Effect of training for dog fear identification on dog owner ratings of fear in familiar and unfamiliar dogs

Scientific studies often assess aspects of dog behaviour, such as fear, via owner reports. More information on how accurate these ratings are for dogs displaying different levels of fear would be valuable. The current study assessed which fear behaviours dog owners are able to reliably recognize and whether provision of training on recognition of fear alters participant ratings of fear in familiar and unfamiliar dogs. Dog owners (n = 735) were asked to identify which dog fear behaviours were present/absent in a series of videos. This survey showed that owners were able to correctly identify the presence and absence of (Sensitivity and Specificity >0.75) lowered body posture, ears back, lowered tail, wagging tail, panting, lolling tongue, yawning, lip licking, avoiding eye contact, and attempts to hide/escape/retreat. These behaviours were used to make a training tool for recognizing fear in dogs. Next, an intervention study was conducted where dog owners (n = 1413) were surveyed and asked to complete the fear and anxiety section of the Canine Behavioural Assessment and Research Questionnaire (C-BARQ) for their own dogs, and then either received training on recognizing fear in dogs (i.e., training), or did not (i.e., control). Both groups were then asked to rate the severity of fear in dogs in nine videos, with three examples each of no fear, mild/moderate fear, and high/extreme fear. Finally, owners were asked to complete the C-BARQ survey for their own dogs a second time to determine if training to recognize fear in dogs altered their responses. Owners were scored as either correct or incorrect in their rating of each video, and this was modelled as the outcome of a mixed logistic regression model with owner as a random intercept, in order to assess the effect of training on owners' ability to correctly rate dog fear displayed in videos. While training was not associated with owners being more likely to correctly identify 'no fear' (OR: 1.01; 95% CI: 0.86, 1.20; p = 0.881), it resulted in owners being more likely to correctly identify 'mild/moderate fear' (OR: 1.60; 95% CI: 1.34, 1.91; p < 0.001) and 'high/extreme fear' (OR: 1.95; 95% CI: 1.50, 2.54; p < 0.001). The effect of training on owner ratings of fear in their own dog were assessed using three mixed linear regression models aligning with three previously identified C-BARQ factors relating to fear (i.e., stranger-, dog- and non-social fear) as the outcomes and with owner as a random intercept. Training was not associated with the outcome in any of the three models, either independently or as part of an interaction with time. High ICC values were obtained for all three fear factors; this suggests excellent test-retest reliability when reassessment occurs in a short time frame (<20 min.). Before recommending the use of training to improve the use of the C-BARQ, further studies are needed to determine why training improves recognition of fear in videos of unfamiliar dogs, but does not alter the owners' rating of their own dogs.