Do dogs rationally infer the cases of failed actions?

Humans regularly reason about the causes of events and actions we observe in the world, both to infer the physical properties and mechanisms of objects, and to understand others’ actions. Evidence for causal reasoning in nonhuman animals is mixed, and may be more easily detected in some contexts than others. Dogs, for example, fail at most tests of causal reasoning pertaining to physical cognition, yet possess sophisticated sociocognitive abilities. In this pre-registered study, we test whether dogs are capable of making rational inferences about the causes of failed actions in two analogous experiments, which differed only in the nature of said failures. Dogs observed human agents either succeed or fail to open two gates, in contexts where their failures could be attributed either to the lack of competency of an agent, or the physical properties of a gate. If dogs are capable of making causal inferences equally in social and physical contexts, they should succeed in both experiments. However, if dogs are more likely to make social rather than physical causal inferences, they should find the competency context more interpretable than the physical one. Dogs failed to make rational inferences in either context, raising theoretical and methodological questions for future work.