The Use of Novel Stimuli as Indicators of Aggressive Behavior in Dogs

Introduction

When a patient is presented for a behavior consultation, prognosis and treatment are based partially on potential risk factors within the household, as well as clinical diagnosis and history. In some cases, the history is incomplete or unknown. There is a widespread concern about interactions between dogs and children and the potential for dog-bite injuries. In the United States, approximately 334,000 humans are admitted to emergency rooms annually with dog bite-associated injuries.¹² Nearly half of those are children under 12 years of age.³⁴ The problem of dog bites directed at children is an issue that is imperative to address. Patronek, et al. (1996), found that dogs with behavior problems were more likely to be relinquished to shelters and that aggressive dogs were highly likely to be relinquished to shelters for euthanasia.⁵ Several temperament tests are used in an attempt to screen dogs for these potential problems. Many shelters have implemented temperament testing in order to determine which dogs are most likely to be aggressive or to bite in the future.⁶ When determining temperament, aggressive tendencies, and adoptability of dogs in shelters, the validity of these tests remains in question. Penny and Reid found that dogs with a history of aggression to children respond differently to a doll stimulus than dogs with friendly behavior.⁷ Netto, et al., found that dogs with a biting history reacted differently to doll stimuli than dogs without a biting history and that these tests did show some validity.⁸

One of the purposes of this study was to determine whether the reactions to two novel stimuli accurately reflected the clinical diagnosis of fear-related aggression, dominance aggression, or the known history with children. Furthermore, if there was a relationship, the accuracy for screening purposes would be important to identify. If there is a relationship between the reactions of the dogs and their known history with children, then these tests might be used to predict future behaviors where there is no known history. This would have an impact on the way dogs are screened in shelters and how the results are interpreted. If the tests are reliable, the results would help to predict future problems so that they can be prevented.

Materials and Methods

Two novel stimuli were chosen for testing in this study. A plastic toddler doll, with black, curly hair, wearing a polka-dotted dress, and measuring 1 meter high, was chosen. The doll had moveable joints and glassy blue eyes [Figure 1]. The doll was used as a stimulus that would approximate the appearance, movements, and actions of an ambulating child. The doll approximated the size of a toddler. The arms could be stretched out to touch the dog. The target age range of children was 1 to 6 years. The other novel stimulus used in testing was an artificial hand.^a This was used to simulate reaching toward the dog and petting the dog—actions frequently directed by humans toward dogs [Figure 2]. The hand was used because its flexibility allowed the tester to palpate the dogs, open their mouths, and remove a bowl of food safely.

The stimuli were meant to approximate children and human hands, not to be exact duplications. Although the doll and the hand were visually realistic, they did not smell, move, or look exactly human. It is possible that other stimuli would have elicited the same response from the dogs. The reactions of the dogs to the actions of the stimuli were what were considered important, rather than the actual stimuli themselves.

Dogs (n=100) were tested during presentation at Cornell University for behavioral consultations. All dogs were presented for a primary behavior complaint, but not all were presented for aggression. A history form was completed by the owner prior to the consultation. During the course of the consultation, the dogs were tested with two stimuli controlled by the tester. There was only one tester to avoid differences in style of handling the stimuli. The tester directly manipulated the doll by the hair and extended it out from the tester’s body. Attention was paid to where the dog’s eyes were fixed to assure that the dog was responding to the doll. The doll was moved quickly toward the dog with arms reaching out; attempts were made to pet the dog on the top of the head, neck, and back, and to touch inside the dog’s mouth. The doll was dropped in front of the dog, was picked up, made to corner the dog, and finally moved quickly away to see if the dog would follow. The tester made the doll approach in a jerky motion.

The artificial hand was used to pet the dog on the top of the head, on its abdomen, and on its hindquarters. The hand was then used to palpate the mouth area, to reach for the collar and leash, and to manipulate the dog’s ears. Canned dog food mixed with chicken baby food was presented to each dog. The dog was allowed to inspect the food. As the dog started to eat the food, the hand was used to pet it. Then the hand was put in the bowl near the dog’s mouth. Finally, the hand was used to drag the bowl of food away from the dog. Each dog was leashed, so the bowl could be removed without the dog following. The reactions to handling and food were originally going to be separate categories. However, in this set of dogs, all of the food reactions corresponded to handling reactions, so one hand reaction score was given to each dog. All dogs except five readily ate the food. Those five dogs that were not food motivated in the consultation were all reported by their owners to be nonaggressive around food. They were all also normal reactors to handling by the hand stimulus. All of the dogs were videotaped during the testing period for later analysis.

The videotapes were then analyzed, and each dog was assessed. For each stimulus, the dog was placed in one of four categories: normal reaction or nonaggressive (NR), fearful reaction (F), fearfully aggressive reaction (FR), or offensively aggressive reaction (A). The criteria for each category were based on objective criteria such as ear, mouth, and tail position; body posture; attempt to flee or lunge at the stimuli; vocal response of the dog; and attempt to bite at the stimuli [see Appendix]. Each dog was grouped into one of the four categories for each of the stimuli separately.

The videos were evaluated by four judges. They did not know the dogs, their diagnosis, or primary complaint at the time of consultation. Two judges were veterinarians with a special interest in behavior, and two were veterinary students in the behavior clinic. Multiple judges were used for two reasons: first, to control for bias in interpreting behavior; and second, to allow for better observation, especially when dogs showed subtle signals. The investigator met with the four judges and showed them videos of dogs not included in the study. They were asked to evaluate the dogs based on a criteria sheet they received. The results were discussed until the investigator was satisfied that they understood how to evaluate body language and the criteria they were using to categorize the dogs’ behavior. For the actual study evaluations, each judge was given a list of each dog’s name, breed, and the same criteria sheet used for the practice cases [see Appendix]. Breed or known breed mix was given to the judges, because ear position was being evaluated, and some dogs may have pricked ears or be incapable of raising their ears. This information was given to the judges because it was not always apparent on the video what breeds were in the mixed-breed dogs. This had to be taken into account when evaluating ear position relative to behavior. Each of the dogs was put into one of the four categories for each stimulus.

A data set was constructed that consisted of signalment, clinical diagnosis of dominance aggression (DA, yes or no) or fear aggression (FA, yes or no) as determined by the behaviorist at the consultation, and known history with children. For the latter, there were four possibilities: no experience with children, good or never bad with children, barks or growls at children, has snapped at or bitten a child. The child involved had to be between the ages of 1 and 6 years of age.

Results

The four judges’ category rankings were compared. In 90 of 100 dogs, there was agreement among the judges. In the other 10 dogs, three out of four viewers agreed, and this was considered sufficient.

Significant associations were found between different combinations of stimuli and diagnosis (all P≤0.02) [Tables 1, 2]. Examination of the distributions of dogs across the cells of the tables (and of the significant chi-square values) indicated that the associations were in the expected directions when these tests were used to aid in making the diagnoses of DA or FA. There were 41 dogs that reacted normally to the hand [Table 1]. In this subset, 31 (75%) were not diagnosed with DA, and 37 (90%) were not diagnosed with FA. There were 25% of dogs diagnosed with DA and 10% diagnosed with FA that also had a normal reaction; these were false negatives. There were 23 dogs that reacted fearfully but not aggressively to the hand. There were 27 dogs out of the 100 (27%) that exhibited fearful aggressiveness to the hand. In this group, 24 of 27 (89%) were diagnosed with FA. The other three were not diagnosed with FA but had a fearfully aggressive response to the hand. There were 18 dogs that were diagnosed with FA that had false-negative results, in that they did not test as being fearfully aggressive to the hand. There was no clear association between being fearfully aggressive to the hand and being offensively aggressive. All of the dogs that were offensively aggressive to the hand (n=9) were diagnosed with DA. There were 31 other dogs that were diagnosed with DA. Ten of these had a normal reaction to the hand, or false negatives. The other 21 were fearful of the hand or fearfully aggressive to the hand, which may or may not be interpreted as false negatives. Dominance aggression is usually aimed at family members, and the dogs may have had other behavioral problems toward strangers. There could have been multiple diagnoses in these dogs, and the test was highlighting stranger-associated aggression versus DA.

There were 37 out of 95 (39%) dogs that reacted normally or nonaggressively to the doll, and 24 (64%) of them had a good history with children. There were five dogs omitted from this group because the past history with children could not be determined. Thirty-eight out of 95 (40%) dogs reacted aggressively to the doll; of these, 22 (58%) had DA, and 28 (74%) were diagnosed with FA. Some dogs in this group were diagnosed with both, and there were some false positives. Thirty-four out of 95 (36%) dogs were aggressive to the doll, and 22 (65%) of these dogs had already bitten or snapped at a child. There was a false-positive rate of 35%, meaning that these dogs were aggressive to the doll but had no history of being poor with children. Although within the parameters of this test they were considered false positives, these dogs did present a possible concern, and this was discussed with the owners so that heightened vigilance could be taken to prevent future problems. There were 36 dogs that had diagnoses other than DA or FA or both [Table 3]. Only three of these dogs were aggressive to the doll, and only one was aggressive to the hand [Table 4]. The dog that showed FA to the hand also showed it to the doll, and the other two dogs that reacted in this group were both judged with FA.

The tests are not close to perfect, giving both false-positive and false-negative results. There were nine of 37 (24%) dogs that reacted nonaggressively to the doll; these dogs had a history of biting or snapping at a child [Table 2]. In addition, four of the 37 (11%) dogs were diagnosed with FA. There were also four (11%) dogs that were offensively aggressive to the doll that had a good history with children, leading to discussions of potential future problems. There were four (10%) dogs that reacted nonaggressively to the hand, but they had been diagnosed with FA; 10 dogs were diagnosed with DA [Table 1]. There were 36 dogs that were not diagnosed with either DA or FA. When these dogs were tested with the doll, 33 (92%) of them were normal or fearful with no sign of aggression. When presented with the hand, 35 (97%) of the dogs reacted nonaggressively.

Discussion

The authors found statistically significant relationships that were in a diagnostically positive direction. Some of the results were more clinically compelling than others. All of the dogs that were offensively aggressive to the hand (n=9) were diagnosed with DA (100%). There were 10 that were offensively aggressive that had a normal reaction. More dogs that were fearfully aggressive to the hand also were diagnosed with FA (89%). Ninety percent of the dogs that showed no fear of the hand had no previous history of FA. The hand was only diagnostic for DA if the dogs reacted offensively aggressive. Dogs that reacted fearfully to the hand did not have a clear pattern of diagnosis. Fearfully aggressive reactions to the hand were only helpful for diagnosing FA, not DA.

More dogs that reacted normally to the doll were not diagnosed with FA (33/37; 89%), and fewer dogs were diagnosed with FA (4/37; 11%). Reacting fearfully to the doll did not predict whether the dog clinically had FA. Of the 34 dogs that reacted aggressively to the doll, 22 (65%) had actually bitten or snapped at a child, and eight (24%) had barked or growled at a child. In the tests involving doll reactions, the categories of DA and FA were combined. This was done because the sample size for offensively aggressive dogs was so small that the chi-square test became invalid. It appears that dogs not presenting for aggressive behavior were very unlikely to be aggressive to the stimuli. Only one of these 36 dogs had a history of biting a child, and she was one of the three that reacted to the doll. For dogs that present with nonaggressive behavior problems, these tests proved to be more accurate than for the population of dogs presenting for aggression. Aggression would appear to be a more complex issue than what these novel stimuli can discriminate. The purposes of this study were to see if these tests would be valid for determining adoptability in a shelter situation and if these tests could be used in the consulting room with diagnostic or prognostic value. The authors felt that the potential for biting was more important to determine than the dog’s motivation. Therefore, all dogs reacting aggressively were put together in one category. The results suggest a trend that could be used as the first step in identifying dogs with potential aggression problems. It does not definitively identify all dogs that are going to be aggressive, nor does an aggressive reaction to the stimuli necessarily mean that all of these dogs should be deemed unadoptable. Even if not predictive, the aggression that was displayed should still be a concern that is taken seriously. It serves as a first step in temperament testing. The results clearly show that one test alone should not be used as the criteria for labeling a dog as adoptable or not, based on its aggressive tendencies. The hope is that it would identify potential aggression to be evaluated further. In a consultation setting, the tests are another tool the behaviorist has at his disposal to evaluate dogs presenting for behavior problems. The trends suggested by the testing implied that dogs reacting normally to the doll tended to be good with children and not clinically aggressive. The trends also indicated that a reaction of FA to the hand was indicative of a clinically fearfully aggressive dog. Van der Borg, et al., investigated behavioral testing as an instrument for characterizing the behavior of dogs in a shelter situation that included a questionnaire about the dog’s behavior 1 to 2 months after adoption.⁶ The testing involved using a doll stimulus. The results were similar to this study in that the trend of being aggressive to the doll did relate to aggression toward children once in a home environment. Their tests predicted aggression toward children 57.1% of the time. The authors’ study accurately predicted the behavior of dogs 64.2% of the time. Van der Borg, et al’s group also reported that five of the seven dogs that tested aggressive toward the doll later proved to be aggressive toward children.⁶ In the authors’ study, 22 of 34 (65%) dogs that tested offensively aggressive to the doll had already bit or snapped at a child. Eight more had histories of growling or aggressive barking. Taking these two groups together, this study identified 88% of the dogs that acted aggressive to the doll as having histories of some level of aggression to children. It is important to note that there were nine dogs that had been aggressive to a child in the past that tested only fearful to the doll, and 13 dogs that tested normal. These false negatives are of concern. The test was not able to reproduce the past reaction the dogs had demonstrated. This could be due to the circumstances of the particular incident; for example, factors could include what the child was doing in relation to the dog, age of the child, and discomfort of the dog. This is why the authors believe that the doll cannot be used as the sole criteria for determining a dog’s reaction to children.

Because the dogs’ reactions to the doll were similar to the reactions to children, the test may be a useful tool in screening potential behavior. Dogs that reacted normally or fearfully to the doll were significantly more likely to have been good with children, and dogs that reacted aggressively or fearfully aggressive were more likely to have growled or bitten a child. This would indicate an inhibition of aggression in the first group of dogs. This is consistent with the results of the Penny and Reid and Netto, et al., studies.⁷⁸

The dogs in this study had known histories in contrast to the van der Borg, et al., study in which only short-term follow-up could be compared to temperament test results.⁶ The objective in this study was to determine how useful these tests were, even when no history is available (such as in a shelter situation). The tests themselves may also have other applications. Within a consultation setting, the tests are a way to observe the owners’ reactions to the dog, the intensity of reaction, and the body language of the dog that the owners had previously described. The intensity of the reaction to the stimuli may help veterinarians in part determine the severity of a dog’s aggression. If there is an association when the history is known, there should also be one when history is not known, thus serving as a predictor of behavior. The results showed some promise with definite limitations. The test did not predict behavior 100% of the time; there were false positives and false negatives in all but hand reaction versus DA [Table 1]. There are many possible explanations for this. There were 114 diagnoses made in 100 dogs. Therefore, 14 dogs had multiple diagnoses. All of these were DA and FA. These dogs had conflicting emotions and motivations, and this could have accounted for some false positives and false negatives. It is possible that the novel stimuli were not similar enough to actual children or hands to elicit the same responses from some of the dogs. An example of this might be that the hand could have been perceived as a toy, and not a threat, whereas a human hand would have been threatening. These dogs were not as afraid of the hand as they were of humans. Conversely, dogs afraid of the hand may have been frightened of it alone, whereas humans did not frighten them. Because the dogs were at a veterinary clinic, they were exposed to the sight, smell, and occasionally, the sounds of animals in pain or distress; so these healthy dogs may have been more inhibited or fearful than in their home environment. The clinic can be busy, so there are always individuals hurrying about outside the consultation room. This may have frightened the dogs. The dogs may have had such subtle changes in posture that the judges and the authors did not perceive the signals. It is logical that the dogs that were fearfully aggressive to the hand were not diagnosed with DA, as fear would not be the motivation expected in the DA subset. There were dogs that were fine with the doll that had bitten children; in other words, these dogs in some way discriminated between the doll and real children. It should be considered that the doll and hand cannot completely reproduce the odor, actions, and sounds of a real child or hand. Another possible explanation is that there were different sets of situations with different stressors that precipitated the bites that were not present in the consultation. Possible stressors were pain, extreme fear, and other dogs in proximity that were not present at the time of testing but had been present during the time in the dog’s history when he was aggressive. Conversely, there were dogs that were aggressive or fearfully aggressive to the doll that had no history of aggressive behavior with children. These dogs may have been artificially stressed by the hospital visit, anxious, or exhibiting play aggression. The doll may have been too novel for them, and children may be an everyday occurrence in their lives and, as such, were not frightening. Some of these dogs may be inherently unpredictable. It was clear that these tests cannot be used to make a yes or no decision about the future of a dog. There are many factors that should go into that decision, and the amount of false-positive and false-negative results indicate that these stimuli would only be one piece of the puzzle.

Conclusion

In a shelter situation, if these tests are used as a part of determining adoptability, then dogs will be missed that are aggressive, and some dogs may be deemed unadoptable that would otherwise be adoptable. The doll and the hand give some information as to the possibility of a dog having a DA or FA problem. Any aggression shown to the hand or doll should be looked at with concern. Whether or not they are false positives, a potential problem is presented that should be considered carefully when making adoption decisions. These tests have shown potential as one component of temperament testing in dogs that should be further evaluated. The follow-up research to this study is to repeat the testing in a shelter situation and then follow-up with new owners 3 months, 6 months, and 1 year after adoption. This will help to further validate the predictive value of these tests.