ABSTRACT— Psychiatric epidemiology has traditionally employed both census and proband-based methods to determine the population prevalence of psychiatric disorders. In this article we outline a potential bias in the proband method. With this method, the probability of ascertainment for an individual depends on the number of available relatives. If the average size of families containing one or more affected individuals (RA) differs from the average size of families with no affected individuals (Ru). then the ratio of the prevalence estimated by the proband method to the true prevalence is approximately [(RA -1)/(Ru -1)]. For certain psychiatric disorders associated with decreased family size (e.g. schizophrenia), the underestimation of prevalence by the proband method may be non-trivial.