Direct and indirect measures of progesterone secretion are the basis for assessing corpus luteum function. Despite methodological advances such as radioligand binding assays for the hormone, the frequency with which corpus luteum defects result in infertility or early abortion remains a matter of contention. The contention arises in part.from the fact that not all defects in corpus luteum function have a common aetiology and in part from failure to adhere to strict criteria for diagnosing or for determining response to treatment in the published accounts of studies of these disorders. In concluding this review, additional consideration will be given to some of these issues. From the point of view of current concepts about the aetiology of dysfunctional corpora lutea, it seems clear that both follicle development leading to ovulation and subsequent corpus luteum function result from a sequence of dose-related gonadotrophin-dependent responses in follicleschosen for ovulation. These responses include granulosa cell proliferation, differentiation of granulosa cells into luteal cells, and hormonally-dependent steroidogenesis by these cells. Failure of any step in the sequence ofevents may be expressed clinically by defective corpus luteum function coupled with failure to establish or to maintain early pregnancy. However, it should be apparent that any single therapeutic regimen directed towards normalizing corpus luteum function would not be equally effective for all aetiologic variants. Moreover, failure to distinguish between aetiologic variants in the population under study complicates interpretation of data purporting to describe incidence or response to treatment. A second problem arises from failure to adhere to uniform criteria for diagnosing defects in corpus luteum function. Before radioligand binding assays became available, clinical assessment of progesterone secretion (and thus of corpus luteum function) was based on cyclic changes in patterns of basal body temperature, on the quantity and quality of endocervical mucus, on endometrial morphology, or on measurements of pregnanediol (the principal urinary metabolite of progesterone), or combinations of these. Clearly each of these tests involves responses which are at least one step removed from progesterone secretion and thus subject to other variables. Thus a population of patients chosen on the basis of reduced levels of progesterone in randomly collected blood specimens might be dissimilar to a population chosen on the basis of endometrial biopsies. A third problem relates to variations in clinical expression before, and responses after, treating defects in corpus luteum function. Thus both failure to establish, and failure to maintain, pregnancy have been shown to occur among women with asynchronous secretory transformation of the endometrium and reduced mean integrated luteal phase serum progesterone levels. Furthermore, giving sufficient progesterone to restore blood levels of the hormone to ‘normal’ or to correct the delay in endometrial differentiation is not uniformly effective in restoring fertility or in preventing abortion. Given these sources of contention, how can the clinician confidently diagnose and treat women presenting with disorders of corpus luteum function? To our knowledge, the comparative reliability of direct and indirect methods for measuring progesterone secretion has not been subjected to vigorous examination. Under the circumstances physicians should familiarize themselves with the virtues and limitations ofa test which will perform reliably for them from patient to patient. For this purpose we recommend endometrial biopsies for use in monitoring corpus luteum function in women undergoing evaluation prior to treatment of infertility or early abortions. Furthermore, we believe that anomalous secretory transformation in this target tissue correlates reliably, albeit imperfectly, with both reduction in progesterone secretion and therapeutic efficacy of replacement therapy with the steroid in women with these disorders. Otherwise, physicians must engage in long and expensive, and often ineffective, empirical trials of therapy.