The trap-tube problem is used to assess whether an individual is able to foresee the outcome of its actions. To solve the task, an animal must use a tool to push a piece of food out of a tube, which has a trap along its length. An animal may learn to avoid the trap through a rule based on associative processes, e.g. using the distance of trap or food as a cue, or by understanding relations between cause and effect. This task has been used to test physical cognition in a number of tool-using species, but never a non-tool-user. We developed an experimental design that enabled us to test non-tool-using rooks, Corvus frugilegus. Our modification of the task removed the cognitive requirements of active tool use but still allowed us to test whether rooks can solve the trap-tube problem, and if so how. Additionally, we developed two new control tasks to determine whether rooks were able to transfer knowledge to similar, but novel problems, thus revealing more about the mechanisms involved in solving the task. We found that three out of seven rooks solved the modified trap-tube problem task, showing that the ability to solve the trap-tube problem is not restricted to tool-using animals. We found no evidence that the birds solved the task using an understanding of its causal properties, given that none of the birds passed the novel transfer tasks.