Abstract: Summary Although animals (particularly tool-users) are capable of solving physical tasks in the laboratory and the degree to which they understand them in terms of their underlying physical forces is a matter of contention. Here, using a new paradigm, the two-trap tube task, we report the performance of non-tool-using rooks. In contrast to the low success rates of previous studies using trap-tube problems , , and , seven out of eight rooks solved the initial task, and did so rapidly. Instead of the usual, conceptually flawed control, we used a series of novel transfer tasks to test for understanding. All seven transferred their solution across a change in stimuli. However, six out of seven were unable to transfer to two further tasks, which did not share any one visual constant. One female was able to solve these further transfer tasks. Her result is suggestive evidence that rooks are capable of sophisticated physical cognition, if not through an understanding of unobservable forces and , perhaps through rule abstraction. Our results highlight the need to investigate cognitive mechanisms other than causal understanding in studying animal physical cognition.