To best understand this article in the context of the Breeds and Behavior literature, please see National Canine Research Council’s complete analysis here.

Duffy, D. L., Hsu, Y., & Serpell, J. A. (2008). Breed differences in canine aggression. Applied Animal Behaviour Science114(3-4), 441-460. doi:

National Canine Research Council Summary and Analysis:

This dog-owner survey is included because it is widely cited in discussions of canine aggression, has received significant media attention, and asserts significant differences between breeds in warning and biting behaviors. Moreover, the Canine Behavioral Assessment and Research questionnaire (C-BARQ) has been described in this study and others as “validated” and used as an instrument to characterize various dog demographic groups in numerous subsequent studies. Some such studies have used the data generated by these authors’ original online survey, others have distributed the survey to other populations, still others have combined portions of it with portions of other instruments. It has been said that “validation is not an event whereby an instrument is deemed valid once and for all – rather, it is a periodic and systematic examination that helps to build a body of evidence” (Williams et al., 2006). In contrast, the reported validation of the C-BARQ consisted of a comparison between owners’ responses to the questionnaire and behavior consultants’ diagnoses of a group of dogs presented to the consultants for behavior modification (Hsu and Serpell, 2003). Such a dog population is different from the general population and thus these results are not clearly generalizable to pet dogs overall. And it would be surprising indeed if owners bringing their dogs to a professional for help with behaviors the owners considered problematic did not give affirmative responses to questions related to those behaviors. Moreover, such a comparison represents only a very small segment of the accepted range of reliability and validity measures expected when evaluating psychometric instruments. A list of just the basic ones would include face validity, test-retest reliability, construct validity, content validity, criterion validity, etc., and nothing like this scope of examination has been done.

Duffy, Hsu, and Serpell (2008) used the C-BARQ to investigate differences in aggression across breeds. The C-BARQ is a 101-item owner survey with questions that sort into 14 categories, fewer in some iterations of the questionnaire. Five of the categories relate to warning or biting behavior toward familiar or unfamiliar people or other dogs and to fear. None includes any information, however, about whether the dog has actually bitten (bites were summarily grouped with snaps and attempts to bite), much less whether it has injured a person or another dog. Respondents rated their dogs on a 5-point scale. To set the stage for their study, the authors critically discussed four common methods for studying breed differences; analyses of bite statistics, behavior consultants’ caseloads, expert opinions, and behavior assessments. Duffy et al. (2008) explain many of the drawbacks of each approach including misrepresentation of larger breeds for treated bites and behavior consults, biases and shared stereotypes among experts, and the lack of predictive validity for behavior evaluations, particularly for behaviors categorized as “aggression.” However, a self-selected sample of dog owners such as people responding to an online survey is no less likely to be influenced by breed stereotypes and poor recall in reporting their own dogs’ behavior than are the respondents in the samples described above. And finally, owners’ expectations based on breed may affect their husbandry decisions which in turn form the dog’s behavior.

Owner participants were recruited in two ways: mailers were sent to 11 breed clubs and an online survey was made available and was circulated by word of mouth. The first recruitment method contains at least some assurance that the dogs are actually members of the breeds reported; the latter does not. Breed identification without verification is well known to be unreliable. Researchers were not able to verify the dogs’ breeds and so it is not known whether the owner knows the dog’s pedigree or is guessing by visual identification. Because the study is meant to identify breed differences, it is critical that the dog’s breed be known with certainty. Over 70% of the dogs in this study were from the online convenience sample (3,791 of the 5,312 total). In other words, in this study on breed differences, breed was only known with any confidence for 30% of the subjects. And neither sample group can be assumed to be free of breed stereotypes affecting behavioral perceptions and husbandry decisions.

As with other owner surveys, researchers never interacted with the dogs in this study; all behavioral data was collected via the C-BARQ. As discussed in the literature review, it is important in such a case to emphasize that the data consist of owners’ reported perceptions of behavior, not the researchers’ direct observations.

Based on the subsets of “aggression” related questions, four categories of aggression were created: stranger-directed (SDA), owner-directed (ODA), aggression towards unfamiliar dogs (DDA), and aggression, or rivalry, towards dogs living in the same household (DR). Breed differences were analyzed separately for the two samples. The online respondents reported 33 different breeds among 3791 pets. Eleven breeds (1521 dogs total) were included in the breed club sample. The data showed significant differences between the four types of aggression studied (SDA, ODA, DDA, and DR among the dogs in the online sample; DR was omitted in the breed club sample). The authors reported effect sizes between subscale scores for the difference between the most and least aggressive breeds and mean aggression scores for the breeds that were common to both the online and breed club sample. In their conclusion the authors discuss the importance of keeping the big picture in mind; though they report some significant differences between breeds, there was considerable variation within each breed indicating that they could not determine the relative roles of genetics and environment on the behaviors of interest.

Abstract and Link to Purchase Full Text of the Original Article:

Additional references:

Williams, V. S., Smith, M. Y., & Fehnel, S. E. (2006). The validity and utility of the BPI interference measures for evaluating the impact of osteoarthritic pain. Journal of pain and symptom management, 31(1), 48-57.

Hsu, Y., & Serpell, J. A. (2003). Development and validation of a questionnaire for measuring behavior and temperament traits in pet dogs. Journal of the American Veterinary Medical Association, 223(9), 1293-1300.

Page last updated July 10, 2019