Turbulence theory is difficult because the Navier-Stokes equations are an unclosed set of statistical relations. This review begins with an account of the problem of closure, and shows that this problem is evaded rather than solved by the methods now used for practical calculation: examples of the use of these methods are given.T h e bulk of the review is devoted to more fundamental procedures whose aim is to find the dependence of the pair correlation function on the separation of the two measuring points in space and time. This is done by making closure assumptions which are, at the least, free from arbitrary constants and which have some claim to uniqueness at the lowest level of complexity. T h e application of these theories to homogeneous isotropic turbulence is described. It is shown that they give a value of the Kolmogorov constant, and a shape for the dissipation range spectrum, which are in good agreement with experiment. The difficulties of applying these more fundamental procedures to practical problems are described, and a way of overcoming them is outlined.The review closes with an account of a problem in nuclear engineering which highlights both the defects of existing methods of calculation for tubulent flows, and the need for the better methods that the more fundamental procedures may one day off er.