Statistical problems relating to sample size are of no consequence when a clinical disease state exists, with its attendant signs and elevated mortality. Under other circumstances, problems of sampHng sufficiency must be addressed. Where detection of carrier individuals is important, or where sublethal infections exist without obvious external indications, an appraisal of population health must rely on more intensive examination of a sample that, hopefully, reflects the status of the population.Inquiry into the presence or prevalence of disease begins with the question of numbers of individuals needed to be adequately representative. The matter of sample adequacy is not a constant, but varies according to the estimation precision required, numbers of individuals in the population, whether the sampling is destructive, and other factors. Sampling costs may also influence sampling magnitude. In different situations these factors are likely to change; consequently, generalizations about the sufficiency of a sample are not particularly useful. Neither is there any convenient way of solving such problems with ordinary calculators, especially for persons who are unable to spend the time needed to master the required calculations.These difliculties were appreciably reduced for persons working with fish diseases when numbers needed to satisfy sampling requirements were published by Ossiander & Wedemeyer (1973) and were later closely approximated in abbreviated form by McDaniel (1979). Although both publications are helpful, accuracy of their tabular data has not been independently confirmed. Furthermore, the scope of both papers is hmited by the fact that only one confidence level (95% probability) is given. More significantly, during the last three years we have received on average more than one request per month for assistance with disease/sample-size problems. Thus, despite the availability of a FORTRAN program from Ossiander & Wedemeyer and existence of the publications noted, there appears to be recurring need for more extensive tabular data, especially where computer access is difficult, or when the FORTRAN language is unfamiliar or incompatible with equipment available.Objectives of the present paper are to test the accuracy of the data provided by Ossiander & Wedemeyer (1973) and by McDaniel (1979), to triple their coverage at the 95 % confidence level, to provide data to include the 90 % and 99 % confidence levels, and to discuss the importance of detection method efficiency.We calculated sample size requirements with a Hewlett-Packard (H-P) 9845