Theories of Aggression

In brief

Why does a person deliberately hurt another? Three research traditions grew up answering that question, and for decades they were taught as rivals. The frustration-aggression hypothesis, stated by John Dollard and his Yale colleagues in 1939, held that aggression always traces back to a blocked goal. Social learning theory, built by Albert Bandura in the 1960s, held that aggression is learned like any other behavior, by watching others and seeing what gets rewarded. Biological and evolutionary accounts hold that the capacity for aggression is inherited hardware, shaped by natural selection and tuned by hormones, genes, and brain circuits. The productive move, made explicit by Craig Anderson and Brad Bushman in 2002, is to stop asking which is correct. Frustration is a situational input, learning is a history, and biology is the substrate on which both operate. They describe different parts of one causal chain.

The full treatment

The problem it answers

Aggression is behavior intended to harm another who is motivated to avoid the harm. That definition, now standard, already rules out accidents and consensual roughness, and it separates the act from the emotion of anger, which need not produce it. The scientific problem is that aggression is wildly variable. The same insult provokes a shrug from one person and a fist from another, and the same person reacts differently drunk, tired, armed, or watched. A useful theory has to explain both the reliable triggers and the enormous individual and situational variation. The three traditions each grabbed a different handle on that problem.

Frustration and aggression

In Frustration and Aggression (1939), John Dollard, Leonard Doob, Neal Miller, O. Hobart Mowrer, and Robert Sears advanced a bold, testable claim: "the occurrence of aggressive behavior always presupposes the existence of frustration," where frustration means the blocking of a goal-directed activity. Aggression was the natural, though not the only possible, response to a thwarted drive. The original wording was too strong, and Neal Miller softened it in 1941: frustration produces an instigation to aggression, but the response can be inhibited or displaced, and other responses compete with it. The theory also introduced displacement, the idea that when the frustrating agent is too dangerous or unavailable to attack (your boss, the economy), aggression is redirected onto a safer target, a mechanism later invoked to explain scapegoating.

The hypothesis nearly collapsed under evidence that frustration does not reliably yield aggression, until Leonard Berkowitz rebuilt it in 1989 as cognitive neoassociation. On his account it is not frustration specifically but aversive experience in general, heat, pain, foul odors, bad news, that generates negative affect, and negative affect automatically primes both fight and flight tendencies along associative networks. Whether aggression follows depends on cues in the environment. Berkowitz and Anthony LePage had already produced the theory's signature finding in 1967, the "weapons effect": angered subjects delivered more shocks when a rifle and revolver lay on a nearby table than when neutral objects did. The gun, Berkowitz argued, was an aggression-primed cue. The effect proved fragile in some replications and its robustness has been debated in the wider replication crisis, which is discussed below.

Aggression as learned behavior

Albert Bandura took the opposite starting point: aggression is not squeezed out by inner pressure but acquired from the social world. In the Bobo doll experiments (Bandura, Ross, and Ross, 1961, 1963), nursery-school children watched an adult attack an inflatable clown doll, hitting it with a mallet, kicking it, shouting distinctive phrases. Children who had seen the aggressive model, later left alone with the doll, reproduced the specific acts and words far more than children who had not, and boys did so more than girls. Crucially, in the 1963 follow-up, children who saw the model punished imitated less, and those who saw the model rewarded or unpunished imitated more, yet when offered incentives to reproduce the behavior, all groups could do so more equally. This split acquisition from performance: children learned the aggressive repertoire simply by observing, and separately decided whether to enact it based on expected consequences. Aggression, on this view, is behavior with a learning history. It is shaped by direct reinforcement, by the models a person watches, and by the outcomes those models experience.

Aggression as evolved and embodied

The biological tradition asks where the capacity itself comes from. Behavioral genetics finds that individual differences in aggression are substantially heritable: twin and adoption studies converge on roughly half the variance in aggressive and antisocial tendencies being attributable to genes, the rest to environment. A specific case became famous: Han Brunner's 1993 study of a Dutch family with a mutation disabling the MAOA gene, whose affected men showed impulsive aggression, and Avshalom Caspi's 2002 finding that a low-activity MAOA variant predicted antisocial behavior mainly in men who had been maltreated as children, a gene-environment interaction rather than a "warrior gene" acting alone. At the physiological level, testosterone correlates with aggression (though the causal arrow is tangled, since winning contests raises it), and damage or dysregulation in the amygdala and prefrontal cortex figures in impulsive violence.

Evolutionary psychology supplies the ultimate why. Martin Daly and Margo Wilson, in Homicide (1988), argued that patterns of lethal violence fit reproductive logic with unsettling precision: killers and victims are overwhelmingly young men, homicide rates track competition for status and mates, and violence against stepchildren far exceeds that against biological children (the "Cinderella effect"). Aggression, on this account, is not pathology but an evolved conditional strategy, deployed when its expected payoff in resources, status, or mating access outweighs its cost.

How the traditions fit together

The General Aggression Model of Craig Anderson and Brad Bushman (2002) makes the integration explicit. A given episode of aggression is the output of person factors (traits, genes, learned scripts, sex) and situation factors (provocation, frustration, pain, heat, aggressive cues) that feed a person's internal state, affect, arousal, and cognition, which then drives appraisal and a decision to act. Frustration is one situational input. Learning history supplies the scripts. Biology sets the reactivity of the whole system. What were three theories become three sources of variables in one process.

Lineage

The intellectual backdrop is the ancient dispute over whether humans are aggressive by nature. Thomas Hobbes cast the natural condition as a war "of every man against every man," while Rousseau and later romantics saw aggression as a corruption of a peaceable original. The Yale group of 1939 stood on Freud's drive theory (aggression as dammed energy seeking discharge) and on Clark Hull's learning theory, splicing the two. Bandura's social learning grew out of, and then broke with, the behaviorism of Skinner by insisting that learning could occur through observation alone, without the observer being reinforced. Konrad Lorenz's ethological On Aggression (1966) pushed the instinct view into the popular imagination and drew heavy fire, before the more disciplined evolutionary psychology of the 1980s replaced its loose energy metaphors with the logic of adaptation. Berkowitz and, later, Anderson and Bushman are the synthesizers who folded the older drive language into a cognitive framework.

The strongest case for it

Taken together, the traditions explain what none could alone. The situational tradition accounts for why ordinary people aggress under heat, pain, crowding, and provocation, and it yields interventions: reduce aversive conditions and remove aggressive cues. Social learning explains transmission, why aggression runs in families, subcultures, and media diets, and why it can be unlearned, and it is the empirical basis for the widely studied finding that exposure to modeled violence raises the probability of aggressive behavior. The biological tradition explains the stubborn individual differences that no amount of situational tinkering erases, and the evolutionary account explains the demographic signature of violence, its concentration among young men competing for status, that many societies exhibit and that a purely learning-based theory struggles to derive. The integrative models are strong precisely because they are cumulative: each input has independent evidence, and they combine so that the same provocation detonates in one person and not another.

The strongest case against it

Each tradition has real critics making real points. The frustration-aggression hypothesis in its original form was falsified; frustration plainly does not always cause aggression, and Berkowitz's rescue, by widening "frustration" to all aversive stimulation, risks unfalsifiability, since almost anything unpleasant then qualifies as a cause. The weapons effect, its showpiece, has been questioned by replication attempts that found weaker or inconsistent effects under controlled conditions, part of the broader replication crisis that has hit social psychology hard.

Bandura's Bobo doll work draws its own fire. A Bobo doll is designed to be hit and cannot be harmed, so critics question whether striking it is aggression at all rather than imitative play, threatening the study's construct validity. The children may have inferred, from the odd laboratory setup, that hitting the doll was expected, a demand-characteristics objection. And generalizing from a padded clown to real interpersonal violence is a leap of external validity. The wider claim that media violence causes real-world aggression remains genuinely contested: Christopher Ferguson and others argue the effect sizes are trivial, confounded by publication bias, and swamped by third variables, against Anderson and Bushman, who defend a real if modest effect.

The biological and evolutionary accounts face the charge of overreach. The "warrior gene" story was oversold: MAOA's effect is small, conditional on abuse, and easily misused to excuse or to stigmatize, and behavioral geneticists warn against reading heritability as destiny. Evolutionary explanations are accused of being unfalsifiable just-so stories, adaptive narratives fitted after the fact, and of naturalizing violence in a way that blurs the line between explaining and excusing. Feminist and cultural critics note that the theories' data skew heavily male and Western, and that cross-cultural variation in violence rates, from near-peaceful societies to feuding ones, is larger than any biological account comfortably predicts. Even the integrative General Aggression Model has been criticized for being so inclusive that it accommodates any result, which is a virtue for a framework and a vice for a theory that wants to be tested.

Where it stands now

The field has largely abandoned the winner-take-all framing. No serious researcher now argues that aggression is purely learned, purely frustration-driven, or purely biological. The working consensus is interactionist: aggression emerges from evolved and heritable predispositions, calibrated by developmental environment and learning history, and triggered by situational states. The live disputes are quantitative and specific, how large the media-violence effect really is, whether the weapons effect survives, how MAOA interacts with early adversity, not about which grand theory reigns. The replication crisis has been clarifying rather than destructive: it forced the classic demonstrations to be re-run with larger samples and pre-registration, and the effects that survive are the ones the applied world, in violence prevention, forensic risk assessment, and developmental intervention, can actually build on.

Test yourself

Bring to mind the last time you snapped at someone, a curt reply, a slammed door, a cutting remark. Try to sort the causes. What in the situation was aversive (tired, hot, thwarted, provoked)? What in your history made that response available (whose temper are you repeating)? And be honest about the part you would rather attribute to circumstance: notice how naturally you explain your own aggression by the situation, while explaining others' by their character. That asymmetry has its own name in this library, and it is one more reason a single-cause theory of aggression will always feel true from the inside and prove false under measurement.