Media violence research

Research into the media and violence examines whether a link between consuming media violence and subsequent aggressive and violent behavior exists. Although some social scientists support this link, methodological and theoretical problems with the existing literature limit interpretation of findings in this area. There is concern among some scholars that media researchers may have exaggerated effects (Ferguson & Kilburn, 2009; Freedman, 2002; Pinker 2002; Savage, 2004).

For example, a 2008 editorial in medical journal the Lancet concluded that discussions of media violence effects were exaggerated. There are many explanations of aggressive behavior which do not emphasize the role of the media. For example, some researchers have suggested that the pathway to aggression is largely biological/genetic (see the work of Hare, 1993, Larsson, Andershed, & Lichtenstein, in Press, among others), while others have suggested that aggression can be explained by principles of evolutionary psychology.

Complaints about the possible deleterious effects of mass media are nothing new. Plato complained about the effects of plays on youth. Various media/genres, including dime novels, comic books, jazz, rock and roll, role playing/computer games  and many others have attracted speculation that consumers of such media may become more aggressive, rebellious or immoral. This has led some scholars to conclude statements made by some researchers merely fit into a cycle of media-based moral panics (e.g. Gauntlett, 1995; Trend, 2007; Kutner & Olson, 2008). The advent of television prompted research into the effects of this new medium in the 1960s. Much of research has been guided by social learning theory developed by Albert Bandura. Social learning theory suggests that one way in which human beings learn is by the process of modeling.

Media effects Theories
The research looking at the theoretical mechanisms that link consuming media violence and aggression has resulted in a handful of processes that some scholars suggest may explain any relationship that may exist. Although vocal support for these theories remains in some quarters, particularly among social psychologists, critics have contended that rhetorical support for these theories has generally outstripped largely absent data (Freedman, 2002; Guantlett, 1995; Olson, 2004; Svage, 2004).

The first hypothesis,briefly mentioned above, includes Bandura's social learning theory, which projects that media characters who serve as models for aggressive behavior may be attended to by viewers and depending upon whether the behaviors are rewarded or punished, would either inhibit or disinhibit imitation of that behavior. This idea of modeling was observed in Bandura's famous Bobo Doll experiments. In this particular study, Bandura showed a child a video of a model beating up a Bobo doll and then put the child in a room with a Bobo doll to see if he/she would imitate the behavior previously seen on the video.

The findings of this experiment would support that through social learning processes there is a causal relationship between consumption of violent media and aggressive behavior. However, Bandura's social learning theory has been revised as it has evolved over time. In 2002, Bandura updated his theoretical perspective in terms of social cognitive theory, which demonstrates how his initial formulation has developed over the years. Furthermore Bandura's experiments have been criticized (e.g. Gauntlett, 1995) on several grounds.

First, it is difficult to generalize from aggression toward a bo-bo doll (which is intended to be hit) to person-on-person violence. Secondly, it may be possible that the children were motivated simply to please the experimenter rather than to be aggressive. In other words, the children may have viewed the videos as instructions, rather than incentives to feel more aggressive. Third, in a latter study (1965) Bandura included a condition in which the adult model was punished for hitting the bo-bo doll by himself being physically punished. Specifically the adult was pushed down in the video by the experimenter and hit with a newspaper while being berated. This actual person-on-person violence actually decreased aggressive acts in the children, probably due to vicarious reinforcement. Nonetheless these last results indicate that even young children don't automatically imitate aggression, but rather consider the context of aggression.

The second hypothesis is priming, an idea formulated by Jo and Berkowitz in 1967, which was then later revised in 1994. The revised formulation of this theory focused on the belief that media violence might prime thoughts of aggressive behavior and, consequently, make actual aggressive behavior more likely. The priming hypothesis has received only weak and inconsistent support by research in the context of media violence (Freedman, 2002; Savage, 2004).

A third hypothesis is Zillman's theory of excitation advancing the notion that the arousal-inducing properties of media violence are important for understanding the potency of emotional reactions that occur immediately after exposure. For example, if a viewer becomes angry at a situation following exposure to a depiction of arousing violent media, this arousal could later transfer to that person's anger and intensify it--making aggressive behavior more likely. Evidence for this theory has generally been absent.

The fourth hypothesis advocated to explain the link between media violence and aggression is the idea of desensitization. According to this belief, with repeated exposure to media violence, a psychological saturation or emotional adjustment takes place such that initial levels of anxiety and disgust diminish or weaken. It is thought that the lower level of negative emotion associated with consistent exposure to media violence may reduce the urgency to respond to violence in the real world. The most recent evidence that supports emotional desensitization was reported by Carnagey, Anderson, and Bushman in 2007 as it related to violent video games. A sample of college students were assigned at random to play either a violent or non-violent video game for 20 minutes. They were then asked to watch a 10 minute video of real life violence. The students who had played the violent video games were observed to be significantly less effected by the a simulated aggressive act than those who didn't play the violent video games. However the degree to which the simulation was "believable" to the participants, or to which the participants may have responded to "demand characteristics" is unclear (see criticisms below). Research does seem to suggest exposure to media violence may desensitize viewers to media violence itself. However, support for the belief that this transfers to real-world violence has been weak (Savage, 2004).

The fifth hypothesis is physiological activation, which suggests that when children watch TV violence, the neurophysiology of a "phylogenetically-old brain system" is mobilized along with activation of limbic and neo-cortical systems that prepare the organism for motor plans associated with fight or flight. The authors, Murray et al (2006), of this theory suggest that responses to media violence may be "preconcious" and have long-term implications that extend beyond the period of exposure.

Some authors have contended that support for these theories have generally been poor (Freedman, 2002, Pinker, 2002, Olson, 2004; Savage, 2004). Thus debate on the merits of these theories is likely to continue.

One alternate theory the Catalyst Model (Ferguson et al., 2008) has been proposed to explain the etiology of violenced. According to the Catalyst Model, violence arises from a combination of genetic and early social influences (family and peers in particular). However media violence is explicitly not considered a causal influence according to this model, considered too weak an influence. Specific violent acts are "catalyzed" by stressful environment circumstances, with less stress required to catalyze violence in individuals with greater violence predisposition. Given that the Catalyst Model specifically deemphasizes media violence, this theory is directly at odds with most learning-focused media violence researchers.

A final theory worth mentioning is "Moral Panic" theory. Elucidated largely by David Gauntlett this theory postulates that concerns about new media are historical and cyclical. Society forms a predetermined negative belief about a new media (typically not used by elder members of society in power). The research which is produced, and statements by scholars and politicians are designed to confirm the pre-existing belief, rather than objectively study the issue with care. Ultimately the panic dies out after several years or even decades, but ultimately resurfaces when yet another new form of media is introduced.

Criticisms of media violence research
Although organizations such as the American Academy of Pediatrics and the American Psychological Association have suggested that thousands (3500 according to the AAP) of studies have been conducted confirming this link, others have argued that this information is incorrect. Rather, only about two hundred studies (confirmed by meta-analyses such as Paik and Comstock, 1994) have been conducted in peer-reviewed scientific journals on television, movie, music and video game violence effects. Critics argue that about half find some link between media and subsequent aggression (but not violent crime), whereas the other half do not find a link between consuming violent media and subsequent aggression of any kind.

Criticisms of the media violence link focus on a number of methodological and theoretical problems including (but not limited to) the following (see Freedman, 2002; Olson, 2004; Tedeschi & Quigley, 1996; Pinker, 2002):


 * 1) Failure to employ Standardized, reliable and valid measures of aggression and media violence exposure.  Although measurement of psychological variables is always tricky at best, it is generally accepted that measurement techniques should be standardized, reliable and valid, as demonstrated empirically.  However, a read of the studies involved notes that the measurement tools involved are often unstandardized, sloppily employed and fail to report reliability coefficients.  Examples include the "Competitive Reaction Time Test" in which participants believe that they are punishing an opponent for losing in a reaction time test by subjecting the opponent to noise blasts or electric shocks.  There is no standardized way of employing this task, raising the possibility that authors may manipulate the results to support their conclusions.  This task may produce dozens of different possible ways to measure "aggression", all from a single participant's data.  Without a standardized way of employing and measuring aggression using this task, there is no way of knowing whether the results reported are a valid measure of aggression, or were selected from among the possible alternatives simply because they produced positive findings where other alternatives did not. Ferguson and Kilburn, in a paper in Journal of Pediatrics, have found that poorly standardized and validated measures of aggression tend to produce higher effects than well validated aggression measures.
 * 2) Failure to report negative findings. Many of the articles that purport positive findings regarding a link between media violence and subsequent aggression, on a closer read, actually have negative or inconclusive results. One example is the experimental portion of Anderson & Dill (2000; with video games) which measures aggression four separate ways (using an unstandardized, unreliable and unvalidated measure of aggression, the Competitive Reaction Time Test mentioned above) and finds significance for only one of those measures. Had a statistical adjustment known as a Bonferroni correction been properly employed, that fourth finding also would have been insignificant. This issue of selective reporting differs from the "file drawer" effect in which journals fail to publish articles with negative findings.  Rather, this is due to authors finding a "mixed bag" of results and discussing only the supportive findings and ignoring the negative findings within a single manuscript.  The problem of non-reporting of non-significant findings (the so-called "file cabinet effect") is a problem throughout all areas of science but may be a particular issue for publicized areas such as media violence.
 * 3) Failure to account for "third" variables. Media violence studies regularly fail to account for other variables such as genetics, personality and exposure to family violence that may explain both why some people become violent and why those same people may choose to expose themselves to violent media. Several recent studies have found that, when factors such as mental health, family environment and personality are controlled, no predictive relationship between either video games or television violence and youth violence remain (Ferguson, San Miguel & Hartley, 2009; Ybarra et al., 2008, Figure 2).
 * 4) Failure to adequately define "aggression."  Experimental measures of aggression have been questioned by critics (Mussen & Rutherford, 1961; Deselms & Altman, 2003). The main concern of critics has been the issue of the external validity of experimental measures of aggression.  The validity of the concept of aggression itself, however, is rarely questioned.  Highly detailed taxonomies of different forms of aggression do exist.  Whether or not researchers agree on the particular terminology used to indicate the particular sub-types of aggression (i.e. relational versus social aggression), concepts of aggression are always operationally defined in peer-reviewed journals.  However many of these operational definitions of aggression are specifically criticized.   Many experimental measures of aggression are rather questionable (i.e. Mussen & Rutherford, 1961; Berkowitz, 1965; Bushman & Anderson, 2002; Deselms & Altman, 2003). Other studies fail to differentiate between "aggression" aimed at causing harm to another person, and "aggressive play" in which two individuals (usually children) may pretend to engage in aggressive behavior, but do so consensually for the purpose of mutual enjoyment. (Goldstein)
 * 5) Small "effects" sizes. In the research world, the meaning of "statistical significance" can be ambiguous.  A measure of effect size can aid in the interpretation of statistical significance. In a meta-analysis of 217 studies by Paik and Comstock (1994), effect sizes for experiments were r = .37 and r = .19 for surveys, which are small to moderate effects.  Most of these studies however did not actually measure aggression against another person.  Paik and Comstock note that when aggression toward another person, and particularly actual violent crime is considered, the relationship between media violence and these outcomes is near zero.  Effects can vary according to their size (for example the effects of eating bananas on your mood could very well be "statistically significant" but would be tiny, almost imperceptible, whereas the effect of a death in the immediate family would also be "statistically significant" but obviously much larger). Media violence studies usually produce very small, transient effects that do not translate into large effects in the real world. Media violence researchers often defend this by stating that many medical studies also produce small effects (although as Block and Crain, 2007, note, these researchers may have miscalculated the effect sizes from medical research).However the social importance of the effect size has been taken into consideration, which is causing more debate on this particular issue.  Sparks et al suggests that because media audiences sometimes number in the millions that even small statistical effects can actually translate into bigger social problems.  For example, if one in several thousand is influenced by a violent TV show to commit a serious act of aggression, the social consequences of several million viewers could be significant.  But given the huge size of media audiences, it can be believed that these small statistical effects are practically unpreventable and therefore, do not deserve much attention.
 * 6) Media violence rates are not correlated with violent crime rates. Ultimately the biggest problem for this body of literature is that for this theory to be true, media violence (which appears to have been consistently and unfailingly on the rise since the 1950s) should be well correlated with violent crime (which has been cycling up and down throughout human history). By discussing only the data from the 1950s through the 1990s, media violence researchers create the illusion that there is a correlation, when in fact there is not. Large spikes in violent crime in the United States occurred without associated media violence spikes during the 1880s (when records were first kept) and 1930s. The homicide rate in the United States has never been higher than during the 1930s. Similarly, this theory fails to explain why violent crime rates (including among juveniles) dramatically fell in the mid 1990s and have stayed low, during a time when media violence has continued to increase, and saw the addition of violent video games. Lastly media violence researchers can not explain why many countries with media violence rates similar to or equal to the U.S. (such as Norway, Canada, Japan, etc.) have much lower violent crime rates. Huesmann & Eron's own cross-national study (which is often cited in support of media violence effects) failed to find a link between television violence and aggressive behavior in most of the countries included in the analysis (including America, and even in studies on American boys).
 * 7) Media violence on TV is a reflection of the level of violence that occurs in the real world. Many TV programmers argue that their shows just mirror the violence that goes on in the real world.  Zev Braun,of CBS, in 1990 argued in a debate on the Violence Bill that, “We live in a violent society.  Art imitates modes of life, not the other way around: it would be better for Congress to clean that society then to clean that reflection of society.”

Researchers' response to criticisms

 * 1) Regarding instruments used to measure aggression, some media researchers argue that better measures are often not readily available Also measuring "violent criminal behavior" in laboratory studies would clearly be unethical, much the same way as experimental studies of smoking and lung cancer would have been unethical (Bushman & Anderson, 2001).
 * 2) Regarding the inconclusive nature of some findings, media researchers often contend that it is the critics who are misinterpreting or selectively reporting studies (Anderson et al., 2003). It may be that both sides of the debate are highlighting separate findings that are most favorable to their own "cause".
 * 3) Regarding "third" variables, media violence researchers acknowledge that other variables may play a role in aggression (Bushman & Anderson, 2001) and that aggression is due to a confluence of variables. Rowell Huesmann has said: "Serious aggressive behavior only occurs when there is a convergence of multiple predisposing and precipitating factors". These variables are known as "third variables" and if found, would probably be mediator variables (which differ from moderator variables). A mediator variable could 'explain away' media violence effects, whereas a moderator variable cannot. For instance, some scholars contend that trait aggressiveness has been demonstrated to moderate media violence effects (Bushman), although in some studies "trait aggression" does appear to account for any link between media violence exposure and aggression. Other variables have also been found to moderate media violence effects (Bushman & Geen, 1990). Another issue is the way in which experimental studies deal with potential confounding variables. Researchers use random assignment to attempt to neutralize the effects of what commonly are cited as third variables (i.e. gender, trait aggressiveness, preference for violent media). Because experimental designs employ random assignment to conditions, the effect of such attributive variables on experimental results is assumed to be random (not systematic).  However, the same can not be said for correlational studies, and failure to control for such variables in correlational studies limits the interpretation of such studies.  Often, something as simple as gender proves capable of "mediating" media violence effects.
 * 4) Definitions of aggression have improved with time. argue some media scholars. Most researchers agree that aggression, as a construct involves wilful intent to cause harm or intent to another person who would wish to avoid the same.  As such the problem may have less to do with the definition of aggression, but rather how aggression is measured in studies, and how aggression and violent crime are used interchangeably in the public eye.
 * 5) Much of the debate on this issue seems to revolve around ambiguity regarding what is considered a "small" effect. Media violence researchers contend that effect sizes noted in media violence effects are similar to those found in some medical research which is considered important by the medical community (Bushman & Anderson, 2001), although medical research may suffer from some of the same interpretational flaws as social science.  This argument has been challenged as based on flawed statistics, however (Bloack & Crain, 2007).  Block & Crain (2007) recently found that social scientists (Bushman & Anderson, 2001) had been miscalculating some medical effect sizes.  The interpretation of effect size in both medical and social science remains in its infancy.
 * 6) More recently, media violence experts have acknowledged that societal media consumption and violent crime rates are not well associated, but claim that this is likely due to other variable that are poorly understood. However, this effect remains poorly explained by current media violence theories, and media violence researchers may need to be more careful not to retreat to an unfalsifiable theory – one that cannot be disproven (Freedman, 2002).
 * 7) Researchers argue that the discrepancy of violent acts seen on TV compared to that in the real world are huge. Some findings looking at reality TV programming appear to support researchers claims.  One study looked at the frequency of crimes occurring in the real world compared with the frequency of crimes occurring in the following reality-based TV programs: America’s Most Wanted, Cops, Top Cops, FBI, The Untold Story and American Detective, (Oliver, 1994). The types of crimes were divided into two categories, violent crimes and non-violent crimes.  87% of crimes occurring in the real world are non-violent crimes, whereas only 13% of crimes occurring on TV are considered non-violent crimes.   However, this discrepancy between media and real-life crimes may arguably dispute rather than support media effects theories.

Media violence and youth violence
Several scholars (e.g. Freedman, 2002; Olson, 2004; Savage, 2004) have pointed out that as media content has increased in violence in the past few decades, violent crimes among youth have declined rapidly. Although most scholars caution that this decline cannot be attributed to a causal effect, they conclude that this observation argues against causal harmful effects for media violence.