Aggression

Abstract

Aggression, which MeSH classifies under aberrant motor behavior in dementia, is behavior directed toward another individual that is carried out with the proximate intent to cause harm. This article develops the construct as cognitive and social psychology now treat it: as the output of a sequence of appraisals and decisions, not a single drive. It distinguishes reactive from proactive aggression, traces the social-information-processing account of how a provocation is read and answered, surveys the frustration-aggression and neoassociationist traditions, and sets out the General Aggression Model that integrates person and situation inputs through present internal state. A worked example reconstructs a fixed-effect meta-analysis of the small but reliable media-violence association, and three interactive demonstrations make the appraisal sequence, the reactive-proactive distinction, and effect-size aggregation concrete.

Keywords: aggression, hostile attribution bias, general aggression model, reactive aggression, media violence

Aggression is behavior directed toward another individual carried out with the proximate intent to cause harm, where the perpetrator believes the behavior will harm the target and the target is motivated to avoid it (Anderson & Bushman, 2002). The definition is deliberately behavioral and intentional rather than emotional: an angry outburst that harms no one and is not meant to is not aggression, while a calmly planned assault is. Violence is the subset of aggression whose goal is extreme physical harm. This precision matters because the lay word folds together a feeling, an intent, and an act that the research literature keeps apart.

The modern treatment is cognitive. Where early theory cast aggression as a drive to be discharged, the dominant frameworks now model it as the product of how a person construes a situation: what they attend to, what intent they attribute to another's act, what responses they can access, and which they choose (Crick & Dodge, 1994). The sections below follow that logic from the reactive-proactive distinction through the social-information-processing sequence to the integrative General Aggression Model, then to the neural and hormonal substrate that constrains it.

What Aggression Is

Aggression is operationally any behavior directed toward another individual that is carried out with the proximate intent to cause harm, given that the actor expects the behavior to harm the target and the target is motivated to avoid it (Anderson & Bushman, 2002). Three terms in that definition do real work. *Intent* excludes accidental harm. *Proximate* fixes the intent at the moment of acting, so that a dentist who causes pain to help is not aggressive. And the *target's motivation to avoid* excludes consensual harm. Anger, by contrast, is an emotional state that often accompanies aggression but is neither necessary nor sufficient for it, and hostility is an enduring attitude rather than an act.

A long-standing division cuts the construct by its function. *Reactive* aggression, also called hostile or affective aggression, is an impulsive, angry response to a perceived provocation or threat, whereas *proactive* aggression, also called instrumental aggression, is a deliberate, unprovoked means to an end such as dominance or resources (Dodge & Coie, 1987). Bushman and Anderson argued that the older hostile-versus-instrumental dichotomy is too blunt, because a single act can serve both an emotional and an instrumental goal, and proposed that the dimensions of a specific episode matter more than a forced binary (Bushman & Anderson, 2001). The distinction nonetheless survives because reactive and proactive aggression have different correlates: reactive aggression tracks hostile attribution and emotional dysregulation, proactive aggression tracks positive outcome expectancies for aggression. At the level of measurement, self-report captures the construct along several facets at once: the widely used Aggression Questionnaire separates physical aggression, verbal aggression, anger, and hostility into distinct subscales, treating the behavioral act, the emotion, and the attitude as related but separable (Buss & Perry, 1992).

Types of Aggression

The Medical Subject Headings vocabulary files *Aggression* (D000374) as a polyhierarchical descriptor, placing it at once under aberrant motor behavior in dementia, under behavioral symptoms, and under social behavior, and lists three narrower descriptors directly beneath it, shown in Table 1. These subtypes are not mutually exclusive: an act of bullying is also agonistic, and a microaggression can be read as a low-intensity form of either. Nor is the list a theory of the kinds of aggression; it is an *indexing classification*, a controlled vocabulary built so that the published literature can be retrieved, not a claim that these three categories carve aggression at its natural joints. It records how the literature is indexed, not what the true types are. None of the three has its own article on this site, so each is named in plain text and glossed from its MeSH sense.

Social Information Processing

The most developed cognitive account of aggression is the social-information-processing model, which treats an aggressive response as the endpoint of a sequence of mental operations applied to a social cue (Crick & Dodge, 1994). A child who is bumped in a hallway must *encode* the cue, *interpret* it by attributing an intent to the other, *clarify* a goal, *access or construct* possible responses, *decide* among them by evaluating their likely outcomes, and *enact* the chosen response. Aggression can arise from a bias or deficit at any step, and the model's power is that it localizes the fault rather than treating aggressive children as uniformly deficient.

The interpretation step carries the best-known finding. Dodge and colleagues showed that aggressive children are disproportionately prone to a *hostile attribution bias*: when another child's intent is ambiguous, they infer hostility more readily than non-aggressive peers, and that inference licenses a retaliatory response (Dodge & Coie, 1987). The bias is specific to reactive aggression, consistent with the reactive-proactive split: reactively aggressive children over-read hostility, while proactively aggressive children instead hold more positive expectations about what aggression will achieve. Because the sequence is explicit, it is also a map of where intervention can act, whether by retraining attributions, expanding the response repertoire, or changing the evaluation of aggressive outcomes.

From Frustration to Cognitive Neoassociation

The oldest systematic theory held that aggression is always the consequence of frustration, the blocking of goal-directed activity. Berkowitz reformulated that frustration-aggression hypothesis decisively: frustration produces aggression not automatically but only to the extent that it generates *negative affect*, and any aversive event, not just a blocked goal, can do so (Berkowitz, 1989). In his cognitive-neoassociation model, aversive stimulation triggers negative affect, which activates an associative network linking anger-related feelings, thoughts, and motor tendencies, priming both fight and flight; higher-order appraisal then shapes which prevails. The model explains why heat, pain, and crowding raise aggression, and why aggression-related cues in the environment, the *weapons effect*, can amplify it.

A second tradition runs alongside this one. Bandura's social-learning account showed that aggression need not be provoked at all to be acquired: in the Bobo doll experiments, children who merely watched an adult model behave aggressively toward an inflatable doll later reproduced that aggression, demonstrating that aggressive behavior is learned through observation and imitation rather than only through frustration or reinforcement (Bandura et al., 1961). This learned-script route is one of the inputs the integrative model below must accommodate.

This is the lineage the General Aggression Model synthesizes. Anderson and Bushman cast person factors such as trait hostility and situation factors such as provocation or aggressive cues as inputs that act through a single present internal state of cognition, affect, and arousal, which feeds appraisal and decision (Anderson & Bushman, 2002; Allen et al., 2018). The model absorbs frustration-aggression, social learning, and social-information-processing as special cases, each naming a different route into the same internal state, and it makes the loop explicit: every aggressive encounter updates the knowledge structures the person brings to the next. An early application to interactive media found that violent video-game exposure was associated with aggressive thoughts, feelings, and behavior in both laboratory and correlational designs, an effect the model reads as the priming and rehearsal of aggressive knowledge structures (Anderson & Dill, 2000).

The Neural and Hormonal Substrate

Aggression has a conserved biological machinery that the cognitive account constrains rather than replaces. Nelson and Trainor reviewed the circuitry: a limbic core centered on the amygdala and hypothalamus generates aggressive responses, and the prefrontal cortex, especially its orbitofrontal and ventromedial sectors, regulates them, so that reactive aggression is associated with heightened limbic reactivity under weak prefrontal control (Nelson & Trainor, 2007). Serotonin is the best-characterized modulator, with low serotonergic function linked to impulsive rather than premeditated aggression, consistent with the reactive-proactive distinction drawn at the behavioral level.

The hormonal story is more nuanced than the folk link between testosterone and violence suggests. Denson and colleagues reviewed aggression in women and showed that the neural and endocrine systems once studied almost entirely in men operate in women too, that the testosterone-aggression relationship is modest and context-dependent in both sexes, and that sex differences in aggression are larger for physical than for indirect forms (Denson et al., 2018). The substrate therefore sets dispositions and thresholds, lowering or raising the probability that a given appraisal culminates in an aggressive act, without determining the act outright.

Worked Example

A recurring empirical question is how strongly exposure to violent media relates to aggression, and the answer comes from meta-analysis, which pools the correlation reported by many studies into a single weighted estimate. The standard procedure transforms each study's correlation *r* to a Fisher *z*, averages the *z* values weighted by each study's precision, and transforms the mean back to *r*. The Fisher transform is *z* = ½ ln[(1 + *r*)/(1 − *r*)], and the inverse-variance weight for a study of *n* participants is *w* = *n* − 3.

Consider five illustrative studies of violent-media exposure and aggression:

The weighted sum of the *z* values is 117(0.213) + 197(0.151) + 87(0.090) + 57(0.245) + 147(0.182) = 103.27, and the sum of the weights is 117 + 197 + 87 + 57 + 147 = 605. The mean Fisher *z* is therefore 103.27 / 605 = 0.171, and transforming back gives a pooled correlation of *r* = 0.17. The standard error of the mean *z* is the square root of 1/605 = 0.041, so the 95% confidence interval on *z* is 0.171 ± 1.96(0.041) = [0.091, 0.250], which back-transforms to a confidence interval on *r* of [0.09, 0.25]. The pooled correlation is small but reliably greater than zero, and it accounts for *r*² = 0.17² ≈ 0.029, under 3% of the variance in aggression (Prescott et al., 2018). The MediaEffectDemo lets the reader add or reweight studies and watch the pooled estimate and its interval update; the values are illustrative of the aggregation procedure, not drawn from a specific meta-analysis.

Discussion

Across a century of theory the center of gravity has moved from drive to appraisal. Frustration-aggression began with a near-hydraulic image of a damming and discharge of aggressive energy; Berkowitz replaced the hydraulics with associative networks and negative affect; social-information-processing localized the decision to read and act on a provocation; and the General Aggression Model folded all of these into a single cycle in which person and situation act through a present internal state (Berkowitz, 1989; Crick & Dodge, 1994; Anderson & Bushman, 2002). The practical payoff of the cognitive turn is that it names modifiable targets: an attribution that can be retrained, a response repertoire that can be widened, an outcome expectancy that can be revalued.

The open problems are substantive. The reactive-proactive distinction is robust in its correlates yet hard to apply to a single act that serves both ends, which is the force of the Bushman and Anderson critique (Bushman & Anderson, 2001). The size and meaning of media-violence effects remain contested even where their statistical reliability is not. And the integration of the biological substrate with the cognitive model is still schematic: naming the amygdala-prefrontal balance is not yet the same as predicting, for a given person in a given moment, whether an appraisal will tip into an act (Nelson & Trainor, 2007).

Current Directions

The General Aggression Model remains the field's organizing framework, and recent work has sharpened rather than displaced it. Allen, Anderson, and Bushman restated the model as a developmental and situational account in which repeated episodes build the knowledge structures, scripts, and attitudes that a person carries into the next encounter, so that long-run personality change and short-run situational effects are two time-scales of one process (Allen et al., 2018). The framing reframes chronic aggression as learned and therefore, in principle, re-learnable.

A second active front is the quantitative one. Prescott, Sargent, and Hull conducted a longitudinal meta-analysis of violent video-game play and later physical aggression and found a small but significant prospective association that held across cultures, bearing directly on the causal question that cross-sectional correlations cannot settle (Prescott et al., 2018). A third front widens the evidence base beyond its historical sample: Denson and colleagues synthesized the behavioral, neural, and hormonal study of aggression in women, correcting a literature that had generalized from men and clarifying where sex differences are real and where they have been overstated (Denson et al., 2018). Together these lines move the field toward longitudinal, mechanistic, and representative evidence rather than one-off laboratory effects.

Common Misconceptions

Aggression is a buildup of pressure that venting safely discharges.

The catharsis idea is not supported; acting aggressively, even on an inanimate object, tends to maintain or increase later aggression rather than drain it, because it rehearses aggressive scripts and leaves anger salient (Bushman & Anderson, 2001).

Frustration always leads to aggression, and aggression always comes from frustration.

The original hypothesis was reformulated: frustration raises aggression only when it produces negative affect, any aversive event can do the same, and the link is probabilistic rather than inevitable (Berkowitz, 1989).

Violent video games have no effect on aggression at all.

Longitudinal meta-analysis finds a small but statistically reliable prospective association across cultures; the live debate is over the effect's size and real-world importance, not its existence (Prescott et al., 2018).

Glossary

Aggression.

Behavior directed toward another individual carried out with the proximate intent to cause harm to a target motivated to avoid it.

Agonistic behavior.

The within-species complex of threat, attack, appeasement, and retreat seen in animal contests over resources, territory, or rank.

Amygdala.

A limbic structure central to threat detection and the generation of aggressive and defensive responses.

Catharsis hypothesis.

The unsupported idea that expressing aggression discharges an aggressive drive and reduces later aggression.

Cognitive neoassociation.

Berkowitz's model in which aversive events produce negative affect that activates a network linking anger-related feelings, thoughts, and action tendencies.

Frustration-aggression hypothesis.

The proposal that aggression follows from the blocking of goal-directed activity, later reformulated around negative affect.

General Aggression Model.

An integrative framework in which person and situation inputs act through a present internal state of cognition, affect, and arousal to drive appraisal and action.

Hostile attribution bias.

The tendency to infer hostile intent from another's ambiguous behavior, characteristic of reactively aggressive individuals.

Instrumental aggression.

A synonym for proactive aggression: a deliberate, goal-directed use of harm as a means to an end.

Meta-analysis.

A statistical method that pools effect sizes from many studies into a single precision-weighted estimate with a confidence interval.

Proactive aggression.

Cold, premeditated, unprovoked aggression used to obtain a goal such as dominance or resources.

Reactive aggression.

Hot, impulsive, angry aggression triggered by a perceived provocation or threat.

Social information processing.

The sequence of encoding, interpretation, goal selection, response access, decision, and enactment by which a social cue leads to a response.

Violence.

Aggression whose goal is extreme physical harm, such as injury or death; the most severe subset of aggression.

Weapons effect.

The finding that the mere presence of aggression-related cues such as weapons can increase aggressive responding.

Key Researchers

Craig A. Anderson

(Iowa State University). Distinguished Professor of Psychology who, with Bushman, developed the General Aggression Model and led much of the experimental work on media violence and aggression.
ORCID - Wikipedia

Albert Bandura

(1925–2021). Professor at Stanford University whose social-learning theory and the Bobo doll experiments showed that aggression is acquired by observing and imitating models.
Wikipedia - Wikidata

Leonard Berkowitz

(1926–2016). Professor at the University of Wisconsin–Madison who reformulated the frustration-aggression hypothesis and developed the cognitive-neoassociation model and the weapons effect.
Wikipedia - Wikidata

Brad J. Bushman

(The Ohio State University). Professor of communication and psychology whose work on human aggression, media violence, and the General Aggression Model includes the critique of the hostile-instrumental dichotomy.
ORCID - Wikipedia

Nicki R. Crick

(1958–2012). Professor at the University of Minnesota who established relational aggression as a construct and co-authored the influential reformulation of social-information-processing theory.
Wikipedia - Wikidata

Kenneth A. Dodge

(Duke University). Professor of public policy and psychology whose social-information-processing model and work on the hostile attribution bias shaped the study of reactive and proactive aggression in children.
ORCID - Wikipedia

Frequently Asked Questions

What is aggression in psychology?

Aggression is behavior directed toward another individual carried out with the proximate intent to cause harm, where the actor expects the behavior to harm and the target is motivated to avoid it (Anderson & Bushman, 2002).

What is the difference between aggression and anger?

Anger is an emotional state, whereas aggression is a behavior; anger often accompanies aggression but is neither necessary nor sufficient for it, since calmly planned harm is aggressive and an angry feeling that harms no one is not (Anderson & Bushman, 2002).

What is the difference between reactive and proactive aggression?

Reactive aggression is a hot, impulsive, angry response to a perceived provocation, while proactive aggression is cold, deliberate, and used instrumentally to reach a goal such as dominance or resources (Dodge & Coie, 1987).

What is the hostile attribution bias?

It is the tendency to infer hostile intent from another person's ambiguous behavior; reactively aggressive children show it more than their peers, which helps explain their retaliatory responses (Dodge & Coie, 1987).

Does frustration always cause aggression?

No. The frustration-aggression hypothesis was reformulated so that frustration raises aggression only when it produces negative affect, and other aversive events can do the same (Berkowitz, 1989).

Do violent video games cause aggression?

Longitudinal meta-analysis finds a small but statistically reliable prospective association between violent game play and later physical aggression across cultures; the size and importance of the effect remain debated (Prescott et al., 2018).

What is the General Aggression Model?

It is an integrative framework in which person and situation inputs act through a present internal state of cognition, affect, and arousal that drives appraisal and the decision to act, with each episode updating the knowledge structures used in the next (Anderson & Bushman, 2002).

What parts of the brain are involved in aggression?

A limbic core centered on the amygdala and hypothalamus generates aggressive responses, while the prefrontal cortex regulates them, and low serotonergic function is linked to impulsive aggression in particular (Nelson & Trainor, 2007).

References