Consistency of shelter dogs' behavior toward a fake versus real stimulus dog during a behavior evaluation

Behavior evaluations are widely used by animal shelters and other organizations that rehome dogs. The dog-to-dog subtest is a common feature of most canine behavior evaluations. The use of model devices such as a stuffed dog during this subtest could be convenient for shelters and increase safety. However, there is little research indicating if a fake dog can be reliably used instead of a live dog. In this study, the consistency of shelter dogs' reactions toward a fake and a real dog during the dog-to-dog subtest was investigated. Forty-five shelter dogs were evaluated using two different stimulus conditions. In one condition, the test dog was confronted with a single plush dog (the same plush dog for all test dogs), and in the other, with a single live dog (the same live dog for all test dogs). A standardized list of behaviors was recorded as observed or absent for both conditions with each dog serving as its own control. To calculate the agreement of individual behaviors between the two conditions, Cohen's Kappa was used. However, since many of the behaviors occurred at very low or high frequency rates, Prevalence-Adjusted, Bias-Adjusted Kappa (PABAK) was used along with Cohen's Kappa due to Cohen's Kappa's sensitivity to high or low prevalence, for which PABAK adjusts. For the purposes of this study, PABAK or Kappa scores greater than 0.61 were considered an indicator of a good degree of agreement between reactions toward the fake and the real dogs. The degree of agreement varied widely across individual behaviors with, Kappa ranging from −0.04 to 0.75 and PABAK from 0.29 to 1. Collapsing individual behaviors into behavior traits (e.g., friendly, aggressive, fearful) revealed a high degree of agreement for the friendly trait (Kappa = 0.60, PABAK = 0.69). However, the aggressive trait did not demonstrate adequate agreement (Kappa = 0.11 and PABAK = 0.38) and the fearful trait demonstrated only moderate agreement between the two stimulus conditions (Kappa = 0.50 and PABAK = 0.51). These results suggest that, while it may be possible to use a fake dog for the dog-to-dog subtest to assess friendly behavior toward other dogs, fearful and aggressive behaviors may not be consistent between the fake and real dogs, thus limiting the usefulness of the fake dog during behavior evaluations. In addition, the results of this study suggest more research is needed into the predictive validity of both fake and real dogs, since it appears the stimulus dog, whether fake or real, can influence the subtest's results.