Several attempts have been made to verify the marginal productivity theory of distribution. None of the recent attempts have given results which are consistent with the theory. In this note we show how the evolution of these studies reflects an attempt to deal with a fundamental methodological problem. The essence of all tests is to assume a functional form and level of aggregation of the factor inputs into a production function, estimate the function at the appropriate level of aggregation, and compare the resulting marginal products with the observed factor payments. Therefore, by necessity, all tests of the distribution theory are intimately linked to tests of the production theory. If estimated marginal productivities are not equal to factor payments, one can only conclude that the distribution theory or the production assumptions are wrong. A door is always left open to salvage either theory. This is the fundamental methodological problem.The only practical way of dealing with this dilemma is to agree on a fair test of the theory before the test is run. This is essential if the marginal productivity theory is to satisfy the criterion that a theory must be refutable. In this paper we offer the outlines of such a test so that it may be known and discussed before the test is performed.The paper is divided into five sections. Section 1 introduces our problem. Section 2 reviews recent attempts to take account of criticisms of previous tests. Section 3 reviews a method developed for dealing with the problem of multicollinearity which arises when several labour categories are introduced into the production function. We offer a qualification to this method and review some evidence which indicates that the qualitative results are not