Basic research

Theories of human aggression have fallen into one of two camps. Social learning theory (Bandura, 1973) stipulates that aggressive behave is an acquired instrumental act that is controlled over time by its consequences. It is learned behavior that is closely governed by rules. In contrast, the frustration–aggression model (Dollard, Doob, Miller, Mowrer, & Sears, 1939) proposes that aggression is an automatic hostile reaction to perceived frustration. The key element is hostility. Indeed, “(a)ggression and hostility are synonymous terms” (Berkowitz, 1963, p. xii). The emphasis of this theory is on the antecedents of the aggressive reaction, particularly goal blocking, heightened anger, threat, and frustrated expectations. Even though Bandura acknowledged the aversive instigative mechanisms of some aggressive behaviors, he emphasized the goal directedness and proactive features of aggression. He noted, “theories generally give disproportionate attention to aversively motivated aggression … (I)ncentives also constitute important impellors of action. A great deal of aggression is prompted by its anticipated benefits. Here, the instigator is the pull of expected success, rather than the push of aversive treatment” (Bandura, 1973, p. 57).

Notwithstanding his human political misguidedness, the ethologist Lorenz (1966) provided a resolution to this debate by meticulously describing the different aggressive behaviors that are naturally displayed by a wide range of species, from fish to primates. He identified numerous kinds of aggression but focused on the major distinction between cold-blooded predation and frenzied anger. The former is a nonemotional behavior that is clearly aimed at a specific goal or outcome, such as food, territory, or dominance. Slow, deliberate planning is evident, and, although the act itself may be vicious, the preparation indicates forethought. The latter behavior, in contrast, is an outburst of anger that appears to be a defensive reaction to goal blocking, provocation, threat, or frustration. For Lorenz (1966), this reactive anger is more violent, less controlled, and less predictable. Other ethologists have also distinguished between instrumental (proactive) aggression and hostile (reactive) aggression and have identified unique psychobiological mechanisms for each behavior (Moyer, 1976; Scott, 1972). It appears that hostile reactive aggression is universal in primates and must be “unlearned” or controlled through development; in contrast, instrumental aggression may be universal in some species but is more likely acquired through reinforcement.

Boyce and Ellis (2005) have integrated the ethological, evolutionary, and psychobiological literatures to articulate a more general theory of reactivity to environmental threats and stressors, of which personal provocation is a particular kind. In their general model, humans have adaptively evolved to respond to environmental endangerment with an integrated system of biological sensitivity to context that include increases in heart rate, metabolic mobilization of cellular nutrients, and “augmenting vigilance to threats and dangers” (p. 272). Hostile attributional bias may be considered a cognitive component of this system of reactivity. Although generally adaptive, this system becomes maladaptive when the response fails to extinguish after the actual environmental threat has dissipated. The cognitive recognition that a provocation is not threatening is crucial to the extinction of the stress response. Most children learn to make these cognitive responses during the first 4 years of life, but some children do not. Ellis, Essex, and Boyce (2005) found in a study of 249 children that very stressful early environments produce chronic heightened reactivity to later stimuli, suggesting a developmental origin for hostile attributional bias to be discussed later in this article.

Both ethologists and psychobiologists emphasize the difference between unemotional instrumental behavior and reactivity. Crucial to this distinction are the different cognitive mechanisms that have been articulated by ethologists when describing these behaviors. The key cognitive aspect of proactive aggression is the expectation that reward will follow. The key aspect of reactive aggression is the response to threat, or, rather, the perception that one is under threat; hence, the term used is retaliatory aggression. Lorenz (1966) noted that sometimes a fish will display what appears to be self-defensive retaliatory aggression even though the instigation is not real. Sudden movement in the water might inspire a self-defensive aggressive response, even though the movement was random. Lorenz speculated that the sensory-perceptual apparatus of the fish “interpreted” the movement as threatening.

Lorenz identified self-defensive aggression as a core behavioral characteristic of virtually every species. Furthermore, he noted that retaliatory aggression in response to threat often moves well beyond the defense of self to destruction of a threatening opponent. Hartup and deWit (1974) summarized the primate studies that indicate that rough-and-rumble play among young primate peers is the primary context for the occurrence of angry reactive aggressive behavior, when some aspect of the play triggers a defensive response. Thus, the contribution of ethology has been to identify a type of aggressive behavior that is an emotionally frenzied self-defensive and retaliatory response to the perception of threat.

Translational research

The distinction between reactive and proactive aggression has been discussed as types of behavior in young adults (Buss, 1966; Feshbach, 1970) and in children (Hartup, 1974; Rule, 1974), but only recently has the distinction been directly observed in children. Price and Dodge (1989) developed a coding scheme to distinguish these behaviors in direct observations of 70 5- and 6-year-old boys on school playgrounds. They found support for the discriminant validity of these behaviors and identified different social contexts that are associated with different probabilities of occurrence of each type. Rough play was the primary context for reactive aggression. Further, they found that individual differences in reactive aggression were correlated with lower play ratings from peers, whereas instrumental use of aggression was not negatively valued by young peers.

Other studies have found that hostile attributional biases are associated with reactive, but not proactive, aggressive behavior, despite the fact that these two aggressive behaviors are positively correlated with each other. Dodge and Coie (1987) identified groups of reactively aggressive, proactively aggressive, combined aggressive, and nonaggressive first- and third-grade African American boys and presented them with a measure of intention-cue detection like that used by Dodge et al. (1984) but adapted with new race-specific stimuli. The two reactively aggressive groups displayed less accuracy at detecting peers' benign intentions, and instead, overinterpreted hostile intent. When the peer's intent was actually hostile, they were highly accurate. When the peer's intent was ambiguous, the reactively aggressive groups were more likely than the other groups to attribute hostile intent. Furthermore, direct observations of these boys in peer interactions indicated that the number of errors of presumed hostility on the laboratory task predicted a boy's rate of overreactive aggression, but not proactive aggression.

Crick and Dodge (1996) replicated the correlation between hostile attributional biases and reactive, but not proactive, aggressive behavior in a sample of 624 9- to 12-year-old boys and girls. Graham and Juvonen (1998) found that hostile attributional biases were correlated with retaliatory aggression among middle scholars, and may have their origin in previous experiences of being victimized by peers. Schwartz et al. (1998) found that hostile attributional biases among 66 8-year-old African American boys predicted acts of reactive aggression but not proactive aggression or dominance as directly observed during contrived peer group interactions. Dodge, Lochman, Harnish, Bates, and Pettit (1997) found that reactively aggressive, but not proactively aggressive, third-grade children displayed higher scores for encoding inaccuracy of peer intent stimuli. An important feature of this finding is that they statistically controlled for teacher-rated attention problems and impulsivity, thus eliminating an alternate hypothesis to explain this pattern. Finally, Dodge, Price, Bachorowski, and Newman (1990) presented 128 adolescent boys in a maximum security prison with videorecorded stimuli and found that hostile attributional biases were positively correlated with the number of interpersonally violent crimes committed (as rated from official records), undersocialized conduct disorder, and the reactive aggression subscale of the Revised Problem Behavior Checklist, but not with the number of nonviolent crimes, socialized conduct disorder, or the proactive aggression subscale. Furthermore, this pattern of findings held even when intelligence, socioeconomic status, and ethnicity were controlled statistically.

Basic research

Recent advances in functional magnetic resonance imaging (fMRI) have enabled basic scientists to identify the brain regions that are implicated in various types of aggressive behavior. This work sheds light on the cognitive mechanisms in retaliatory aggression. The amygdala has been well documented as the site for the much of emotion processing and aggressive behavior, but only recently has it also been implicated as the site for the detection of threat, and the processing of the source and meaning of threatening stimuli. Adams, Gordon, Baird, Ambady, and Kleck (2003) presented clearly threatening and ambiguously threatening facial stimuli to adults while measuring amygdala responses through fMRI. They found that left amygdala responses varied significantly between these stimuli, with greater signal intensity change occurring during the harder “work” of interpreting the ambiguous stimulus.

Other fMRI studies have revealed that the paralimbic cortex and other limbic regions associated with the midbrain dopamine system are implicated in impulsive, immediate-reward behavioral decisions such as retaliation, whereas frontoparietal activity is associated with reasoned behavioral choices such as restraint from impulsive aggression (McClure, Laibson, Loewenstein, & Cohen, 2004). The authors interpret this finding as suggesting that “human behavior is often governed by a competition between lower level, automatic processes that may reflect evolutionary adaptation to particular environments, and the more recently evolved, uniquely human capacity for abstract, domain-general reasoning and future planning” (p. 506).

De Quervain et al. (2004) used positron emission tomography to examine brain regions associated with aggressive revenge during a contrived laboratory game among adults. What they found is extraordinary. They found that the dorsal striatum is activated in anticipated satisfaction from punishing wrongdoers, a region that is also activated during anticipation of monetary gains and pleasant tastes. Thus, humans do not require extrinsic rewards for retaliatory aggressive behavior; merely engaging in revengeful aggression brings pleasure to the brain, even when it brings other extrinsic costs (Knutson, 2004). The tendency to engage in retaliation brings its own reward and is deeply rooted in the evolution of the brain.

Given that the amygdala is an “older” brain region than the frontal cortex in evolutionary terms, the tendency to make hostile interpretations of experienced negative outcomes may be an “older” phenomenon than the “newer” capacity to distinguish hostile from nonhostile intent in others. The capacity to understand that others may be acting benignly is crucial for social cooperation and may be associated with the development of civilization.

Translational research

What is intriguing about this work is that retaliatory and justified aggression are core human behaviors, with long evolutionary histories and deep-rooted neural mechanisms. It is the development of higher order brain structures that enable humans to discern when this behavior is justified and when one should refrain from inappropriate aggressive retaliation. The mechanisms of this restraint are still to be discovered, but they probably involve the frontal cortex and its executive control, capacity for complex thinking, and delay of gratification. This work suggests that humans may be born with the aggressive retaliatory response in their repertoire and may acquire the capacity to inhibit that response during the course of development.

Basic research

Whereas ethology identified the link between threatening stimuli and retaliatory aggressive responses, it was social psychology that articulated the human tendency to infer cause to the events around oneself, the mediating role of the attribution process in the link between external events and personal behavioral responses, and the conditions under which different kinds of attributions are made. These processes concern psychological epistemology (Jones et al., 1971). Heider (1958) created “naive” psychology, which describes how humans act as lay scientists in making attributions of intention and inferring cause. Social psychologists conducted numerous studies following this formulation, to understand the rules that humans follow in inferring cause and then responding behaviorally, Epstein and Taylor (1967) completed an important study to demonstrate that retaliatory aggressive behavior occurs as a direct function of the conditions leading to a victim's attribution of hostile intent by a provocateur, which were experimentally manipulated. Further, the retaliation occurred as a function of the degree to which the victim actually attributed hostile intent.

Kelley (1971) and Jones and Davis (1965) are credited with the basic tenets of attribution theory, which are that

1. human beings have evolved to be motivated to assign a cause for important behaviors displayed by the self and others;
2. the assignment of causes follows systematic rules; and
3. the attribution of cause has important consequences for subsequent affect and behavioral responses.

The first of these tenets is hardly disprovable, but its importance has been supported by the testing of attributional hypotheses in a wide variety of domains ranging from depression to aggression.

The second tenet has been tested by laboratory experiments that led to the articulation of numerous rules of how causal attributions are reached. Individuals use logical principles when making attributions about others' intentions, such as covariation (the tendency to attribute an outcome to a cause with which it covaries), distinctiveness (the tendency to increase confidence in an attribution to the extent that other plausible causes are eliminated), and personalism (the tendency to attribute hostile intent if the actor had not behaved in this manner to anyone except the self). Kelley (1971) proposed that humans act like good scientists in making causal attributions. In making “correspondent inferences,” they consider covariation over time, examine possible discounting factors, and question the constancy of a relation. Specifically regarding provocation stimuli in which the cause of a frustration is at issue, Pastore (1952) found that arbitrary frustrations, in which a plausible alternative explanation cannot be identified, are more likely to be attributed to hostile intentional behavior by the frustrator than are nonarbitrary frustrations for which plausible alternate causal explanations are obvious. Plausible alternate factors include a widely recognized physical cause for the frustration, a compelling institutional rule that must have guided the frustrator, or even mental illness in the frustrator. Jones and Davis (1965) proposed that the perceiver may even consider his or her own responsibility for the provocation: “If the perceiver believes he has done something to earn attack, insult, or rejection, he will presumably be less inclined to appraise his attacker negatively than if the attack was unreasonable or arbitrary” (p. 249).

Jones and Davis (1965) proposed that high hedonic relevance, or motivational significance, will increase an individual's likelihood of making correspondent inferences about an act. So, if a child suffers a provocation in a very important realm (such as peer status, disrespect, or relationship threat), the child will be more likely to attribute hostile intent to the peer provocateur than if the provocation occurred in an unimportant realm.

The third tenet is that the factors that determine the perceived cause of a stimulus also influence the perceiver's behavioral affect and response to the event. Indeed, aggressive behavior is a more likely response when the provocation is intentional than accidental (Rule & Duker, 1973) and foreseeable than unforeseeable (Dyck & Rule, 1978). Costanzo, Grumet, and Brehm (1974) found that aggression is a more likely response when the provocation was experimentally manipulated to be freely chosen than constrained, and Darley, Klossen, and Zanna (1978) found that pairing the provocation with a mitigating circumstance reduced the likelihood of aggressive retaliation.

Translational research

Basing his hypotheses on attribution theory in social psychology, Dodge (1980) found that when a child made an attribution that a peer acted with hostile intent, the probability that the child would respond with retaliatory aggression was .60, whereas when the same child made an attribution that the peer had acted benignly (as in an accident), the probability of an aggressive response was .24. Although the attribution that a peer provocateur has acted with hostile intent is the most obvious inference that could lead to retaliatory aggression and the attribution that a peer acted accidentally is the most obvious inference that could lead to restraint from retaliation, in practice, persons learn to make a wide variety of attributions about others. Attributions that have been associated with responsibility, culpability, and therefore meriting retaliation or sanction, include the inference that a peer acted with willful neglect of the self, that a peer paid insufficient attention to the self, that a peer placed her or his own perspective above that of the self, that a peer failed to plan ahead to see the potential consequences of his or her actions, and that a peer acted irresponsibly (e.g., by drinking, getting out of control) in a way that led indirectly to the self's harm. These inferences of “indirect responsibility” are similar to many of the principles that guide legal responsibility and justify legal means of retaliation in both criminal and civil court (Dodge, 1991). The principle seems to be that if the peer's actions lead to the self's harm, the peer is, by default, culpable unless a plausible mitigating explanation is identified. Thus, inferring nonhostile intent may take more cognitive “work” and be more developmentally advanced than inferring hostility. This point has important implications for differences in attributional tendencies due to age, cognitive capacity, and temporary incapacitation due to fatigue or stress, which will be discussed later.

Basic research

Among the most important contributions of social psychology is that the “systematic” rules followed in making causal attributions include some universal nonrational tendencies that provide a window into the problematic behavior of some youth. Petty, Wegener, and Fabrigar (1997) have summarized the variety of processes that lead to biased conclusions, ranging from objective-inadequate information reasons to personal-motivational reasons. These biases are universal, and overcoming these biases takes more cognitive “work” and, presumably, greater cognitive development and emotional control.

The objective–cognitive bases for these tendencies begin with the simple fact that the array of possible information to use in making a causal attribution in any instance is overwhelming, especially during ongoing interaction when inferences must be made online in real time, leading humans to adopt heuristic shorthand rules in reaching conclusions. Tversky and Kahneman (1973) identified the “availability” heuristic as a common rule to make inferences quickly. Other heuristics include “salience” and “accessibility,” in which more recent information is weighted more heavily (Petty et al., 1997).

In most cases, heuristics provide efficient ways of reaching accurate conclusions, but in some cases, the use of heuristics leads to clearly nonrational and inaccurate judgments. For example, Jones and Nisbett (1971) reported a series of laboratory experiments to demonstrate the fundamental attribution error that “there is a pervasive tendency for actors to attribute their actions to situational requirements, whereas observers tend to attribute the same action to stable personal dispositions” (p. 80). The different attributions are due to differences in the salience of stimuli to the different parties during information processing. Applied to attribution of a provocation by a peer, the bias is that victims will tend to attribute dispositional volition (that is, hostile intent) to the provocateur, whereas the provocateur will tend to identify situational causes for the same behavior. The result is a recipe for a conflict between parties having different perspectives on the same event.

Pepitone and Sherberg (1957) identified another nonrational attributional tendency, “a general cognitive tendency to match motives with acts” (p. 761). This tendency was discovered in experiments in which participants were asked to play games in which they were to bargain with opponents, who were secretly constrained in their options. After the game, participants regularly underestimated the degree to which opponents had been externally constrained and overestimated the degree to which opponents had acted willfully. Their explanations alluded to the behavior displayed by the opponent, ignoring the possibility of constraint. This tendency has been found to be enhanced even more when the opponent gains rather than loses from a constrained action.

Another nonrational tendency is parsimony, which is the tendency to accept as sufficient the first cause that becomes salient, leading to termination of information search processes that might reveal a more accurate explanation (Simon, 1967). Yet another nonrational bias is the tendency to weight negative information more strongly than positive information (Kanouse & Hanson, 1971). Kogan and Wallach (1967) presented hypothetical life dilemmas to laboratory participants and found that “the deterrence value of costs for failure exceeds the attraction value of gains from success in affective risk-taking” (p. 133). This concept forms the basis of prospect theory (i.e., “losses loom larger than gains”), which earned Kahneman and Tversky the Nobel Prize (Kahneman & Tversky, 1982). Why might this tendency have developed in humans? First, negativity stands in contrast to a predominantly positive world, and thus may have greater salience by its uniqueness. Second, a single instance of negativity may signal danger and threat to survival, whereas a single instance of positivity rarely is essential to survival. Thus, negativity and vigilance to negative stimuli may be evolutionarily adaptive. Third, in judging a complex stimulus, a negative component of the stimulus may “trump” or “preempt” (Costanzo & Dix, 1983) subsequent processing and may be a simpler strategy in cognitive computation than weighing both positive and negative cues equally.

On the favorable end of the continuum, Zajonc (1965) has identified the mere exposure effect to become more favorable toward a stimulus merely based on frequency of previous contact. Thus, strangers will be attributed more hostile intent than will friends (Dodge, 1980). This tendency is exacerbated if the initial exposure led to a favorable judgment and outcome, which then is repeated over time through self-affirming cognitive processes that have been called a halo effect.

At the same time that social psychologists were articulating attribution theory, cognitive psychologists were describing how individuals make judgments and decisions (Dawes, 1964; Simon, 1957). Kahneman and Tversky (1973) described several heuristics that individuals use to make judgments when presented with too much information, notably representativeness and availability. The former is the tendency to classify a stimulus as belonging to a particular category when the stimulus has features that are similar to the features of members of the category. For example, a child might have a cognitive construct in memory of “mean peers,” which includes male gender, large size, and unfamiliarity as features. Upon confronting a new peer, the child might be biased toward inferring that the peer is a member of the mean peer group simply because he has these characteristics, that is, they are representative. The availability heuristic is the tendency to make a judgment consistent with a category simply because that category has been used frequently in the recent past. So, a child who has been the object of frequent hostile provocations in a given week may be likely to attribute hostile intent to peers in future circumstances simply because that category is highly available.

These heuristics have developed in the human species because they lead to accurate decisions and effective behavior in most circumstances. However, they can also lead to erroneous judgments and poor decisions. As an example, Kahneman and Tversky (1973) found that the representativeness heuristic leads to ignoring base rates when making classification judgments. The availability heuristic describes the tendency to overestimate the actual likelihood of events that have been reported in the media (Slovic, Fischoff, & Lichtenstein, 1979).

Another major process in biased attribution is motivational. Dissonance theory (Festinger, 1957) proposes that inconsistency in elements in one's cognitive system produces dissonance, which is reduced through elimination of disconsonant elements. Individuals thus display confirmation biases, in which they engage in attributional processes that are biased toward confirming preexisting schemas or hypotheses despite contradictory evidence (Darley & Fazio, 1980; Ross, 1977). The confirmation occurs through discounting of contradictory cues and overweighting of supporting cues (Gurwitz & Dodge, 1975). In cases in which an action by another produces a negative outcome for the self, any information that is inconsistent with the negative outcome (such as a benign intent cue) is more cognitively complex, and is likely to be discounted unless highly compelling.

In addition to these general factors, situational factors have been found to affect the attribution made about a cause. For example, a perceiver's mood state has been shown in numerous studies to affect the perceiver's causal attributions, mostly in a direction that is consistent with the mood state (Wegener & Petty, 1996). Thus, a perceiver who is in a hostile mood or agitated state is more likely to attribute hostile intent to a provocateur than is a perceiver in a benign mood state (Mayer & Hanson, 1995). Both direct effects of mood-congruent attributional tendencies and indirect effects through mood effects on vigilance, search, and cognitive load have been proposed to describe this effect (Petty et al.,1997).

Translational research

Based on the social psychological finding that attributional biases are more likely to occur in contexts of emotional importance, MacBrayer et al. (2003) found that provocations that strike at a child's relationships are more likely to be attributed to hostile intent than are provocations that involve overt physical contact. Other factors, such the prior behavior of the actor and the relationship between the actor and the perceiver, also have an impact (Clore, Schwarz, & Conway, 1994). Dodge (1980) found that children are much more likely to attribute hostile intent to peers who have a reputation for aggressiveness than to nonaggressive peers. Children are more likely to discount provocations that are initiated by acquaintances than by unfamiliar peers, except when the acquaintance is a reputed aggressive child (Steinberg & Dodge, 1983).

The context of the interaction with a peer also has an impact, as the salience heuristic of attribution theory would suggest. Lochman and Dodge (1998) experimentally manipulated the context as competitive versus cooperative, and fond that competitive contexts elicited more hostile attributions than did cooperative contexts. Interestingly, nonaggressive children were able to shift attributional tendencies to match the new context as the context shifted, whereas aggressive children were unable to shift out of a competitive context and hostile attribution tendency into a cooperative context with a nonhostile attribution tendency. Thus, situational constraints may interact with individual differences to lead to particularly vexing circumstances for aggressive children.

Based on the social psychological observation that attributions involving the self are especially prone to biases, Dodge and Frame (1982) found that hostile attributional biases by aggressive children are restricted to circumstances in which the self is the object of the provocation. When children are asked to imagine peer A provoking peer B, no hostile attributional bias tendencies by aggressive children were found. This finding is important because it suggests that the phenomenon is truly related to the interpersonal context and not to a cognitive bias to attribute behavior to intentional action.

The importance and specificity of the interpersonal context was examined by Hubbard, Dodge, Cillessen, Coie, and Schwartz (2001), who observed 11 groups of six boys interact with each other in contrived peer groups on a daily basis for 5 consecutive days. In between play sessions, boys responded to hypothetical provocations involving each of their peer group members. The Social Relations Model by the social psychologist Kenny (1994) predicts that both behavior and social cognitions will be modeled more accurately at the dyadic level than the individual difference level. Using Kenny's variance-partitioning procedure for round-robin data, Hubbard et al. (2001) found that the behaving child accounted for 9% of the variance, the peer partner accounted for 6% of the variance, and the dyadic relationship accounted for 16% of the variance in reactive aggressive behavior that was directly observed in play sessions. Likewise, in terms of hostile attributions made by children, the actor accounted for 19% of the variance, the peer provocateur accounted for 8% of the variance, and the dyadic relationship accounted for 21% of the variance. Finally, a mixed-model regression analysis revealed that reactive aggressive behavior was significantly predicted from hostile attributional biases at the dyadic level. A full model of the relation between hostile attributions and reactive aggression would include individual difference features of the child (i.e., aggressive children make more hostile attributions), the peer (i.e., children make more hostile attributions about known aggressive peers than nonaggressive peers), and the dyadic relationships (that is, certain relationship contexts are especially volatile).

Yet another influence is the social context of anxiety and threat, separate from the stimulus. Attribution theory predicts that hostile attributions are more likely to be made under ambient conditions of threat, due to salience. Dodge and Somberg (1987) experimentally manipulated these conditions by exposing children to fictitious peers who were supposedly in a room next door. These peers led the child to believe that they would, or would not, threaten the child in future interactions. Immediately afterward, the child attended to the experimenter's hypothetical videorecorded stimuli and made attributional judgments about peer provocateurs. The condition of threat to the self had a significant impact on reducing the accuracy of children's intent inferences and increasing hostile attributional biases, but only among aggressive children. Nonaggressive children were relatively immune to this situational context, whereas it had a discernible impact on furthering the biases of aggressive children.

Basic research

The contributions of social psychology describe and explain the cognitive processes that are implicated in hostile attributions and individual acts of aggression, but they do not help us understand how chronic and stable patterns of aggressive personalities are related to these processes. Traditional personality psychologists (e.g., Goldberg, 1993; McCrae & Costa, 1997) who have studied such “traits” have espoused a view that runs counter to the notions that personality characteristics develop, change, and dynamically interact with the social environment. Instead, their view is that static, biologically driven characteristics present at birth drive all individual differences in behavior. Such a notion would not allow for the hypothesis that hostile attributional biases develop and guide behavior; instead, it would dismiss these biases as epiphenomena that are more parsimoniously explained by an underlying hostility trait that causes both attributional biases and aggressive behavior. Contemporary personality has evolved an alternative perspective, however, that supports and amplifies the model being proposed here.

The social–cognitive theory of personality (Cervone & Shoda, 1999) is based in the formulations by Mischel (1999) and Bandura (1999) that personality coherence is guided not by traits but rather by situation-specific processes of social cognition. Instead of exogenous traits causing behavior in all situations, patterns of stable cognitive processing guide behavior in specific situations. Because these cognitive patterns are stable, behavior patterns become stable and coherent “from the bottom up” rather than from the top down. It is stable cognitive processes that lead us to infer stable personality traits, rather than the other way around. Zelli and Dodge (1999) cite an analogy to behavioral traits of cities. They note that cities are described in personality terms, such as bawdy (New Orleans), rude (New York), and flaky (San Francisco). However, it is erroneous to conclude that the trait causes the city and its people to act as they do; rather, microprocesses must operate to lead people to behave in particular ways, and the coherence of the behavior patterns lead to higher order inferences about the cities' traits. The traits are a descriptor rather than a cause; the processes provide the causal engine.

So, too, it is with human personality. The major units of personality coherence are cognitive structures and processes. Coherence develops because individuals organize their memories for past events in the form of internal mental representations that make sense and tell a story about how the world operates (Higgins, 1990; Stromquist & Strauman, 1992). These memories have been called knowledge structures (Cervone & Shoda, 1999), schemas (Kelley, 1971; Wyer, 1981), heuristics (Kahneman & Tversky, 1973), scripts (Huesmann, 1988), stereotypes (Mackie & Hamilton, 1993), stories or narratives (Shank & Abelson, 1995), and working models (Bowlby, 1973, 1980, 1982). These schemas efficiently summarize past experiences, although they necessarily leave out details and include biases. These latent cognitive structures guide the processing of future information, including attributions that one makes about future events. Schemas develop as an efficient organizing frame to store memories of past events (Cantor & Kihlstrom, 1982; Markus, 1977). Children who develop hostile schemas, through whatever means, are likely to use these schemas to perceive and interpret future stimuli through a process of perceptual readiness (Hochberg, 1970). Prior exposure to a biased set of events may lead an individual to become perceptually ready to interpret future events in ways that are consistent with the past events, through selective vigilance to certain cues and selective recall of schemas that are consistent with those cues. Snyder, Tanke, and Berscheid (1977) found that selective recall of past events or schemas can shape attributions that are made about current stimuli.

Translational research

The work in social–cognitive theories of personality suggests that aggressive children have may hostile schemas that are stored in memory. Burks, Dodge, Price, and Laird (1999) administered a version of the Children's Social Constructs Assessment (CSCA; Stromquist & Strauman, 1992) to 259 children followed longitudinally. In the CSCA, children are asked to speak open-endedly about peers whom they like and dislike, and transcribed responses are later scored for the kinds of constructs that had been invoked. The CSCA scores accounted for 26% of future externalizing problem scores and predicted growth in these scores even when past externalizing behavior had been statistically controlled. Burks, Laird, Dodge, Pettit, and Bates (1999) adapted two instruments, the Sentence Completion Task and the Assessment of Schema Typicality, to measure schemas in a sample of 585 eighth-grade children, and found that hostile schemas strongly predicted later externalizing problems, with a path coefficient of .59. Zelli, Dodge, Lochman, Laird, and CPPRG (1999) assessed normative beliefs about aggressive behavior in a sample of 387 elementary school children and found that they predicted teacher, parent, and self-ratings of aggressive behavior. Thus, measured in several different ways, hostile schemas robustly correlate with aggressive behavior and predict growth in aggressive behavior across development.

Another major proposition of social–cognitive personality theory is that patterns of social cognition are situation specific. Dodge, Laird, Lochman, Zelli, and CPPRG (2002) utilized confirmatory factor analysis to conclude that measures of hostile attributions are, indeed, organized within types of situations; that is, measures within peer group entry situations cohere more tightly with each other than with measures in peer provocation situations, and vice versa. Furthermore, a measure of social knowledge structures (in this case, positive and negative emotion understanding) predicted these hostile attributions, and hostile attributions mediated the effect of knowledge structures on aggressive behavior.

Following from the work that links hostile schemas to attributional processes, Zelli et al. (1999) hypothesized that attributional processes would mediate the effect of hostile schemas on children's aggressive behavior. They assessed hostile beliefs in Grade 2, attributional processes in Grade 4, and aggressive behavior in Grade 5 in a sample of 300 children and found support for this hypothesis. Attributional processes accounted for about 50% of the effect of hostile beliefs on later aggression.

How do hostile schemas affect attributional processes? Dodge and Newman (1981) presented aggressive and nonaggressive boys with a variety of social cues about a hypothetical provocation that they imagined to be directed toward them. Selective recall of hostile cues over benign cues predicted a hostile attribution about the provocateur. Furthermore, rapid responding without regarding to the full array of available cues predicted a hostile attribution, but only among the aggressive group of boys, who presumably called on biased schemas about hostile peers to make their inferences.

One consequence of acquiring self-schemas that involve peers directing threatening stimuli to the self is heightened focus on the self. Individuals who hold schemas about being threatened will become focused on threats to the self. In turn, self-focused attention is likely to lead an individual to react quickly and aggressively to provocations, even before waiting to determine whether the provocation is hostile or benign. Indeed, Fenigstein (1979) found that individuals who are high in self-focused attention are more likely to respond negatively to an ambiguous rejection by a peer than are individuals who are low in self-focused attention. Dodge and Tomlin (1987) translated these principles into the hypotheses that aggressive children would be more likely than nonaggressive children to focus on and utilize self-schemas in response to ambiguous provocations (rather than objective information in the stimulus array) and that utilization of self-schemas would predict hostile attributional responses. When aggressive and nonaggressive children were presented cues surrounding ambiguous provocations, both hypotheses were supported.

Yet another process that follows from the acquisition of a hostile schema is resistance to new information that would contradict the schema (Cantor & Kihlstrom, 1981). Lochman and Dodge (1998) presented aggressive and nonaggressive children with actual dyadic discussion tasks and then experimentally manipulated hostile and benign peer-intent information to change across time. Consistent with the hypothesis, once information was presented that confirmed a hostile schema, aggressive children were resistant to new information that would force a change of interpretation. Nonaggressive children were more flexible in response to changing cues.

Costanzo and Dix (1983) identified the process of making overly quick and biased attributions in response to specific cues as preemptive processing. In this circumstance, the individual abandons rational processes of examining information objectively and responds “automatically” with predetermined and inaccurate attributions and behaviors. Lochman, Lampron, and Rabiner (1989) and Rabiner, Lenhart, and Lochman (1990) found that aggressive children engage in such preemptive processes in social problem-solving situations that require rapid responding, and Dodge and Somberg (1987) found that experimentally manipulated ambient threat triggers preemptive processing by aggressive children.

Basic research

Although personality psychology has provided the theory that accounts for cross-time stability in hostile attributional biases and aggressive behavior, it is developmental psychology that has provided the basis for understanding when and how these patterns develop in childhood.

Reviews by Hay (2005) and Lemerise and Dodge (2000) reveal that aggressive behavior is universal in human beings, and begins early in life. Recognizable signs of the capacity for anger appear in the first year of life. Resistance and protest in response to being thwarted predate retaliation, but all occur in the first year. Consistent with the thesis that retaliatory aggression requires relatively little cognitive processing, the onset of retaliatory aggression occurs in the first few years of life, before cognitive development of attributional processes.

It appears that the capacity to infer intent develops in the third and fourth years of life (Schult, 2002; Wellman, Phillips, & Rodriquez, 2000), when children begin to develop a theory of mind (Flavell & Miller, 1998). Nonetheless, the press to match the outcome of an act with the intent of the actor remains strong for the next several years. Understanding that a peer's act that leads to a negative outcome for oneself (e.g., knocking over one's building blocks) might be nonintentional and benign is a more complicated cognitive endeavor than simply matching the outcome to the intent (e.g., negative outcome means negative intent). Björkqvist and Österman (2001) found that 7-year-old children are more able to distinguish act and actor than are 4-year-olds. Following from this developmental work, Dodge and Newman (1981) found that the tendency to infer hostile intent in response to an ambiguous provocation decreases across these years. Dodge and Price (1994) found that this tendency is largely accounted for by advances in cognitive skills of encoding relevant cues and accurately recognizing others' intent in stimulus circumstances of certain intent.

Early adolescence brings another stage of cognitive development when contemplation of alternatives and thinking about hypothetical possibilities grows with the transition to formal operations (Neimark, 1982). This period thus represents an opportune time for intervention to alter patterns of attributional biases.

Translational research

The development of individual differences in attributional styles is a function of more than cognitive maturation and inherent neuropsychological deficits. Deficits such as impulsivity and cognitive limits on the ability to examine alternatives have been found to correlate only weakly with hostile attributional tendencies, although it is plausible that they could interact with life experiences, much in the way that genetic risk factors interact with adverse life experiences to produce conduct disorder (Caspi et al., 2002). The model proposed here suggests that the key to understanding individual differences in attributional patterns is understanding the acquisition of the tendency to attribute benign intent given a provocation stimulus and adverse personal outcome. How children acquire this uniquely human pattern has been informed by developmental psychology.

Several principles in socialization contributed by developmental psychology undergird the proposed model. First, in general, early influences have greater impact than later influences. The primacy of the first several years of life was emphasized by the National Academy of Sciences study, From Neurons to Neighborhoods (Shonkoff & Phillips, 2000), based on the discovery that neural development is rapid and ongoing in these years. Experiences in these years also tend to shape which later environments are experienced; that is, children begin to self-select their environments, and thus restrict their later experiences. Therefore, the impact of the early years is felt both directly in shaping social–cognitive patterns but also indirectly in narrowing the future stimulus field. Second, social information processing patterns become more differentiated across situations as the child gets older (Werner, 1948). As a result, behavior becomes more situation specific and thus more complex with advancing age. Third, change in attributions and behavior patterns remains possible across development, but change involves layering new response patterns on top of the former patterns. It is not clear that early patterns are ever truly lost.

Five mechanisms in socialization have been proposed for the development of benign (versus hostile) attributional biases. The first mechanism is social learning through imitation of models who overtly display benign or hostile attributional tendencies in their interactions with children. MacBrayer et al. (2003) presented evidence that children's attributional bias patterns are similar to those of their mothers. Consistent with social learning evidence, they found that this transmission process is stronger within gender than across gender. Nix et al. (1999), in turn, found that mothers' hostile attributional biases predicted their own harsh discipline behavior toward their child, which mediated the child's growth in externalizing behavior problems. Given that aggressive children are more likely to display hostile attributional biases when a provocation is directed personally rather than toward a third peer, the social learning of attributional biases probably follows from personal rather than observed experiences. Thus, social learning of cognitive styles through modeling has preliminary support.

The second mechanism in socialization is acquired perceptual readiness based on the accumulation of past experiences. Processes of selective attention and heuristic categorization of stimuli are shaped by early experience (Posner & Rothbart, 2005). Children who repeatedly experience parents' prosocial behavior toward them may become perceptually ready (Bruner, 1957) to attribute positive intent to others in new situations, to discount discrepant information that would favor a hostile attribution, and to attend selectively to informational cues that are consistent with a growing schema of others as benign. Likewise, children who experience physical abuse in the first several years of life may become likely to become hypervigilant to future cues that indicate threat. Indeed, experiments in developmental psychopathology indicate that physically abused children attend selectively to hostile cues (Pollak & Tolley-Schell, 2003). Longitudinal studies indicate that they display hostile attributional biases in response to hypothetical provocations by peers (Dodge et al., 1990).

The third mechanism is facilitation of prosocial working models through a secure attachment relationship with one's caregiver (Bowlby, 1973, 1988). Bretherton (1999) asserted that early attachment relationships influence later social information processing patterns, including attributions of others' intent, and that insecure early attachment relationships could distort later processing. Cassidy, Kirsh, Scolton, and Parke (1996) found that young children with histories of insecure attachment later held defensive representations of peer relationships. Ziv, Oppenheim, and Sagi-Schwartz (2004) presented hypothetical mother–child interaction stories to children with varying attachment histories, and found that children with a history of insecure attachment were more likely than children with a secure attachment history to interpret the mother's behavior as indicating anger or insensitivity.

The fourth mechanism derives from social psychology as much as developmental psychology, and is based on studies indicating that ambient positive mood and recent task success foster attributions of prosocial intent, through both direct effects on mood-congruent attributional tendencies and indirect effects on selective attention to mood-congruent cues (Petty et al., 1997). Thus, it is posited that recurrent social, academic, and life success will be associated with the later tendency to attribute benign intent to others, and that recurrent social, academic, and life failure will be associated with later hostile attributional biases. Longitudinal evidence supports the relation between early peer social failure and later attributional tendencies (Dodge et al., 2003) and between peer victimization and later hostile attributions (Graham & Juvonen, 1998).

The fifth mechanism derives from a series of ingenious social psychological experiments in cultural studies by Nisbett and Cohen (1996). They sought to understand why the American South has much larger rates of violence than the North. They hypothesized that culture affects social–cognitive responding and aggressive behavior by placing relative value on disrespect, honor, and the sense of threat to one's personal integrity. A culture that highly values personal honor will encourage hostile attributions in response to personal threats and provocations. They provide evidence that the American South is characterized by these values, in contrast with the North. In laboratory experiments, the authors had confederates provoke southern and northern college students by bumping into them and calling them profanities (in an adult version of the same experimental manipulations in the aforementioned studies with aggressive and nonaggressive children). Over 80% of southerners responded to this provocation with anger toward the provocateur, in contrast with less than 40% of northerners. After this provocation, when presented with a hypothetical scenario depicting insult, 75% of southerners answered with an aggressive response, in contrast with 41% of northerners. Physiological measures collected at the same time indicated that southerners responded to the provocation with higher increases in cortisol and testosterone than northerners. In yet another experiment in which participants were provoked or not, southerners were more likely than northerners to attribute malevolent intent to the provocateur and to believe that the provocateur thought of the participants as “wimpy.” The authors conclude that a major cause of southern males' violence is the “sense of threat to one of his most valued possessions, namely, his reputation for strength and toughness” (p. xv).