The psychology of aggression
In this chapter, definitions and theories of aggressive behaviour are discussed. Several factors are important in aggressive behaviour and these are discussed under three broad headings: interpersonal factors, external factors, and individual differences.
TERMINOLOGY
Aggression is a word that in ordinary language encompasses a wide range of behaviours. However, these sometimes diverse behaviours seem to have a common thread and the aim of a psychological definition is to explain this. There have been many definitions of aggression. For example, Harre and Lamb (1983, p.13) catalogued more than 250. These various definitions seem to share the idea that aggression involves inflicting harm or damage but beyond this, there are a number of differences. For example, does the damage have to be inflicted on a living creature or does damage to objects or property also count as aggression? Does the damage have to be physical as argued by Zillman (1978) or does psychological harm such as saying something hurtful or damaging someone's reputation also constitute an aggressive act? Given that these distinctions may be important, one solution is to define aggression broadly but then describe different types of aggression. This approach has been taken by Buss (1961) who said aggression could be physical or verbal, active or passive, direct or indirect. Combining these factors yields eight types of aggression altogether.
Another issue, perhaps less easily resolved, concerns whether the damage or harm must be inflicted intentionally for the act to be considered aggressive. On the one hand it seems desirable to exclude inflicting harm accidentally, e.g. accidentally hitting someone with an object or saying something tactless without thinking. On the other hand, intentions cannot be observed and it may cause problems to use a definition that rests on unobservable behaviour. This was the view of Buss (1961) and Bandura (1973) whose definitions ignore the intentions that may lie behind the "response that delivers noxious stimuli to another organism". Despite these difficulties, many authors have included intention in their definitions of aggression. For example, Dollard, Doob, Miller, Mowrer, and Sears (1939) defined aggression as "an act whose goal-response is injury to an organism". Berkowitz (1974) also argues that intention is crucial. This was taken up by Geen (1976) whose working definition can be summarised in three points: (1) "Aggression consists of the delivery of noxious stimuli by one organism to another"; (2) "The stimuli are delivered by the former with the intent to harm the latter"; (3) "The one delivering the stimuli expects that the probability of the stimuli reaching the source is greater than zero". Note that this definition does not specify any particular emotional state, such as anger, or any particular attitude towards the victim, such as dislike or hostility. It does specify intention to harm but this should not be taken to exclude other intentions or motives.
This definition then implies that aggression is a form of social interaction in which the perpetrator intends to harm the victim in some way. Baron (1977) also proposes that the victim must be motivated to avoid such treatment in order to exclude sado-masochistic behaviour. Aggressive behaviour may also be classified into subtypes.
AFFECTIVE AND INSTRUMENTAL AGGRESSION
Behaviours characterised as aggressive may have different antecedents and intervening processes. Thus, aggressive behaviours are different from each other with respect to what motivates them, although they may be accommodated heuristically under the terms of affective or hostile and instrumental (cf. Baron, 1977). "Affective" aggression is that accompanied by a strong negative emotional state (see Table 1.1). In this case an intervening process—anger—instigates and guides the aggressive behaviour of which the main goal is injury or harm to the provocateur (Feshbach, 1964). Johansson (1981) lists its main characteristics: activation of the hypothalamus, increased blood flow to the musculature, heightened blood pressure and pulse rate, pupillary dilatation and decreased flow of blood to the viscera. Anger is inferred as a necessary but not sufficient condition—its presence does not inevitably produce affective aggression. However, the closeness between anger and the actual aggressive behaviour sometimes reinforces the idea of a cause-effect relationship.
TABLE 1.1
Characteristics of different types of aggression
Another type of aggression has been dubbed by Tedeschi (1984) "instrumental", and is related to behaviour whose main goal is not intent to harm but to establish social and coercive power over others through aggressive means. In this case, an aggressive option has been judged as most likely to be successful in achieving a favourable outcome to a conflict of interests. This form of aggression may occur when people defend property or rights and is found in war and social conflicts: Following a command from a person in authority some peaceful individuals are capable of committing extreme acts of violence (Milgram, 1963). Bandura (1973) argues that both affective and instrumental aggression are directed towards the attainment of specific goals and are therefore instrumental. As a result of this, other terms have been coined (see Table 1.1). Zillmann (1979) refers to annoyance-motivated and incentive-motivated. Dodge and Coie (1987) have proposed the terms reactive and proactive. Reactive aggression concerns retaliation to a perceived threat, whereas proactive aggression is used to obtain some other goal, e.g. dominance. However, an aggressive act often has elements of both types.
THEORETICAL PERSPECTIVES
Having defined aggressive behaviour, we now need to consider the reasons why it occurs. Many theoretical perspectives exist but Baron and Richardson (1994) divide them into four categories, which we will briefly discuss.
Aggression as an instinct
According to the theories of instinctive aggression, aggression occurs because humans are genetically programmed for such behaviour. These theories have generally arisen from animal work and assume that aggression provided biological advantages to our prehistoric ancestors. Lorenz (1966) proposed an evolutionary perspective. He suggested that the aggressive instinct is a source of energy that builds up over time and needs to be discharged. The greater the amount of energy accumulated, the weaker the stimulus necessary for overt aggression to occur. Freud (1920) also proposed that aggression stemmed from an innate, self-directed death instinct (thanatos) that is turned outward towards destruction of others. Both these theories regarded aggression as inevitable: The energy would build up until it could be released in some way either to attain the original goal or in some form of displacement activity. Lorenz believed that aggressive energy could be dispersed by many minor hostile but noninjurious acts, thus decreasing the likelihood of violent outbursts. According to instinctive theories, aggression is an unavoidable, integral part of human nature.
Aggression as a drive
The theories of aggression as a drive assume that it is a motivational force induced by some kind of deprivation to terminate or reduce that state. This perspective was elaborated by Dollard et al. (1939) in the frustration—aggression hypothesis: "aggression is always a consequence of frustration and, contrariwise, that the existence of frustration always leads to some form of aggression". Frustration, i.e. the blocking of some form of goal-directed behaviour, induces an aggressive drive, which facilitates aggressive behaviour. This theory initially gained widespread support but it eventually faltered on the certainty of both premises. First, the tenet that frustration always leads to aggression has been challenged by many empirical studies showing that frustration may lead to numerous alternative responses, e.g. despair. Thus, Miller (1941) refined the model to include other possible forms of behaviour. Second, there is much evidence that aggression may occur as a result of factors other than frustration. Indeed, frustration does not seem necessary for instrumental aggression such as self-defence.
Berkowitz (1974) further modified the drive theory to include contextual cues. He still accepted frustration as an important antecedent for aggressive behaviour but stressed the role of mediating factors in the environment. He therefore postulated that frustration led to negative affect or anger but this only led to aggression in the presence of aggressive cues. Both people and objects may become aggressive cues by previous association with anger arousal.
Cognitive models
The instinctive and drive theories of aggression ignore both emotion and cognition. Newer theories have attempted to revise the frustration theory to include these important elements. Thus, Berkowitz (1989) revised his earlier theory to state that frustration or other aversive stimuli (external factors) instigate aggressive reactions by creating negative affect. The response would then be determined by the individual's interpretation of the negative affect. If it was interpreted as anger, then aggressive tendencies would be likely to be aroused and more attention might be paid to aggressive cues. This was termed the "cognitive neoassociation model".
Zillman (1988) argued that cognition and excitation (arousal) are truly interdependent. Under very high levels of excitation, there may be interference with the cognitive processes that inhibit aggression, leading to impulsive aggression. At moderate levels of excitation, cognitive processes would operate to reduce aggression. In the same way, in a normal state of arousal, cognitions about an event are likely to influence the extent of arousal. If mitigating circumstances are recognised, there is likely to be a decay of any excitation. According to cognitive theories, people might be able to reduce aggressive behaviour by learning new ways of coping with provocation.
Social learning models
In contrast to aggression perceived as an instinct or a drive, Bandura (1977) has postulated that aggressive behaviour is acquired and maintained in a similar way to other forms of social behaviour. He does not discount biological or motivational factors, but postulates that the specific form, frequency, location, and target of aggressive behaviour are largely determined by social learning. Thus, reinforcing or rewarding aggressive behaviour increases the likelihood of repetition. Although direct experience may be important, work on children has shown observation and imitation (modelling) to be very influential. Bandura has suggested that there are three principal sources of aggressive models: the family, the subculture, and the mass media. If we observe aggressive behaviour being rewarded and we identify with the aggressor, we are more likely to use it ourselves. We may in fact learn new (aggressive) ways of responding. This behaviour can, however, be modified by self-imposed consequences. Thus, habitually aggressive people may reward themselves for successful intimidation of others, whereas people who disapprove of aggressive behaviour may experience guilt if they behave in the same way. Social learning theory then suggests that aggression only occurs under appropriate social conditions and we can change these conditions to reduce it.
INTERPERSONAL FACTORS IN AGGRESSION
Aggressive behaviour occurs within an interactive situation. Several possible behaviours may occur in an interpersonal exchange and therefore much research has been devoted to examining the conditions that facilitate an aggressive response. These studies have focused on frustration, provocation, attack, and instigation from others.
Frustration
Frustration, defined as the thwarting by one individual of another's goal-directed behaviour, has been assumed to be a strong elicitor of aggressive behaviour since the original hypothesis (Dollard et al., 1939). A large number of experiments seem at first to confirm that frustration is indeed an important antecedent of aggression. However, Baron and Richardson (1994) have pointed out that frustration has often been confounded by other factors in these early investigations. Studies that have carefully isolated different antecedents do confirm that frustration is one important factor in the induction of aggression (Berkowitz & Geen, 1967; Burnstein & Worchel, 1962). In one study it was shown that both frustration and personal insult increased aggression (Geen, 1968). Moreover, frustration did not need to be a result of the actions of others. Subjects who became frustrated through their own inability to accomplish a task also showed more aggression.
Other studies have found that frustration does not necessarily lead to an aggressive response (Buss, 1966; Kuhn, Madsen, & Becker, 1967). In these studies no differences were found between the frustrated and control groups. Together these studies cast some doubts on the early formulations that frustration is always a cause for aggression. Perhaps even more surprising than these results, however, is the suggestion in several additional experiments (Gentry, 1970; Rule & Hewitt, 1971) that exposure to strong frustration may sometimes serve to reduce rather than enhance later aggression. Frustration may elicit behaviours other than aggressive ones that attenuate the connection between frustration and aggression (cf. Buss, 1961). Barker, Dembo, and Lewin (1941, cf. Buss, 1961) contend that frustration may also produce useless, maladaptive, or regressive behaviour. Mandler (1972) argues that helplessness and anxiety, not aggression, may follow as a consequence of a person's response to feeling frustrated. Thus, the sequence of behaviours would depend on the subject's understanding of the responses that are available.
The current position seems to be that frustration sometimes facilitates aggression depending on certain factors. Baron and Richardson (1994) cite four important mediating factors: the magnitude; the presence of aggressive cues; the extent to which it is arbitrary or unexpected; and the emotional and cognitive processes of the frustrated individual. Frustration can be increased by intervening when the subject is close to his or her goal, by blocking expectation of success, and by making the subject's reasons for the action less legitimate (Harris, 1974; Kulick & Brown, 1979). Berkowitz (1989) has argued that frustration only produces a readiness for aggressive behaviour. The occurrence will depend on the presence of aggressive cues, i.e. stimuli associated with anger arousal. Such cues may be heightened by viewing violent films. Frustration that can be predicted or expected rarely produces aggression. However, if it is seen to be arbitrary or unjustified, it is more likely to result in aggressive behaviour. As well as aggressive cues, Berkowitz (1989) has emphasised the current feelings of the individual. Interpersonal or external factors that lead to negative affect or irritation increase the likelihood of aggressive behaviour.
In his reformulation of the relationship between frustration and aggression, Berkowitz (1989) discerns an associative network between negative affects, frustration, and aggression. In his model, the intensity and arbitrariness of the frustration leads to negative affect, and then aggressive cues may intensify the instigation to aggress. Attributions and judgements may facilitate or inhibit this process.
Provocation and attack
Attack is the key instigator to retaliatory aggression. Interpersonal aggression takes place in an interactive situation between two people. Therefore Bandura (1973, p.153) suggests that "if one wished to provoke aggression, the most dependable way to do so would be simply to physically assault another person, who would then be likely to oblige with a vigorous counterattack". Many studies support this idea. Persons who have been insulted or physically attacked are likely to respond with verbal abuse (James & Mosher, 1967), or with physical counterattack (Baron, 1972; Berkowitz, 1974). Moreover, the intensity of aggression depends on the intensity of the initiating attack (Epstein & Taylor, 1967; O'Leary & Dengerink, 1973). Verbal provocation can often lead on to aggressive actions. In research examining police reports of criminal violence, Felson (1982, 1984) found a typical pattern to start with an insult leading to verbal retaliation, escalating to an argument, threats, and finally physical assault. This work emphasises the interactive nature of aggression. Several investigators have compared frustration (blocking of ongoing behaviour) with attack (delivery of noxious stimulus) as antecedents of aggression (e.g. Buss, 1963; Geen, 1968; Geen & Berkowitz, 1967). They conclude that attack is a more effective manoeuvre in eliciting aggression. Diamond et al. (1984) have also shown that attack is a more powerful source of arousal than is frustration. Baron (1977), however, argues that although frustration seems to be a weaker stimulus for aggression than either physical or verbal attack, comparable outcomes between the two variables have sometimes been misleading due to the ignorance of the subjective effects of each on the individual. The relevant question is whether or not a mild insult would always be more effective in eliciting subsequent aggression than extremely strong frustration. Somewhat more unexpected, however, is evidence that individuals often react aggressively to indications of aggressive intention on the part of others, even when they are not actually attacked by these persons. Greenwell and Dengerink (1973) found experimentally that although attack is an important instigator of aggressive behaviour, symbolic elements that are incorporated in that attack may play a major role.
These results have received support from many other empirical studies. Epstein and Taylor (1967) also showed that intent to harm is in fact a more powerful determinant of retaliation than is the absolute intensity of an attack. Dodge, Murphy, and Buchsbaum (1984) showed that subjects tend to be more aggressive when they perceive a hostile intention in the other party, whether or not the person is actually hostile. The absence of malice or lack of intention in the attacker constrains victims not to retaliate (Johnson & Rule, 1986; Zillmann & Cantor, 1976). The role of perceived intentionality in aggression also seems to be influenced by personality characteristics (Dodge, 1980). Attribution of hostile intent by aggressive boys, even in ambiguous circumstances, has been explained by Nasby, Hayden, and DePaulo (1979) as a consequence of a hostile attributional bias resulting from a generally aggressive disposition.
A factor related to the relationship between attack and aggression is that of apology. Schwartz, Kane, Joseph, and Tedeschi (1978) and Darby and Schlenker (1982) studied the belief that apologies can mitigate a victim's anger and aggression. These authors found that the subjects tended to judge the transgressors as less liable to punishment if they made apologies. Schlenker and Darby (1981) suggested that people tend to use a more complex apology involving a larger number of components as the harm gets more serious. When the harm is severe, they argue, more intense apologies may be needed to alleviate the conditions of the victims. It seems likely that the severity of harm may in some way control the inhibitory effects of apologies on aggression.
Instigation from others
Aggression may not just be an interaction between two people. The presence of others may influence behaviour. This may be either by emphasising restraint or conversely by encouraging increased aggression. This is related to anticipated approval or disapproval of onlookers (Borden, 1975). Figures of authority have much more power to influence events. Many acts of aggression, especially those carried out by the police or armed forces, stem from commands from superiors, not from provocation or frustration. Obedience to these figures of authority may not be surprising because such figures have the support of society at large. However, some influential research has shown that even a relatively powerless source of authority can induce individuals to be very aggressive. Milgram (1974) conducted a series of well-known experiments and found that the majority of people (65%) were prepared to punish a supposed learner with increasingly painful shocks when instructed to do so. Two factors seem to be effective in counteracting this obedience: First, having to take responsibility for your own actions, and second, exposure to disobedient models.
EXTERNAL FACTORS IN AGGRESSION
Various adverse environmental factors may increase the likelihood of aggression, e.g. heat, noise, crowding, pollution. However, Bandura (1973) claims that this is only true when such behaviour represents a dominant behavioural tendency. When other responses are dominant, then adverse conditions may inhibit aggression. Much laboratory work has attempted to elucidate the relationship between excessive heat and aggression. These investigations have in fact indicated that the relationship is curvilinear. The chance of aggression increases as negative affect (induced by high temperatures) rises only up to a certain point. After this point is reached, the likelihood of aggression decreases as other responses such as escape from the situation dominate (Baron & Bell, 1976). Unfortunately this relationship is not supported by archival studies. In these a significant linear relationship is demonstrated between increasing temperature, violent crime, and other aggressive behaviour. It is likely that this relationship is much more complex, e.g. hot weather encouraging more people on to the streets and more consumption of alcohol, leading to the possibility of more aggression.
Noise seems to have a clearer relationship to aggression than temperature. Laboratory studies using loud and unpleasant noise have found that it facilitates the occurrence of aggression (Geen, 1978; Mueller, 1983). This especially applies when the noise is unpredictable or uncontrollable.
Work on overcrowding originated in studies of animals where a link with aggressive behaviour has been found. Although there is evidence that both high population density and crowding may be perceived as unpleasant and stressful to humans, there is no clear link to increased aggression. Another environmental factor is air pollution with smoke, solid wastes, and noxious gas. Laboratory studies have revealed that subjects required to breathe secondary cigarette smoke were more aggressive to the experimenter regardless of his behaviour to them (Zillman, Baron, & Tamborini, 1981). Other forms of air pollution may be more associated with malodour. An experiment investigating the effects of unpleasant odour (Rotton, Frey, Barry, Milligan, & Fitzpatrick, 1979) found more effects on aggression when it was moderately unpleasant than when it was extremely obnoxious. It is likely then that malodour is similar to temperature in having a curvilinear effect. The likelihood of aggression may decrease when the source of discomfort can be clearly identified and when escape becomes the dominant response.
As well as environmental factors, other external components may contribute to aggressive behaviour. A substantial body of work has linked violence in the mass media to aggressive feelings and behaviour. Several studies have found that viewing violent as opposed to nonviolent films increases aggressive behaviour on a task afterwards (e.g. Geen & O'Neal, 1969). This work has been criticised as lacking a no-film control. It has been claimed that viewing a nonviolent film may actually lower aggression compared to not seeing a film. Zillman and Johnson (1973) investigated this hypothesis and found that viewing a neutral, nonviolent film did lower aggression compared to both a violent film and no-film conditions. Berkowitz (1984) suggests that media effects are due to aggressive cues resulting in priming. If people are asked to list thoughts after viewing various films, they produce more aggressive thoughts as their rating of the level of violence increases (Bushman & Geen, 1990). Although much research has investigated the facilitating effects of violence shown in the media and a positive relationship has been found between watching television violence and aggression in adolescence, it is important to recognise that people who habitually display aggressive behaviour also show an increased preference for viewing violence (Huesmann & Eron, 1986; Wartella, 1995). Violence in the media may also increase the autonomic arousal of the viewer. Increased arousal may strengthen any responses, one of which may be aggression depending on cues, or it may be aversive to the subject leading to negative mood, which may be interpreted as anger, which in turn may be a precursor to aggressive behaviour.
Pain may also serve as an external factor increasing the likelihood of aggression. The prime goal may be escape but where this is blocked it may lead to aggression. Berkowitz (1983) proposed that pain generates negative affect in a similar way to heat and that this may facilitate an aggressive response.
Substance use is a common factor in aggression of varying degrees of severity. Habitual alcohol and drug use has been reported for repeat murderers (Adler & Lidberg, 1995; Tiihonen & Hakola, 1994). Both alcohol use and aggression have shared antecedents such as family pathology and childhood experiences of neglect or abuse (Fagan, 1990; White, Brick, & Hansell, 1993), which could partially account for such a relationship. However, in nonclinical populations, alcohol use is often reported prior to violent crime (Collins & Schlenger, 1988). Therefore, both acute and chronic use of alcohol has been shown to contribute to increased aggression. Laboratory research has confirmed this association (Bushman & Cooper, 1990) but has indicated that the relationship is complex, both contextual cues and individual characteristics being important. Further discussion of the effects of alcohol can be found in Chapters 5 and 6.
Arousal, cognition, and anger
Many external variables may lead to increased arousal, which may be cognitively interpreted as anger. These processes may be described as intervening variables between the provoking stimulus and the aggressive response. Aggressive behaviour itself may also be associated with activation of the sympathetic nervous system leading to physiological arousal expressed in bodily symptoms. Arousal increases the probability of aggressive behaviour under certain conditions. It has been argued (Berkowitz, 1969) that frustration and provocation lead to a state of arousal that in turn energises aggressive responses that the person is disposed to make. Arousal, however, contributes to aggression not only by energising responses. Its combination with specific thoughts generated within the situation can create a state of increased anger, which predisposes the person to aggression. Schachter and Singer's general two-factor theory of emotion (1962) is the original proposition of this idea. According to their model, a central determinant of how individuals behave is the way in which they label a state of heightened physiological arousal as an affective experience.
The link between physiological arousal and emotion is reducible to three propositions: (1) when an individual has no immediate physiological explanation for a state of physiological arousal, s/he will label the state and describe her/his feelings according to the cognitions available to her/him; (2) when the individual has a completely appropriate explanation for her/his physiological arousal, no evaluative needs will arise and the individual will label her/his feelings accordingly; (3) the individual will react emotionally and will report emotions only to the extent that s/he experiences the state of physiological arousal. The widely cited experiment (Schachter & Singer, 1962) was designed to test this theory. Physiologically aroused persons who were uninformed or who were misinformed about the various physiological effects of adrenaline injections were more verbally hostile in the presence of an angry companion, or were more euphoric in the presence of a happy companion, than were persons cognisant of the physiological effects of the drug. These labelling effects were consistent with the experimental context.
Other studies have shown similar contextual and labelling effects. Inducing subjects to label arousal produced by means other than frustration as anger, mediates an increase in aggression (Zillmann, 1979) and showing subjects bogus physiological evidence indicating that they are angry increases their level of aggression (Berkowitz & LePage, 1967). Several studies have reported findings consistent with Schachter and Singer's initial results (Ross et al., 1969). Thus, frustrated subjects who are induced to attribute their arousal to sources other than the frustration do not become as aggressive as subjects who are not so induced (Baron & Bell, 1975; Geen, Rakosky, & Pigg, 1972; Harris & Huang, 1974; Nisbett & Schachter, 1966).
The cognitive labelling perspective has also been challenged by some results. Marshall and Zimbardo (1979) compared the emotional states of subjects who received an injection of either placebo or adrenaline. Both groups were misinformed about possible bodily effects from the injections and then exposed to an euphoric confederate. Reports of affect did not differ according to the information provided about their injections. Their inadequately explained psychological arousal did not make them more susceptible to changes in affect. Maslach (1979) employed amnesia for hypnotically induced arousal. Subjects either did or did not experience unexplained arousal in the presence of a confederate who exhibited either happy or angry emotions. The results indicated that subjects with unexplained arousal reported negative emotions regardless of the confederate's mood. The intensity of the arousal and mood manipulation may therefore be important variables.
The two-factor theory of emotion has been extended by one research programme to conditions different from those of Schachter and Singer (1962). Zillmann (1978) proposed an "excitation transfer model", according to which residues from an earlier arousal state intensify a subsequent arousal state. Investigating this proposition, Zillmann, Hoyt, and Day (1974) demonstrated that when provoked subjects could attribute their arousal to immediately preceding strenuous physical exertion, they administered less shock to their provokers than when a delay was interposed that presumably reduced the cues linking their arousal to exertion and led them to mislabel it as anger provoked by the behaviour of a confederate. Bryant and Zillmann (1979) showed that misattribution is more likely when the source of the extraneous arousal is ambiguous. Any activity that increases arousal can, when combined with sufficient provocation, serve as an antecedent of aggression. Because of the anger-labelling that follows excitation transfer, an arousing condition may lead to aggression far beyond the time at which the arousal itself has dissipated. Thus, the misattribution process has implications for long-term behaviour. The labelling of an emotional state may initiate scripted sequences of behaviour: Deciding that one is angry then predisposes one to play out an aggressive script (cf. Abelson, 1981; Schank & Abelson, 1977).
Certain competing positive emotions, e.g. happiness, may actually decrease anger. Bandura (1973) argues that behaviours energised by joyful or euphoric experiences and consistent with happiness do not provoke aggression, even in angry people. Baron and Ball (1974) reported that in subjects who had been provoked to anger by another person, exposure to humorous (nonhostile) cartoons led to significantly less retaliatory aggression than in similarly provoked subjects not shown cartoons. Baron (1978) also found that humour related to sex served to render provoked subjects less aggressive than they would otherwise have been. Thus, excitation transfer may be less likely when arousal is elicited by stimuli that evoke responses that are so different from aggression that they interfere with aggression.
Research using self-reports of anger has consistently found a positive relationship between anger and aggression (Baron, 1972; Gentry, 1970; Rule & Hewitt, 1971). In addition, Megargee, Cook, and Mendelsohn (1967) have found that people who report that they control their anger tend to be less aggressive across a variety of situations than those who do not. It is generally recognised that insult and attack are more powerful and more reliable instigators of aggression than frustration (see Chapter 3). Attack increases emotional reactivity as reflected by increases in heart rate (Ferguson, Rule, & Lindsay, 1982), basal skin conductance (Shortell, Epstein, & Taylor, 1970; Taylor, 1967), and blood pressure (Ax, 1953; Funkenstein, King, & Drolette, 1954). Ferguson et al. (1982) conclude that such conditions promote an increase in anger and that this is revealed in enhanced emotional reactivity and in accompanying verbal reports of anger. However, research that has employed physiological measures of anger has not supplied consistent support for this relationship. Fehr and Stern (1970) argue that it is extremely difficult to distinguish physiologically between emotional states. As a consequence, the relationship between physiologically based measures of anger or aggression may be confounded by other types of physiological arousal, such as fear. Although anger apparently mediates hostile aggression, it may not be a sufficient condition for this response. The link between anger and aggression seems not to be invariant. A variety of other factors influence the relationship. Thus, the circumstances under which increased arousal and anger lead to aggression are determined not only by the prevailing environmental circumstances but also by prior experience and learning.
Attribution is an important cognitive process affecting aggression. The determination, for instance, of how much an action frustrates an individual depends on how the individual perceives the action (e.g. Pastore, 1952). Thus aggression is greater when the subject perceives an act as intentional rather than as accidental (Greenwell & Dengerink, 1973; Nickel, 1974). For example, subjects prevented from performing a task by a confederate who was supposed to have hearing problems were much less aggressive than a similar group who had their task intentionally impeded by a confederate with normal hearing (Burnstein & Worchel, 1962). Judgement of intentionality seems also to be related to the characteristics of the person doing harm. If a person is known to have done a lot of harm in the past, any harmful act by that person will have a strong likelihood of being judged intentional (Ferguson & Rule, 1980). Therefore, stereotypes based on membership of categories such as race, sex, and physical attractiveness also enter into judgements of intent behind harming (Duncan, 1976). The judgement of intentionality will be more likely to result in retaliation when the victim decides that the intent was malicious (Rule, 1978). On the other hand, when frustration is attributed to external factors it provokes less aggression than when it is perceived to be more under the personal control of the subjects (cf. Kulik & Brown, 1979). Zillmann and Cantor (1976) suggest that one reason why attributing the harmful acts of another person to external factors may inhibit aggression is because this type of attribution inhibits anger. People differ both in how easily they are provoked to anger and in their attributions of hostile intent, and these will be discussed under individual differences.
INDIVIDUAL DIFFERENCES AND AGGRESSIVENESS
Defining and measuring individual differences in aggressiveness, as a trait, has not always been deemed successful (cf. Edmunds & Kendrick, 1980; Megargee, 1985). Some extreme views even dismiss the whole notion of personality's relevance to the analysis of behaviour (e.g. Krasner & Ullmann, 1973; Mischel, 1968). Campbell, Bibel, and Muncer (1985) argued that aggressiveness did not have consistency across situations, but some experimental work has refuted this. Deluty (1985) found aggressiveness consistent across times and situations in a range of naturally occurring activities in boys aged between 8 and 10, but not among girls. In addition, Olweus (1979), after reviewing a large number of longitudinal studies with children, found evidence for stability in aggressive behaviour. His findings indicated that general aggressiveness among children is a good predictor of aggressive behaviour. He argues that aggressive behaviour among children "is often maintained irrespective of considerable environmental variation and in opposition to forces acting to change this same behaviour". Evidence to support this comes from a study showing that the higher the levels of aggression reported by children at one point in time, the higher their levels of aggression five years later (Botha & Mels, 1990). Olweus (1979) draws attention to emotional reactions to and cognitive appraisal of the situations as well as other tendencies that may inhibit aggressive responses. In line with this, stability in aggressiveness among children has been explained in terms of cognitive representations of their environment or "schemata" (Huesmann & Eron, 1984). These representations encoded in memory are maintained through rehearsal and can become self-perpetuating, emerging whenever conflict arises. It might be that, when interpersonal conflict occurs, there is an increased likelihood of retrieving violent "scenarios" already rehearsed for that particular occasion. Some of the disagreement concerning the consistency of aggressive behaviour can be resolved by the careful perusal of longitudinal and cross-sectional data. Moffitt (1993) has delineated two distinct categories of adolescents engaging in antisocial behaviour. The larger group only commit such acts during adolescence as a result of the gap emerging between biological and social maturity and imitation of influential antisocal role models or peers. In contrast a small group of individuals exhibit "life-course-persistent" antisocial behaviour, i.e. they engage in such behaviour at every stage of life. Research has taken two directions in this field. First, attempts have been made to identify the characteristics that predispose normal individuals towards aggressive behaviour, and second, the characteristics of people known to be exceptionally violent and accepted as psychologically disturbed have been studied.
Personality traits related to aggression
One aspect of the personality of normal subjects that plays an important role in situations with the potential for aggressive behaviour is impulsivity Hynan and Grush (1986) found a positive correlation between high levels of impulsivity and aggression toward an experimental partner among male subjects in whom a negative affective state had been induced. If impulse control is an important element for inhibition of aggression, in some cases individuals with chronic overcontrol can behave in an opposite manner. Megargee (1966) argues that these individuals can be characterised by strong defences and inhibitions against aggressing. Thus, when provocation occurs, the associated tension could summate over time until it exceeds the levels of control. Presumably these subjects are habitually passive and thus gain no experience of being mildly aggressive and relieving tension. Belief in one's own ability to control events has also been studied. Subjects who believe they can readily influence events have been named "internals" in contrast to "externals" who feel powerless (Feshbach, 1970). Internals are more likely to engage in aggressive behaviour than externals, although both may exhibit it under extreme provocation.
Two other important differences among individuals in relation to aggressiveness are irritability and emotional susceptibility. Caprara et al. (1985) define the latter characteristic as "the tendency ... to experience feelings of discomfort, helplessness, inadequacy, and vulnerability". High scores on this variable were found in aggressive subjects in comparison with controls (Caprara, Renzi, Alcini, D'Imperio, & Travaglia, 1983). The variable, however, does not interact with provocation in the same way as does irritability. Emotional susceptibility may reflect a more generalised tendency to experience negative affect, which per se can be an antecedent of aggression (Berkowitz, 1989). On the other hand, it has been suggested that irritability is a more direct moderator of aggression, leading subjects to react offensively to minimal provocation (Caprara et al., 1983). Older studies suggested that anxiety inhibited the expression of aggression. However, research discriminating different kinds of anxiety has found that this only relates to social inhibition (Taylor, 1970). Fear of social disapproval may prevent the expression of aggressive behaviour except under extreme circumstances.
The personality characteristic of hostility has received a lot of attention from health behaviour research into coronary-prone or Type A behaviour patterns. Briefly, people with this behaviour pattern exhibit three major characteristics. They are competitive, impatient, and hostile (Rosenman & Friedman, 1974). There is a growing body of research showing that people with this behaviour pattern are also more aggressive. They are more likely to exhibit hostile aggression in response to provocation (Carver & Glass, 1978) and to report more interpersonal conflict at work (Baron, 1989). Recently this "hostility syndrome" has been linked to brain serotonin systems (Williams, 1994) (see Chapter 2).
Finally, there has been some recent interest in shame-proneness. This has been distinguished from guilt as it involves negative evaluations of the self (Tangney, 1990). People who report more shame-proneness on questionnaires also report more anger and aggression (Tangney, Wagner, Fletcher, & Gramzow, 1992).
Thus, many different personality characteristics are important in the occurrence of aggression. Social learning theorists would claim that situational factors are much more important but evidence is accruing that personality may influence the learning of behaviour and shaping of overt responses (Clarke & Hoyle, 1988): People may interpret situational factors in different ways. Thus, Dodge (1993) has identified different types of processing styles in children. Severely aggressive boys are not only more likely to interpret ambiguous or accidental acts as having hostile intentions, they are also more likely to demonstrate actual errors in the interpretation of interpersonal behaviour when the cues are clear (Dodge et al., 1984). This "hostile attributional bias" is correlated with undersocialised, interpersonal, aggressive behaviour (reactive aggression) but not with nonviolent crime or socialised delinquency (Dodge, Price, Bachorowski, & Newman, 1990a).
Pathological aggression
A small proportion of people are responsible for a very large number of extremely aggressive acts and so the study of their characteristics may help us to elucidate the components of aggression. Although the term "aggression" is not specifically defined in the fourth edition of the American Psychiatry Association's Diagnostic and Statistical Manual of Mental Disorders (APA, 1994), it overlaps with certain subtypes of personality disorder and other abnormal mental conditions, e.g borderline and antisocial personality disorder. Tyrer (1988) defines personality disorder as "a persistent abnormality of personal and social functioning that is independent of mental integration". In this connection several subtypes of personality disorder are associated with aggressive behaviour in varying degrees. Violence among these patients is a stable characteristic present from an early age (Benezech, Bourgeois, & Yesavage, 1980; Kermani, 1981), not associated with emotional turmoil or any specific symptom cluster (Craig, 1982; Kermani, 1981), and resistant to change (Kermani, 1981). The history of violence usually starts before or around puberty; when it appears later it is much more likely to be due to an identifiable brain disease or psychosis (Kermani, 1981). As personality disorders in general form a heterogeneous group without clear-cut neurological or psychiatric disorder, efforts to define a specific personality disorder for aggressive patients have failed (cf. Kermani, 1981) and there has been much discussion on the use of traits rather than behaviour or categories (Stone, 1993; Widiger, 1992). However, some consistency is seen in the history of such individuals. Perinatal difficulties may play a role in predisposing individuals to aggression but only when combined with an unstable family environment (Mednick, Brennan, & Kandel, 1988). Not infrequently a history is found of parental discord, often with divorce, separation, or death, as well as parental alcoholism, (Bach-Y-Rita & Veno, 1974; Faretra, 1981). Physical abuse in childhood seems to be a major risk (Dodge, Bates, & Pettit, 1990b) and has been found to be more predictive of aggressive behaviour in young adult men than paternal alcoholism (Pollock et al., 1990). Moreover, a deficient rearing environment is much more commonly found for violent individuals with a diagnosis of personality disorder than in those with a diagnosis of psychosis (Benezech et al., 1980; Kermani, 1981). Other characterological difficulties such as poor self-image (Kermani, 1981) and depressive traits may be present (Bach-Y-Rita, 1974).
The antisocial, "psychopathic", or "sociopathic" personality is the one most frequently linked to aggression. Monahan (1981), however, has noted that in studies of violence these diagnoses are often ill-defined and may not be independent of the history of criminal behaviour. He averred that the essential feature of this disorder is a continuous history of behaviour that violates the rights of others. Aggression would be one element of a pattern of lying, stealing, truancy, fighting, substance abuse, and disrespect for authority manifest through childhood and adolescence. The more flagrant antisocial behaviour diminishes during adult life, although the deficits in interpersonal relationship skills are sustained. Such persons often perceive others as being hostile toward them. Millon (1981) has criticised the term "antisocial personality disorder", pointing out that a type of behaviour cannot be used to define a disorder of personality. He preferred the term "aggressive personality", defined by traits of hostile affectivity, social rebelliousness, vindictiveness, and disregard for danger. Blackburn (1993) has argued that psychopaths may be construed as individuals showing a more extreme coercive style, but people displaying pathological aggression are not a homogeneous group.
Two investigators have published representative studies describing extremely aggressive individuals (Megargee, 1966; Toch, 1980). Toch (1980) arranged long interviews with 77 prisoners and recent parolees in penal institutions. He described three main types of aggressive personality. The "self-indulgent compensator", the most common type, were those individuals with both a very low self-opinion and a fear that others would come to share this view. In order to protect these susceptibilities, they responded aggressively to even the slightest insult or provocation. The "self-indulger" represented those individuals who clung to the infantile view that others exist simply to satisfy their needs and wants. Violence would emerge when people did not pander to their whims. A third group, the "self-defender", was made up of those who displayed an intense fear of others attacking them and therefore attacked pre-emptively.
Individuals who exhibit chronic violence or whose violence is momentary but exceedingly brutal have also been studied by Megargee (1966). Following extensive interviews, he delineated two types of violent individuals. The first type he called "undercontrolled", individuals with a lack of control or inhibitions that normally prevent people from engaging in aggression. The second type, whom he called "overcontrolled", included men who normally inhibit aggression but commit extremely violent acts when their frustration is no longer containable. Undercontrolled offenders may have a history of many moderate acts of aggression, but overcontrolled offenders usually have been convicted of single acts of extreme violence. These people possess strong inhibitions against displaying aggression and therefore show extreme restraint in the face of repeated provocations. However, the latter mount up until the inhibitions are finally overcome and they erupt into sudden violence. These contrasting personality types have been investigated in the field of domestic violence (Subotnik, 1989). It was found that men who battered their wives could have either personality. Those classified as undercontrolled had a history of violence towards others, approved of aggression, experienced little guilt but felt impulsive and powerless. In contrast, those classified as overcontrolled had not been violent to others, disapproved of aggression, and were low in impulsivity. They indicated that the incidents had occurred after they had abused alcohol and they felt very guilty.
Blackburn (1971) investigated Megargee's hypothesis and found four instead of two groups among mentally disordered offenders. To investigate these further, he constructed a questionnaire (Special Hospitals Assessment of Personality and Socialisation (SHAPS); Blackburn, 1982). Cluster analysis on the results of the SHAPS has consistently reproduced four categories of personality deviation in the special hospitals population (Blackburn, 1986): primary psychopaths (impulsive, aggressive, hostile, extraverted, self-confident, low to average anxiety); secondary psychopaths (hostile, impulsive, aggressive, socially anxious, withdrawn, moody, low self-esteem); controlled (defensive, controlled, sociable, very low anxiety); and inhibited (shy, withdrawn, controlled, moderately anxious, low self-esteem). Blackburn and associates (Blackburn, 1993) then went on to look at the social expectations of these groups. They isolated three factors: others could be viewed as challenging, attentive, or yielding, and the groups differed in their expectations. Primary psychopaths do not expect others to be challenging but they expect them to be both attentive and yielding, whereas secondary psychopaths expect them to be challenging but not attentive or yielding. The controlled and inhibited groups did not expect challenge or attention.
Clinical features
In contrast to media accounts, the vast majority of those with severe psychiatric disorders are not violent or dangerous. However, a small subgroup may be violent both in hospital and in the community (Torrey, 1994). Violent behaviour should be understood in relation to other clinical features present in the underlying disorder. Thus, the severity of psychopathology appears to correlate with frequency of violent acts in psychiatric patients. Violent in-patients were more impaired than nonviolent ones with regard to delusions, hallucinations, inappropriate affect, and bizarre behaviour (Tardiff & Sweillam, 1980). Another study of violent schizophrenics reported that aggression was accompanied by delusions and hallucinations, and typified by unprovoked destruction of property and self-injury in addition to explosive, murderous onslaughts. These attacks were accompanied by retrograde amnesia, leaving the patients perplexed by their actions (Plananski & Johnston, 1977). Similarly, Noble and Rodger (1989) compared in-patients who committed violent assaults with matched controls during one year in the Bethlem Royal and Maudsley Hospital. A significantly higher proportion of patients with delusions and hallucinations and an increased incidence of schizophrenia was found among the patients with assaultive behaviour, confirming the results of previous similar studies (Fottrell, 1980; Pearson, Wilmot, & Podi, 1986; Tardiff, 1981). Violence has also been reported in association with paranoid delusions or ideation, whether in paranoid schizophrenia, paranoia, or paraphrenia. Benezech et al. (1980) conducted a study in a French hospital for the criminally insane and found the largest number of murderers was among the patients with paranoid delusions. Most of the violence of these patients was directed at a specific person, usually a significant individual in the patient's life. The same researchers studied patients in several other French psychiatric institutions (Addad, Benezech, Bourgeois, & Yesavage, 1981), contrasting the crimes committed by paranoid schizophrenics with those carried out by chronic undifferentiated schizophrenics. Violence occurring in the context of more disorganised psychotic behaviour (undifferentiated schizophrenics) was less focused, less planned, and often less dangerous.
The relation between violence and depression has been commented on in the literature, although the incidence of actual violence in major depression is rare. The expression of depression in violent patients varies from transient dysphoria accompanying the violent outburst to chronic depressive traits (Bach-Y-Rita, 1974; Bach-Y-Rita & Veno, 1974; Faretra, 1981; Planansky & Johnston, 1977; Tardiff & Sweillam, 1982). These chronic depressive traits are prominent in certain categories of violent patients with personality disorder and appear to be part of their characterological makeup (Bach-Y-Rita, 1974; Bach-Y-Rita & Veno, 1974; Kermani, 1981). A common biological mechanism has been advanced in the literature with regard to the role of the neurotransmitter serotonin (5-HT) and its putative dysfunction in aggressive behaviour (see Chapter 2). Impulsivity and depressed mood are also associated with 5-HT imbalance, and anxiety disorders have been linked to the same biochemical substrate (cf. van Praag et al., 1987). More practically, clinical data support the view that depressed mood, anxiety, and aggression are closely related (Apter et al., 1990; Kahn, van Praag, Wetzler, Asnis, & Barr, 1988; Kotler et al., 1993), e.g. a subtype of depressed patients show not only increased suicidality but also have high hostility and aggression ratings (Farmer, 1987) and may exhibit anger attacks (Fava, Anderson, & Rosenbaum, 1990).
Genetic factors
Genetic influences on aggression in lower animals are well recognised. Lagerspetz (1979) bred mice selectively for aggressiveness, with inherited aggressive tendencies emerging as early as the second generation. The biological basis and the phytogenetic history of antagonistic behaviours seems unquestionable (cf. Fromm, 1973). However, degrees of freedom of the organism are also inherent in its encoded genotype, only a portion of which can come to potential expression, partly as a result of internal and external environmental events. This constitutes the effective genotype, which interacts further with developmental events to determine the final phenotype (Lindzey, Loehlin, Manosevitz, & Thiessen, 1971; Vale & Vale, 1969). Fisher (1955, cited in Selmanoff & Ginsburg, 1981), for example, found that particular pedigrees of terriers had the potential for extreme aggression, depending on conditions of early rearing.
Has similar evidence been found for human beings? One initial difficulty is methodological in nature. Human reproduction cannot be controlled through selective breeding in the same way as that of lower animals. Consequently, twin studies have been the main tool in human genetic research.
Rushton, Fulker, Neale, Nias, and Eysenck (1986) presented evidence that aggressiveness is partially hereditary in humans. Five personality variables (altruism, empathy, nurturance, assertiveness, and aggressiveness) were measured in 573 pairs of monozygotic and dizygotic twins. The monozygotic group had higher correlations in each personality variable. In studies of twins reared apart, impulsivity as a personality trait has been shown to be partly heritable in both healthy (Coccaro & Bergeman, 1993) and psychiatric (Torgerson, 1984) populations. Although borderline personality disorder itself does not appear to be inherited (Torgerson, 1984), relatives of patients with this disorder are more likely to display traits of impulsivity and affective instability (Silverman et al., 1991). These studies indicate that positive and negative personality traits related to aggression may be genetically connected.
Another area of genetic studies of human aggression concerns the XYY karyotype. Although the first discovery of a man with an extra chromosome (Sandberg, Koeph, Ishinara, & Hauschka, 1961) found that he was not especially aggressive or criminal, extensive research was then undertaken in an attempt to establish links between genes, criminality, and violence. Price and Whatmore (1967) showed that aggressive behaviour that was related to an extra Y chromosome was present even among males raised in family environments of no more than average aggressiveness. Moreover, the ratio of XYY males to normals in prisons and other security institutions ranges from about 1:35 to 1:100, whereas that in the population at large has been estimated as approximately 1:550 (Court-Brown, 1968). These findings seem to implicate a defect in the nervous system that is genetically determined (Jacobs, Brunton, Melville, Brittain, & McClermont, 1965). The extra Y chromosome has also been connected with higher male sexual hormone levels, which could lead XYY males to be differentiated from average people (Selmanoff & Ginsburg, 1981). This could account for their potential aggressiveness as well as for their unusual height (Daly, 1969).
On the other hand, the correlational nature of the data allows other possible explanations. Schiavi, Theilgaard, Owen, and White (1984) found no association between XYY or XXY chromosomal abnormalities and violence outside correctional institutions. Witkin and his associates (1976) found that the XYY karyotype was related to both criminal behaviour and low intelligence, but not to overall aggressiveness. Heritability of criminality is not the same as heritability of aggressiveness. Bohman, Cloninger, Sigvardsson, and Von Knorring (1982) in a study of adopted men failed to show that their violent crimes were related to violence in their biological or adoptive parents, but rather to other causes, such as alcohol abuse. However, nonviolent petty property crimes do seem to have a genetic predisposition, a finding confirmed by a positive relationship between the conviction rate of adoptees and their biological father but not their adopted father (Mednick, Gabrielli, & Hutchings, 1987). Any association with arrest for crimes is probably linked to other factors, such as low intelligence.
Cognitive correlates
Individual variations in the level of cognitive development also influence aggression in response to both provocation and the observation of violence being carried out by others (cf. Parke & Slaby, 1983). In addition, poor cognitive development may be associated with high levels of delinquency and with a generally negative self-image (Feshbach & Price, 1984; Pitkanen-Pulkinen, 1979). However, these cognitive deficits may be small and therefore are often disregarded. They are not believed to be deterministic but rather part of a complex interaction (Moffitt, 1990). Comparing cognitive function between prisoners remanded for violent crimes and those remanded for nonviolent crimes, Robertson, Taylor, and Gunn (1987) found that although the violent group were of slightly lower general ability, this was not significant. Some research has attempted to link impaired frontal lobe function with antisocial behaviour, but the evidence supports a general relationship with the impaired regulation of complex social behaviour rather than aggression per se, and even then is not clear-cut (Kandel & Freed, 1989).
Attitudes and aggression
Attitudes or beliefs may also influence behaviour of all kinds. This is most likely to occur when external influences are minimal, the attitude is specific to the behaviour, and we are conscious or made aware of our attitudes (Myers, 1994). Attitudes about aggressive behaviour are likely to influence its expression in many differing ways. Most research in this area, however, has focused on prejudice, i.e. negative attitudes toward members of a particular social group. It has generally been found that highly prejudiced people tend to be more aggressive than less prejudiced ones regardless of the target. It has therefore been suggested that both prejudice and aggression may be products of "a hostile disposition".
Gender differences in aggression
Gender is another variable that has been implicated in aggression, although the underlying cause is disputed. Maccoby and Jacklin (1974, 1980) argue that in most societies men are generally more aggressive than women. They contend that male–female differences in aggression begin too early in life to be attributed confidently to learning and that a hormonal basis for sex differences in humans as well as other species may play a crucial role (see p.36 ). This evidence has been criticised by other investigators, who argue that the case for a genetic basis for sex differences in aggression has not been properly established (Tieger, 1980). Hyde (1984) reviewed 143 studies on sex differences in aggression, including those studies previously reviewed by Maccoby and Jacklin (1980), which involved subjects ranging in age from 2.5 to 35 years, and found smaller sex differences related to aggression. Moreover, aggression diminished with age. Eagly and Steffen (1986) in another review, with adult subjects, corroborated Hyde's findings. However, some of those who propose a genetic link to sex differences concede that a large part of male–female differences is traceable to social learning (cf. Frodi, Macaulay, & Thome, 1977). The idea that men prefer physical aggression and women favour verbal means of aggression (Bandura, 1973; Buss, 1963) is weakly supported, and women are more likely than men to experience guilt and anxiety over aggressing (e.g. Brock & Buss, 1964; Wyer, Weatherly & Terrell, 1965). Other sex differences occur at a more cognitive level. Men and women have somewhat different ways of interpreting provoking situations and may differ in aggression as a result (cf. Duncan & Hobson, 1977; Frodi et al., 1977). The two sexes also tend to react to observed violence in contrasting ways, and this can be attributed to different values placed on aggression as a way of reacting to conflict (Cantor, Zillman, & Einsiedel, 1978).
SUMMARY
In this chapter we discussed definitions of aggression and concluded that intention to harm is an important element whether this is motivated primarily by affect or by some other goal. Historically, theories of aggression have developed from instinct and drive to broader approaches incorporating emotions and cognitions and recognising the importance of learned methods of responding. Several factors increase the likelihood of aggressive behaviour and these can be divided into three broad categories: interpersonal, external, and individual differences. Aggression occurs within an interpersonal exchange and thus much research has been devoted to elements such as frustration, provocation, attack, and instigation from others. Attack is the most reliable way of inducing aggression, but even so the response can be modified by perceived intentionality or apology. Adverse environmental factors may also increase the likelihood of aggression. This may be linked to increased arousal or negative affect. Other external factors such as pain, substance abuse, and the mass media have effects on aggression. Watching violence on television is associated with aggression in children. It is likely that this effect is mediated through social learning and that children with a hostile disposition or in an adverse family environment are most at risk. Individual differences in aggressiveness are important and may be partially hereditary. Various personality characteristics seem to be important in the normal population but a small proportion of people are responsible for a very large number of extreme aggressive acts. The most prevalent diagnosis among these individuals is personality disorder. The pattern of aggressive behaviour starts early in life and may be associated with both perinatal difficulties and an unstable family environment. These individuals may also show a hostile attributional bias, i.e. they expect provocation from others and therefore perceive it in neutral situations.