Overview of Canine Aggression--Part 1: Breed, Genetics, and Behaviors in Selecting a Dog
World Small Animal Veterinary Association World Congress Proceedings, 2006
Karen L. Overall, MA,VMD, PhD, DACVB, ABS Certified Applied Animal Behaviorist
Center for Neurobiology and Behavior, Psychiatry Department - Penn Med Translation Research Laboratory, Philadelphia, PA, USA

The breed selection paradigm: Hundreds of years of artificial selection has developed a degree of canine size and shape variability that exceeds thousands of years of natural selection on wolves (1-3). Much of the physical variation in dog breeds is a consequence of overt selection for specific behavioral suites (e.g., herding vs. retrieving behaviors). Accordingly, the manner in which dogs communicate to each other and people is likely to be influenced by these selected traits, and when dogs develop behavioral problems for which a diagnosis can be made, some of the traits for which humans selected are likely to manifest themselves in the way in which these problems are expressed.

All breeds share characteristics with humans that have rendered dogs so compatible for joint working and social relationships: they have extended and extensive parental care, other family members contribute to the care and social development of offspring, they are socially mature after they are sexually mature, social systems are based in deference, and rules governing it so that signaling is often redundant, and most signaling or affirmation of signaling is non-vocal rather than vocal. Unfortunately, these similarities may lead people to under-rate subtleties of canine behavior and to anthropomorphize or anthropocentrize. For example, a dog who wags his tail may or may not be happy; a wagging tail is indicative of a willingness to interact and stiff tail whose tip is wagging is common in confident, aggressive dogs. If the dog has a problem aggression, staring at or reaching for the dog may be sufficient to trigger further agonistic behavior and frank aggression.

In the case of most aggressions and anxieties related to social maturity, the condition manifest by the dog has actually been changing because of changes in the interactive social environment. Most aggressive dogs are clinically behaviorally abnormal; the abnormality is usually progressive and is influenced by the social environment, so the signs noted by the client and clinician have been changing. We can easily understand such progressive changes in infectious and non-infectious disease, and so should be able to understand them as conditions that manifest as behavioral illness, yet regular screening for behavioral propensities is not a common part of routine veterinary examinations.

Role of selection for breed-associated traits: Before breed-based legislation gains credibility, we need to ask if such an impetus is warranted and substantiated by biological and scientific knowledge. To do this we need understand the role that we have played, both actively and passively, in selecting for inappropriate or aggressive behavior. The data previously discussed suggest that demographics of breeds involved in dog bites may co-vary with popularity of the breed, and that these changes may have relevance for the one group of children (males between 5-9 years) who are over-represented in dog bite data. Behaviors of breeds do not remain static as the breed become popular--these behaviors change in ways well known to population geneticists. Examples of these types of changes follow.

If a breed was bred and selected to display a relatively narrow suite of behaviors that were considered acceptable and individuals outside the bounds of acceptability were either culled or not bred. Twenty generations later when the breed got popular, two things happened: a) first, selection was relaxed, and because there was still underlying additive genetic variance, the less favorable traits were expressed, and b) the individuals expressing these traits and behaviors were not selected against; rather, they were highly desired because the dogs were "hot", or "tough", or "sexy" or "sharp". In this case the owners are tolerating, selecting for, and enhancing inappropriate out-of-context behaviors (see Table 1 for some idea of how popularity influences frequency of reported bites, and variable popularity of different breeds is over time and space).

If the breed was bred and selected to look a certain way or to do a suite of behaviors (e.g., guarding, or herding) and the breeders deliberately wanted to move the mean of the population to a slightly more "perfect" dog--forgetting that, in the absence of selection against undesirable traits, the entire normal distribution shifts, and the proportion of dogs with undesirable traits or behaving inappropriately also increases.

Other than these broad population level associations, we do not understand much about canine behavioral genetics even when normal behavior is involved (4). Studies designed to elucidate heritable components of breed specific performance traits have not produced definitive results. Tracking and scenting ability in German wire-haired pointers appears moderately heritable (having high additive genetic variance) and so should respond rapidly to selection. For a more complex behavioral suite, such as sheep herding, the mode of inheritance and the extent to which any of the behaviors comprising the style of approach and instinct are heritable is arguable and complex. Investigations of unpredictable aggression have produced no firm results regarding heritability although for many breeds in which dominance aggression is a relatively common diagnosis, each generation may contain affected individuals suggesting that simple dominant inheritance may be operating and should be further explored. Shyness or lack of exploratory behavior has been investigated in pointers; while these behaviors appear to run in lines, environmental factors can not be eliminated. Some evidence exists to link temperament and the probability of developing hip dysplasia in German shepherds; however, this assessment involved a scored system for rating temperament that may have obscured individual behaviors. The genetics of normal and abnormal behaviors requires intensive work.

Applicability for choosing pets: Every study that has looked at "temperament"--a still poorly defined concept--has found, regardless of breed and purpose for which dog was bred and was being raised--that between individual variability is greater than that for litters, families, and breeds. This finding should give us pause. It means that unless some fabulous or hideous trait has been deliberately selected, the behaviors of any dog are largely a manifestation of that individual dog's 'individuality'. Breed will shape the form any "normal"--and some "abnormal" behaviors will take. Accordingly, if clients need a silent dog, beagles are a poor choice. If clients live on a lake and don't want to deal with wet dogs, Labrador retrievers, Newfoundlands, et cetera are poor choices. If clients do not want to stimulate their dogs and redirect herding behaviors, border collies and Australian shepherds--particularly those from working lines--are seriously poor choices.

The point is simply this: we have selected dogs to be good at jobs. Until breed fancy valued looks over performance and when we were more mercenary in our relationships with animals who co-habited with us, we also indirectly selected for dogs that were "good family dogs". That selection criterion has largely been ignored for the past 50 years, and there was never a time when a breed was developed especially to "be good with children". If clients need a risk-free dog, the best bet is a stuffed one. If clients want the closest we can currently come to a dog who was bred to be "good with children", they need to go the town square in any small town in an impoverished country and buy any of the dogs who so gratefully and quietly hang around hoping for any cast-off love or food. Those dogs survive because they excel at reading human behavior and have posed no overt risks to humans. However, once that dog is removed from that environment, the dog is removed from a rule structure. American lifestyles impose a very different rule structure. Only if the new rule structure is clear, humane, and cognizant of the fact that their new dog is a cognitive, sovereign individual, will the dog retain the behaviors which the clients originally so valued.

Table 1. Dog bite data from various studies that specify breed data for at least 3 breeds; top 3 breeds in each study listed (references can be found in (5))



1993 breed distribution of bite quarantine dogs in a WI, USA humane shelter compared with total dogs of the listed breeds received at the shelter for other reasons

3 top breeds

(+ / - level of significance)

Chow chow

13 / 170 (7.6%)

P < 0.05; Gadj = 9.49 dagger

Cocker spaniel

23 / 316 (7.3%)

P < 0.05; Gadj = 15.543 dagger

Lhasa apso

14 / 203 (6.9%)

P < 0.05; Gadj = 8.261 dagger

daggerlog-likelihood 2, our statistics


Incidence rates of dog bites by breed at 2 USAF bases, January 1976-December 1977

3 top breeds

Bites /100 animals/ year

Relative risk compared to mixed breed (all P < 0.01)




G. Shepherd



Cocker spaniel




Predominant breed in case controlled study of 178 non-biting and 178 biting dogs

3 top breeds

Biting N (%)

Non-biting N (%)

Chow chow

31 (17.4%)

9 (5.1%)

P < 0.001; Yates corrected 2 test

G. shepherd

34 (19.1%)

13 (7.3%)

P < 0.001; Yates corrected 2 test


8 (4.5%)

1 (0.6%)

P = 0.04; Fisher's exact test, 2 tailed


Breed in prospective study of dog bite related injuries seen at CHOP, 1989; 156 dogs of identifiable breeds plus those identified as mixed breeds

3 top breeds

Bites N (%)


G. shepherd

35 (20.8%)


Pit bull ***

33 (19.6%)



8 (5.4%)


*** of these biting dogs, pit bulls were over-represented in unprovoked bites and those inflicted by roaming dogs; p < 0.01 or better, 2 test


Bites seen in 1975 at UCLA emergency department referred for surgery

3 top breeds

Bites N/total (%)



41/135 (31%)


G. shepherd

28/135 (37%)


Terrier (unspecified)

5/135 (7%)


Survey of 455 families in a Denver pediatric practice comparing breeds that had bitten and those owned by families in practice

3 top breeds

Bites N (%)

Owned by family
N (%)

G. shepherd + shepherd mixes

34 (17.5)

21 (11.0)*

Mixed breed > 30 lbs

24 (12.4)

27 (14.2)


20 (10.3)

27 (14.2)


*p < 0.05; 2 test; based on distribution of breeds when known


Retrospective study of breeds involved in 835 dog bites reported in Norfolk VA USA, 1 January-30 June 1971
No statistical analysis possible

3 top breeds

Bites N (%)



350 (41.5%)


G. shepherd

211 (25%)



37 ( 4.4%)


Retrospective study of 16 severe / fatal dog bites in 5 SC USA counties,
1 July 1979-30 June 1982
No statistical analysis possible

3 top breeds

Bites N


Am Staff **



St. Bernard



Cocker spaniel



** possibly--"pit bulls"


Retrospective study of 250 dog bites recorded by Guelph Health Unit, 1986-1987

3 top breeds

Relative risk*


Am staff



Duck tolling retriever



St. Bernard



* relative risk = biting rate of breed (# licensed biting in breed / total # licensed x 1000) / biting rate of all other breeds

dagger Gadj is the log-likelihood ratio2 test statistic. * Indicates P < 0.05.


1.  Wayne RK. Molecular evolution of the dog family. Trends Genetics 1993; 9: 216-224.

2.  Wayne RK, Vilà C. Phylogeny and origin of the domestic. In: The Genetics of the Dog, edited by A. Ruvinsky and J. Sampson, CABI International, New York, 2001: 1-14.

3.  Wayne RK. Consequences of domestication: morphological diversity of the dog. In: The Genetics of the Dog, edited by A. Ruvinsky and J. Sampson, CABI International, New York, 2001: 43-60.

4.  Houpt KA, Willis MB. Genetics of behaviour. In: The Genetics of the Dog, edited by A. Ruvinsky and J. Sampson, CABI International, New York, 2001: 371-400.

5.  Overall KL, Love M. Dog bites to humans: demography, epidemiology, and risk. J Am Vet Med Assoc 2001; 218: 1-12.

Speaker Information
(click the speaker's name to view other papers and abstracts submitted by this speaker)

Karen L. Overall, MA, VMD, PhD DACVB
Center for Neurobiology and Behavior
Penn Med Translation Research Laboratory
Philadelphia, Pennsylvania, USA

MAIN : Animal Welfare : Canine Aggression
Powered By VIN