studied are small the choice of an appropriate control group may often be decisive in order to assess their presence and magnitude. Although in many cases corrections can be applied to allow for obvious differences, a subtle difference may remain unrecognized and may thus alter to an unknown extent the interpretation of the data. In all cases differences between the control and the test sample add variability to the data and uncertainty to the conclusions. 332. In retrospective studies the accuracy of the records is often a problem. For some groups (physicians, radiotherapy patients) the high standard of the medical care makes the records on causes of death very useful and well documented. But in other instances records are poor and causes of death only ap- proximately known. Uncertainties may well apply only to some and not to all causes of death and the ability of the epidemiologist lies in identifying these sources of errors and properly allowing for them. Concerning the radiation dose records, they are mostly uncertain or not known at all, as for occupational exposures where presumptive evidence must often be used instead of more precise statements of dose. In these cases no analyses of dose-response rela- tionships are possible, but only contrasts of broad categories of exposed versus unexposed groups. At the other extreme, doses are very well known for radio- therapy patients. Exposures in Hiroshima and Nagasaki have also posed dosime- tric problems but the newest results on the T65 doses and shielding conditions, together with clinical evidence about the presence of early symptoms of irradiation, make the assessments fairly reliable [02]. 333. Radiation dose distribution in time is often unknown and variable within the ascertained or presumptive period of occupational exposure; the radiation beams are often of very low energy and therefore likely to be absorbed superficially; irradiation of the hands, arms or upper part of the body makes the sample of occupationally exposed individuals very inhomogeneous. And, in ad- dition to the above-mentioned factors, the acute, fractionated or chronic con- ditions of the exposures make it difficult to compare the results of the various series.