Literature Review (3)

Visual Breed Identification: A Literature Review

People commonly believe they can identify the breed of a dog or the breeds in a dog’s ancestry by simply looking at the dog. They further imagine that if they know the breed or breeds, this will provide them with meaningful information about the individual dog: Is she smart? Will he be big or small? Will he bark a lot? Does she need a lot of exercise? While such beliefs are problematic at best (even purebred dogs with documented pedigrees have been selected primarily for appearance for more than a century) they are rendered effectively irrelevant by the lack of accuracy of the breed determination itself, particularly when it is based on visual identification rather than actual documentation like a pedigree.

Researchers have known for decades that even first generation crossbreeds often look dramatically different than either parent (Scott & Fuller, 1960). Recent research has consistently shown that visual breed identification is often inaccurate when compared with DNA analysis (Voith et al., 2009; Voith et al., 2013; Gunter et al., 2018; and Olson et al., 2015). Moreover, research also indicates that even experts have very little agreement when visually determining the background of dogs of unknown origin (Voith et al., 2013; Olson et al., 2015; and Hoffman et al., 2014).

The most recent study that examined visual breed identification and objective DNA results (Gunter et al., 2018) used the largest sample of dogs to date (all the dogs entering two large limited admission shelters over a year-long study period) to look at the breed composition of the dogs in the shelters and in the case of one of the shelters, the accuracy of visual breed identifications given by shelter staff. DNA-based breed results were compared to the visual breed identification provided by staff at the San Diego Humane Society for the study period. DNA analysis revealed a population with an overwhelmingly mixed breed heritage. Less than 5% of the dogs at both shelters were purebred, and only 1.3% were identified as having 2 purebred parents of different breeds. The vast majority were mixed-breed dogs with 3 or more breeds identified in their genetic makeup. Visual breed identification was better in this study than in previous reports (though there were limitations that must be considered), however staff were still completely wrong for 33% of the dogs in their shelter—that is, one out of three dogs were labeled with a breed that could not be found in their DNA even at the great-grandparent level. Based on their findings, the authors conclude, “when we consider the complexity of shelter dog breed heritage and the failure to identify multiple breeds based on visual identification coupled with our inability to predict how these breeds then interact within an individual dog, we believe that focusing resources on communicating the physical and behavioral characteristics of shelter dogs would best support adoption efforts.” (Gunter et al., 2018).

Earlier studies generally looked at smaller populations of dogs but often larger samples of people doing the visual breed identifications. The two seminal studies are Voith et al. (2009 and 2013), which used the same group of 20 mixed-breed dogs. The 2009 study compared breed identifications provided to owners by the 17 agencies the dogs were adopted from to DNA analysis. Only 31% of the dogs showed any DNA evidence of the adoption agencies’ named breeds anywhere in the dogs’ ancestry, and over 90% of the dogs identified as having one or two specific breeds in their ancestry did not have their visually identified breeds as the predominant breed in their DNA analysis. The 2013 study asked dog professionals to visually identify the same dogs, and found that fewer than half of 923 participants were able to correctly identify any breed in the dogs DNA analysis for 14 (70%) of the dogs. The experts were also not able to agree with one another on their identifications – for only 7 dogs (35%) could even half of the observers agree on a predominant breed.

Two other studies focused on different aspects of visual breed identification of dogs labeled “pit bulls,” as such labeling is likely to have dire consequences for the dogs (Hoffman et al., 2014; Olson et al., 2015). Hoffman et al. (2014), considered only inter-rater reliability, not accuracy, and found that shelter and adoption agency staff and volunteers (416 from the U.S., 54 from the U.K.) could not agree on which of a group of 20 dogs they considered “pit bulls”: a majority of U.S. participants identified 7 of the 20 dogs as “pit bulls” but for U.K. respondents only 1 of the 20 was thought to be a “pit bull.” Olson et al. (2015) queried 16 shelter personnel who regularly provided breed labels for the dogs in their care to label 120 dogs. Fully 1/3 of the dogs that did not have DNA profiles showing evidence of the breeds the study designated as “pit bull-type” (even at the great-grandparent level) were none-the-less labeled by at least one assessor as what the study had grouped as “pit bull-type.” It is important to note that there is no agreed upon classification for “pit bull” or “pit bull-type” dogs – not in science, the law, kennel clubs, nor animal shelters. While researchers in various studies have defined “pit bull” or “pit bull-type” to have a specific meaning in their samples, the authors create those definitions, a curious practice also employed in Gunter et al. (2018) above, but any such categories are by definition arbitrary and so effectively meaningless in other contexts than the individual studies. This importance cannot be overstated.

The discrepancies between visual breed identification and DNA analysis have strong implications for any research that links breeds and behavior. Yet shelter workers, owners, veterinarians, police, and witnesses to dog bites are among the people who may claim to know a dog’s breed simply by having seen the dog. Further, there are many scenarios where those same people’s opinions regarding a dog’s presumed breed end up in records that can seem authoritative — from arrival at a shelter, veterinary visits, licensing, housing and insurance applications, medical visits following dog bites, and the subsequent media reporting of such incidents. Eventually these unverified breed attributions make their way to databases that are then used in retrospective research studies to make claims about canine behavior. Many studies have used unverified breed identifications to relate breed to behavior, to injurious dog bite incidence, and even to dog bite-related fatalities. Such misinformation is particularly pervasive in studies on dog bites authored by human health care providers (Arluke et al., 2017). These studies influence public opinion and policy regarding particular breeds. Inaccurate data regarding breed are being propagated, studied, and used to drive policy. All studies completed so far that specifically address visual breed identification (links to summaries and analyses below) bring into question the findings of any such study and demonstrate a need for eliminating visual breed identification as a data source for ongoing canine behavioral studies. In fact, recommendations have been made that veterinarians cease attempting to visually identify their clients’ dogs of unknown origin by breed, in recognition of the confusion and potentially harmful biases this can evoke (Simpson et al., 2012).

The following papers are referenced in the above literature review; and each has a link to a National Canine Research Council Summary & Analysis, in addition to their descriptions above:

Gunter, L. M., Barber, R. T., & Wynne, C. D. (2018). A canine identity crisis: Genetic breed heritage testing of shelter dogs. PloS one, 13(8), e0202633.

Voith, V. L., Ingram, E., Mitsouras, K., & Irizarry, K. (2009). Comparison of Adoption Agency Breed Identification and DNA Breed Identification of Dogs. Journal of Applied Animal Welfare Science12(3), 253-262. doi:10.1080/10888700902956151

Voith, V. L., Trevejo, R., Dowling-Guyer, S., Chadik, C., Marder, A., Johnson, V., & Irizarry, K. (2013). Comparison of visual and DNA breed identification of dogs and inter-observer reliability, American Journal of Sociological Research3(2) 17-29. doi: 10.5923/j.sociology.20130302.02.

Hoffman, C. L., Harrison, N., Wolff, L., & Westgarth, C. (2014). Is that dog a pit bull? A cross-country comparison of perceptions of shelter workers regarding breed identification. Journal of Applied Animal Welfare Science: JAAWS17(4), 322-339. doi:10.1080/10888705.2014.895904

Olson, K. R., Levy, J. K., Norby, B., Crandall, M. M., Broadhurst, J. E., Jacks, S., Barton, R. C., & Zimmerman, M. S. (2015). Inconsistent identification of pit bull-type dogs by shelter staff. The Veterinary Journal, 206, 197-202.

Simpson, R. J., Simpson, K., & VanKavage, L. (2012). Rethinking dog breed identification in veterinary practice. Journal of the American Veterinary Medical Association241(9), 1163-1166.

Additional References:

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.

Scott, J. P., & Fuller, J. L. (1965). Genetics and the social behavior of the dog. Chicago, IL: University of Chicago Press.

Page last updated May 15, 2019