Factors influencing a hand-touch learning task outcome in the domestic dog (Canis lupus familiaris)

Abstract

Dog cognition research often requires dogs to meet initial screening and preliminary behavioural criteria to proceed with subsequent experimental task phases. Reasons for the excluded dogs’ inability to meet the criteria, especially in the absence of major stressors or procedural errors, remain uncertain, whether stemming from variations in cognitive capacities, reward motivation, or lack of specific experiences in dogs’ learning history. Our objective was to identify predictors of failure in an owner-administered hand-touch learning task, with the aim of highlighting the characteristics of dogs more likely to succeed in this specific task and the potential sampling bias this success may introduce to the final study populations in similar cognitive studies. A total of 150 pure-bred dogs performed a hand-touch learning task with their owners, using food rewards as a reinforcer. The task consisted of different phases, each necessitating the dog to fulfill specific criteria before progressing to the subsequent phase. Failure was defined as the inability to meet the specified task criteria after three days of attempting the task. Binomial logistic regression was used to evaluate the probability of failure, incorporating dogs’ demographic characteristics, training history, reward responsiveness, and impulsivity scores, along with factors specifying owners’ previous experiences with dogs as independent variables.

Results indicated a higher likelihood of task failure associated with advancing age, lower food responsiveness scores, being a non-sporting dog, receiving non-food rewards in previous trainings, lack of familiarity with the hand-touch behaviour, along with a lack of dog training knowledge and a history of owning fewer dogs by the owner. While findings are specific to the hand-touch training and may not generalize to other types of canine cognitive tasks or broader contexts, they highlight potential selection biases in similar cognitive research, where certain groups of dogs may demonstrate a higher likelihood of success, and thus, be disproportionately represented. This possibility warrants further exploration across a broader range of cognitive tasks and contexts.