As anyone who has been in a long-term relationship can testify, minor conflicts between loved ones happen from time to time. Even the strongest couples experience spats on occasion. The human-canine relationship is no different. Conflict happens, but luckily the vast majority of such encounters are benign; a dog bone and a face lick will resolve most issues. In the academic literature, there are many studies and reports that attempt to quantify conflict behavior as well as identify preventable factors. There exists a great challenge in interpreting and effectively using the data from many of these studies due to sampling issues and to the broad grouping of diverse conflict behaviors. Most studies do not distinguish injurious dog bites from non-injurious behaviors, and seldom define the severity of injuries. Behaviors as different as growling and bites resulting in hospital treatment are grouped and counted together. Reports on “aggressive behavior,” lump data on growling, snarling, snapping, dog bites generally, non-fatal dog bites, bites that require medical care, bites that require treatment in an emergency department, and bites that result in “restricted activity.” There is, of course, overlap among several of these categories and any given behavior may be categorized differently depending on the study.
Further confusing the issue, alarmist language is often used in discussions of dog bites, either unconsciously or perhaps to elevate the importance of these incidents to the level of a serious public health issue. An analysis of 156 published papers on dog bites authored by human health care professionals (Arluke et al., 2017) found frequent instances of clear-cut factual errors, misinterpretations, omissions, emotionally loaded language, and exaggerations based on misunderstood or inaccurate statistics, in addition to failures to validate secondary sources resulting in the inclusion of non-existent data presented as fact.
In part because there is no standard method to measure or report these data, and because each researcher approaches dog bite data from a different perspective with varying goals and interests, there is a wide variety of factors to consider. Data include, but are not limited to, human demographics (gender, age, physical ability), canine demographics (sex, reproductive status, size, presumed breed, age, training), prior canine behavior, circumstances surrounding bite incident including whether it occurred on or off the owner’s property, severity of injury, location of bite on the victim’s body, and familiarity between dog and victim. One common approach for organizing data is to divide it in a binary fashion; for example behavior is described as defensive or offensive, bites are defined based either on presumed motivation or objective behavior, injuries to children are contrasted with those to adults, and bite location is tallied as head and neck or extremities. Other reports use statistical models to identify which factors the authors believe best predict bite incidence. The purpose of this review is to organize the existing data on dog bites and aggression in a meaningful way that will be accessible for academics, journalists, policy makers, and the general public. An overview of the literature is provided including bite incidence, and human and canine demographics. The relevant papers are then individually summarized and analyzed with respect to methodology, sample quality, and the appropriateness of the conclusions drawn. Strengths and weaknesses are identified.
The literature shows that conflict behavior is common between pet dogs and human beings, but this is often misconstrued as injurious bites being common. The two should be recognized as distinct events. Injurious bites are not common. To best incorporate the data from all studies, we will consider all bites regardless of motivation; however, behavior reported as playful mouthing—when the data makes this distinction—will not constitute a bite. For purposes of this analysis, injurious bites are defined as those for which medical treatment was sought and a bite is deemed severely injurious if it resulted in hospitalization.
Several studies have attempted to quantify the number of annual dog bites in the U.S., along with numbers of bites sustained over a person’s lifetime. Sacks et al. (1996) and Gilchrist et al. (2009) reported on extrapolations from randomized telephone interviews, the first and second “Injury control and risk surveys” (ICARIS-1 and 2). These were general injury risk surveys conducted by the Centers for Disease Control and Prevention (CDC) in 1994 and from 2001-2003 respectively. The first reported 4.7 million total (non-injurious and injurious combined) bites annually. The follow-up 12 years later showed a decrease to 4.5 million, with most of the decrease attributable to a 47% drop in bites to children 14 and under. Beck and Jones (1985) surveyed a group of school children between the ages of 4 and 18, 45% of whom reported having been bitten by a dog at some time during their lives. All three studies relied on subjects’ long term memory and combined injurious and non-injurious bites. Studies also exist which have attempted to establish incidence through local dog bite reporting systems (usually animal control departments). However, there is wide agreement that an unknown, but large, percentage of bites are never reported to such agencies. We include only a few studies derived from community bite reports here and only those that describe findings other than overall incidence, such as human demographics or context of the bites. The only consistently maintained record of injurious bites is the one that can be retrieved from the CDC's Web-based Injury Statistics Query and Reporting System (WISQARS), which records injuries presented for treatment at emergency departments around the country. Since 2001, the WISQARS-reported annual rate has averaged 339,000 dog bite cases per year (or ~1 person in 900). Approximately 1 in 65 dog bites treated in emergency departments are severely injurious (requiring hospitalization). Both rates are much lower than those recorded for other types of recorded injury causes, from falls to automobile accidents to bicycle mishaps.
Other studies (Guy et al., 2001a; 2001b; 2001c; Matos et al., 2015; Casey et al., 2014) look beyond biting behavior and attempt to measure canine aggression in general, including behaviors such as growling, snarling, snapping, lunging, and barking in addition to biting. In the literature, aggression is often not adequately defined, or is used so broadly that data based on the term become almost meaningless. For example, Casey et al.’s (2014) U.K. based study included barking as a measure of aggression, but barking is the dog's primary mode of vocalization and, thus, a common behavior with many motivations. Conversely, Guy et al. (2001a) and Matos et al. (2015) excluded barking in their definitions. Some studies included lunging, while others replaced lunging with “snarling.” This lack of a standard for what constitutes aggression makes identifying trends particularly difficult; it is nearly impossible to compare data across studies, or to meaningfully aggregate and interpret aggression data. We attempt to identify trends, but by and large we consider and discuss each study individually.
In the 2001a study by Guy et al., 41% of the Canadian owners reported that their dog had growled at household members and 20.6% reported that their dog had growled or snapped when a family member tried taking away food or toys, including incidents that the owners categorized as play or accidental. Casey et al. (2014) found different results in their convenience sample. The U.K. owners in this study reported lower levels of the group of behaviors they surveyed (barking, growling, lunging, and biting), even though this included more respondants than those captured in Guy et al. (2001a) and included more human demographics. Approximately 3%, 7%, and 5% of owners reported their dogs expressed aggression towards family members, unfamiliar persons entering the home, and unfamiliar persons outside the home, respectively.
There are several possible explanations for the wide discrepancy in incidence rates between these studies. The Casey et al. (2014) study may be an example of self-selection bias. The low response rate (~26% even among people who agreed to accept a questionnaire) may well indicate a sample of respondents more inclined to perceive their dogs’ behavior in a positive light. Moreover, more than 50% of the Casey sample were dog specialists of sorts, either dog walkers or people at dog shows, again with a possible positive bias regarding their dog’s behavior. Guy et al.’s (2001a; 2001b; 2001c) group of owners included only clients presenting dogs for treatment at 20 Canadian veterinary services, probably a more typical population of dog owners, particularly since the very high response rate (~86%) virtually eliminated self selection as a confound. However, this less behaviorally sophisticated group may have captured respondents more likely to mistake play behavior for aggression, or to generalize from the “even if you think it was in play” part of the questionnaire regarding biting to the other behaviors queried. In either case, it is clear that there is nothing approaching a reliable estimate in the literature for the rate of these behaviors among pet dogs.
Canine demographics correlated with biting
The three Guy et al. studies (2001a; 2001b; 2001c) included data on demographics of dogs who had bitten members of their household and risk factors for biting behavior. To date, these studies, since they contain the least potential for sample biases, are the most comprehensive and reliable on incidence and demographics correlated with biting. In the first study (2001a), they surveyed clients bringing dogs to veterinary clinics and achieved an 81.4% response rate. The follow up studies (2001b; 2001c) tracked all the dogs from the first study who met the researchers’ criteria as having bitten a familiar person. Based on the sum of the data, it appears that the correlation of breed with biting in the full population studied disappeared when closer owner questioning revealed that many of the initially identified biters had actually only engaged in mouthing, often playful. Further, breed gave way as a correlative when size was taken into account (Guy et al., 2001b; 2001c ). Smaller dogs bit more often, especially in interactions related to protecting resources, being handled, and being shouted at or struck. And size surprisingly did not correlate with the injury severity, measured by whether medical treatment was sought.
Multiple studies (Guy et al., 2001c; 2001b; Matos et al., 2015) indicate a link between fear and threatening behaviors (e.g., growling, snarling snapping) in addition to biting. Guy et al. (2001c) found that resource guarding (which the authors labeled “dominance aggression”) and self-defensive behaviors were common among dogs who have bitten, and that the same dogs were often fearful of various stimuli. Dogs who were afraid of men, children, and strangers were more likely to bite than dogs who did not fear these populations (Guy et al., 2001b). Other attempts to look at motivation, however, often lump all threatening and biting behavior together, so can offer no insight into the behavioral processes influencing any biting behavior, or to the liklihood of injurious bites. Matos et al. (2015) reported that fear behavior increased the probability of growling, snarling, snapping and biting in their samples. Conversely, Casey et al. (2014) categorized retreating and hiding as motivated by fear, and barking, lunging, growling, or biting as aggression and found the two groups of behaviors to seldom co-occur, ignoring the basic behavioral axiom that threat behaviors are often motivated by a perceived need for self-defense and thus are motivated by fear. Casey et al. (2014) also found that individual dogs did not behave consistently across contexts, suggesting that threatening and biting may be a product of particular circumstances rather than particular dogs.
Familiarity between dogs and bite victims has also been studied, particularly with regard to bites to children. Chiam et al. (2014) is an example of a recent finding that most dogs (78%) in their South Australian sample were familiar to the children they had bitten, or at least assumed to be so as the dog was identified as belonging to a friend or neighbor. However, this “familiar” classification must be taken with caution, as in this study, it is not established whether the dog and child actually knew each other, but rather that the dog was not a stray of unknown ownership. Other studies differentiate between dogs belonging to the victim’s household and dogs otherwise known by the victim (Matthias et al., 2015). These findings that dogs most frequently bite people they know are in line with earlier studies including Beck and Jones (1985) and Wright (1990). Data from bite incidents reported to authorities sometimes reveal higher proportions of unfamiliar or stray dogs, but some authors have speculated that this may be because the dog’s vaccination status is unknown in those situations and thus medical treatment may be more likely to be sought out of concern about zoonotic disease (Beck & Jones, 1985). In almost all U.S. municipalities, medical treatment triggers reporting of the incident, regardless of whether the victim chooses to do so himself. The exception is bites to children, which are reported at fairly consistent rates whether the child is familiar or unfamiliar to the dog.
Analyses of reproductive status, the dog’s sex, and the dog’s age have yielded a wide range of results, often times with conflicting data. Hsu and Sun (2010) reporting on dog bites in Taiwan, found that male dogs, old dogs (> 10-years-old), and intact dogs were more likely to exhibit “aggression” (barking, growling, baring teeth, snapping, lunging, biting, or attempting to bite) towards their owners. Guy et al. (2001c) reported that dogs who caused an injurious bite were older at the time of the first bite incident than those who caused a non-injurious bite. Young dogs (< 1-year-old) bit at the highest frequency of any demographic grouping, while intact females who were at least 1-year-old bit at the lowest frequency. Neutered males in this sample were more than 3 times as likely to have bitten someone as intact females. When weight was accounted for, larger dogs were associated with reduced odds of biting (Guy et al., 2001a). Matos et al. (2015) found that neutered dogs were less likely to exhibit aggressive behavior. Wright (1990) reported that male dogs were responsible for 70.5% of the bites in their sample, but Guy et al. (2001b) reported the opposite; female dogs were nearly 3 times more likely to have bitten than male dogs. Casey et al. (2014) analyzed barking, lunging, growling, and biting (all lumped together and called “aggression”) and found that spayed female dogs were less likely to exhibit these behaviors than neutered males. In contrast, Matos et al. (2015) and Casey et al. (2014) reported that intact dogs of either sex did not differ from neutered males. Guy et al. (2001a) reported that neutered male dogs were most likely to have bitten and intact females were least likely. The mixed data do not reveal a reliable pattern between sex, neuter status, and biting behavior.
Casey et al. (2014) eloquently summarize the current data on canine demographics and aggression, “Importantly, for all types of aggression, the variables measured explained a relatively small amount of the variance (<10%) between aggressive and non-aggressive animals, suggesting a much greater importance of factors specific to the experience of individual dogs in the development of aggression. These data suggest that although general characteristics of dogs and owners may be a factor at population level, it would be inappropriate to make assumptions about an individual animal’s risk of aggression to people based on characteristics such as breed.”
Human demographics correlated with biting
Several studies note that children are the most common victims of dog bites (Centers for Disease Control and Prevention, 2003; Matthias et al., 2015) and that they require proportionally more medical attention from dog bites than adults (Chiam et al., 2014; Sacks et al., 1996). The prevalence of this assumption has resulted in multiple studies that focus on children exclusively such as Beck and Jones (1985). However, the second ICARIS study, the largest survey ever conducted on dog bites, revealed that when the rates of dog bites to adults and children are compared according to percentage of the population in the U.S. (counted as persons affected per 1,000), adults were actually more likely to be bitten and the rates of medical treatment sought for adults and children were nearly equal (Gilchrist et al., 2008). Gilchrist et al. (2008) also found that a slightly higher proportion of adults in their sample had been bitten than children (1.6% versus 1.4%) and the bite rate among children had decreased 47% since the 1996 study (Sacks et al., 1996). The slightly elevated frequency of medical treatment for children is not surprising – parents decide whether their child will have medical attention and they may be more likely to seek care for their child than for themselves. In other words, there may or may not be severity differences between adults and children, but children are more likely to be medically treated and have their bites reported. In fact, in the Guy et al. (2001c) case series study, most victims were adults. Examining age from a different perspective, Matos et al. (2015) reported that younger owners were more likely than older owners to have aggressive dogs.
It should be noted that inconsistent definitions of children are used throughout the literature; some studies include teenagers up to 17 years (Chiam et al., 2014), others only count children up to 14 (Gilchrist et al., 2008; Sacks et al., 1996), or 12 (Guy et al., 2001c), and still others defined their sample of children as grade school students (preschool to 12th grade) (Beck & Jones, 1985). The wide variation makes it impractical to make cross-study comparisons or contrasts.
In addition to age, gender differences are often evaluated. Studies indicate that boys might be more likely to be bitten (Chiam et al., 2014; Matthias et al., 2015). These recent studies corroborate historical findings as well. Overall and Love’s 2001 review article summarized the (then-current) data on gender effects on dog bites. Though the quality of data varied across studies, the trend was that males were bitten significantly more often than females for all age groups. Moreover, familiarity of the dog may interact with gender; Matthias et al. (2015) reported that females were 1.5 times more likely than males to be bitten by the family dog while males were bitten more frequently by acquaintance and unfamiliar dogs. Context also plays a role, as Matos et al. (2015) found that the approach of an unfamiliar male was the situation most likely to elicit growling, lunging, snapping, or biting. Finally, race, ethnicity, urbanicity, SES, and region have been studied less often (Sacks et al., 1996), but significant differences were not found among these factors.
It is very common in the literature regarding growling, snarling, snapping, and biting behavior in dogs to cite sources in the introduction to support the notion of dog bites as a serious public health problem. Such citations are often presented in either misleading or uncritical ways, however, ignoring the limitations and flaws of the studies cited. In the analyses of the specific papers for this topic, we have chosen to point out some, although by no means all, of the examples of such weak literature review practices.
The following papers are referenced in the above literature review, and each has a link to a National Canine Research Council Summary & Analysis:
Owner Risk Factor Surveys
Guy, N. C., Luescher, U. A., Dohoo, S. E., Spangler, E., Miller, J. B., Dohoo, I. R., & Bate, L. A., (2001a). Demographic and aggressive characteristics of dogs in a general veterinary caseload. Applied Animal Behaviour Science, 74, 15-28.
Guy, N. C., Luescher, U. A., Dohoo, S. E., Spangler, E., Miller, J. B., Dohoo, I. R., & Bate, L. A., (2001b). Risk factors for dog bites to owners in a general veterinary caseload. Applied Animal Behaviour Science, 74, 29-42.
Guy, N. C., Luescher, U. A., Dohoo, S. E., Spangler, E., Miller, J. B., Dohoo, I. R., & Bate, L. A., (2001c). A case series of biting dogs: characteristics of the dogs, their behavior, and their victims. Applied Animal Behaviour Science, 74, 43-57.
Casey, R. A., Loftus, B., Bolster, C., Richards, G. J., & Blackwell, E. J. (2014). Human directed aggression in domestic dogs (Canis familiaris): Occurrence in different contexts and risk factors. Applied Animal Behavior Science, 152, 52-63.
Matos, R. E., Jakuba, T., Mino, I., Fejsakova, M., Demeova, A. & Kottferova, J. (2015). Characteristics and risk factors of dog aggression in the Slovak Republic. Veterinarni Medicina, 60(8), 432-445.
Sample of Reported Dog Bites Study
Sample Medical Report
Chiam, S. C., Solanki, N. S., Lodge, M., Higgins, M., & Sparnon, A. L. (2014). Retrospective review of dog bite injuries in children presenting to a South Australian tertiary children’s hospital emergency department. Journal of Pediatrics and Child Health, 50, 791-794.
Early Literature Review
Errors Reported in Medical Literature
Arluke, A., Cleary, D., Patronek, G., & Bradley, J. (2017). Defaming Rover: Error-Based Latent Rhetoric in the Medical Literature on Dog Bites. Journal of Applied Animal Welfare Science, 1-13. doi: https://doi.org/10.1080/10888705.2017.1387550
Centers for Disease Control and Prevention, Web-based Injury Statistics Query and Reporting System (WISQARS), Non-fatal injury reports, 2001-201. Available at: http://webappa.cdc.gov/sasweb/ncipc/nfirates2001.html