D projector screen while wearing 3D glasses. Students who played violently reported higher levels of anger afterward than those who bowled, and this angry response was strongest among violent players using the 3D display.
Statistical analysis indicated that this amplified effect operated through “presence,” a psychological state in which players feel physically located inside the virtual environment rather than aware of the screen in front of them. Stereoscopic 3D graphics, which present slightly different images to each eye to create an illusion of depth, increased feelings of presence and intensified anger, though only when the gameplay was violent.
Yet critics of short-term laboratory studies have long questioned whether momentary spikes in anger persist beyond a few minutes after a gaming session ends. Many researchers argue that immediate post-game shifts reflect “priming,” a temporary psychological effect in which recent exposure to violent words or images briefly makes related concepts easier for the mind to access.
To test whether repeated exposure produces lasting changes, Scott Kie Zin Teng and colleagues conducted one of the first longitudinal laboratory experiments using the franchise, published in 2011 in Cyberpsychology, Behavior, and Social Networking. The authors recruited 135 university students in Singapore who normally consumed less than two hours of violent media per week and randomly assigned half to play Grand Theft Auto IV in a campus lab for 12 hours spread across three weeks.
When participants completed questionnaires during a separate session after the three-week period, the researchers found no statistically significant differences in overall trait aggression or empathy between players and the no-game control group. Players did show a small, statistically significant increase in pro-violence attitudes, specifically endorsing harsher views regarding criminal justice and interpersonal violence.
Using data from the same three-week Singapore experiment, Yew Mun Gabriel Chong and colleagues examined whether Grand Theft Auto IV altered players’ views of social reality, publishing their findings in 2012 in the Journal of Social and Clinical Psychology. They tested cultivation theory, an idea originally developed for television research proposing that heavy exposure to a fictional world gradually shapes how people perceive the real world.
Compared to non-players, students who played Grand Theft Auto IV gave higher numerical estimates for the percentage of real-world deaths caused by car accidents while driving and by drug overdoses, both recurring themes in the game’s missions. However, the players did not estimate higher rates of serious crime or physical assault, nor did they report greater personal fear of becoming a crime victim.
Unexpectedly, on one of the five attitude measures, the effect ran in the opposite direction. Even though stealing vehicles takes seconds in Grand Theft Auto IV, students who played the game were less likely than control participants to believe that car theft is easy or common in real life. The authors suggested that stark differences between the fictional city and low-crime Singapore may have reinforced players’ real-world beliefs instead.
The most extensive experimental test of long-term play came from Simone Kühn and her team in Germany, who published a randomized controlled trial in 2019 in Molecular Psychiatry. The scientists recruited 90 healthy adult non-gamers, and 77 completed the two-month intervention. Those participants played Grand Theft Auto V (25 people), the nonviolent life simulation game The Sims 3 (24 people), or no game at all (28 people) for roughly 30 minutes a day.
Participants were evaluated before the intervention, at least one full day after the two-month period ended, and again two months later across 52 outcome variables drawn from questionnaires and computer tasks. These measures covered aggression, hostile thinking, empathy, impulsivity, sexist attitudes, mental health, and cognitive control, while the one-day delay separated lasting psychological change from brief priming.
Across 208 statistical comparisons, only eight tests reached conventional thresholds before adjustment, fewer than the roughly 10 false positives expected purely by chance, and just three of those eight ran in the direction of harm. None survived standard corrections for multiple testing, which are statistical adjustments that tighten the threshold for significance when dozens of tests are run at once to prevent random flukes from looking real.
The authors also applied Bayesian statistics, an analytical method that estimates how strongly the data support the absence of an effect rather than merely failing to detect one. Those calculations provided evidence favoring no detrimental impact of playing Grand Theft Auto V.
Does virtual mayhem numb human empathy?
Even if violent games do not turn adults into aggressive individuals, another widespread concern involves desensitization. In psychology, desensitization refers to the gradual weakening of emotional, cognitive, or physical reactions to distressing events after repeated exposure.
In a 2015 clinical review published in Child and Adolescent Psychiatric Clinics of North America, Jeanne Funk Brockmyer noted that national survey data from the Kaiser Family Foundation indicated more than half of American youth aged 8 to 18 had played a game from the Grand Theft Auto series. Reviewing questionnaire, behavioral, and physiological research, Brockmyer concluded that exposure to violent video games tends to increase the relative risk of desensitization to violence, which may in turn increase aggression and decrease prosocial behavior.
At the same time, Brockmyer emphasized that media exposure does not act in isolation. Protective factors, such as parental monitoring, conversations about the difference between fictional and real violence, and nonviolent problem-solving at home, can buffer children against potential risks.
Because much of the earlier brain imaging work relied on cross-sectional comparisons between longtime gamers and non-gamers, scientists could not tell whether violent games blunted empathy or whether less empathic individuals were simply drawn to violent games in the first place. To address this causal puzzle, researchers turned to longitudinal experiments using functional magnetic resonance imaging, or fMRI, a brain-scanning technique that tracks changes in blood flow to estimate neural activity.
In a 2018 study published in Neurosignals, Kühn and her colleagues reported brain imaging results from their two-month German trial comparing Grand Theft Auto V, The Sims 3, and a passive control group. Inside the scanner, participants viewed photographs and video-game-style drawings of hands interacting with everyday tools like knives or hammers, either safely or in painful accidents.
At baseline, seeing painful scenes reliably activated a well-established brain network linked to empathy for pain, including the anterior insula and the anterior cingulate cortex. After two months of daily Grand Theft Auto V play, and again two months later, the scientists found no evidence of reduced activity in these empathy-related brain regions compared to either control group.
A research team led by Lukas Leopold Lengersdorff and Claus Lamm at the University of Vienna sought to tighten experimental control even further in a 2023 study published in eLife. One limitation of Kühn’s earlier work was that Grand Theft Auto V allows players to explore peacefully, meaning participants playing at home might have avoided violent acts, and comparing it to The Sims 3 introduced differences beyond violence alone.
Lengersdorff and colleagues recruited 89 young adult men with no violent video game experience in the previous year and brought them into the laboratory for seven one-hour sessions over two weeks. Half played a modified version of Grand Theft Auto V where they earned points for killing as many characters as possible, averaging 2,845 virtual kills per participant.
The control group played the exact same game with all violence stripped out, earning points for taking an average of 3,055 photographs of characters. Before and at least 24 hours after the two-week period, participants underwent fMRI scans while observing a research actor (whom they believed was another volunteer) receive painful or non-painful electric shocks, and also viewed real and game-based violent images.
Using Bayesian analyses, the authors found substantial evidence for the absence of a desensitizing effect on behavioral empathy ratings and emotional discomfort toward violent pictures. They also found weak-to-moderate evidence for no change in empathy-related brain areas or the amygdala, a region tied to emotional processing.
In an accompanying 2024 commentary in eLife, Nicolas Roy and Michel-Pierre Coll noted that the Austrian experiment overcame many longstanding design weaknesses by isolating violent content inside the same game engine and supervising play. Lengersdorff’s team did acknowledge important limits, including low session-to-session reliability in the brain scan measurements and the possibility that children, vulnerable individuals, or habitual gamers who play far more than seven hours might respond differently.
A 2024 Canadian fMRI study by Mary B. Ritchie and colleagues in Social Cognitive and Affective Neuroscience examined both immediate gameplay and cumulative gaming history among 46 healthy adults. Participants played either violent Grand Theft Auto V missions involving shootouts and torture or a nonviolent modified version involving racing, flying, and skydiving for one hour before completing an emotional empathy task in the scanner.
Neither acute violent play nor recent months of habitual violent gaming showed a statistically significant effect on empathy ratings or brain activity in core empathy regions such as the amygdala, anterior insula, or inferior frontal gyrus, with Bayesian analyses again supporting the null hypothesis.
However, exploratory whole-brain scans pointed to an unexpected interaction between gameplay and “coldhearted” personality traits in parts of the motor and somatosensory cortices when viewing positive emotional scenes. That preliminary pattern hinted that individual personality differences could subtly shape how people process emotional cues after gaming.
In a 2022 companion analysis from the same Grand Theft Auto V experiment, published in Social Neuroscience, Shannon A. H. Compton and colleagues examined brain regions involved in mirroring others’ actions. Acute violent gameplay did not alter simulation-related activity between groups, although a heavier cumulative history of violent gaming was associated with lower activity in part of that network.
Morality, self-control, and helping others
Beyond aggression and empathy, psychologists have investigated whether navigating a virtual criminal underworld influences everyday moral judgment, honesty, and willingness to help strangers. Several studies have focused on moral disengagement, a psychological process described by Albert Bandura in which people mentally reframe unethical behavior so that their usual moral standards feel temporarily suspended.
In a 2012 cross-sectional survey of 385 Italian high school students published in the Journal of Adolescence, Alessandro Gabbiadini, Luca Andrighetto, and Chiara Volpato measured exposure specifically to Grand Theft Auto IV. Teens who had played the game reported slightly higher moral disengagement scores than non-players, even after accounting for age, gender, and overall gaming time.
Among players, how recently they had played predicted greater moral justification, diffusion of responsibility, and minimizing of harmful consequences. Because the data were collected at a single point in time, the authors noted that causality could not be established.
To test cause and effect, Gabbiadini teamed up with Paolo Riva, Andrighetto, Volpato, and Brad J. Bushman for a 2014 experiment in Social Psychological and Personality Science. The researchers randomly assigned 172 Italian high school students to play either Grand Theft Auto III or Grand Theft Auto: San Andreas, or one of two nonviolent games (Pinball 3D or Mini Golf 3D), for 35 minutes after a 10-minute practice period.
During the session, a bowl containing 100 grams of chocolate candies sat beside the computer, and students were reminded that eating large amounts of candy quickly is unhealthy, turning the bowl into a behavioral test of self-control. Afterward, students could cheat on a self-scored logic puzzle to grab extra raffle tickets for prizes like an iPad, and they completed a competitive reaction-time task in which they could blast an unseen opponent with unpleasant noise.
Adolescents assigned to play Grand Theft Auto ate more chocolates, took more unearned raffle tickets, and delivered stronger noise blasts than those who played pinball or mini golf. Notably, these effects depended heavily on pre-existing personality traits: the gap between violent and nonviolent players grew wider among students who scored high in moral disengagement, while students with very low moral disengagement scores showed little difference.
Another short-term experiment examined whether criminal gameplay alters automatic self-perceptions. In a 2015 study published in Simulation & Gaming, Julia Shaw, Kimberly Crosby, and Stephen Porter randomly assigned 136 Canadian undergraduates to four 20-minute activities.
Two groups watched either neutral news clips or crime-focused news clips, while the other two groups completed either noncriminal tasks (like shopping and bowling) or criminal tasks (like robbing a cash register and assaulting characters) inside Grand Theft Auto IV. Self-report questionnaires showed no statistically significant differences in conscious, explicit criminal thinking across the four groups.
However, participants also completed an Implicit Association Test, a timed computer sorting task designed to measure split-second automatic mental links between concepts like “me” and “crime” that people may be unaware of or reluctant to admit. Students who played Grand Theft Auto IV in a criminal manner became faster at pairing crime-related words with themselves than participants in all three other groups, including those who watched crime news.
Because watching criminal news did not produce the same shift, the authors suggested that actively performing virtual crimes may briefly prime criminal self-associations in ways passive viewing does not. Whether such shifts last beyond the laboratory remains untested.
Other experiments and surveys examining prosocial behavior and moral maturity have produced contrasting results. In a 2013 paper published in PLOS ONE, Morgan J. Tear and Mark Nielsen ran three laboratory experiments in Australia measuring whether students spontaneously helped an experimenter who “accidentally” spilled a handful of pens on the floor.
Across two experiments involving 128 undergraduates, playing 20 minutes of Grand Theft Auto IV did not produce a statistically significant drop in helping behavior compared to playing Portal 2, Call of Duty: Black Ops, or World of Zoo. Instead, social context was far more influential: 75 percent of participants helped pick up pens when the spill happened in the middle of the experiment, compared to just 31 percent when participants believed the session had already ended and were walking out the door.
Similarly, a 2019 cross-sectional survey of 166 English adolescents aged 11 to 18 by Sarah E. Hodge, Jacqui Taylor, and John McAlaney in Frontiers in Psychology examined moral reasoning development using short-answer questions about moral values. Playing a broader variety of video game genres was a statistically significant predictor of higher moral reasoning scores, while listing Grand Theft Auto as a favorite game showed only a small negative trend that was not statistically significant.
Attitudes toward women and a contentious scientific debate
No aspect of Grand Theft Auto has sparked fiercer criticism than its portrayal of women, who frequently appear as strippers or sex workers whom players can hire and subsequently harm. Researchers have repeatedly tested whether these depictions foster sexist beliefs or increase “rape myth acceptance,” a term psychologists use for false, stereotypical attitudes that excuse sexual violence or shift blame onto victims.
An early experiment by Victoria Simpson Beck, Stephanie Boys, Christopher Rose, and Eric Beck, published in 2012 in the Journal of Interpersonal Violence, tested 141 American college students using a Solomon four-group design. This four-group setup includes subsets of participants who skip the initial pretest survey so scientists can verify that answering questions beforehand did not tip off volunteers or alter their post-game scores.
Students in the treatment groups watched a researcher play Grand Theft Auto IV, during which the male protagonist visited a strip club, hired a sex worker, and then shot her to reclaim his cash, while control students watched a baseball video game. Across the full sample, watching the game produced no statistically significant overall effect on rape myth acceptance, and prior violent gaming habits showed no correlation with rape-supportive attitudes.
When the researchers looked separately by gender among pretested students, however, male participants showed a small, statistically significant rise in rape myth scores from pretest to posttest, whereas female participants showed no change. Because participants merely watched someone else play and were not tracked afterward, the authors noted that actual interactive gameplay and longer-term follow-ups were needed.
Four years later, a high-profile study by Gabbiadini and colleagues in PLOS ONE reignited the controversy. The team assigned 154 Italian high school students to play either a violent-sexist game (Grand Theft Auto: San Andreas or Vice City), a violent-only game (Half-Life 1 or 2), or a nonviolent game (Dream Pinball 3D or Q.U.B.E. 2) for 25 minutes before viewing photographs of a teenage girl who had been physically beaten by a boy.
Playing Grand Theft Auto had no direct, statistically significant effect on empathy toward the female victim. Instead, the authors tested a “moderated mediation” model, a statistical approach that checks whether a treatment influences an outcome indirectly through a middle step, and whether that chain applies only to specific subgroups.
Based on this model, Gabbiadini’s team reported a three-way interaction, meaning a statistical pattern in which the effect of one factor (the game played) depended simultaneously on two other factors (the player’s gender and how strongly they identified with their character). Among male students who strongly identified with their in-game character, playing Grand Theft Auto was associated with stronger endorsement of traditional masculine norms, which in turn correlated with lower empathy for the female victim.
Because PLOS ONE requires open data sharing, psychologists Christopher J. Ferguson and M. Brent Donnellan downloaded the dataset and published a detailed reanalysis in 2017 in the Journal of Youth and Adolescence. Their reanalysis confirmed the absence of any direct effect on empathy and uncovered a major procedural problem.
Despite the original paper stating that participants were randomly assigned, all 22 fifteen-year-olds in the study had ended up in the Grand Theft Auto condition. As a result, that group was substantially younger than the other two groups.
Ferguson and Donnellan also discovered that the dataset contained two additional measures of character identification that were not reported in the original paper. When they substituted those alternative identification scales, or tested other theoretically plausible statistical models, the reported three-way interaction was no longer statistically significant.
In their view, the original findings were fragile and pointed to analytic flexibility in the absence of preregistration, the practice of publicly locking in a study’s hypotheses, measures, and statistical plans before looking at the data. Without preregistration, critics worry researchers can unintentionally explore multiple statistical pathways until one crosses the threshold for significance.
In a 2017 response published in the Journal of Youth and Adolescence, Gabbiadini and his coauthors pushed back, arguing that Ferguson and Donnellan had built a “straw person” critique because the original study never claimed a direct effect on empathy. They also defended classifying Grand Theft Auto as sexist by pointing to European and American content rating descriptors and mandatory strip-club missions.
Gabbiadini’s team explained that high school scheduling forced them to assign classrooms rather than individuals. They also ran follow-up analyses showing that their three-way interaction still held when combining two identification subscales and using all 15 masculinity items in the dataset.
Ferguson and Donnellan replied in a 2017 rejoinder in the Journal of Youth and Adolescence, pointing out six shifting accounts of the randomization process across the original article, online journal comments, and emails. They noted that classroom-level grouping creates statistical dependencies that cannot be corrected without recording classroom IDs, and quoted university press releases to argue that public framing had implied direct harm.
Adding an unexpected twist to this literature, Victoria Beck and Chris Rose published a follow-up experiment in 2018 in the Journal of Interpersonal Violence designed to correct the flaws of earlier studies, including their own 2012 paper. Sixty male upper-level university students were individually randomized to play either Grand Theft Auto V or the football game Madden NFL 12 for up to an hour, with rape myth acceptance measured two weeks before play, immediately afterward, and two weeks later.
To ensure every man in the Grand Theft Auto V group actually witnessed sexual objectification and violence against women, each participant first watched a confederate (an undercover research assistant posing as another student) visit a strip club, hire a sex worker, and kill her. Afterward, the participant took his turn playing however he chose.
Not a single one of the 30 male participants copied the confederate’s violence against female characters. None solicited or harmed a sex worker, only one requested a virtual lap dance, and several described the confederate’s gameplay as “heartless.”
Even more surprisingly, rape myth acceptance did not increase in the Grand Theft Auto V group. Instead, it declined after gameplay and dropped further by the two-week follow-up, creating a statistically significant difference in which the Grand Theft Auto V group scored lower than the football control group. The authors suggested that witnessing blatant virtual mistreatment may have provoked sympathy for the female victim rather than victim blaming.
Racial stereotypes and viewer identity
Researchers have also begun exploring how racialized character portrayals in violent games interact with a viewer’s own racial identity. Content analyses have noted that video games often cast Black characters in stereotypical roles linked to crime or street violence, raising questions about whether pairing racial cues with virtual aggression activates social stereotypes.
In a study published in late 2025 in the International Journal of Psychology, Tailson Evangelista Mariano and colleagues in Brazil and Portugal conducted an online experiment with 140 adults (87 identifying as White and 53 as Black) who had minimal prior gaming experience. Participants watched a one-minute clip of violent Grand Theft Auto V gameplay showing either a Black or a White character carrying out identical assaults and weapon attacks.
Afterward, volunteers completed a race-based Implicit Association Test measuring automatic pro-White versus anti-Black bias, alongside a self-report scale measuring explicit biological racism. Overall, participants who watched the violent Black character displayed stronger implicit pro-White and anti-Black bias than those who watched the violent White character, with no overall difference on explicit racism.
The researchers then tested participant race as a moderator, which is a variable that alters the strength or direction of a relationship between two other variables. Those statistical interactions fell just short of conventional thresholds for statistical significance, meaning the subgroup differences should be interpreted cautiously.
Still, follow-up simple slope tests, which evaluate the effect of the video separately within each racial subgroup, indicated a diverging pattern. White viewers showed higher implicit bias after seeing the violent Black character, while Black viewers showed no shift in implicit bias and reported lower explicit biological racism.
The authors interpreted these preliminary patterns through evaluative conditioning (forming automatic associations when two stimuli appear together) and racial identity theory, suggesting that stereotyped violent imagery may reinforce automatic cultural associations among majority-group viewers while triggering protective pushback among ingroup viewers. Because participants passively watched a one-minute video clip rather than actively playing the game, the authors noted that future experiments using hands-on gameplay and longer timeframes are needed.
Real-world crime and what players actually do in the sandbox
One of the biggest questions facing policymakers is whether any psychological effects observed in university laboratories scale up to real-world violent crime. To test this, researchers have compared major video game sales and blockbuster releases against official police and court records.
In a 2015 study published in Psychology of Popular Media Culture, Patrick M. Markey, Charlotte N. Markey, and Juliana E. French conducted four time-series analyses using FBI statistics on homicides and aggravated assaults in the United States. Time-series modeling is a statistical technique that strips away long-term historical trends and seasonal cycles, such as violent crime rising in summer and video game sales spiking during winter holidays, so researchers can see whether unexpected spikes in gaming align with unexpected changes in crime.
The authors examined annual video game sales from 1978 to 2011, monthly sales from 2007 to 2011, and Google searches for violent game walkthrough guides from 2004 to 2011. They also tracked the North American release dates of Grand Theft Auto: San Andreas, Grand Theft Auto IV, and Call of Duty: Black Ops. Across all four analyses, they found no evidence of a positive link between violent video games and real-world violent crime.
Instead, several analyses pointed in the opposite direction. Higher monthly game sales aligned with concurrent drops in aggravated assaults, while spikes in walkthrough searches aligned with drops in assaults and homicides two months later. In addition, homicides dipped three and four months after those blockbuster releases.
Markey’s team raised catharsis, the idea that players vent aggression virtually, as one possible explanation while noting that prior research offers little support for it. They favored what they called a simpler explanation: individuals predisposed to aggression are drawn to violent games and spend time playing them at home, removing them from streets or social venues where they might otherwise commit violent acts.
A 2017 study in the Netherlands by Marinus G. C. J. Beerthuizen, Gijs Weijters, and André M. van der Laan, published in the European Journal of Criminology, tested whether a crime dip could be detected at the level of a single game launch. The authors analyzed daily Dutch court records covering more than 419,000 registered offenses by young men aged 12 to 25 from the start of 2012 through the end of 2014.
They then modeled the surge and gradual decline of active players following the September 2013 release of Grand Theft Auto V. Even after controlling for days of the week, holidays, seasons, and reporting lags, the release of Grand Theft Auto V was tied to a small drop in registered offenses among both teenage boys and young adult men. The association was statistically significant, though it accounted for less than one percent of daily variance.
When the researchers checked other major releases, an independent negative association also appeared for Call of Duty: Ghosts, though not independently for Call of Duty: Black Ops II. They concluded the pattern held for at least one other popular game, and perhaps two.
Beerthuizen and colleagues framed their findings around “voluntary incapacitation,” an idea rooted in routine activity theory in criminology. According to this framework, street crime requires motivated offenders and potential targets to cross paths in physical space without supervision, so an absorbing video game played at home can temporarily keep both potential offenders and potential victims off the streets.
Qualitative criminological research also challenges the assumption that players automatically embrace chaos when handed a lawless virtual city. In a 2018 study in the British Journal of Criminology, Jack Denham and Matthew Spokes tested a cultural criminology concept known as the “murder box,” which argues that open-world games serve as lawless arenas where people unleash repressed sadistic desires.
The researchers observed 15 British young adults playing Grand Theft Auto V as they normally would and interviewed them during and after play. All 15 participants engaged in violence at some point during the session, with one player enthusiastically going on a shooting spree and others stumbling into shootouts because of tricky driving controls or an unruly in-game dog attacking police.
Even so, player behavior varied widely and frequently included everyday, nonviolent activities. Several participants chose to follow the rules of the road, get haircuts, customize cars, or go shopping, and one participant worked as a virtual taxi driver to earn money for a trip to the funfair.
Moral restraint also surfaced when game missions pushed players toward violence. One player who was forced by a mission to punch a rival gang leader felt genuinely distressed when the character accidentally drowned in a river.
Another participant asked to opt out of a shooting mission altogether because a real-world school shooting had occurred the day before. Together, these observations suggest that players carry their internalized everyday social restraints into the virtual world rather than abandoning them.
Risk-taking, driving, and substance use
While links between Grand Theft Auto and real-world violent crime appear absent or slightly negative, a separate branch of health psychology research has explored whether risk-glorifying games correlate with reckless driving, substance use, and other adolescent risk behaviors.
In a 2006 laboratory experiment published in the Archives of Pediatrics & Adolescent Medicine, Sonya S. Brady and Karen A. Matthews randomly assigned 100 male undergraduates aged 18 to 21 to play 10 minutes of Grand Theft Auto III or the lower-violence driving game The Simpsons: Hit and Run. Men who played Grand Theft Auto III showed larger blood pressure increases during play and less cooperative behavior in a subsequent point-sharing task.
They also reported more permissive attitudes toward alcohol and marijuana use, though not toward violence or unprotected sex. Although the Grand Theft Auto III mission featured a drug dealer, the mission itself took a punitive stance against drugs, since players hit the dealer with a baseball bat. As a result, the authors suggested that other mechanisms, such as violent content broadly condoning risk-taking and rule-breaking, may have shifted substance attitudes.
In addition, systolic blood pressure rose sharply during Grand Theft Auto III play only among young men who reported high lifetime exposure to real-world violence in their homes and communities. That interaction indicated that personal background can shape physiological reactions to violent media.
Moving outside the laboratory, Jay G. Hull, Ana M. Draghici, and James D. Sargent conducted a four-year longitudinal telephone survey of American adolescents, starting with 5,019 teens at an average age of 13.8 and tracking 2,718 through an average age of 17.9. Publishing their driving findings in 2012 in Psychology of Popular Media Culture, they measured whether teens played mature-rated, risk-glorifying games, combining parental permission for mature games with play of Grand Theft Auto III, Manhunt, and Spider-Man 2.
After controlling for age, gender, race, family income, parental education, and parenting style, playing these mature risk-glorifying games at baseline predicted higher self-reported rates of speeding, tailgating, weaving through traffic, being pulled over by police, automobile accidents, and willingness to drink and drive years later. Statistical modeling indicated these relationships were mediated by increases in sensation seeking and rebelliousness.
In a companion 2014 paper using the same national dataset in the Journal of Personality and Social Psychology, Hull, Timothy J. Brunelle, Anna T. Prescott, and Sargent expanded the analysis to alcohol consumption, cigarette smoking, physical fighting, delinquency, and risky sexual behavior. Among adolescents who played video games, higher play of mature risk-glorifying titles was associated with steeper increases in drinking and smoking over time as well as higher rates of aggression, delinquency, and risky sex.
These relationships were again partially linked through sensation seeking, rebelliousness, favorable attitudes toward deviance, and friendships with peers who drank or smoked. Because video games rarely reward drinking or smoking directly, Hull’s team proposed an “identity simulation” hypothesis: role-playing anti-social, rule-breaking protagonists may encourage teens to adopt a broader rebellious self-image.
Supporting the idea that character values matter, playing Grand Theft Auto III showed stronger statistical correlations with deviant behaviors than playing Spider-Man 2, an action game featuring a law-abiding superhero. Yet two nuances in Hull’s data stand out.
First, simply playing video games in general was not linked to deviant behavior; in fact, teens who played non-mature games reported lower rates of drinking, smoking, and sexual risk-taking than teens who did not play video games at all. Second, the authors emphasized that longitudinal surveys cannot establish causation and cannot rule out what they called the “bad-seed” hypothesis, the possibility that youths already heading toward risk-taking behavior are simply drawn to mature, rule-breaking games early on.
A similar caution applies to a 2016 British study by Jo Cranwell, Kathy Whittamore, John Britton, and Jo Leonardi-Bee in Cyberpsychology, Behavior, and Social Networking. Surveying 1,094 British adolescents aged 11 to 17, the team found that teens who played at least one of 17 bestselling games containing tobacco or alcohol imagery had more than twice the odds of reporting that they had tried smoking or drinking alcohol.
When the researchers coded 217 five-minute intervals of storyline cut scenes from the five most popular titles, alcohol and tobacco depictions appeared most frequently in Grand Theft Auto V and Grand Theft Auto IV, complete with fictitious beer and cigarette brands. Yet none of the 17 games carried alcohol or tobacco warnings under Europe’s official Pan European Game Information rating system. The authors called for updated content descriptors while acknowledging their cross-sectional survey could not establish whether the games caused teens to experiment with substances.
Methodological debates and the replication crisis
Why do studies of Grand Theft Auto and similar games yield such sharply different conclusions, ranging from laboratory shifts in cheating or implicit thinking to zero effects in two-month brain imaging trials? Reviewing the broader violent video game debate in 2010 in the Review of General Psychology and again in 2020 in Current Opinion in Psychology, Christopher J. Ferguson placed these contradictions inside psychology’s wider replication crisis.
Ferguson argued that public fears about new media, from ancient Greek theater and early cinema to Grand Theft Auto, repeatedly follow cycles of “moral panic” in which societal anxieties outpace evidence. Within academia, he pointed to publication bias, the tendency for journals to publish statistically significant results while leaving null findings unpublished.
He also highlighted unstandardized laboratory aggression tasks and poorly matched control games, such as comparing a complex action title to pinball or puzzles. In survey research, he noted that failing to control for third variables like gender, family violence, or pre-existing personality traits can inflate apparent links.
As Ferguson noted in his 2020 review, when researchers began adopting preregistration, experimental and longitudinal studies of violent and sexualized games largely began returning null results. At the same time, Ferguson pointed out in 2010 that an exclusive focus on harm often overlooks potential positive applications of fast-paced action games, such as modest improvements in visuospatial attention, online social connection, and educational engagement.
Repurposing Los Santos for road safety
Outside psychology, engineers have started harnessing the realistic open world of Grand Theft Auto V for practical safety research. In a 2023 study published in Sensors, Julius Schöning, Jan Kettler, Milena I. Jäger, and Artur Gunia turned the game’s bustling virtual traffic system into a low-cost cycling simulator.
By mounting steering, wheel-speed, and braking sensors onto a real stationary bicycle and wiring them to a microcontroller acting as a game controller, the team allowed riders to pedal safely through Los Santos while virtual trucks passed close behind them.
Using custom modification scripts to track nearby vehicles, the engineers piloted two early-stage cognitive enhancement safety systems: vibrating handlebar alerts and a smartphone visual distance display designed to warn cyclists of approaching trucks without putting human volunteers in real-world traffic danger.
What two decades of research suggest
Taken as a whole, two decades of research on Grand Theft Auto illustrate how difficult it is to reduce complex interactive media to a simple story of harm or harmlessness. Controlled experiments indicate that brief violent gameplay can temporarily elevate anger, blood pressure, or automatic crime-related associations, while two large adolescent survey projects have observed that teenagers who gravitate toward mature, rule-breaking titles also report more risk-taking behaviors outside the screen.
Yet scientists have also tracked participants over weeks or months of play in randomized controlled trials, scanned their brains for signs of emotional numbing, observed their behavior inside the virtual sandbox, and examined national violent crime records following blockbuster releases. In those studies, the evidence points away from lasting aggressive transformations.
Because the longest randomized trials tested adults rather than teenagers, open questions remain about younger or more vulnerable players. Even so, rather than acting as a behavioral blueprint that overrides a player’s character, Grand Theft Auto appears to function as an intense, customizable virtual playground whose psychological impact depends heavily on who is holding the controller and the real-world social context around them.