SUMMARYPrediction as a rejection test for theories is limited and unnecessarily severe. Useful knowledge frequently does not fall into this mold. Also the prediction test does not follow from the view that science provides coherent explanations of experience. The inapplicability of prediction as a sharp test is especially likely to be true for macro‐economic development theories. Thus it may not be meaningful to say that a given theory is correct or incorrect in a strict sense. Where the phe‐nomena that a theory covers is inseparable from a large system in which it is imbedded then it may be impossible in principle to discover the true relations between variables. Although no single simple test may exist for a class of theories there may be a number of considerations that help us judge the adequacy of some theories.Development theories may be looked upon as sets of partially connectedsamplerelations that operate within a changing environment. Such relations should be credible. Sample relations based on hypotheses that constantly flout experience are likely to be poor samples. Thus the realism of assumptions matter. Theories of the type considered may be useful in that the sample relations are helpful in the discovery of specific relations in specific contexts that we hope will work frequently in a loose fitting servo‐mechanistic type of arrangement so that (1) diagnoses of existing difficulties can be made, and (2) corrective policies fashioned which, while they cannot be expected to work universally, do work sufficiently frequently in terms of directional changes to be of int