Does increased task difficulty reveal individual differences in executive function in the domestic dog?

Abstract

Pet dogs are carnivores that inhabit a largely human-dominated context, in which certain normal canid behaviors (e.g., resource-guarding, barking, mounting) are considered undesirable and even dangerous. Safety and welfare implications of human-dog interaction have recently led researchers to take an interest in canine executive function. Two tasks have become particularly popular in this area of study: the cylinder task and the A-not-B task. Because canine cognition tasks are not typically subjected to the same scrutiny as those used in human research, it is unclear whether these tasks indeed measure what researchers expect them to. Even though they ostensibly measure canine inhibitory control, task performance seldom correlates between the two, and researchers have suggested that they might be too easy to reflect effortful processes. Further complicating the matter are lack of reliability estimates and frequent use of under-powered samples. In this study, I evaluated the reliability and construct validity of the cylinder task and A-not-B task. Across two experiments, I tested modified forms of the cylinder task to make it more difficult and thus more reflective of individual differences in executive function. In Experiment 1, subjects completed the cylinder task under normal conditions and following self-control exertion. In Experiment 2, subjects performed the cylinder task either with or without practice retrieving a treat from an opaque apparatus. Subjects in both experiments performed the A-not-B task with removal of ostensive human cuing. Performance on behavioral tasks was compared to owner-reported measures of impulsivity, inattention, behavioral regulation, responsiveness, and aggression. In Experiment 1, performance was negatively affected by self-control exertion, but only to the degree that dogs exhibited self-control. This suggests that the cylinder task reflects an effortful, limited-capacity process. In Experiment 2, subjects performed worse when practice was omitted, suggesting that cylinder task performance partially reflects the ability to transfer the strategy learned during practice to the test trials. Across both studies, performance during the cylinder task and A-not-B task was uncorrelated. Further, the cylinder task showed high reliability whereas the A-not-B did not. Implications of these results and suggestions for future directions are discussed.