Problems of ascertainment bedevil the investigation of the etiology of congenital anomalies in singletons and in multiple births by epidemiological methods. It is shown that the definition of the population of affected births is ambiguous and that the problem of tracing etiology is complicated by the systematic way in which anomalies may be missed at birth. The available methods of dealing with problems of ascertainment are reviewed. Methods of adjusting for possible bias of ascertainment of affected births, by fitting statistical models to data from several sources, have been employed in some previous studies. In these methods, it is assumed that there are no errors of diagnosis or of recording. However, it is shown that there may be discrepancies in recorded diagnoses between sources, rendering this assumption untenable. In these methods, it is also assumed that the model which is the best fit to the data on the ascertained cases in also the best model for the cases which were not ascertained. This assumption is tested indirectly in a retrospective analysis of data from Aberdeen and Belfast collected concurrently through routine recording systems. It is demonstrated that there was a social process in ascertainment which renders the methods of adjusting for bias of ascertainment at best very complicated and at worst inapplicable.