General discussion and conclusions
The incidence of both physical and verbal aggression is of wide concern. In the minds of the mass media and the general public it is often associated with mental illness. However, although violence can be associated with florid symptoms of a psychiatric disorder, this is in fact rare. More commonly, severe aggression is associated with a diagnosis of personality disorder and with alcohol or substance abuse. The personality disorders most commonly associated with hostility and aggression are antisocial or borderline. These diagnoses imply a malfunctioning of learned social behaviour as well as experiencing extreme anger or responding aggressively to minimal provocation. Such behaviours or feelings are perhaps better understood in dimensional rather than categorical terms. Thus, the dimension of hostility may range from someone who is difficult to please to someone who can be extremely violent. There is evidence that traits of aggressiveness and hostility are partially inherited and correlational studies have linked traits to biological indices. The clearest relationship to be isolated so far has been that of diminished 5-HT metabolism in the central nervous system (CNS) with the dysregulation of aggression. A hostility syndrome with many biobehavioural components has been proposed (Williams, 1994). In this model, a deficient CNS 5-HT system is seen to underlie increased anger and irritability, increased sympathetic nervous system (SNS) reactivity, decreased parasympathetic nervous system activation, and increased alcohol consumption, as well as increased eating and smoking behaviour. Alcohol has a long history of association with aggression but other substance abuse has also been implicated. There is some evidence that alcohol and/or substance abuse and adult aggression have common antecedents. Thus, as well as personality traits, factors such as perinatal difficulties, family conflict, childhood neglect or abuse, and school failure are very important in the history of habitually aggressive people.
THE MEASUREMENT OF AGGRESSION
Aggressive behaviour is both multicausal and multifaceted, but in order to study it, we have to break it down into components and find ways in which to measure these. It is important to examine both what people say they do and how they behave in a particular situation. The majority of work on aggression has concentrated on questionnaire data, but laboratory tasks allow the precise measurement of aggressive responding while eliminating the possibility of actual harm. The current work then describes a specific experimental technique to measure behavioural aggression in the laboratory. We modified an already existing laboratory task both to make it ethically more acceptable and to make it more like everyday experience. Noise is now a common noxious stimulus, especially when viewed as a threat, but shock is rarely encountered.
The first question we posed was would behavioural responding on this task relate to aggressive behaviour outside the laboratory, i.e. did it demonstrate external validity? Our major finding was that it did. Forensic patients with a history of aggression responded more aggressively on the task than either healthy controls or psychiatric patients without a specific history of aggression. Thus, people who display aggressive behaviour in ordinary life also show it in the laboratory.
We examined factors that might have contributed to this behaviour, such as mood and physiological state of activation. Berkowitz (1993) has proposed that negative affect is an important influence on aggressive behaviour. However, the behaviour of the forensic patients could not be explained by negative mood alone as although the psychiatric patients showed more dysphoria, anxiety, and irritability than the healthy subjects, the forensic patients did not differ from the control patients on these measures. Nor could the behaviour be explained by physiological differences in levels of activation. Some work has suggested that habitual criminals show a history of underarousal on both central and peripheral measures from adolescence (Raine et al., 1990). However, this work examined the incidence of all serious offending. In this context aggressive behaviour is likely to have been primarily instrumental rather than affective (see p.2). Concentrating on affective aggression, Zillman (1988) has claimed that high sympathetic excitation may impair cognitive appraisal and guidance of behaviour leading to a greater likelihood of aggression. Both our patient groups exhibited higher levels of physiological activation before and during the task but there were only minor differences between groups. In fact, the forensic patients were more similar to healthy controls in their reactivity to trial events on the CRT. Thus, their levels of SNS excitation were no higher than the other groups. However, this does not preclude them from being somehow more sensitive to similar levels of dysphoric mood and both central and peripheral arousal. In contrast to patients with panic disorder, who are hypothesised to interpret such symptoms as indicating impending danger to their own physical or mental health, aggressive patients may be keen to seek an external or interpersonal cause for their negative feelings and physiological activation so that they can attribute blame, and attack.
The second question posed was would a questionnaire measure of aggression differentiate between the groups? The BDHI was selected as the best validated measure (Edmunds & Kendrick, 1980) and it was found to discriminate between them; the forensic patients admitted to exhibiting more "motor aggression" and feeling more suspicious than both other groups, confirming earlier findings with other aggressive populations (Lothstein & Jones, 1978; Renson et al., 1978). The motor aggression factor of the BDHI also correlated with trait measures of anxiety and anger. Therefore, the aggressive patients generally felt both more threatened by others and more hostile towards them. This kind of aggressive disposition is compatible with the expression of a hostile attributional bias postulated to develop early in life (Dodge, 1993; Nasby et al., 1979). Aggressive response patterns are learned and encoded in memory in childhood. These "scenarios" are rehearsed (Huesmann & Eron, 1984) and therefore readily retrievable in situations in which provocation or attack are perceived. The CRT created these conditions in the laboratory, and in combination with dysphoric mood and high levels of peripheral and central activation, led to increased aggression by the forensic group. This is in line with Berkowitz's (1993) theory that people with a hostile disposition and who have learned patterns of aggressive behaviour are more likely to be aggressive under conditions leading to increased arousal and anger.
We attempted to control the current social status of the subjects so that the groups did not differ on demographic features. The differences in aggression are therefore likely to be more deep-seated, due to hereditary or early developmental influences. The fewer years of education in the forensic group may have been due either to pre-existing interpersonal difficulties or to poor cognitive development, but in any case add to the complex interaction between hereditary and environmental factors leading to habitual aggressive behaviour.
THE PSYCHOLOGICAL EFFECTS OF ALCOHOL
As mentioned before, alcohol has long been recognised as a contributory factor in aggressive behaviour. Our groups did not differ on current alcohol intake and few had a history of alcohol or substance abuse and so these factors could not have accounted for their behaviour on the task. Nevertheless, alcohol may have played a part in past aggressive acts. Having shown that the CRT was sensitive to individual differences in aggressive disposition, we wanted to see if it would prove sensitive to the effects of alcohol in healthy volunteer subjects. Although the consumption of alcohol has been persistently linked to aggressive behaviour, such behaviour is not an automatic consequence of drinking and the actual factors involved in alcohol-induced aggression are difficult to pinpoint. One of the important psychological factors appears to be the perception of threat or provocation. Thus, an early study that examined the effects of alcohol on performance on the Buss aggression machine (see p.53) found that it did not increase aggressive responding (Bennett, Buss, & Carpenter, 1969). In contrast, studies using the CRT with increasing levels of provocation (shock delivery) have shown alcohol to consistently increase aggressive responding (Shuntich & Taylor, 1972; Taylor & Gammon, 1975; Taylor, Schmutte, Leonard, & Cranston, 1979). These results were found to be related to dose as only the high dose increased aggression, and only male subjects were used. In the first experiment reported here (Chapter 5), we confirmed that a moderate dose of alcohol did indeed increase aggressive responding compared to placebo. As in previous experiments (Taylor & Chermack, 1993), we found that intoxicated subjects initiated aggressive behaviour. Both male and female subjects on the higher dose of alcohol set moderate levels of noise for their opponents on the first trial, pre-provocation, and both continued to set these levels for the first block of trials despite receiving much lower levels themselves. Males and females then showed a different pattern. Females kept to moderate settings, whereas males continued to escalate the levels they administered. Previous studies have not examined concurrent mood. We found that not only did alcohol exert some calming, anxiolytic effect before the task when the subjects rated themselves as more calm, tranquil, and friendly than those on placebo, but the increase in aggression was not accompanied by corresponding increases in hostile or dysphoric mood and alcohol showed a stress response dampening effect on cardiac activity. It is important to note that these subjects were normal, healthy social drinkers with no history of aggressive behaviour, yet they behaved very aggressively on this laboratory task.
There are many theories relating to alcohol's effects on aggression (Gustafson, 1993). These results do not support the disinhibition model as the subjects' behaviour was not totally disinhibited. They in fact increased their aggression as the provocation increased. Neither do the results accord with the arousal model. Although physiological arousal was generally increased by alcohol, responses to particular points of stress within the experiment were in fact decreased. The expectancy model does not appear to account completely for our results. Most of our subjects believed that they had consumed some alcohol but not all groups behaved aggressively. Also, if subjects expect to behave aggressively after consuming alcohol, then they might also be expected to rate themselves as more aggressive. Our results offer most support to cognitive theories that suggest that alcohol is consumed to reduce self-awareness (Hull, 1981) or that it reduces attentional capacity, so that attention is directed to the most salient external cues and away from subtle inhibitory cues (Taylor & Leonard, 1983). A similar explanation has been termed "alcoholic myopia" by Steele and Josephs (1990) who suggest that alcohol both restricts attention to immediate situational cues and reduces the ability to process this information fully.
Despite a lack of direct evidence on some aspects, a number of researchers now feel that their results fit most clearly into this kind of explanation (e.g. Gustafson, 1993). Alcohol is postulated to impair higher-order information processing (Peterson, Rothfleisch, & Zelazco, 1990), which may in turn reduce inhibitory control and attention to inhibitory cues. Alcohol has in fact been shown to make subjects report being more externally orientated on a state locus of control scale (Gustafson, 1993). In our experiment, by reducing both negative mood and physiological reaction to negative events, alcohol rendered the subjects less aware of themselves and less responsive to situational norms of appropriate behaviour. They then concentrated on task variables such as increasing provocation. It has been postulated that this effect may be greater for groups who are habitually more likely to express aggression during the provocation stage (Taylor & Chermack, 1993). In groups in whom sanctions against aggressive behaviour are stronger, it may be more difficult to overcome inhibitory cues and this may account for the lesser effect in women (Gomberg, 1993).
THE PHARMACOLOGICAL EFFECTS OF ALCOHOL
The theories mentioned so far concentrate on the psychological effects of alcohol but attempts are now being made to link these in with its pharmacological actions (Pihl, Peterson, & Lau, 1993). Alcohol has effects on subtypes of 5-HT, NMDA, and GABA receptors (Miczek et al., 1993). It is not known what the respective roles of these actions are in heightening aggression and it is likely to be due to a complex interaction. However, the anxiolytic effect of alcohol is probably due to its action on GABA. By enhancing the inhibitory action of GABA, alcohol reduces the inhibitory effect that fear exercises on behaviour in response to threat or possible punishment (Gray, 1987). This in turn may encourage the expression of aggressive behaviour that is normally inhibited and this effect is likely to be strongest in individuals with readily available aggressive response patterns.
THE EFFECTS OF ANXIOLYTICS AND ALCOHOL
Benzodiazepines bind to receptor sites on the inhibitory GABAA-chloride complex. They maximise the normally occurring GABA-mediated inhibition and therefore exert powerful anxiolytic effects. However, whereas work with alcohol has shown consistent aggression-enhancing effects under conditions of provocation (Bushman & Cooper, 1990), work with benzodiazepines has revealed differences among them. Oxazepam, in particular, has not been shown to increase aggression (Salzman et al., 1975). We therefore set out to investigate if we could replicate these effects on the CRT. We compared two very similar compounds and confirmed that oxazepam did not increase aggression, neither dose differing from placebo. At the low dose, the chemically very similar lorazepam also showed no effects but the high dose increased aggression more than any other condition. Neither benzodiazepine exerted a stress-response dampening effect and they did not differ in their effects on mood; so increased aggression was not related to any anxiolytic effect nor to increased anger or excitation. This disproportionate effect of lorazepam is not confined to aggressive behaviour. It has been shown not only to have more general effect on memory function (Curran et al., 1987) but also to impair different functions, e.g. priming (Curran & Gorenstein, 1993; Sellal et al., 1992), and to cause more withdrawal problems (Tyrer et al., 1981). Although lorazepam has a very similar structure to oxazepam, it is 15 times more potent. Other high-potency compounds have also exhibited more problems, with adverse effects such as amnesia, disinhibition, and aggression, especially at high doses (Drug and Therapeutics Bulletin, 1991; Martinez-Cano et al., 1995). We therefore went on to examine the effects of another high-potency benzodiazepine.
Alprazolam is a newer triazolobenzodiazepine compound of high potency. It is not only used to treat generalised anxiety but in high doses also to combat panic and this has led to reports of increased hostility and aggression (Gardner & Cowdry, 1985; Pyke & Kraus, 1988; Rosenbaum et al., 1984). We found a low dose of alprazolam to have only minimal effects on aggression and hostility, but when combined with a low dose of alcohol, the picture changed dramatically. The combination increased aggressive responding more than any of the other treatments. This effect was not accompanied by any increase in reported anger or hostility, and anxiety was decreased compared to the other conditions. The combination also exerted a stress-response dampening effect on both cardiac and electrodermal activity. These effects were then similar but more powerful than the previous findings with a moderate dose of alcohol. The combination of alcohol and a high-potency benzodiazepine may therefore reduce both self-awareness and attention to inhibitory cues to a greater extent, thus facilitating the expression of aggressive behaviour. It has been shown that the aggression of intoxicated subjects can be controlled by redirecting their attention away from instigative cues towards normative, socially appropriate standards of behaviour (Jeavons & Taylor, 1985) but it is not known if this would be possible with benzodiazepine-facilitated aggression.
It has been suggested that benzodiazepines only increase aggression in people with a predisposition to hostility. However, both verbal and behavioural aggression have been reported in a proportion of patients with anxiety disorders with no previous history of such behaviour (Rosenbaum et al., 1984). Our study with patients with panic disorder confirmed these findings. These patients had no history of aggressive behaviour or increased hostility yet the patients on alprazolam showed significantly more behavioural aggression on the CRT than those taking placebo. As previously, subjects on the benzodiazepine did not exhibit more behavioural aggression pre-provocation or when they received the minimal levels of provocation. They responded aggressively to provocation but as with lorazepam, this behaviour was not accompanied by any reported increase in anger or hostility. This kind of dissociation then confirms a lack of insight or reduced self-awareness. In fact, three factors seem to be important in recognising benzodiazepine-linked aggression. First, it occurs in response to provocation; second, it is recognised by others and not complained of by the patients themselves; and third it is more likely to occur with higher doses. An additional fourth factor may be important but needs further investigation: Reports of increased hostility and aggression tend to occur more often after taking high potency compounds like alprazolam, triazolam, lorazepam, and clonazepam.
Anxiety disorders are reported to be more prevalent among females (Reich, 1986) and so benzodiazepines are more commonly prescribed for women. Most of the patients in the alprazolam study were female but this did not affect the results. In fact, we found no sex differences in any of the studies with benzodiazepines in contrast to alcohol. The lesser effect of alcohol on aggression in women has been explained by sex differences in the expression of aggression (Eron, 1987). When presented with situations designed to elicit anger (e.g. strong criticism), women are more likely to express feelings of hurt or disappointment (Brody, 1985). It may be that by decreasing these alternative feelings, benzodiazepines allow anger to be expressed, or it may be that taking a pill to lessen anxiety and reduce self-awareness legitimises aggressive behaviour.
Differences have been shown between the effects of alcohol and benzodiazepines on the CRT. Benzodiazepines did not increase aggression pre-provocation in contrast to alcohol. Healthy control subjects who are untreated or on placebo are not impaired cognitively and are thus aware of implicit norms of acceptable behaviour. They both assume that the opponent will not try to harm them on the initial trial and realise the potential consequences of provoking the opponent. They therefore set low levels of noise. During the provocation stage, as the intensity of noise given escalates, the (nonintoxicated) subjects become sensitive to instigative cues and attribute hostile intent to their opponent. They then increase their settings gradually but rarely exceed the level administered to them, in order to avoid provoking the opponent further. Subjects with a hostile disposition, like the forensic patients described here, are not excessively aggressive when no threat cues are present (pre-provocation) but respond to the perceived hostile intent of their opponent and attempt to affect their behaviour by attacking more. Subjects on benzodiazepines presumably feel quite calm and unthreatened initially and so set low intensities of noise. However, when they perceive hostile intent they ignore situational norms of appropriate behaviour and also attack. In contrast to these groups, intoxicated subjects initiate aggressive behaviour on the first trial. This behaviour is not totally disinhibitory, only moderate levels are set, but they do not reduce in response to very low levels administered to them but selectively attend to any provocation and attack. There is some evidence that healthy subjects with a hostile disposition, as measured on the BDHI, attack more under provocation when they have consumed alcohol (Bailey & Taylor, 1991). Alcohol then seems to exert most effect in people who have patterns of aggressive behaviour readily available and under conditions that minimise awareness of inhibitory social cues. The combination of alcohol and a benzodiazepine, however, results in both initiation of attack and continued escalation of aggressive behaviour. The influence of hostile disposition is less apparent, perhaps because expectancy is less important, and so sex differences are eradicated.
ANIMAL MODELS
Recent work in animals has attempted to examine the pharmacology behind alcohol's effects on aggression. Research has followed two separate strands, investigating GABA and 5-HT. It has been found that mice with high or low levels of aggressive behaviour as a result of selected breeding can be further differentiated by benzodiazepine receptor binding at the GABAA receptor complex (Miczek et al., 1993). Confirming our findings in humans, aggression induced by low doses of alcohol has been found to be potentiated by BZD agonists in mice, but in addition, BZD antagonists have been shown to be capable of reversing it. It may be, therefore, that BZD antagonists, like flumazenil, would also be able to reverse aggression induced by BZD agonists. However, there is some controversy over whether they are able to reverse fully the cognitive effects of some compounds (Kapczinski, Curran, Gray, & Lader, 1994) and the picture may be similar with aggressive behaviour.
The aggression-enhancing effects of benzodiazepines and alcohol may not be primarily mediated by GABA. Benzodiazepines have also been shown to reduce 5-HT function by enhancing dendritic release of 5-HT (Soubrie, Bias, Ferron, & Glowinski, 1983) and to decrease 5-HT turnover and release (Pratt, Jenner, & Marsden, 1985). Abnormalities of the serotonin system have been shown to relate to hostility and aggression. Most of the human work linking deficient levels of central 5-HT with aggressive behaviour has been correlational, relating biochemical indices to aggressive traits or past behaviour (see p.34) and there has been very little work examining current behaviour. Depletion of the 5-HT precursor, tryptophan, allows us to examine, by implication, the effects of depleted central 5-HT on current feelings and behaviour, but it may be necessary to both manipulate trait characteristics and provoke aggressive behaviour. Confirming this, a recent study has found a tryptophan deficient amino acid drink to increase both feelings of hostility and aggressive responding on the CRT in subjects with high scores on the BDHI (Cleare & Bond, 1995).
Animal studies have set out to examine the effects of manipulating 5-HT on alcohol-induced aggression. It has been found that tryptophan can partially block alcohol-induced aggression in mice (Wagner, Fisher, Pole, Borve, & Johnson, 1993). Serotonergic drugs such as fluoxetine and fenfluramine were also found to reduce substantially aggressive behaviour in the same series of studies. Little work has so far examined the role of 5-HT in alcohol-induced aggression in humans. However, a recent study investigated the effects of a moderate dose (0.8g/kg) of alcohol on both tryptophan concentrations and the ratio of tryptophan to the sum of the five amino acids that compete with tryptophan for the same cerebral uptake mechanism (Badawy, Morgan, Lovett, Bradley, & Thomas, 1995). Both total serum tryptophan and the ratio were significantly decreased by alcohol. Circulating tryptophan available to the brain was therefore reduced, resulting in reduced central 5-HT synthesis. The authors present their results as a possible biological explanation of the incidence of aggressive behaviour following the consumption of alcohol.
CONCLUSION
In summary, cognitive theories of aggressive behaviour emphasise the emotional and cognitive processes underlying such behaviour. In order to study these processes in more detail, we need to develop new methods of measuring interpersonal aggression in the field and in the laboratory. Aggression is primarily a social behaviour and the way in which people appraise a situation will influence how they feel and behave. Drugs that impair an individual's ability to process socially relevant information may lower the threshold at which aggressive behaviour is likely to occur. Serotonin is involved in the regulation of complex, intelligent, adaptive behaviour and is thought to be an important mediator of control over premature responding. Deficiencies in the central serotonergic system, whatever their origin, may lead to both increased hostility and an increased likelihood of responding aggressively to minimal provocation, i.e. before appropriate evaluation of all the information available. Newer, more specific serotonergic drugs may be able to reverse this by decreasing hostility and facilitating delay, thus allowing full appraisal of the situation. However, other neurotransmitter systems may also be involved. It is only by continuing to study the complex interaction between pharmacological and psychological factors that we will gain more insight into the mechanisms underlying aggressive behaviour.