a Unless otherwise indicated, data are expressed as number (percentage) of participants. Percentages have been rounded and may not sum 100.
b Scores range from 0 to 2, with higher scores indicating more frequent aggressive behavior.
c Scored by multiplying frequency of engagement with media score (range, 1-3, with higher scores indicting greater frequency) and media rating score (range, 1-12.8, with higher scores indicating more violent content).
d Scores range from 0 to 60, with higher scores indicating stricter attitudes toward guns.
e Scores range from 0 to 4, with higher scores indicating more interest.
a Scores range from 0 to 4, with higher scores indicating greater agreement.
b Calculated using 2-tailed, unpaired, or independent t tests.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Dillon KP, Bushman BJ. Effects of Exposure to Gun Violence in Movies on Children’s Interest in Real Guns. JAMA Pediatr. Published online September 25, 2017171(11):1057–1062. doi:10.1001/jamapediatrics.2017.2229
What are the immediate effects of exposure to movie characters with guns on children’s unsupervised play with guns?
This randomized experiment included 104 children aged 8 to 12 years who were tested in pairs. Children who viewed a PG-rated movie containing guns played with a real gun longer and pulled the trigger more times than did children who viewed the same movie not containing guns.
The connection shown in this experiment is a compelling start to a broader conversation on the various factors that can increase a child’s interest in guns and violence, including gun violence in movies.
More US children die by accidental gun use than children in other developed countries. One factor that can influence children’s interest in guns is exposure to media containing guns.
To test whether children who see a movie containing guns will handle a real gun longer and will pull the trigger more times than children who see the same movie not containing guns.
Design, Setting, and Participants
One hundred four children aged 8 to 12 years recruited through advertisements were randomly assigned in pairs to watch a 20-minute PG-rated movie containing or not containing guns in a university laboratory. Children then played with toys and games in a room for 20 minutes while being video recorded. A cabinet in the room contained a real (disabled) gun with a sensor counting trigger pulls. Recordings were coded for the time spent holding the gun and in aggressive play. Data were collected from July 15, 2015, through January 1, 2016, and analyzed using generalized estimating equations (Tweedie log-link for time spent holding the gun; Poisson log-link for pulling the trigger).
Main Outcomes and Measures
The 2 main outcomes were time spent holding the gun and the number of trigger pulls. Control variables included sex, age, trait aggressiveness, exposure to violent media, interest in guns, and number of guns at home.
Among the 104 study participants (62 boys [59.6%] and 42 girls [40.4%]; mean (SD) age, 9.9 [1.5] years), the adjusted median number of trigger pulls among children who saw the movie containing guns was 2.8 (interquartile range [IQR], 0.2-2.8) compared with 0.01 (IQR, 0.01-0.2) among children who saw the movie not containing guns (adjusted odds ratio, 22.3; 95% CI, 6.0-83.4; P < .001). The adjusted median number of seconds spent holding the gun among children who saw a movie containing guns was 53.1 (IQR, 35.5-53.1) compared with 11.1 (IQR, 10.7-16.7) among children who saw the movie not containing guns (adjusted odds ratio, 3.0; 95% CI, 0.9-9.9; P = .07). Qualitative analyses on 4 pairs from each condition found that children who saw the movie containing guns also played more aggressively and sometimes fired the gun at people (ie, self, partner, or passersby on street).
Conclusions and Relevance
Children in the United States frequently have access to unsecured firearms and frequently consume media containing guns. This experiment shows that children who see movie characters use guns are more likely to use guns themselves.
clinicaltrials.gov Identifier NCT03220412
Nearly 60% of US households with guns do not secure them.1 If children find these guns, the consequences can be deadly. Most unintentional gun shootings happen at home, typically as a result of children playing with a loaded, unlocked gun.2 Children in the United States are 10 times more likely to die by unintentional gun shootings than children from other developed countries.3
Many factors can influence children’s interest in guns. Previous research has shown that children exposed to movie characters who smoke are more likely to smoke4 and that children exposed to movie characters who drink alcohol are more likely to drink alcohol.5 This experiment focuses on exposure to movie characters with guns. We hypothesized that children exposed to movie characters who use guns will be more likely to use guns.
Gun violence in movies is increasing, especially in movies that target younger viewers. An analysis of top-selling films found that the depiction of guns in violent scenes in films rated PG-13 (parental guidance suggested for viewers younger than 13 years) has more than doubled since 1985 when the rating was introduced, increasing from the rating levels of G (general audience) and PG (parental guidance suggested) films to the rating level of R films (for viewers aged ≥17) by 2005.6 A follow-up study found that the amount of gun violence in PG-13 films continued to increase through 2015.7
This experiment tests the hypothesis that children who see movie characters use guns are more likely to use a real gun. Specifically, we hypothesized that children aged 8 to 12 years who see movie scenes containing guns will hold a real gun longer and will pull the trigger more times compared with children who see the same movie scenes not containing guns.
Participants included 104 children in 52 pairs aged 8 to 12 years (Table 1). To be included in the experiment, the participant was required to bring another child aged 8 to 12 years (ie, sibling, step-sibling, cousin, or friend). Participants were recruited using advertisements in university employee newsletters, community websites, public spaces, and through ResearchMatch.org, a national registry for research studies.8 Each participant was compensated $25 for the hour that they spent in the laboratory. All methods, measures, and materials were approved by the institutional review board of The Ohio State University, Columbus. Parents were made aware of the true purpose of the experiment before giving written informed consent (rate, 100%); children also provided assent (rate, 100%). We did not prospectively register this randomized trial because there was no institutional requirement to register it as a randomized clinical trial prior to publication, but we have now done so at the request of the journal. The trial protocol is available in the Supplement.
Study data were collected from July 15, 2015, through January 1, 2016. Participants were told: “This study is about what kids like to do in their spare time, such as watching movies, and playing with toys and games.” Before viewing the movie, participants completed a 9-item measure of how often they engaged in aggressive behaviors such as yelling, arguing, kicking, and hitting (0 indicates never; 1, sometimes; and 2, often; Cronbach α = .76).9,10 They also completed a 15-item measure of their attitudes toward guns (eg, “Carrying a gun makes people feel safe,” “I wish everyone would get rid of their guns [reverse scored],” where 0 indicates strongly disagree and 4, strongly agree; Cronbach α = .80).11,12 Next, participants listed their 3 favorite movies, television shows, and video games and indicated how often they watched or played them each week (1 indicates only once in a while; 2, a lot but not always; and 3, many times). Ratings on each media were standardized using the Motion Picture Association of America film ratings system13 for movies, Common Sense Media14 for television shows, and Entertainment Software Rating Board15 for video games, in which lower numbers indicated more permissive ratings (eg, 1 for video games rated E for everyone) and higher numbers indicated more violent content (eg, 4 for video games rated M for mature players aged ≥17 years). We multiplied the frequency of engagement with media rating and then calculated the mean values for the 3 platforms (Cronbach α = .71).
Next, each pair was randomly assigned (using a random number generator before arrival) to watch a 20-minute edited version of a PG-rated film (The Rocketeer  or National Treasure ) containing or not containing guns on a 99-cm (39-in) screen. The gun version contained portions of the film as they were distributed. Scenes showing guns were edited out for the no-gun version, but the action and narrative of the film was not altered. Two different movies were used to increase the generalizability of findings. After watching the film clip, participants indicated whether they had seen it before (no or yes) and rated how exciting, boring, fun, and violent it was; how much they liked it; whether they felt like they were part of the action; and whether they wanted to see the rest of the movie (0 indicates not at all; 4, very much). While the children watched the movie, parents completed a form with demographic information (eg, race, sex, and age), their child’s presumed interest in guns (0 indicates not at all interested; 4, very interested), and how many firearms were in their home.
Next, participants went to a different room with a cabinet containing toys (eg, Lego bricks [The Lego Group], Nerf guns [Hasbro]) and games (eg, checkers). They were told they could play with any of the toys and games in the room. One of the cabinet drawers contained a real semiautomatic 9-mm handgun that was modified so that it could not fire. The gun’s hammer and trigger were still functional to allow an individual to pull the trigger, but no other components of the gun were functional. The magazine held the infrared wiring to count the number of times the trigger was pulled with sufficient force to discharge the gun. The gun looked and felt as it would normally. Parents were informed about the gun before giving their consent. The campus police chief examined and gave written approval for the modified gun to be used.
Participants were given 20 minutes to play in the room together with the door closed. Their parents and the experimenter secretly watched the children from a separate room via a hidden camera. A research assistant was always in the exterior greeting room in case the participants had questions. If participants opened the laboratory door and told the research assistant about the gun or gave it to the research assistant, the gun was removed and the children were permitted to continue playing. After the play period ended, participants were thoroughly debriefed and given an opportunity to ask questions. Information for counseling resources and gun safety training were made available to all participants and parents.
A generalized estimating equation (GEE) with an exchangeable correlation structure was used to contrast between movie conditions, with pairs as units. A log-link model with Poisson distribution was used to analyze the number of trigger pulls. A log-link model with Tweedie distribution was used to analyze time spent handling the gun (in seconds). The main independent variable in each model was movie condition (guns vs no guns). The first set of models included movie condition (guns vs no guns), movie (National Treasure vs The Rocketeer), and the participant’s sex (male vs female). A second set of models included movie condition, movie, sex, age, number of guns in the home, attitudes toward guns, trait aggressiveness, consumption of violent media, and parental estimation of their child’s interest in the gun. All analyses were performed using SPSS software (version 22; SPSS, Inc). P < .05 indicated significance.
Two trained research assistants blinded to movie condition and hypotheses independently watched and transcribed videos of each play session and recorded type of play (ie, aggressive, gentle, independent, together, or taking turns), type of verbal communication (ie, swearing, complimenting), any interpersonal aggression between participations (ie, punching, wrestling, or shooting a Nerf gun at partner), and pointing the real gun at self, partner, or others. One of us (K.P.D.) conducted random quality assurance checks to confirm codings. From these recordings, 4 videotaped sessions from each condition were randomly selected and interpreted using grounded theory.16
Among the 104 participants (52 pairs), mean (SD) age was 9.9 (1.5) years. Sixty-two participants were white (62 [59.6%]) compared with 42 nonwhite participants (40.4%). Sixty-two were boys (59.6%), and 42 were girls (40.4%).
Children randomly assigned to watch a movie clip containing guns did not significantly differ from children randomly assigned to watch a movie clip not containing guns in any covariates (Table 1). The sex difference was significant (38 of 54 paticipants [70.4%] who watched movie clips containing guns were boys compared with 24 of 50 [48.0%] who watched movie clips not containing guns), but we randomized by pair rather than by sex. Sex was included as a covariate in the analyses.
As expected, participants rated the movie clips containing guns to be more violent than the movie clips not containing guns (mean [SD], 2.76 [1.40] vs 1.69 [1.16]; P < .001). However, the 2 types of movie clips did not differ on the other rating dimensions (Table 2).
Of the 52 pairs of participants, 43 pairs (82.7%) found the gun in the cabinet drawer. Fourteen pairs (26.9%) gave the gun to the research assistant or told him or her about it. In total, 22 pairs (42.3%) had 1 or both participants handle the gun. The type of movie clip (ie, containing or not containing guns) did not influence whether participants found the gun (40 of 54 [74.1%] vs 35 of 50 [70.0%]; P = .40) or handled it (17 of 54 [31.5%] vs 15 of 50 [30.0%]; P = .52).
As hypothesized, on average, participants who saw movies containing guns pulled the trigger more times than those who saw the movie not containing guns. The adjusted median number of trigger pulls among participants who saw a movie containing guns was 2.8 (interquartile range [IQR], 0.2-2.8) compared with 0.01 (IQR, 0.01-0.2) among participants who saw the movie not containing guns. In the GEE model using only movie condition, movie, and sex as in the model, the adjusted odds ratio (AOR) was 17.1 (95% CI, 4.2-70.7; P < .001). In the GEE model using all covariates, the AOR was 22.3 (95% CI, 6.0-83.4; P < .001).
As hypothesized, children who saw movies containing guns held the gun longer when at play than children who saw movies not containing guns. The adjusted median number of seconds spent with the gun among participants who saw a movie containing guns was 53.1 (IQR, 35.5-53.1) compared with 11.1 (IQR, 10.7-16.7) who saw the movie not containing guns. This difference was significant in the first model only. In the GEE model using only movie condition, movie, and sex as risk factors, the AOR was 3.2 (95% CI, 1.1-9.3; P = .03). In the GEE model using all covariates, the AOR was 3.0 (95% CI, 0.9-9.9; P = .07).
When we kept condition and all other covariates equal, sex affected trigger pulls but not the time spent with the gun. Being male was associated with an increased number of trigger pulls compared with being female (AOR, 19.8; 95% CI, 2.5-155.6; P = .005). Sex was not associated with any differences in the number of seconds spent handling the gun (AOR, 2.0; 95% CI, 0.6-6.6; P = .25). When we kept condition and all other covariates equal, attitudes toward guns were found to influence trigger pulls but not time spent with the gun. More positive attitudes toward guns, as indicated by higher mean scores on the Attitudes Toward Guns and Violence Questionnaire,11,12 was associated with an increased number of trigger pulls (AOR, 3.6; 95% CI, 1.7-7.3; P = .001). Attitude toward guns was not associated with time spent with the gun (AOR, 2.2; 95% CI, 0.8-6.3; P = .15). None of the other covariates significantly influenced trigger pulls (P > .67) or seconds spent handling the gun (P > .13).
To provide a robust description of the data, qualitative analyses were conducted on a random selection of 4 pairs who saw a movie containing guns and 4 pairs who saw a movie not containing guns. All of the randomly chosen pairs (pairs 49, 20, 1, and 52) found the gun during the play session. Pair 49 found the gun and immediately opened the door and gave it to the research assistant. Afterward, the pair vocalized regret, wishing they had kept the real gun to play with. They proceeded to play violently, pretended their Nerf guns were M14s (semiautomatic rifles), and talked about killing zombies and “blowing heads off.” Pair 20 found the gun but did not touch it. This pair opened the door and asked for permission to play with the Nerf guns but did not mention the real gun. They played with the Nerf guns for nearly the entire session.
The 2 remaining pairs chose to play with the gun. In pair 52, 1 child held the gun for 18 minutes of the 20-minute play period and pulled the trigger 26 times. He also pulled the trigger at the temple of his partner’s head. For 1.5 of the remaining minutes, he held the Nerf gun and shot 12 Nerf bullets at his partner. His partner appeared to be disinterested in the real gun and the Nerf gun, playing along with whatever the other child instructed. In pair 1, 1 boy pulled the trigger 35 times, pointed the real gun out the laboratory window at people in the street, threatened to hit his friend with the gun, and attempted to steal toys and games from the play room. This pair also used aggressive language (eg, “I told you don’t mess with me b—!” and “Are you dumb as f—?!”) The other boy held the gun for 30 seconds and pulled the trigger 4 times. For 18 of the other 18.5 minutes, he held the Nerf gun and shot it 66 times at his partner.
Four randomly selected pairs in the no-gun condition (6, 24, 42, and 48) also underwent analysis. All but pair 48 found the gun during the play session. Both children in pair 6 found the real gun, but neither handled it nor told an adult about it. They instead played with checkers and Legos for most of the play session. Pair 24 found the gun shortly after the play session began and later opened the drawer 4 more times to look at it. Neither child touched the gun during the play session until one handed it to the experimenter at the end the session. When playing with the Nerf guns, this pair mainly fired at a pretend target on the wall rather than at each other. When they did fire the Nerf gun at each other, they avoided headshots and gave each other time to reload. Pair 42 found the gun almost immediately, and 1 child stated, “uh-uh, uh-uh, no, no, no” before taking the gun and opening the door to give it to the research assistant. They played with the Nerf gun for nearly 14 minutes but mainly shot at a pretend target on the wall. Similar to pair 24, when they fired the Nerf gun at each other, they avoided headshots and gave each other time to reload.
Previous research has shown that children who see movie characters smoke are more likely to smoke5 and children who see movie characters drink alcohol are more likely to drink alcohol.4 The present research extends previous research by showing that children who see movie characters use guns are more likely to use guns. The children in this experiment were randomly assigned to watch a 20-minute movie clip containing or not containing guns and then were given the chance to play with the same toys in the same room, which also contained a real gun hidden in a drawer. Viewing a movie containing guns was associated with higher rates of trigger pulls, longer time handling the gun, and playing more aggressively compared with viewing a movie not containing guns. The results from this experiment suggest that exposure to gun violence in movies increases interest in guns in the real world. Of note, the movies that we showed children were age appropriate (rated PG) and not very graphic in terms of gun violence. The effects might be greater with newer films containing more graphic gun violence.
Like all studies, this study has limitations. Only 1 modified handgun was available for the participants to find and with which they could engage, but 2 Nerf guns were available. This may have inadvertently invited coplay with Nerf guns but not with the real handgun. Although the hidden camera could record the entire room, it was stationary and unable to capture all actions of all participants. Finally, most participants were from suburban or urban neighborhoods, possibly limiting their exposure to and training with firearms. Rural households tend to have a greater likelihood of gun ownership,17 and exposure and experience with firearms may negate the novelty of finding a hidden gun, resulting in less curiosity and play. We also did not ask parents or participants about previous gun safety training, such as the National Rifle Association’s Eddie Eagle program. Although these programs have been found to be helpful in teaching children to repeat verbal commands,18 they have also been found to increase the likelihood of children handling found guns.19 Future research should take into account previous experience and training as a covariate.
Additional research on this topic is necessary. Future research could examine other types of media that contain guns, such as video games. Future research in a more natural setting (eg, a real home) would enhance the generalizability of these findings. Longitudinal research with similar dependent variables would allow researchers to examine the long-term effects of exposure to gun violence in the media. Regardless of design, research on this topic could be life-saving.
Every day in the United States, nearly 40 children are shot.20 More than 1.7 million children live in a home with unsecured guns. Nationwide campaigns by organizations such as Everytown for Gun Safety21 and the National Rifle Association22 aim to educate parents and homeowners on responsible gun safety. Organizations such as the Motion Picture Association of America and the Entertainment Software Rating Board provide information to parents about media content using ratings. However, these ratings are often inaccurate for violent content.23 The present experiment aimed to understand the connection between exposure to gun violence in the media and interest in and playing with guns in the real world. We believe that these data are a compelling start to the conversation on the various factors that can increase children’s interest in guns and violence.
Corresponding Author: Brad J. Bushman, PhD, School of Communication, The Ohio State University, 154 N Oval Mall, 3016 Derby Hall, Columbus, OH 43210 (email@example.com).
Accepted for Publication: May 30, 2017.
Correction: This article was corrected on November 6, 2017, to fix the caliber of the gun, and on May 6, 2019, to fix a typo in the Results section.
Published Online: September 25, 2017. doi:10.1001/jamapediatrics.2017.2229
Author Contributions: Both authors had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Bushman.
Acquisition, analysis, or interpretation of data: Both authors.
Drafting of the manuscript: Both authors.
Critical revision of the manuscript for important intellectual content: Both authors.
Statistical analysis: Both authors.
Administrative, technical, or material support: Both authors.
Study supervision: Both authors.
Conflict of Interest Disclosures: None reported.
Additional Contributions: Joseph Szymczak, BS, The Ohio State University, modified the gun for this experiment, for which he was compensated. Bridget Potocki, PhD, Zheng Wang, PhD, Hyunjin Song, PhD, and Robert Bond, PhD, The Ohio State University, assisted with data analysis. Rose Hallarn, MS, with Research Match, funded by award UL1TR001070 from the National Center for Advancing Translational Sciences, helped with recruitment opportunities. None of these contributors were compensated for these contributions. We thank the undergraduate research assistants who helped collect and code our data and the anonymous reviewers of this manuscript for the helpful suggestions.
Create a personal account or sign in to: