Researchers using the same Dog Aging Project data found substantially lower rates of aggression, fear, and separation-related behavior, but say the true prevalence remains uncertain.
A
Researchers at the University of Pennsylvania School of Veterinary Medicine now say those figures were overstated by how the data were analyzed. In a
The original study, by Bonnie V. Beaver, DVM, MS, DSc, DPNAP, DACVB, DACAW, professor of small animal clinical sciences at the Texas A&M University College of Veterinary Medicine and Biomedical Sciences, used owner responses to a behavioral questionnaire completed when dogs enrolled in the Dog Aging Project, a long-term community science study of aging in pet dogs.1 Beaver concluded that essentially all US pet dogs have at least one moderate to serious behavioral issue and suggested that these issues typically go unaddressed by veterinary practitioners.
James A. Serpell, PhD, and Lauren R. Powell, PhD, authors of the reanalysis, wrote that Beaver's paper appeared to be the first to produce national prevalence estimates from the survey tool. "Given the important policy implications of such estimates, and the implausibly high levels of problems reported, the methods used to derive these estimates therefore merit particular scrutiny," they wrote.4
Both analyses relied on the mini Canine Behavioral Assessment and Research Questionnaire (mini C-BARQ), an owner-completed behavioral survey in which most items are rated on a 0-to-4 scale. The instrument uses factor analysis to group correlated questions into 14 behavioral subscales, such as stranger-directed aggression and dog-directed fear. Subscale scores are calculated by averaging the items within each one.4,5
According to Serpell and Powell, the original analysis departed from that structure in 3 ways. It treated individual survey items as independent measures and counted a dog as aggressive if it scored 2 or higher on any 2 of 10 aggression items, rather than averaging items within the 4 separate aggression subscales. It also pooled fear items from 3 distinct subscales into a single fear/anxiety category.
The largest discrepancy involved separation-related and attachment behaviors. The original analysis combined 2 separate subscales: separation-related behavior and attachment/attention-seeking, which includes behaviors such as following owners from room to room and sitting close to them. Scores for separation-related behavior clustered near zero, while attachment scores skewed high. Analyzed separately, 9.3% of dogs scored in the moderate to serious range for separation-related behavior and 88.9% for attachment/attention-seeking. The authors argued that “most owners do not perceive attachment/attention-seeking as problematic, except at the most extreme ends of the frequency spectrum.”
The reanalysis also questioned the original threshold of 2 for labeling a behavior as problematic. On the mini C-BARQ aggression scales, a score of 2 sits under the label "moderate aggression," which the survey describes as growling, barking, or baring teeth, while a 4 indicates snapping or biting.4 Under that threshold, a dog that growls at an approaching delivery worker would be classified as showing moderate to serious aggression, a behavior the authors wrote that "most dog owners would probably regard this as relatively typical canine behavior."4
When Serpell and Powell raised the cutoff to 3, prevalence dropped further: 11.0% for dog-directed aggression, 9.0% for fear of other dogs, 4.2% for fear of strangers, and 2.3% for separation-related behavior.
Some figures remained substantial under either method. Nearly a third of dogs (30.4%) scored in the moderate to serious range for dog-directed aggression at the lower threshold, which the authors described as "still concerning." Persistent barking when alarmed or excited was reported in 46.9% of dogs, and a tendency to escape the home or yard in 29.8%. The authors said the revised estimates were more in line with earlier surveys of canine aggression, fear, and separation-related problems.
Serpell and Powell acknowledged that researchers have used a range of cutoffs to classify C-BARQ scores and that there is no standard rule. They argued that the instrument measures how often and how intensely dogs display behaviors, not whether owners consider those behaviors a problem, which varies by household and culture.
The authors also noted that the Dog Aging Project recruits through media and word of mouth, which makes it a convenience sample and may limit its representativeness of the broader US dog-owning population.
"Representative samples of dog owners are therefore needed to establish the true prevalence of major behavioral problems in any dog population," they wrote.4
The reanalysis does not establish a definitive national prevalence of canine behavior problems. Instead, Serpell and Powell argue that the original estimates were sensitive to how survey responses were grouped and what threshold was used to define a behavior as problematic.
Serpell disclosed that he and the University of Pennsylvania hold the copyright to the C-BARQ. The study received no external funding.
References
