It is exciting to see that ‘cognition in the wild’ is beginning to attract a significant level of research interest.  There is a wide diversity of animals (especially birds) and contexts being investigated, and an ever-richer natural history of cognitive abilities of ‘real’ animals is becoming established.  The reign of the lab pigeon and rat, and tests conducted within the confines of the four walls of the laboratory, is diminishing.  Increasingly, we will be able to determine not only what a wide range of birds can do we will also be able to describe the circumstances under which they put those abilities to work.  Currently, however, there is considerable enthusiasm to test problem-solving abilities or to test rather generic cognitive abilities (e.g. associative learning), where the expectation seems to be that animals that solve problems are ‘smart’ and that any test of cognitive ability is appropriate.  Here I will suggest that some caution might prove useful, firstly because we need to understand what cognitive abilities are being tested in problem-solving experiments and how to quantify variation in problem solving.  Secondly, I will suggest that progress in investing cognition in the wild is likely to be made more rapidly if the choice of cognitive test is based on a sound working knowledge of the behavioural ecology of the species under test.  Finally, I will advocate the advantages of taking a modular approach to investigating the role of cognition in the lives of real birds by specifying the specific abilities from which one’s chosen species might be likely to benefit.