Johan M.G. van der Dennen


Miller's (1941) retraction of the claim in frustration-aggression theory (Dollard et al., 1939) that frustration necessarily induces aggressive drive, which in turn necessarily results in some form of hostility or aggression, seems to have led the way in the dismissal of the concept of specific aggressive drive. In contemporary psychology, this concept - as originally employed in the frustration-aggression hypothesis - generally has been abandoned. This, however, does not mean that all characteristic features of a hypothetical aggressive drive have been discarded nor that the alternative of a nonspecific diffuse drive has been universally accepted. As the drive concept in frustration-aggression theory (and more generally, in behavior theory) was challenged, investigators became understandably reluctant to commit themselves to a nomenclature that could make their theoretical positions vulnerable. The drive concept was henceforth treated with caution. The concept was neither abandoned nor necessarily altered. The word 'drive', however, was largely shunned. If used, it was frequently placed in quotation marks, and when avoided, substitute terms were freely drawn upon. These practices unfortunately have not helped to clarify the reasoning on aggression, since the changes were more apparent than real. Many investigators maintained elements of aggressive drive in their theories, but they did so without clearly identifying them as such (Zillmann, 1979).

1. Appetitive Drive

Aggressive drive has been conceived of, essentially, as an internal state of energy that: (a) is produced by aversive stimulation, (b) compels the organism to aggressive action, and (c) is diminished or terminated after the execution of such action. The drive state is thus reactive rather than spontaneous, and it has appetitive properties that define particular appropriate consummatory acts. Thus conceived, the notion of aggressive drive has continued to be used mainly in the writings of Berkowitz (e.g., 1965a) and Feshbach (e.g., 1970).
In his reinterpretation of the frustration-aggression relationship, Berkowitz (1969a) expressed the conviction that frustration induces a motivational state that favors aggression over alternative response routes. "Basically", he said, "I believe a frustrating event increases the probability that the thwarted organism will act aggressively soon afterward, and that this relationship exists in many different animal species, including man". This statement, in fact, largely reverses Miller's (1941) retraction of the claim that frustration always fosters aggression. Apparently, frustration is viewed as making aggression the likely reaction most of the time. However, more importantly here, the statement attributes appetitive properties to the frustration-induced state - 'drive', 'force', 'anger', or whatever label is used to denote it. In Berkowitz's (1955a, 1969a, 1970) conceptualization, then, aggressive drive predisposes the organism to aggression. It constitutes a state of aggression readiness, sensitizing the organism to releasing cues that are said to be necessary for aggression actually to occur (Zillmann, 1979).
With regard to consummatory aggressive reactions, Berkowitz has taken an extreme stand. He has posited a 'completion tendency' (1965a) for aggressive behavior. Specifically, he proposed that for the individual in an acute state of aggression readiness, the infliction of injury upon the tormertor is the ultimate goal reaction and that only this goal reaction can terminate the state of aggression readiness. He furthermore suggested that such a completion tendency exists even though the response readiness may be purely autistic. Once the organism is "primed to aggress" and "aggressive responses are set into motion, even if only implicitly in the person's thoughts", he wrote, "then an individual will not attain completion until the goal object has been aggressively injured" (Berkowitz, 1965a). Bandura (1973a) has taken issue with this proposal as far as the autistic function is concerned, and the interested reader is referred to his discussion of it. Here, we continue to concentrate on the completion tendency as such.
In contrast to frustration-aggression theory, which permits the substitution of targets, Berkowitz's argument implies that in order to terminate an aucute state of aggressive drive, aggression need be directed against the appropriate target. Berkowitz hastened to point out that because of man's conceptual abilities, the injury need not be inflicted personally by the instigated individual and that it need not be inflicted in exactly the way in which it was contemplated. In other words, aggressive agents and means may be substituted, but targets and intended injuries may not.
This argument is of critical theoretical consequence. Most importantly, it categorically denies catharsis through displaced aggression. Berkowitz's reasoning, generally speaking, adheres very closely to frustration-aggression theory, but with regard to catharsis, it obviously does not. Only aggression against the tormentor is assumed to function as drive-reducing consummatory behavior. As a consequence, aggression cannot be displaced in the sense of taking the place of another, orginally motivated attack. An alternative target that draws an attack because - according to Berkowitz's cue model - it is associated with aggression-triggering stimuli thus cannot substitute for the original tormentor. Attacks upon alternative targets, then, simply constitute excess violence motivated by the readiness to aggress against a specific tormentor (Zillmann, 1979). In this connection, Berkowitz (1965a, 1973b) has developed an interesting rationale that predicts that if anything, attacks upon alternative targets will heighten the probability of an assault upon the actual tormentor. He proposed that since the infliction of injury on the tormentor constitutes the ultimate goal reaction for the aggressively instigated individual, failure to hurt the intended victim is frustrating in itself. This frustration adds to the torment suffered initially, thus increasing the individual's aggression readiness.
Although Berkowitz does not suggest this, his argument can be applied recursively: Every increment in aggression readiness makes the failure to inflict injury all the more frustrating, adding further to aggression readiness, and so forth indefinitely. Since the proposal fails to indicate a mechanism by which this drive escalation can be disrupted, aggression against the tormentor seems largely to be inevitable and only a matter of time. What starts as an 'aggressive thought' must ultimately become overt violence - a challenging, controversial idea indeed (cf. Bandura, 1973a).
Attacks upon alternative targets relate to this escalation model in two ways. First, the emphasis on goal achievement rather than goal response - that is, on hurting rather than hitting - led Berkowitz (e.g. 1973b) to propose that especially for the acutely instigated person, the infliction of injury generally reinforces aggression. Attacks upon alternative targets thus make further attacks, including those directed toward the original target, more likely.
Second, attacks upon alternative targets are viewed as lowering 'aggression anxiety', and this anxiety reduction leads to the same expectations. Finally, Berkowitz (1970, 1973b) suggested that the desire to hurt motivates the aggressively instigated person to enjoy witnessing others inflict injury upon others and that mere exposure to violent activities reinforces aggression. The angry person is said to develop an appetite for injurious activities, and the observation of such activities is considered aggression enhancing because it elicits 'aggression-related anticipatory goal responses' (Berkowitz, 1973b). In other words, witnessing aggression reminds the angry person of his or her desire to hurt the tormentor, and this reminder further instigates the desire to hurt that tormentor.
In summary, Berkowitz projects a highly specific aggressive drive that, once set in motion, is strengthened by additional frustrations, by attacks made upon irrelevant targets, and by exposure to violent actions. Further, the drive dissipates only after the object toward which it is directed has been harmed or injured (Zillmann, 1979).
Feshbach (1955, 1961, 1964, 1970), an outspoken proponent of the notion of aggressive drive, has arrived at a substantially different proposal. In agreement with Berkowitz, he views frustration as the major source of aggressive instigation. His view of the drive-motivated goal activities is also very similar to that of Berkowitz. The infliction of pain and injury upon the tormentor is seen to terminate the drive state. Additionally, the mere observation of the infliction of this pain and injury is seen to be reinforcing for the individual in whom aggressive drive is activated (Feshbach, 1970; Feshbach, Stiles & Bitter, 1967). In sharp disagreement with Berkowitz's position, however, Feshbach (e.g. 1970) proposed that any activity, overt or covert, that 'functionally relates' to the goal behavior constitutes a substitute act and that the performance of substitute acts diminishes aggressive drive. The consequence of this diminution of aggressive drive is, of course, a reduced likelihood of aggression against the original tormentor and against all other targets. The important aspect of this cathartic process is that it is best attained by aggressive activities. Engaging in hostile fantasies, deliberately anticipating aggressive acts, witnessing violence, assaulting an inanimate object, and hurting a defenseless victim - all these activities are expected to reduce aggressive drive because they relate to the thwarted goal reaction more closely than do nonaggressive activities. In the Lewinian nomenclature (e.g. Lewin, 1951) preferred by Feshbach (e.g. 1970), only aggressive activities have 'substitute valence' for goal responses the aggressively instigated person is motivated to perform, and consequently, only such activities can absorb and reduce specific aggressive drive (Zillmann, 1979).

2. Diffuse Drive

The principal weakness of drive theories of aggression is their inability to account for the fact that (a) aggressive drive can facilitate nonaggressive activities (cf. Bandura, 1973a), and (b) nonaggressive drive can facilitate aggressive behavior (cf. Zillmann, 1978). Aggressive drive, it seems, can readily be discharged through nonaggressive actions. Moreover, aggressive actions apparently feed not only on aggressive drive but on other sources of drive as well. Under these cicumstances, as Bandura (1973a) pointed out, the involvement of an aggressive drive in theories of aggression has created more conceptual problems than it has resolved. To deal with recent findings, it apparently has become necessary to abandon the assumed one-to-one correspondence between aggressive instigation, aggressive drive, and aggressive behavior.
In an effort to accomodate findings that show that frustrations potentially facilitate a multitude of behaviors - nonaggressive as well as aggressive activities - Feshbach (1970) conceptually separated anger from aggressive drive. Anger is considered to have drive properties "in the sense of an energizer of ongoing behavior", and it is made clear that this response-energizing function is not specific to aggression. Feshbach thus views anger as a diffuse drive, whereas aggressive drive is seen as specific. Such a dual system helps to cope with findings that cannot be accounted for in terms of aggressive drive, but it does so only in a postmortem analysis. For explanatory purposes, the differentiation between anger and aggressive drive is too ambiguous to permit the formulation of clear-cut predictions (Zillmann, 1979).
Most contemporary investigators of aggression have shown more willingness than Feshbach to abandon the notion of specific aggressive drive altogether. The major impetus for their readiness to accept the alternative concept of drive as a diffuse, universal energizer seems to have come from behavior theory. Hull (1943, 1952) had proposed a generalized drive, that is, a nonspecific, undifferentiated drive state that integrates components of drive from various sources. In his view, the strengths of simultaneously active drives combine into an effective drive state. The entire accumulated force of this state then energizes the behavior that is prepotent in the habit structure. Thus, there is no one-to-one linkage between a particular drive and associated behaviors. Dependent on prevailing stimulus conditions that control habit, elements of drive a can facilitate behavior associated with drive b, and vice versa. In principle, any behavior can be facilitated by irrelevant drive - that is, by energy that in the past has not been connected with the behavior it is energizing.
The application of these propositions to aggression is not as straightforward as it may appear. Hull dealt primarily with deprivation-based drive states such as hunger and thirst. These states are associated with well-defined consummatory reactions. The situation is obviously very different for aggression. Aggression is not based on deprivation; nor are there unquestionable, highly specific, and ultimately drive-reducing associated activities. These differences, which are all too easily overlooked, have been acknowledged by Brown & Farber (1951), who undertook the task of applying Hullian theory to aggression. Specifically, Brown & Farber conceived of emotions as intervening variables, and they dealt with frustrations in this context. They posited that "frustration-produced drive has the functional status of an irrelevant drive". This amounts to saying that frustration induces a drive state that, in the absence of preestablished drive-specific habits, potentially energizes aggressive and nonaggressive behaviors alike. In short, conceiving of frustration as an irrelevant drive removes all elements of aggression specificity from frustration, making it a source of purely diffuse drive.
For some reason, Brown & Farber's efforts have attracted little attention. Investigators have drawn directly on Hull's model, ignoring the problems deriving from differences in the status of drive. Hull's notion of generalized drive was simply transformed into the proposal that drive (or arousal) will facilitate the behavior that holds a prime position in the response hierarchy (e.g., Geen & O'Neal, 1969) or the proposal that arousal (or drive) enhances any and every behavior a person comes to perform (e.g., Tannenbaum, 1972). Such formulations remove the burden for the causation of aggression from drive and place it on habit. Under given circumstances, established aggressive habits are activated, and the strength of drive present at that time comes to energize the activated habit. Consequently, drive is not viewed as pushing toward a specific behavior, but once a specific behavior is determined - mainly by environmental stimuli that trigger a learned reaction - this behavior is seen as being powered by prevailing drive. This position on drive has been concisely expressed by Hebb (1955), who treated generalized drive and arousal as synonymous: Drive "is an energizer, but not a guide; an engine but not a steering gear". This view contrasts sharply with Miller's (1959) conception, which, as will be remembered, assigned strong stimulus properties to a drive state.
The view that prevalent drive diffusely energizes any and every behavior performed has a significant corrolary that is often overlooked: The degree to which behavior will be energized is a monotonic function of the strength of diffuse drive. In the proposed additive integration of various sources of drive, the specificity of these sources is entirely lost. The notion of partial energization, due to incompatible elements of drive, would be contradictory to the very notion of a generalized drive. The full force of drive must thus come to play. With regard to aggression, the theoretical expectation is not ambiguous, then: Diffuse drive operates as a unitary force, intensifying - in proportion to the magnitude of drive or arousal prevailing at that time - any aggressive reaction performed (Zillmann, 1979).

3. Drive Reduction

The principal features of drive theory are, of course: (a) that increases in drive level increase the motivation to seek out conditions under which consummatory activities can be performed and enhance the motivation to perform such activities; and (b) that the performance of these consummatory activities effects a reduction in drive level and consequently decreases the associated motivation. In the hungry animal, for example, the motivation to seek and consume food is assumed to increase with drive, which in turn is assumed to increase with time of food deprivation. Food consumption, that is, the performance of the consummatory response, is considered to lower drive and consequently to relax the animal's foodgathering activity until drive levels rise again, and so forth. Whatever the merits of employing the drive concept in this context, the analogy with aggression must appear forced. As has been pointed out earlier, and notwithstanding Lorenz's convictions to the contrary, aggression is not based on deprivation. Because it is not, consummatory actions that would eliminate a deficiency, and that would thereby necessarily reduce the motivation to aggress, cannot be specified. To consider the infliction of pain or injury upon an antagonist a consummatory, drive-absorbing act thus means overextending a model designed to explain behavior that differs fundamentally from aggression. Nonetheless, presumably because the drive model proved to be so very plausible for behaviors that are motivated by bodily changes resulting from deprivation, the analogy of specific drive in aggression seems to have been broadly accepted. People tend to consider it a truism (a) that the aggressively instigated person is out to hurt his or her tormentor, and (b) that ultimately, only a successful attack will make that person feel better. As discussed earlier, Berkowitz (e.g., 1965a, 1969a) has formally expressed this popular view.
In terms of drive reduction, this view is extreme in suggesting that once aggressive drive is activated, it will remain active until the specific goal to be achieved by the goal reaction has been accomplished, that is, until injury has been inflicted. In fact, drive must be assumed to be maintained at its original level until injury has been inflicted.
Frustration-aggression theory, as will be remembered, has promoted a different view. Dollard et al. (1939) held that the expression of any act of aggression is drive reducing in that it lowers the instigation to all other acts of aggression. This makes any hostile or aggressive act, no matter what target it is directed against, a partial or possibly total substitute for the initially motivated aggressive goal response. The aggressive goal response is partially deprived of its associated drive and hence is less likely to occur. This drive reduction through substitute hostility or aggression is of course the thrust of the catharsis proposal (Zillmann, 1979).
Feshbach (1970) has more narrowly defined the conditions under which a reduction of specific aggressive drive can occur. He stipulated that there "needs to be a functional relationship between the goal behavior and the substitute act", but he failed to specify what exactly is meant by this functional relationship. Feshbach suggested that a high degree of similarity between the substitute act and the goal response for which it serves as a substitute would ensure drive reduction, but the basis for similarity was left unclear. Presumably, a high degree of similarity in the patterns of motor activity associated with the goal response and its substitute would meet the stipulation. But would incidental stimulus similarities between the original and the substitute target also qualify?
In the light of such conceptual and ultimately operational difficulties with the variables upon which the very prediction ef catharsis is based, one might expect that the explanatory power of this notion would be considered limited. This is not the case, however. In Feshbach's reasoning, the discharge of aggressive drive through substitute activities is treated largely as an undeniable truism. This becomes quite clear in Feshbach's recent discussion of the research evidence pertaining to catharsis theory.
Feshbach (1970) proposed that substitute aggression may have several independent effects that together determine the level of subsequent aggressiveness. First, such aggression should reduce the level of aggressive drive and, in turn, aggressiveness. At the same time, however: (a) it may lower the strength of inhibitions upon later aggression; (b) it may be reinforcing through the attainment of feedback relating to the motivated goal response (e.g., the substitute attack may produce the victim's expression of pain, which is considered reinforcing because it is part of the originally intended goal reaction); and (c) it may be reinforcing because of its instrumental value (e.g., it may prove to yield incentives). A finding in which the performance of aggression is followed by reduced aggressiveness can thus be viewed as a case in which a reduction in aggressive drive dominated the aggression-enhancing factors. A finding in which the performance of aggressive acts fails to alter subsequent aggressiveness can be seen as a case in which the effect of drive discharge was counteracted and neutralized by the opposing effects indicated. Last but not least, a finding in which aggression is followed by increased aggressiveness can be interpreted as a case in which the aggression-enhancing forces dominated the impact of drive drainage. Clearly, no matter what effect is observed, it can be accounted for post facto by such a model. Even in the face of what would appear to be counterevidence, a claim can be made that catharsis occurred. Within Feshbach's model, catharsis theory eludes the possibility of falsification. As long as the effect on aggression of the various presumed subprocesses cannot be isolated and assessed independently, the catharsis hypothesis cannot be subjected to decisive empirical testing (cf. Bandura, 1973a).
We now turn to the consideration of the energization of aggressive behavior by diffuse drive and its implications for drive reduction. It will be remembered that the concept of generalized drive is based on the assumption that elements of relevant and irrelevant drive inseparably combine into an effective drive that diffusely energizes behavior associated with a prime location in a hypothetical hierarchy of responses. In this theoretical framework, a reduction in drive level is considered to come about with the performance of energy-absorbing activities. It is secondary here that in behavior theory, such a reduction in drive is said to reinforce the behavior that precipitated the drive discharge. What is important is the view that a reduction in drive level can be brought about by any energy-absorbing activity. It is of little moment whether or not the energy-absorbing behavior is aggressive. Aggressive and nonaggressive activities alike effect a reduction in drive as a function of energy discharge. As a consequence, aggressive behavior that consumes little energy should have little if any impact on subsequent aggressiveness, and nonaggressive acts that absorb considerable energy should reduce subsequent aggressiveness.
A major factor in the discharge of energy through behavior is the involvement of skeletal-motor activity. Generally speaking, the amount of energy discharged increases with the number of skeletal-motor structures involved in a response sequence, with the vigor of the responses in the sequence, and with the length of time the response sequence is performed. The reduction of diffuse drive should thus be more pronounced, the more strenuous the behavior performed (Zillmann, 1979).
This reasoning has interesting implications for aggression. As already indicated, it leads to expectations that sharply contrast with those deriving from the notion of a specific aggressive drive. First, the infliction of injury does not necessarily reduce drive level. Pulling a trigger, for example, consumes so little energy that it can hardly be considered to effect a notable discharge of diffuse drive. Consequently, shooting someone should fail to lower drive level by very much, and if aggressive responses remain in a prime location in the response hierarchy, diffuse drive should be carried along to energize such responses. Also, no matter how much harm is inflicted by hostile activities, these activities should generally not reduce drive notably. Dropping a devastating remark, for example, is simply not very strenuous. A vigorous aggressive struggle, in contrast, should constitute an optimal condition for drive discharge, whether or not it leads to the infliction of injury. Second, and of equal importance, the performance of vigorous nonaggressive activities must be expected to reduce the level of diffuse drive, thus depriving subsequent behavior of much of the irrelevant drive that could have energized it. This expectation relates to the suggestion by Freud (1915) and Hartmann, Kris & Loewenstein (1949) that intense skeletal-motor activities can purge the organism of aggressive urges. It seems also to relate, although more loosely, to Lorenz's (1963) conjecture that participation in sports can help curb violence by draining off a hypothetical reservoir of spontaneously increasing aggressive energy.
In summary, then, models of aggression that explicitly or implicitly endorse the notion of a specific aggressive drive tend to emphasize the importance of goal attainment (the infliction of injury) in the reduction of drive. Drive reduction has been expected, in principle: (a) with the actual goal attainment, (b) with the performance of the goal reaction against a substitute target, and (c) with the performance of a modified, yet inherently similar, goal reaction against a substitute target. The performance of responses that fail to inflict harm or injury of some sort is considered to leave the level of aggressive drive unaltered. Aggressive drive is thus assumed to persist at given levels for indefinitely long periods of time. Models that embrace the notion of a diffuse, generalized drive, by contrast, stress the energetic aspects of behavior. Drive reduction is expected to follow both aggressive and nonaggressive activities that are energy consuming. Since diffuse drive is unrelated to specific goal reactions, the infliction of harm or injury in and of itself is not considered to effect drive discharge. Aggressive goal reactions reduce drive level only to the extent that they are precipitated by and directly involve strenuous activities. Finally, since diffuse drive is viewed as varying freely with energy expenditures, drive levels are generally not considered stable for any length of time (Zillmann, 1979).

4. From Drive to Arousal

In behavior theory (e.g., Hull, 1943, 1952; Spence, 1956, 1960), drive has, strictly speaking, the status of a hypothetical construct. Hull (1943) conceived of drive as an intervening variable that is 'never directly observable'. On many occasions, however, Hull suggested a physical embodiment of drive, although he was relatively imprecise on the matter. He suggested, for example, the following: "Most, if not all, primary needs appear to generate and throw into the blood stream more or less characteristic chemical substances, or else to withdraw a characteristic substance. These substances (or their absence) have a selective physiological effect on more or less restricted and characteristic portions of the body... which serves to activate resident receptors. This receptor activation constitutes the drive stimulus" (Hull, 1943).
Suggestions of this kind tended to equate drive with activity in the autonomic nervous system. The assessment of drive in terms of such activity has remained an unresolved issue, however (cf. Cofer & Appley, 1964).
With the advent of activation theory (e.g., Duffy, 1957, 1962; Lindsley, 1951, 1957; Malmo, 1959), the emphasis changed to activities in the brainstem reticular formation. The ascending reticular activating system had been shown to project diffusely to thalamic, hypothalamic, and cortical regions (e.g. Magoun, 1954; Moruzzi, 1964; Moruzzi & Magoun, 1949). The fact of essentially diffuse projection, which has survived later modifications and refinements in the delineation of the specific structures involved (cf. Thompson, 1967), was interpreted as inconsistent with the assumption of specific, motivating drives. Activation was conceived of as a nonspecific, behavior-energizing force that could be measured through the electroencephalogram. Most importantly, activation was treated as a measurable unitary variable, ranging from death through coma, deep sleep, light sleep, drowsiness, relaxed wakefulness, and alert attentiveness to strong, excited emotions (e.g., Lindsley, 1957). All these levels of activation were shown to be associated with characteristic wave patterns and rhythms in the encephalogram. The concept of activation was readily incorporated in theories of motivation. Brown (1961) accepted the possibility of assessing drive through arousal in the reticular formation, and many others did likewise (e.g., Berlyne, 1960; Bindra, 1959). It should be clear, however, that drive was not simply equated with such arousal. Rather, arousal in the reticular system was treated as a reliable, convenient index of activation at large. It was considered to reflect the degree of excitation of all organismic structures: "the extent to which the organism as a whole is activated or aroused" (Duffy, 1934). Arousal in the reticular formation, although it was considered a superior index of the level of motivation (e.g., Malmo, 1965), was thus assumed to correlate highly with arousal in other structures (cf. Duffy, 1957).
The changing status of drive, from a construct to a measurable entity, is similarly evident in aggression theory. The explicit use of the drive concept has become comparatively rare. The concept of arousal, on the other hand, seems to have become successively more popular. In one way or another, all contemporary theories of aggression try to explain the phenomenon in terms of an interaction of cognition and arousal. The student of this topic may thus readily come to the conclusion that the theories are very similar, at least as far as arousal is concerned. Such an impression is quite erroneous, however. Confusion arises from the fact that the concept of arousal is used very broadly and assumes different meanings in different theories - occasionally even in the same theory.
In Berkowitz's reasoning, for example, arousal designates both a state of acute physiological excitation (e.g., 1970) and an energizing force that apparently only loosely corresponds with excitation (e.g., 1965a). The latter use of the arousal concept plays a part in the proposal of a completion tendency in aggression (Zillmann, 1979).
In his social learning theory of aggression, Bandura (1973a) proposed that aversive treatments produce "a general state of emotional arousal that can facilitate a variety of behaviors, depending on the types of responses the person has learned for coping with their stress and their relative effectiveness". This proposal is entirely consistent with Hullian behavior theory, as Rule & Nesdale (1976a) have pointed out. Emotional arousal is conceived of as a universal, diffuse energizer of behavior controlled by established stimulus-response connections. Concerning this concept of emotional arousal, Bandura has not been very specific, however. He has acknowledged the response-potentiating properties of acute arousal without specifying exactly what comprises such arousal or how it exerts its behavior-modifying influence. Regarding the latter, he has not developed a mechanism by which arousal can facilitate, as proposed, nearly every kind of behavior. The specific operation of arousal, then, remains unclear. In particular, it is difficult to see how a high-activation state can facilitate a low-activation response to provocation. How can, for example, acute arousal 'energize' withdrawal, resignation, or apathy - all these reactions to frustration being used as illustrations by Bandura?
Bandura's analysis of arousal as a facilitating force in aggression has several distinguishing characteristics. First, in sharp contrast to other positions in which the treatment of arousal largely parallels that of drive, Bandura stressed the short-lived nature of acute arousal. He directed attention to the noncathartic dissipation of arousal and to factors that influence and control this dissipation. Second, Bandura felt it necessary to qualify to some extent the notion of arousal as a universal energizer. Whereas arousal from adverse experiences is expected to energize socially constructive behaviors as well as hostile or aggressive activities, arousal deriving from joyful or rewarding experiences is expected to inhibit rather than facilitate hostility and aggression. This expectation, which severely restricts the generality of arousal as an energizer, is based on the contention that euphoric arousal is "incompatible with hostile actions" (1973a; italics added).
In Bandura's view, then, only arousal deriving from adverse experiences can function as the universal energizer referred to in behavior theory. Arousal from pleasant experiences may facilitate similar experiences, but it is seen as incapable of combining with noxious arousal in the facilitation of aggression. In short, the impact of noxious arousal crosses hedonic lines; the impact of pleasant arousal does not. Finally and maybe most importantly, Bandura views the role of arousal in aggression as entirely secondary. He feels that aggression is almost totally controlled by reinforcement contingencies, independent of the arousal prevailing at the time of instigation or at the time of performance of aggression. "By arranging social learning determinants such that arousal-linked aggression is negatively sanctioned but aggression without arousal is well received, one could undoubtedly reverse the relationship between physiological state and action" (Bandura, 1973a).
Arousal is thus merely a cue that may coincide with aggressive inclinations. If, as presumably is the case in most if not all human cultures, arousal facilitates aggression, it does so only because at high levels of arousal, aggression has tended to have greater reward value. If we were consistently rewarded for aggressing when calm and punished for aggressing when aroused, we should come to be greatly inhibited by the arousal that prevails when we are aggressively provoked. It should be clear at this point that Bandura considers the relationship between arousal and aggression to be entirely learned (Zillmann, 1979).
Notwithstanding Bandura's approach to arousal, in recent years investigations of the function of arousal in aggression have shown a strong tendency to assess arousal rather than assume it to exist at particular levels. Arousal has been measured in motor activity (e.g., Gallup & Altomari, 1969; O'Neal, McDonald, Hori & McClinton, 1978), in task performance (e.g., Burgess & Hokanson, 1964; Doob & Kirshenbaum, 1973), and in various physiological reactions - mainly cardiovascular changes (e.g., Doob & Kirshenbaum, 1973; Hokanson & Burgess, 1962b). It appears that this tendency will persist and ultimately lead to a compelling separation of the concept of arousal from the concept of drive as a force that remains elevated after activation has dissipated.
In recent research on aggression, then, the concept of drive, especially aggressive drive, has been replaced by the concept of arousal. At times, arousal is used as a synonym for aggressive drive; at others, for generalized drive. More frequently, however, it is employed in the sense of an acute state of physiological excitation with diffuse, energizing properties. The arousal-aggression relationship has been considered by some to be partly inherent and partly learned and by others to be entirely learned. Arousal has been operationalized in a variety of ways, and with increasing frequency it has been subjected to measurement (Zillmann, 1979).

Aggression as Subsidiary Instinct(s)

An innate ever-accumulating aggressive energy has its dysfunctional side. It is like having a self-destruction device built into the organism (Scherer, Abeles & Fischer, 1975), like a lysosome into the cytoplasm. Unless it is properly dealt with, it constantly threatens to explode. How is this obviously dysfunctional aspect neutralized before it destroys the species?
Lorenz (1966) offers a rather dialectical explanation for this probleme He argues that the "two great constructors of evolution" (mutation and selection) solve the problem in the following manner: "... the generally useful, indispensable drive remains unaltered, but for the particular case in which it might prove harmful, a very special inhibitive mechanism is constructed ad hoc". He points to a large number of examples of this principle in the fighting behavior of animals, where dominance struggles between rivals are often settled by the most impressive threat display rather than a fight-to-the-death. If fighting does occur, it is often a rather harmless nature, like a tug-of-war.
Lorenz claims that there is a direct relationship between the effectiveness of the weapons of a species and the reliability and strength of such inhibiting mechanisms. In short, when dysfunctional consequences are likely to result from the generally functional aggressive drive, special inhibitory mechanisms act to 'turn off' the aggression.
In humans, however, the invention of artificial weapons destroyed the balance between aggression and its innate inhibitors. Lorenz believes that weapons, particularly long-distance (e.g. guns) or remote-control weapons (e.g. guided missiles) "screen the killer against the stimulus situation which would otherwise activate his killing inhibitions".
Of course this hypothesis is highly speculative and untestable without a time-machine. It is certainly possible and even likely that seeing the consequences of one's aggressive acts, such as the victim's pain, may inhibit further aggression. But it is also likely that they would encourage greater aggression by informing the assailant of his attack's success.
In addition, it is hard to justify the weight that Lorenz places upon the invention of weapons in explaining human aggression. The invention of weapons was probably accompanied, if not preceded, by many other dramatic changes in human evolution, particularly the development of language and cognition. These capacities enabled human beings to transcend the here-and-now, to plan, to expect, and to predict. Such abilities may have been much more influential in determining the nature and frequency of human aggression in the course of human evolution and may be primarily responsible for the differences between human and animal aggression.
In terms of scientific theorizing, it does not seem very parsimonous to postulate a particular mechanism just because there are corrective factors (e.g. inhibitions) that keep it from going wrong. For example, we do not justify our car's acceleration by the fact that there are brakes. Inhibitory mechanisms would be necessary even if we assumed that aggressive behavior is motivated not by one instinct of combat, but by several subinstincts of aggression.
In this view, proposed by Scherer, Abeles & Fischer (1975), aggression is not a general instinct in its own right, but rather a part of more general instincts such as reproduction, feeding, or defense. In this respect, it is comparable to a subroutine in a computer program, which may be called from several other main routines.
It seems more reasonable to assume aggressive subinstincts in the service of important species-preserving instincts than to posit a general aggressive instinct, whose consequences would be dysfunctional during much of the individual's life span. The notion of aggressive subinstincts would also answer two important questions: (1) Why is aggressive behavior, in many animals, more frequent during certain seasons (e.g. the mating season) than in others, and (2) Where does the energy or excitatory potential originate?
In an authoritative survey of animal behavior, Hinde (1966; 1970) pointed out that territorial fighting (which is often used as evidence for a single dynamic instinct of aggression) occurs in many species mainly during the mating season. A number of observations show that cyclical physiological processes cause the gonads to release sex hormones into the bloodstream during that season, which, in turn, have a strong influence on aggressive behavior.
Again, it seams reasonable to have an extra supply of aggressive energy available at a time when fights for sexual partners and territory are likely. The existence of such a well-timed and directed increase in aggression as part of the reproductive instinct makes us even less likely to accept the idea of a single, always active, instinct of aggression (Scherer, Abeles & Fischer, 1975).
"The rarity of seeking for fights in animals is not altogether unexpected. Fighting is basically a means of competing more effectively for any commodity in short supply - food, water, nesting sites, mates, or space. Unless something is gained, fighting is at best a waste of time. At worst it engenders the possibility of distraction from other dangers, or of injury, or of death. An endogenous tendency to seek out fights would thus have hazardous consequences" (Marler & Hamilton, 1966).
"[I]f we accept that a process of natural selection can act on a certain form of aggression that serves - for each of the individuals displaying it - a certain purpose in particular circumstances, it is difficult to see how evolution could have given birth to an 'all-purpose' aggressive drive (since it is supposed to express itself in a variety of forms of aggression which serve a multitude of purposes under very varied circumstances, and since, when no fellow-creatures are within reach, some animals are reduced 'to discharge their anger on other objects'). An aggressive drive of this kind rather looks like a real biological nonsense, liable to lead the animal kingdom rapidly to its disappearance" (Karli, 1991).
"In short, there is no evidence that a widespread unitary aggressive instinct exists" (Wilson, 1978).
"In the rut, red deer stags engage in escalated fights involving roaring contests, parallel walking, and head clashes; but how far any fight goes, who initiates the first direct attack, and who eventually gives way, are all dictated by what one contestant does in response to the moves that are being made by the other (Clutton-Brock & Albon, 1979). Just imagine the fate of some poor psychohydraulic stag who, upon entering such a contest in his first season as a mature animal, was unfortunate enough to be burdened with a bucketful of aggression (i.e., a full Lorenzian aggression reservoir). Responding massively to the slightest provovation from older and stronger animals all around him, such a stag would almost certainly suffer severe injury in his very first contest; and his chances of living to fight another day, let alone of ever having any offspring, would be slim indeed. This, presumably, is one very good reason why Lorenzian aggression reservoirs have never evolved" (Klama, 1988).
Hinde (1966; 1970) has further questioned the emphasis that Lorenz puts on the internal motivation to aggress. Specifically, he attacks the notion of the damming up of aggressive energy, which leads to threshold lowering and vacuum activities. Hinde stresses the importance of social release stimuli and argues that the influence of the internal condition of an organism, such as the hormonal balance lies not so much in its effects on the intensity of behavior, but on the evaluation of external stimuli.

Is Aggression Reinforcing?

Rather than a means to an end, it is sometimes argued that aggression is an end in itself and serves as its own reinforcement. A number of experiments have been carried out with animals to see if killing can be considered a goal in itself rather than instrumental behavior for some other goal. This is generally done by observing whether a response will be learned which provides an opportunity to attack and kill (R.N. Johnson, 1972).
Myer (1964) found that some rats, if given the opportunity, will continue to attack and kill mice even though they are given no conventional reinforcement for doing so. If placed in a T-maze with mice in one goal box and baby rats in the other, mouse-killing rats will choose the arm of the maze where they can attack a mouse, while non-mouse-killers prefer the company of babyrats (Myer & White, 1955). The opportunity to attack a mouse can also be used to teach rats a spatial discrimination and a reversal (Tellegen, Horn & Legrand, 1969). If monkeys are given a painful tail shock, they will learn to perform a response which gives them access to a canvas ball they can attack (Azrin, Hutchinson & McLaughlin, 1965). Rats given inescapable electric shock will choose the arm of a T-maze containing another rat with which they can fight (Dreyer & Church, 1970), and cats under the influence of hypothalamic stimulation will seek out a rat to attack (Roberts & Kiess, 1954). Fighting fish (Thompson, 1963) and fighting cocks (Thompson, 1964) will perform an operant response in order to be rewarded with visual exposure to a conspecific.
It is also possible to shape lever- and key-pressing responses in a Skinner box using a live mouse as reinforcement (van Hemel, 1972). Rats will perform operant responses for the reward of killing mice (van Hemel & Myer, 1970) or frogs (Huston, DeSisto & Meyer, 1959) with little apparent satiation, although some experimenters have reported evidence of satiation (Kulkarni, 1958). In an operant situation, killing latencies decrease with practice, just as with conventional reinforcers, and prior killing experience makes such behavior more resistant to the suppressive effects of punishment (Myer, 1967).
As pointed out by R.N. Johnson (1972) the question whether aggression is rewarding for its own sake is more difficult to answer than might be supposed, for it is necessary to show that killing is not instrumental. It is possible that some animals kill because they 'enjoy' killing, but it is also possible that they are defending a quasi-territory, killing defensively because they feel cornered or threatened, or responding to the possibility of a tastier meal than Purina lab chow. The reinforcement may come, not from killing, but from the cessation of distracting stimuli or from the removal of a potential competitor. Killing might also be the result of being trapped in a foreign environment such as a Skinner box where nearly everything is irrelevant to the natural behavior of the animal. It might also be a neurotic reaction resulting from abnormal rearing conditions.
While some experimenters have tried to show that aggression is rewarding in animals, others have tried to show that 'altruism' is rewarding. Rice & Gainer (1962) suspended rats in the air, and found that other rats would learn to press a bar which would lower the distressed rat. Lavery & Foley (1963) argued that the effect might be due to arousal rather than altruism, for rats will learn to make a response which will terminate either distress calls or white noise. Greene (1969) found that rats would rescue a conspecific in distress only if the rescuer himself had previously experienced the distress (e.g. electric shock).
On the other side of the coin, animals will sometimes not only fail to rescue, but may actively torment a conspecific in distress. Mice, termites, and piranha fish are known to attack and kill conspecifics which are injured. In the laboratory, Sidowski (1970) strapped a monkey to a cross to see if other monkeys would aid the helpless victim. At first the other monkeys withdrew, but later they ignored the obvious distress of the 'crucified' monkey and began to pull his hair, gouge his eyes, and lick his genitals. It should be noted that the animals in this experiment were partially or totally isolated from birth and, unfortunately, no control group of normal monkeys was tested. Distress experiments may not be critical tests of altruism, for they are usually carried out in an artificial social situation. A restrained animal is prevented from communicating with gestures so that other animals might feel threatened by the victim's apparent refusal to participate in rituals. Certainly an upright position on a cross easily could be mistaken for a posture of threat (R.N. Johnson, 1972).
In the long run, it is doubtfu.l that it will ever be established that animals in general are either altruistic or aggressive. Individual aggressive and helping behavior is likely to vary with social, developmental, situational, and species differences. There is some evidence that consummatory behaviors such as eating, drinking, or copulating are reinforcing in their own right, or in terms of Glickman & Schiff's (1967) model, the neural activity underlying such consummatory behavior may be reinforcing. Predatory or defensive killing may be a form of such consummatory behavior, but offensive killing for its own sake appears to be the exception, at least in nonhuman animals. Aside from a few clever laboratory demonstrations, the natural aggressive behavior of animals appears to be confined to instrumental aggression (R.N Johnson, 1972). This was stated many years ago by Craig (1921, 1928).
"Fundamentally, among animals, fighting is not sought nor valued for its own sake; it is resorted to rather as an unwelcome necessity, a means of defending the agent's interests... The animal fights in order to gain or to retain possession of that which is of value to him, such as food, mate or nest...Two animals fight only when their interests conflict. This is the fundamental fact in regard to inf ra-human fighting.
Animals do not enjoy fighting for its own sake. Unless his anger is aroused, the agent's behavior indicates that he has no appetence (Craig, 1918) for the fighting situation; he does not seek it; when in it he does not endeavor to prolong it; and he reveals by his expressions that he does not enjoy it. On the contrary, fighting belongs under the class of negative reactions or aversions (Craig, 1918); it is a means of getting rid of an annoying stimulus. As McDougall says, the stimulus of the instinct of pugnacity is the thwarting of some other instinct" (Craig, 1921).

Feshbach's (1964) flowchart model of aggressive behavior

Feshbach (1964) presented a fairly simple flowchart model, consisting of four Choice Points, for aggressive behavior. Choice Point 1 denotes the classification of a stimulus as obstructing, threatening, habitually noxious - all defined as aversive or as not aversive. Where objective stimulus conditions are constant, the habit of classifying a member of certain stimulus subsets into the subjective subset of 'noxious stimuli' may be viewed as being equivalent to the hostility syndrome, while the likelihood of perceiving a stimulus as threatening or obstructing might be seen as related to, respectively, self-esteem, and the ability to perceive alternative goals or alternative goal paths to the original goal.
The role of autonomic arousal in general, not necessarily labeled as anger, at this point consists simply in eliminating alternatives of classification. More important than the autonomic state appears to be a person's momentary perceptual orientation at the time he is faced with the stimulus. He may be anticipating noxious, threatening, or obstructing events, and thereby increase the likelihood of classifying them accordingly. The effect of arousal at this point is, then, that of a probabilistic operator, which changes the likelihood that a stimulus will be classified in a certain way.
Choice Point 2 represents the initiation of a goal response. It is assumed that a stimulus categorized as not aversive has an initial probability of zero of evoking an aggressive fractional anticipating goal response, while noxious, obstructing, or threatening stimuli all have equal probabilities greater than (or equal to) zero of initiating an aggressive response. It is at this point that the construct ' aggressive tendency' or 'aggressive habit' becomes relevant.It plays the role of a probabilistic operator and, for reasons of economy, may be viewed as identicall across classifications though, of course, differing across subjects.
Stimulus classification (at Choice Point 1), affected by both dispositional and situational determinants, can be conceptualized as standing in a multiplicative relationship to aggressive habit. Where a not aversive classification has been made or where aggressive habit is extremely low, the likelihood of an aggressive response being initiated becomes very small.
At Choice Point 2, anger may be present for two of reasons. It may be an initially classically conditioned, later operant-arousal state associated with a response choice and subsequently labeled anger (see Shapiro, Crider & Tursky, 1964), or it may be the result of deliberate self-stimulation, reinforced through previous, successful response activation; that is, the person talks himself into a rage (see Brehm, Back & Bogdonoff. 1964). The role of autonomic arousal here is thus seen as simply energizing rather than as affecting the likelihood of a choice.
Choice Point 3 deals with the dispositional inhibitory variables such as aggression anxiety. Such aggression anxiety is seen as a conditioned inhibitory response which may either, in the tradional Hullian sense, detract from the reaction potential of the ongoing aggressive response, or, through a feedback loop, result in the selection of an alternative response. (The selection of an alternative response can also be accounted for in the traditional paradigm by referring to response hierarchies, where a new response becomes dominant if another one is inhibited. However, the notion of feedback and cognitive reevaluation of an ongoing response is a useful one in this instance.)
Choice Point 4 refers to the situational, including the social, determinants of the situation. The functions of these determinants may be viewed as permitting the person to mediate anticipated rewards or punishments which appear to be likely outcomes of his present ongoing response. The effects of these determinants is either to 'clear' the ongoing response for completion, or to bring about a return to Choice Point 2 for selection of an alternate response. This process may take place where the original ongoing response is aggressive, that is, the person perceives the inappropriateness of his response and modifies it, or it may in some instances lead to the initiation of an aggressive response, even though the original response was nonaggressive. In the case where a stimulus had been classified as not aversive at Choice Point 1 and initiated a nonaggressive response, we should have the equivalent of attack against a target for purely social or similar reward expectations. It is important to note that repeated social aggression of this nature through classical conditioning may soon lead to a greater probability that a set of stimuli will be classified as noxious at Choice Point I.
Choice Point 4 may also result in a return to Choice Point 1, that is, a reevaluation of the stimulus itself. Again, this reevaluation may take place by classifying a stimulus initially perceived as not aversive into one of the three aversive categories, or vice versa.
Once the aggressive or nonaggressive response has been completed, the subject's evaluation of the outcoue may be assumed to take place. It is at this late stage that autonomic arousal or anger may occur or endure as a dissonance or anxiety reducer.
The present classification suggests that the dispositional determinants of perceiving stimuli in a certain manner (Choice Point 1) and strength of aggressive habit (Choice Point 2) should be susceptible to retraining, but such a retraining process could be quite lengthy.
Building up inhibitions at Choice Point 3 may be a questionable strategy, since it implies less aggressive behavior, where aggressive response tendency is held constant, but also greater conflict. The opportunities represented by Choice Point 4 consist of making the likelihood of a favorable social or other payoff for aggression lower, or, preferably, of raising the expectation of a favorable payoff for a nonaggressive response. The general approach here has been either to induce a reevaluation of the stimulus, discussed by Pastore (1952), Pepitone (1958) and Feshbach (1964), or to encourage a new response to a stimulus which need not necessarily be perceived as less objectionable than before (Kaufmann & Feshbach 1963a,b). Such a modification of an ongoing response may, but need not, be 'frustrating' , in that it is perceived as a threat or a path obstruction. As the second Kaufmann & Feshbach (1963b) study indicated, the determining factor may be whether a person refrains from aggression because he fears punishment or guilt, or because the rewards of nonaggression promise to exceed those of aggression.

Feshbach's aggression model

Feshbach (1971) uses the terms aggression and violence similarly, violence being reserved for the more severe forms of physical aggression. At a descriptive level, the label aggression might be applied to any behavioral sequence that results in injury to or destruction of an animal, human or inanimate object. At the level of construct, or mediating process, this definition is much too broad. One obvious distinction that needs to be made is between acts that are 'intentional' or 'motivated' and acts that accidentally lead to injury. By unintentional aggression, we refer to acts that, although resulting in injury, were not contingent upon their injurious consequence.
Motivated aggressive acts can be further subdivided into instrumental aggression that is directed toward the achievement of nonaggressive goals and aggressive drive for which the goal response is injury to some animate or inanimate object.
It is possible, according to Feshbach, through the appropriate analysis of antecedent events and reinforcing stimuli, to distinguish between aggressive acts that are predominantly instrumental in character and aggressive acts that are mediated by aggressive drive. The concept of aggressive drive does not, in this context, presuppose an instinct theory, or even a modified instinct theory such as the frustration-aggression hypothesis.
The ethological studies suggest that what may be innate or instinctive in human aggression is the evocation of aggressive responses by as yet to be specified stimuli. Thus, an aggressive response, to use an old Hullian concept, may be high in the innate habit family hierarchy elicited by particular frustrating stimuli. However, the concept of an instinctive aggressive reaction is quite far from the notion of aggressive drive-mediated behavior, the goal of which is the infliction of injury. Feshbach observes that intraspecies killing in animals is much rarer than in humans, and the pursuit of a rival over space and time is a peculiarly human phenomenon.
To complete the theoretical model, a further distinction is required between aggressive drive and the affective response of anger. The set of autonomic and motoric responses which we describe as anger is usually a component of the innate aggressive reaction to an eliciting stimulus. However, since there appear to be innate aggressive reactions in which 'anger' is minimal (Moyer, 1967), it is useful to seperately categorize the anger response. Anger and aggressive drive are often used synonymously, and clinical observation would indicate they are intimately related. Feshbach suggests that this relationship is a highly probable one in the usual course of development, but it is not inevitable, and there are circumstances in which the anger response can be detached from the motivation to inflict injury. One can obtain satisfaction from the infliction of injury without being angry, and one can be angry without wishing to injure the provoking agent.

The acquisition of aggressive drive.
It has been argued elsewhere (Feshbach, 1964) that the secondary reinforcement hypothesis does not provide a satisfactory explanation of the origins of aggressive drive. The motivation to inflict injury is rare in animal species other than man, and, in addition, aggressive behaviors in young children are frequently followed by punishment to the aggressor rather than reward. As an alternative or supplementary hypothesis, Feshbach suggests that aggressive drive arises from the pairing of the infliction of injury with counteraggression. This pairing is reinforced by cultural norms which state that retaliation is the appropriate response for an injured party. The child learns that the aggressor must be punished, not merely for the purpose of inhibiting subsequent aggression, but to redress the injury. The lex talionis becomes the basis, as it were, for the internalization of aggressive drive.
Other cultural standards and psychological processes add to the satisfactions that humans derive from the infliction of injury. For example, we have said little about the conditions under which aggression becomes a stimulus for sexual gratification. Of greater importance is the relationship between self-esteem and aggression. Violations to self-esteem through insult, humiliation, or coercion are powerful elicitors of hostility, probably the most important source of anger and aggressive drive in humans. Laboratory studies of aggression (Berkowitz, 1962; Buss 1961) and clinical studies of violent men (Toch, 1969) consistently point to this relationship.
Implicit in threats to self-esteem are the impotence and diminished status of the injured party. One way of achieving a restoration of status and demonstrating one's power is to injure the provoking agent. Thus, an instrumental element is introduced into the satisfactions associated with the infliction of injury.
The connection between status and aggression is especially exaggerated for males. The warrior definition of the male image, so characteristic of preliterate societies, is not entirely absent from our own. Feshbach asserts that the relationship between manliness and aggressiveness may not be solely a matter of arbitrary cultural definition.
If sex and aggression are closely related, there is the possibility that inhibition of aggression may result in a diminution of male potency. Freud maintained that inhibition of aggressive impulses would result in a reduction of sexual pleasure, and clinical observations can be cited to support this proposition. A study conducted by Feshbach & Jaffe (1970) indicated that the experimentally induced inhibition of anger and aggression in male college students resulted in a significant decrement on a subsequent measure of sexual arousal in response to erotic stimuli.

Criticism of Drive and Energy Models of Aggression

Hinde (1960) criticized the motivational models of Freud, McDougall, Lorenz and Tinbergen. These are only four of many in which energy concepts are used, but in them the energy analogy is made explicit in terms of a mechanical model, instead of being merely implied by a 'drive' variable which is supposed to energize behavior.
In the psychoanalytic model (Freud, 1932; 1940) the id is pictured as a chaos of instinctive energies which are supposed to originate from some source of stimulation within the body. Their control is in the hands of the ego, which permits, postpones or denies their satisfaction. In this the ego may be dominated by the super-ego. The energy with which Freud was particularly concerned - the sexual energy or libido - is supposed not to require immediate discharge. It can be postponed, repressed, sublimated, and so on. The source of this energy lies in different erogenous zones as the individual develops, being successively oral, anal and phallic, and it is in relation to these changes that the individual develops his responses to the external world. The instinctual energy is supposed to undergo various vicissitudes, discussions of which often imply that it can be stored, or that it can flow like a fluid. It may become attached to objects represented by mental structures or processes (libidinal cathexes) and later withdrawn from them in a manner that Freud (1940) likened to protoplasmic pseudo-podia; it has also been compared with an electric charge. Thus some of the characterisitics of the energy depend on its quantitative distribution.
McDougall (1913) envisaged energy liberated on the afferent side of the nervous system, and held back by 'sluice gates'. If the stimuli necessary to open the gates are not forthcoming, the energy 'bubbles over' among the motor mechanisms to produce appetitive behavior. On receipt of appropriate stimuli, one of the gates opens, and the afferent channels of this instinct become the principal outlet for all available free energy. Later (1923) he used a rather more complex analogy in which each instinct was pictured as a chamber in which gas is constantly liberated. The gas can escape via pipes leading to the executive organs when the appropriate lock(s) is opened. The gas is supposed to drive the motor mechanisms, just as an electric motor is driven by electrical energy.
The models of Lorenz and Tinbergen have much in common with McDougall's. Lorenz' 'reaction specific energy' was earlier (1937) thought of as a gas constantly being pumped into a container, and later (e.g. 1950) as a liquid in a reservoir. In the latter case it is supposed that the reservoir can discharge through a spring-loaded valve at the bottom. The valve is opened in part by the hydrostatic pressure in the reservoir, and in part by a weight on a scale pan which represents the external stimulus. As the reservoir discharges, the hydrostatic pressure on the valve decreases, and thus a greater weight is necessary to open the valve again.
Tinbergen (1951) pictured a hierarchy of nervous centers, each of which has the properties of a Lorenzian reservoir. Each center can be loaded with 'motivational impulses' from a superordinated center and/or other sources. Until the appropriate stimulus is given the outflow is blocked and the animal can show only appetitive behavior; when the block is removed the impulses can flow into the subordinate center or be discharged in action.
The way in which the properties of the model may be confused with those of the original have been discussed for Freudian theory by Meehl & MacCorquodale (1948). Concepts like libido or super-ego may be introduced initially as intervening variables without material properties, but such properties have a way of creeping into discussion without being made explicit. Thus Meehl & MacCorquodale point out that libido may be introduced as a term for the 'set of sexual needs' or 'basic strivings', but subsequently puzzling phenomena are explained in terms of properties of libido, such that it flows, is dammed up, converted, regresses to earlier channels, and so on. Such properties are introduced surreptitiously as occasion demands, and involve a transition from admissible intervening variables, which carry no existence postulates, to hypothetical constructs which require the existence of decidely improbable entities and processes.
Such difficulties are especially likely to occur when a model which purports to be close to the original, like that of Tinbergen, develops out of an 'as if' model, like that of Lorenz. This case has been discussed by Hinde (1956).
In all energy models, the energy is supposed to build up and subsequently to be released in action. McDougall, Lorenz and Tinbergen, all of whom were influenced by Wallace Craig, compare the releasing stimulus to a key which opens a lock. This apparent dichotomy between releasing and motivating effects is a property of the model, and may not be relevant to the mechanisms underlying behavior.
In all these theories, the cessation of activity is ascribed to the discharge of energy - the behavioral energy flows away as a consequence of performance. These models are misleading because they are too simple - energy flow is supposed to control not only what happens between stimulus and response, but also the drop in responsiveness when the response is given. In practice, these may be due to quite different aspects of the mechanisms underlying behavior: for instance the energy model leaves no room for inhibition (Kennedy, 1954). Further, even if the cessation of activity is in some sense due to the performance, many different processes may be involved: the mechanism is not a unitary one, as the energy model implies.
Secondly, if activity is due to the accumulation of energy and cessation to its discharge, the organisms should come to rest when the energy level is minimal. In fact, much behavior serves the function of bringing the animal into conditions of increased stimulation. This has been shown dramatically with humans subjected to acute sensory deprivation - the experimental conditions are intolerable in spite of the considerable financial reward offered (Bexton, Heron & Scott, 1954). Energy models are in difficulty over accounting for such 'reactions to deficit' (Lashley, 1938).
In these energy models, each type of behavior is related to the flow of energy. Increase in strength of the behavior is due to an increased flow of energy, decrease to a diminished flow. The strength of behavior is thus related to a single mechanism. It is, however, apparent that changes in responsiveness to a constant stimulus may he due to many different processes in the nervous system and in the body as a whole - for instance, the changes consequent upon performance may affect one response or many, may or may not be specific to the stimulus, and may have recovery periods varying from seconds to months. Energy models, by lumping together diverse processes which affect the strength of behavior, can lead to an over-simplification of the mechanisms underlying it, and distract attention from the complexities of the behavior itself. Similarly, energy models are in difficulty with the almost cyclic short-term waxing and waning of such activities as the response of chaffinches to owls, the song of many birds, and so on.
Kubie (1947) has emphasized this point with reference to the psychoanalytic model. Changes in behavior are referred to quantitative changes in energy distribution, but in fact so many variables are involved (repression, displacement, substitution, etc.) that it is not justifiable to make easy guesses about what varied to produce a given state. Similar difficulties in relation to other models have been discussed by Hinde (1959).
Tinbergen's model translated the Lorenzian reservoir into nervous 'centers'. Changes in response strength are ascribed to the loading of these centers. Now for many types of behavior it is indeed possible to identify loci in the diencephalon whose ablation leads to the disappearance of the behavior, whose stimulation leads to its elicitation, and where hormones or solutions produce appropriate effects on behavior. There is, however, no evidence that 'energy' is accumulated in such centers, nor that response strength depends solely on their state. Indeed the strength of any response depends on many structures, neural and non-neural, and there is no character-by-character correspondence between such postulated centers and any structure in the brain.
Another difficulty which arises from the use of energy models, though by no means peculiar to them, is due to the emphasis laid on the independence of different activities. Lorenz & Tinbergen (1938) write "If ever we may say that only part of an organism is involved in a reaction, it may confidently be said of instinctive action". Activities are interpreted as due to energies acting in specific structures and not as responses of the organisms as a whole. Both types of attitude carry disadvantages, but an overemphasis on the independence of activities leads to a neglect of, for instance, sensory, metabolic or temperamental factors which affect many activities.
The central nervous system is not normally inert, having to be prodded into activity by specific stimuli external to it. Rather it is in a state of continuous activity - a state supported primarily by the non-specific effects of stimuli acting through the brainstem reticular system. Factors such as stimuli and hormones which affect specific patterns of behavior are to be thought of as controlling this activity, of increasing the probability of one pattern rather than another. Changes in strength or threshold can thus be thought of as changes in the probability of one pattern of activity rather than another, and not as changes in the level of energy in a specific neural mechanism (Hinde, 1960).

The prolonged and often acrimonious controversy over whether aggression is an innate characteristic of animals is based in large part on the failure to make a distinction between the motor patterns by which aggression is expressed and the stimuli that elicit aggression. The motor sequence might be innate for any given species in the sense that it is similar in all members of that species and does not require learning. For any given animal, however, the probability that a particular stimulus or situation will elicit aggression is quite variable and is clearly dependent on previous learning (Lore & Flannelly, 1977).