Dog bite expert & animal behavior specialist

Richard H. Polsky, PhD, CDBC
Los Angeles, California

“Bringing the science of animal behavior to attorneys”

Dr. Richard Polsky - The Dog Expert


Predatory dog bite aggression

This peer-reviewed article authored by Dr. Polsky was published in the peer-reviewed  journal Behavioral Biology. The research reviewed in this article is applicable to all mammalian species, including dogs and cats, although examples discussed in the paper are take mainly from the research conducted with laboratory animals. Nonetheless, the principles and conclusions drawn from this research are still valid today, and have direct applicability to the real world situation with companion dogs and cats, since many dogs and cats continue to regularly display strong predatory tendencies (e.g. chasing motorcycles, chasing cars, chasing skateboards), and such tendencies are occasionally the impetus underlying attacks on humans. A basic question addressed in this paper: Does an animal have to be hungry in order to engage in a predatory behavior? Generally speaking, the answer is no. The conclusion drawn is that predatory behavior is self-sustaining and an animal does not have to be hungry to be motivated to engage in this kind of  behavior. Since the publication of this paper, this conclusion has been drawn by others in papers published in other academic journals.


Research on mammalian predatory aggression is reviewed with special attention being paid to the effects hunger and prey feeding have on the development and maintenance of the killing response. The findings are dichotomized into those which suggest a positive relationship between these two variables and killing, and those which negate a relationship. Those findings which suggest no relationship are drawn from the neprophysiological literature as well as from research which shows that experienced killers will kill in surplus, that the act of killing does not potentiate feeding, that nonkillers cannot be induced to kill through starvation, that experienced killers may not eat the prey after their first kills, and that experienced killers do not have to feed on the prey in order to maintain the killing response. These findings suggest that killing is self-reinforcing and studies are reviewed which buttress this belief. Studies which support the view that hunger, feeding, and killing are positively related come from research in which naive animals were either starved or fed dead prey prior to the initial test for prey killing. It was concluded that hunger and feeding are not needed in order to maintain the killing response in the experienced killer, but both can serve as potentiators for the induction of killing in the naive subject. Implications and limitations of this conclusion were then briefly discussed.

Research within the last two decades on the now well-studied phenomenon of rodent predatory aggression has produced, over-all, inconsistent findings concerning the roles hunger and prey feeding play in the development and maintenance of the killing response. Karli (1956), for example, in his pioneer study, concluded that hunger played only a partial role, and likewise the findings of Heimstra and Newton (1961), Whalen and Fehr (1964), and Heimstra (1965) have been eqUivocal in nature. On the other hand, others claimed the relationship is unclear (Moyer, 1968) while others have outrightly denied any relationship between killing and hunger (Karli, Vergnes, and Didiergeorges, 1969). Still others have suggested that killing, prey feeding, and hunger relate to one another only in certain circumstances (paul, 1972; Paul, Miley, and Mazzagatti, 1973). Further conflicting opinion comes from the recent review of O’Boyle (1974) who cogently argued that a positive relationship between prey feeding and killing does in fact exist.

Theoretically, it would be desirable to clarify the relationship in view of the Widespread interest behavioral scientists have in the phenomenon of predatory aggression. First, if hunger induced killing and caused a predator to feed on the prey which it killed, then this would be one basis on which to separate out predatory aggression from the other types of aggression (O’Boyle, 1974, or see Moyer, 1968). Second, if hunger, prey feeding, and killing were related, then a fairly parsimonious explanation for the killing response could be made; i.e., predators kill because they are hungry, or they kill so that they can feed on what they killed (e.g., the prey). Alternatively, if there was no association between hunger, feeding, and killing, or a weak association, then other factors would have to be postulated to explain the motivational basis of killing. This critique Will, therefore, attempt to reform and systematize the findings, mainly from the rodent literature, pertinent to this issue. In addition, relevant findings from other mammalian species will be considered for the sake of breadth and for comparative purposes.

Functional aspects of the killing response

Functionally, it is not difficult to discern the value of predatory killing; the behavior in all likelihood is primarily concerned with food procurement. Labeling it as a “‘food getting” behavior seems quite appropriate (Denny and Ratner, 1970) for it is known that even laboratory predators will eat the prey after a kill if given the opportunity. For example, experienced rat killers will eat the frogs and mice they kill (BandIer and Moyer, 1970; Paul, 1972; Paul and Posner, 1973; Thome, Aaron, and Latham, 1973), and likewise cricket killing by mice is often followed by consumption (Thomas, 1969, 1972). Because of the obvious functional value, and because of the close sequential relationship between killing and eating, one might well conclude that these two events were related or one caused the other, or perhaps even be tempted to hypothesize a direct relationship between the tendency to kill, the tendency to feed, and a predator’s level of hunger.

Evidence which negates the relationship between hunger, prey feeding and prey killing

However, as might be expected, the relationship between the killing of prey, the consumption of it, and hunger is not as straightforward and simple Presumably then, predators are capable of killing far more prey then they could eat or would need to eat in order to satiate their hunger drive. Prey killing will remain stable and constant in fonn even if a predator is denied the opportunity to feed on the prey which it has killed. Myer (1967, 1969, 1971) has repeatedly demonstrated this in several of his studies; with mouse-killing rats. According to Myer the act of killing is self·reinforcing in itself and reinforcing a rat by allowing it to feed on the mouse is not needed in order to maintain the behavior.

Layhausen (1973) also noted that “once established, the killing bite will continue to develop its own appetite.” Experienced nonkillers cannot be induced to kill through starvation. For example, Karli (1956) found that rats which never killed mice could not be made to do so even if subjected to extreme food deprivation. In fact, KarIi reported that some of his rats starved to death in the presence of the proy. likewise, Kuo (1930) in’ his classic study with cats reported that hunger had little effect on the rat killing response of nonkillers.

A predator after its first few kills may not consume the prey which It has killed. Paul, Miley, and Baenninger (1971) reported thatt initially, rats occasionally showed hesitancy about eating the prey. Karli (1956) also found that after the first few kills rats tended not to eat the mice or eat only after a great delay. With deer mice, Thomas (1971) found that the interval between killing and eating was often several minutes. Moreover, Leyhausen notes that consumption of the prey will not automatically follow a kill. According to Leyhausent a predator has to Jearn the ‘connection” between killing and eating. Therefore, it seems that these two behaviors are unrelated at first but subsequently become sequentially linked through the process of association.

Several variables which influence the probability of eating have been found to have little effect on prey killing. For example, it is well-known that whether hungry or not an experienced mouse killer will kill if given the killed. Further, the act of killing does not signal or serve as a cue to the predator to begin eating the prey which it has killed. Rats given the choice between a piece of chocolate and the prey immediately after a kill were just as likely to eat the chocolate as they were the prey (paul and Posner, 1973).

Different sites in the brain govern eating and killing. For instance, King and Hoe,bel (1968) reported that electrical stimulation in several sites of the rat’s hypothalamus would elicit killing, but not eating. In the study of Panksepp (1971) the reverse was found (i.e., stimulation which elicited eating would not elicit killing). In addition, it has been reported that stimulation of a rat’s lateral hypothalamus will produce intensive oral activities, which resemble eating, but never attack and killing (Karli, Vergnes, and Didiergeorges, 1969). In another study, using cats, Hutchinson and Renfrew (l966) found that, although attack and killing could be elicited from the same hypothalamic sites, different intensities were reqUired’ for each of the behaviors; rs; attack required more intense stimulation for its elicitation than did Further. an intensive research program by Flynn and associates (re- Viewed in Flynn, 1967, or Flynn, Vanegas, Foote, and Edwards, 1970) has produced conclusive evidence indicating that attack, killing, and feeding are neurophysiologjcally distinct. Granted, Flynn argued, the fmdings of Hutchinson and Renfrew (cited above) are correct in that they substantiate the fact of definitive areas within the hypothalamus which, when stimulated, will elicit’ both attack and feeding.

However, according to FIynn, they still do not obviate the likely possibility that different sites may also be involved. Like several others he drew on evidence which shows that stimulation to a particular hypothalamic site, known to elicit attack and killing, will not elicit feeding. Five examples are presented to support this contention.’ First, he cites one of his early studies (Wasman and Flynn, 1962) in which cats were stimulated in a hypothalamic area known to reliably elicit attack, but only in the presence of a dish of food (no prey was present). Under these circumstances, Flynn found that his cats would sniff at the food, savagely bite it, and then prowl around the cage (apparently, an appetitive search for the prey) with the food often falling out of the mouth. In no instance was the food ever ingested. Second, Flynn found that, if stimulation which elicited attack was prolonged beyond the attack itself, this would not induce a cat to start feeding on the prey. The underlying assumption of this finding being, namely, that if killing and feeding were related neurophysiologically, then the site which elicited killing should have likewise elicited eating.

Next, Flynn reported an experiment in which cats were presented either horsemeat or an’ anesthetized rat concomitant with stimulation to several selected sites in the hypothalamus at different levels of intensity. The intensity of stimulation in this experiment was raised in increments until a subject either ate the horsemeat or attacked the rat. Flynn found, in five of his seven cats tested, that stimulation which elicited attack would not elicit feeding. Moreover, the more intense the stimulation was (it ranged from .10 to .60 rnA) the more readily attack was elicited, and in the two subjects in which attack and eating were elicited from the same sites, more intense stimulation was needed to elicit eating than attack. Fourth, Flynn cited an experiment in which cats were stimulated both in the presence of horsemeat and a rat. During all presentations, however, the food was always placed closer than the rat to the cat for it was known from previous research (Hutchinson and Renfrew, 1966) that wheth&r a cat would attack or feed depended, to an extent, on which object was closer. Thus, in this situation, with stimulation to the aame Slte, one would have expected most cats to eat rather than attack, assuming that attack and eat were both under the control of the site being stimulated. However. the results clearly showed that rather than eat, most cats attacked. last, Flynn noted that if cats were continuously starved for 3 days, given food and then shortly afterwards a rat, they would break off eating to attack. the rat when stimulated.

Taken together, the evidence drawn from theabove examples suggests that stimulation in certain sites of the hypothalamus will evoke predatory attack (of the quiet biting type) but not eating. Additional weight for the theory of seperate -neural centers for killing and eating has come from the research of Karli and associates (cited in Karli, Vergnes and Didiergeorges, 1969). These authors claimed that they successfully abolished both killing and eating in rats with bilateral lesions in the hypothalamus; however, they subsequently found that the recovery of killing invariably preceded the recovery of eating. According to these authors, “the question arises as to whether or not hunger or some selective appetite are essential factors in building up the motivational state underlying the killing response. We feel that this is not the case for the following reason: if the animal bearing lateral hypothalamic lesions….recovers oriented behavioral activities, the recovery of the killing response invariably preceding the recovery of feeding behavior; the reappearance of interspecific aggression may thus occur even though that animal still happens to be in a state of complete adipsia and aphagia, never eating anything of the mouse it kills,” The interested reader should consult Roberts and Kiess (1964) for additional evidence that different anatomical sites in the brain govern the eating and killing responses.

Killing as a self reinforcer

Thus, what the evidence reviewed so far suggests is that prey killing in the experienced predator is governed by a motivation which is separate and distinct from the motivation which governs feeding, This being the case, a number of investigators have reported experiments which show that the act of killing itself can serve as a reinforcer. Myer and White (1965), and Kilby, Moore, and Harris (1973), for example, have demonstrated that the opportunity to kill mice or frogs was a sufficiently strong incentive to maintain discrimination learning by rats. In .both of these studies, rats which were experienced killers learned to enter the arm of aT-maze that led to prey which they could kill. In similar fashion, Roberts and Kiess (1964) reported that cats during stimulation of the hypothalamus learned to enter the ann of a V-maze in order to gain access to a rat which they could iall. Rats have even been taught the operant response of bar pressing in a Skinner box for the delivery of a reward-a mouse which could be killed (Van Hemel, 1972; Van Hemel and Myer, 1971) or a frog which could be killed (DeSisto and Huston, 1971).

If the act of killing is reinforcing in its own right and further has motivational prope!ties of its own, then one wo.uld eventually expect the behavior to satiate after it has been performed so many times. Kulkarni (1968) has gathered evidence which shows that this is in fact what. happens. In his experiment three groups of 12 experienced killers were presented .seven mice in succession at intervals of IS, 30, or 60 min, respectively. Kulkarni found that half the rats in the IS-min group stopped killing during testing as opposed to only three ‘in the 3Q..min group, and only one in the 60-min group. Thus, whether or not a rat stopped killing seemed to depend’ on the interval between presentations. Kulkarni argued that the· waning of mouse killing was due to the exhaustion or habituation of the behavior and he used the term “action specific exhaustibility” to explain his findings. Additional evidence along these lines has come from Moyer (l971) who found that the killing behavior of an experienced rat killer would satiate if it was presented between 5 and 10 mice in succession at intervals of 1 min each. Moyer observed that when this occurred a satiated rat would allow an exploring mouse’to walk over it and even nestle with it. Further. Moyer noted that a rat’s tendency to kill frogs also waned after its mouse-killing behavior was satiated. thus suggesting that both the killing response to frogs and mice were governed by a similar motivation.

Evidence which supports the relationship between killing, hunger, and prey feeding

Up to now the discussion has been solely concerned with those animals that were regarded as uexperienced”; that is. experienced in terms of killing or not killing. This distinction was necessary for it helps explain some important data recently collected by Paul and colleagues (paul. 1972; Paul, Miley, and Baenninger. 1971). What these authors found was that food deprivation served to greatly facilitate the initiation of mouse killing in naive rats. Hunger, in one study (paul et al.1971) was induced through a 2-wk period of cyclic food deprivation prior to the initial mouse-killing test, and in another study (paul, 1972) through continuous starvation for 7 days. In fact, it was found that just the experience of being maintained on the cyclic schedule (and later tested when food satiated) increased the chances of killing considerably. These findings are important and above all reliable for they have been replicated by these authors in a series of experiments (also see Paul, Miley. and Mazzagatti, 1973).

At first they may appear discrepant with the earlier-cited work. and especially with the finding of Karli (1956). However, Paul and her colleagues argued that if one attempts to explain the differential effects of starvation ‘in terms of the past experience of the animal being starved, then their fmdings do not conflict with Karli’s. Kar!i. they asserted. exposed his rats to the potential prey both before and dUring the course of food deprivation while in their experiments rats were first exposed to the mice only after a substantial period without food. Accordingly, then. such prior experience “interfered with subsequent killing when the rats were quite hungry” (paul. ‘1972). More to the point, what· is inferred is that the prior exposures Kar!i’s animals received (when they were food satiated) reinforced ~abits incompatible with killing, or simply strengthened the habit of not killing per se. These habits, in tum, interfered with and suppressed whatever potentiating effects starvation might have had.

Paul (I972) conducted an elegant experiment to test this hypothesis. Rats (all naive) were assigned to four groups and housed either individually or with the prey species (a single mouse). Further, half the rats in each group were continuously starved for 7 days prior to the first mouse killing test, or maintained for 7 days on a cyclic feeding regimen. Thus, half the rats were exposed to the prey during ‘the course of starvation and half were not. When tested, those subjects which were housed with the prey killed in significantly fewer instances compared to those subjects which were starved but without such exposure. MoreoverI rats from both the exposed and nonexposed groups which were continuously starved showed a greater incidence of killing than those subjects which were maintained on the cyclic schedule. Further, Paul let all subjects feed ad lib. for 3 days after the last mouse-killing test. They were then subsequently tested and it was found that every rat which killed when hungry continued to kill when food satiated. Thus, hunger did not seem to be a necessary condition to maintain killing initially induced through starvation. Evidence suggesting some relationship between feeding and killing has also come from the work of Paul and associates (see Paul and Posner, 1973).

In one experiment these authors starved naive rats for 4 days ~nd then proffered to them a dead mouse (killed by another rat) which they were allowed to feed on for 30 min. Mouse killing tests were then conducted 30 min later and compared to those of rats which were tested first without prior eating. These eat·first subjects showed a greater incidence of killing (76% vs 51%) and killed with a significantly shorter latency. This finding thus suggests that eating dead prey potentiates killing in the naive predator. Studies conducted in this author’s laboratory (polsky, in preparation) with another rodent species, the golden hamster, also suggest that hunger and prior feeding on dead prey are strong potentiators of predatory “aggression. In several experiments naive hamsters were continuously starved for 3, 4, or 5 days and when first tested for locust capture, deprived subjects captured significantly more often than food-satiated controls. In another experiment naive hamsters were allowed to feed on’ dead locusts prior to their first exposure to a live locust. Again this treatment significantly increased the chances of capture. Prior starvation, however, seems to have little effect on this species. In one experiment, hamsters were starved early in life (prior to weaning) and then tested shortly after weaning when food satiated. Compared to controls these subjects showed no significant difference in the incidence of capture.

This finding thus differs from the results obtained by Paul et al. (1971) with rats. Other studies have also been conducted by this author on the hitherto’ neglected phenomenon of hamster predatory aggression. One report has already been published (polsky, 1914) and othe.s in preparation are concerned with the effects of age and experience, the reinfercing effects of e~ting after capture, priming, water deprivation, prey size, social isolation, and genotype.


In summary, review of the evidence suggests that a subject’s past experience with the prey is of paramount importance in determining the effects hunger and prey feeding have on pr~datory aggression. It seems that these two variables have relatively little influence on the maintenance of killing in the experienced killer but. on the other hand, a positive influence on the initiation of killing in the naive subject.

The fact that prior dead feeding has positive consequences, and that the drive for killing and that of hunger become separate through experience makes sense, biologically speaking. Feeding on dead prey or partially killed prey brought in by the mother could be one means by which the young inexperienced predator familiarizes itself with novel prey. Prey-killing reo sponses could then be practiced and the young predator could learn that what it was feeding on was in fact an edible and palatable food substance. Observations do, in fact, substantiate the belief that a mother often assists in introducing the young to their first prey; this happens, for example, in domestic cats (Ewer, 1968), tigers (Schaller’, 1967), cheetahs (Eaton, 1970; Kruuk and Turner, 1967); grasshopper mice (Ruffer, 1966), and golden hamsters (author’s personal observations).In the more specialized predators, such as the Carnivora, The fact that an experienced predator will continue to kill even though it may not be hungry is one means by which it could assure itself, or its companions (in the case of group-living predators), of an adequate supply of food. Prey which was not eaten after the kill could be passed on to a conspecific or cached for later consumption. T)le sharing of prey could be a means of maintaining organization within a social group, and caching could prove advantageous to predators who hibernate or to predators who do not readily have access to prey the year round (see Ewer, 1968, pp. 54-55 for a brief discussion on this point, or see Kruuk, 1972).

Further, one must realize that many other factors besides hunger and prior prey feeding have been found to facilitate the onset of killing in the naive subject. These include the genotype of the individual (Butler, 1973), prior competitive experience (Heimstra and Newton, 1961), social isolation (Bernstein and Moyer, 1970; Johnson, DeSisto, and Koenig, 1972; Kuo, 1960), type of prey species used (Bandler and Moyer, 1970), rank in a dominance hierarchy (Leyhausen, 1973), observational learning (Kuo, 1930), and the physiological changes· brought on by pregnancy and lactation (FIandera and Novakova, 1971).

Lastly, another shortcoming of the majority of studies reviewed in this paper stems from the fact that most researchers have assumed that hunger is a unitary concept when in fact it is not (Deutsch, 1971). Many specific hungers exist (Rozin and Ka1at, 1971) and it could well be that a rat, for example, that is apparently well-satiated on laboratory chow still has a specific hunger for mice (or perhaps some specific part of a mouse, such as the brain); hence it could be just this type of hunger and not hunger in a general sense which drives it to kill. The fact that eating often follows a kill in the predator satiated on laboratory chow certainly does suggest that a specific hunger may be present; however, few researchers have taken this variable into account as a causal factor. Because of this, additional research is needed to ascertain if a specific hunger for mice (or frogs) exists, and if so, ‘what effect it has on the killing response. Until then one must remain somewhat skeptical of any theory concerned with the relationship between hunger, prey feeding, and killing, or on the motivation for killing in general.