Do Dogs Understand Human Perspective? An Analysis of Construct Validity in the Guesser-Knower Paradigm

Understanding whether animals are capable of attributing knowledge to others or whether this ability is uniquely human, remains a central but controversial question in research on Theory of Mind. Earlier studies used the Guesser-Knower paradigm to examine visual perspective taking in dogs. Nonetheless, concerns regarding the construct validity of this method and the presence of alternative explanations for previous findings, such as behaviour reading or context dependent learning, remain. Therefore, the goal of the present study was not only to replicate earlier findings but also to critically evaluate the construct validity of the Guesser-Knower paradigm as a method to measure visual perspective taking in dogs. Overall, 93 dogs participated in three sessions. Dogs were presented with two experimenters, one of whom knew the location of hidden food (Knower), and the other did not (Guesser). The dogs could observe that one experimenter had visual access to the food hiding process while the other did not, but they could not see themselves where the food was hidden. The test consisted of four conditions in which the Guesser’s visual access was manipulated: “Guesser Present” (control condition), “Guesser Absent”, “Guesser Back Turned” and “Guesser Looking Away”. Pooling together the results across all sessions and conditions, the results show a significant preference for the Knower, the informed person. At the group level, dogs chose the Knower significantly above chance in the “Present”, “Absent” and “Back Turned” conditions, but not in the “Looking Away” condition. At the individual level, no consistent correlations were found between conditions, and results show a large inter-individual variability, that is not stable across conditions. The results thus only partially support the claim that the Guesser-Knower paradigm as currently implemented with dogs is an appropriate test for the valid measurement of visual perspective-taking in dogs. While dogs seem to differentiate between conditions with clear knowledge (but also behavioural) differences, the absence of a significant preference for the Knower in the “Guesser Looking Away” condition and the significance preference for the Knower in the control condition suggests that they might pass the task relying mainly on behaviour-reading. The lack of consistency in individual differences across conditions, and the lack of correlations between individual performances across conditions are critical points that must be considered. Overall, results suggest that the observed performance in this task may reflect context-dependent strategies rather than stable perspective-taking, which calls into question the construct validity of the task and highlights the need for developing tests of perspective-taking with stronger psychometric properties and more stringent controls for behaviour-reading.