46 EFFECTS OF IONIZING RADIATION were sufficient in number to allow all samples to be taken in rapid succession, andtime intervals were rigidly controlled. 4.2 Methods of Treating Data, Control Groups Pre-Exroscre Bioop counts were not available on the exposed Marshallese or Americans; hence the individuals could not be used as their own controls. In orderto estimate the severity of the hematologic response it was necessary to establish control groups as comparable as possible with respect to age, race, sex, background and habits. A control group of 115 Marshallese from Majuro atoll (Control Group A), comparable with respect to age and sex to exposure Group I was obtained duringtheinitial observation period.* For comparison with the exposed .Americans, blood counts were done on approximately 85 American men on duty at Kwajalein. All who had not been on duty in the tropics for more than 2 months were excluded, since the exposed Americans had been in the area for that period of time before exposure. In addition, several who wererecently associated with radioactive materials were excluded. The resulting smaller group of 67 was used as the Kwaj-American control group. Data from the control group A were examined to determine the age and sex dependency of the several hematological determinations. To obtain valid comparisons within and among the various exposure groups, the age and sex dependencies noted for the control groups were taken into account. Although each individual in all groups was studied hematologically, those Marshallese with serious long-standing diseases were omitted from the analysis. A total of two *A second control group of 82 Marshallese from Majuro atoll (control Group B) were obtained during the 6 month medical resurvey. While data from these individuals are given in this report, they are not used Cc rt for comparisons because of a measles epidemic during the resurvey. from Group A and two from control Group B were omitted on this basis. In the following descriptions and comparisons of the data, findings in the exposed groups are frequently expressed in terms of percent of the appropriate age and sex control group. It should be noted, however, that in observational studies of this kind, wnknown factors coud possibly account for part of the differences noted between the control and exposure groups even though all possible measures were taken to select com parable control groups. Tn addition, it was not possible to obtain more than a single blood sample on each control individual. For these reasons, statistical tests of significance were applied mainly to time changes within an exposure group, andnot to differences between control and exposure groups. For the purpose of detecting significant changes in the hematological pattern, nonparametrictests (1. e., statistical tests for whichit is not necessary to specify the functional distribution of the variate under study) were used (2-7). The advantages of nonparametric methods have been summarized by Moses (8). 4.3 Hematological Findings, General In TaBue 4.1 are shownforcontrol group A, by age and sex, the meanvalues for the total white, neutrophile, lymphocyte and platelet counts, as well as for the hematocrit. The age and sex breakdown used for comparisons among exposure groups is shown in Table 4.2. Inthis breakdown the age and sex dependencies noted for the Marshallese control groups were taken into account insofar as was practicable. It should be noted that the Group B control values (Table 4.1) agreed closely with the Group A control data. To allow additional comparison between effects on children and adults, the neutrophile counts were arbitrarily separated into the age groups used for the lymphocyte counts. Monocytes and eosinophiles were broken down also into the same age groups. The age and