It is first shown that den Boer's (1985) claims that competition should be rejected as a factor explaining community structure are not warranted by his tests on temporal niche differentiation in carabid beetles. The power of these tests is so low that they are unlikely to detect even the most extreme pattern of niche differentiation. A major factor that affects their power is the procedure of building up an artificial species pool from the local communities to which the test is to be applied. Alternative tests are then presented, and applied to data on annual activity cycles of carabids in three forest biotopes in Belgium. Three factors are shown to strongly affect their outcome: the unit into which the time scale is divided, the criteria used to incorporate constraints, and the set of species selected. Nevertheless, they provide some evidence that there is a growing niche differentiation from the successional to the climax forest, as predicted by competition theory. Other hypotheses, however, could explain the observed pattern.