It has been suggested that domestic dogs—like young human children—have a ‘gravity bias’; they expect an unsupported object to fall straight down, regardless of any obstacles that redirect or halt its path. In the diagonal tube task this bias is revealed by a persistent tendency to search the incorrect location directly beneath the top of the tube the item is dropped into, rather than the correct location attached to the bottom of the tube. We presented dogs (N=112) with seven different versions of the diagonal tube task, replicating and extending previous research, to examine what factors influence their search behavior for an object dropped down a diagonal tube, and investigate their physical reasoning skills more generally. Contrary to previous claims, we found no evidence for dogs exhibiting a persistent, or even a trial 1, gravity bias. However, in line with previous reports, dogs were also unable to search correctly for the reward, even when it could be heard rolling through the tube, though they succeeded when the tube was transparent (Exp. 1a-c). Exp. 2 suggested that dogs might search on the basis of proximity, but Exp. 3a-b ruled this out and showed that they prefer to commence searching at the center of the apparatus. Finally, when potential sources of bias were eliminated from the task (Exp. 4), dogs’ performance was improved, but still not above chance, suggesting that they are unable to reason about the tube’s physical-causal mechanism. We conclude that, on current evidence, the gravity bias might be unique to some primate species.