Obstacles to historical epidemiologic research are formidable, especially for periods prior to the 19th century. Linguistic and paleographic difficulties are compounded by haphazard contemporary documentation and subsequent archival wastage, making real variations in disease frequency difficult to distinguish from record loss. This report investigates the application of a capture-recapture model, adapted from work on mobile animal populations, to this problem or archival loss. With the use of this technique, the original size of a psychiatric case population is estimated from incomplete extant enumerations. The cases are derived fron the Court of Wards and Liveries (1540-1660), the first statutory institution in England responsible for protecting persons legally incompetent due to mental disability. Documentation on this population survives in at least 10 record collections, the most extensive containing at best 82% of all original cases, the least extensive, 1%. Calculations based on extant files indicate a fourfold increase in the incidence of mental disability certifications between 1540-1549 and 1630-1639. The results of applying the Bernoulli census method to these materials eliminate biased archival loss as a plausible explanation for this rise. The origin of these changes in altered court policy or procedure, population growth, or increasing rates of mental disability awaits further investigation. While the Bernoulli census computations with historical data are mathematically equivalent ot those employed in recent epidemiologic applications, this specific data set prompts a conceptual revision of the Bernoulli model from capture to loss of random samples.