Examining the test-retest reliability of the Guesser-Knower task with dogs

The cognitive abilities of animals and even more so animals who live in close proximity to us humans are subjects of many studies. An important aspect of social cognition is Theory of Mind which describes the ability of individuals to attribute mental states to other individuals and themselves (1). Domestic dogs (Canis familiaris) are especially interesting for this kind of research as they show exceptional abilities at reading human cues (2,3). However, there is still no general consensus about how well animals understand other individuals and how well current tasks are suited for testing for this ability. There has been very little research about the psychometric properties of nonverbal tests used to draw inferences about animal cognition and none with dogs. Catala et al. (4) performed an experiment with 16 domestic dogs using the Guesser-Knower task in order to investigate the ability of dogs to take another individual’s perspective, and in particular the ability to distinguish between a knowledgeable and ignorant experimenter who indicate correctly and incorrectly, respectively, where a piece of food is hidden (5). In their study they found a significant preference for the knowledgeable individual (“Knower”) across several conditions, pointing to the possibility that dogs are able to take another individual’s perspective and follow their gaze geometrically. We not only replicated their study with 93 dogs but also introduced a new condition to further investigate dogs’ ability to take another individuals perspective. We could replicate some results by Catala et al. (4) but could also show that the preference for the Knower does not hold up in a larger sample size in a condition in which both informants behave identically, by looking towards the same direction, but because of their position in the room one looks in the direction of the food hiding and the other away from it while the food is hidden by a third experimenter. Our findings speak against the possibility of dogs being able to follow gaze geometrically. Furthermore we tested dogs’ performances across several sessions with several weeks between appointments. We found dogs’ performances to be highly variable across sessions, indicating a low test-retest reliability.