infant and childhood mortality varies appreciably from urban to urban-rural areas in Japan and (b) establishing proper controls for exposed persons who had left the two cities prior to childbirth would be extremely difficult.

2. Honseki was not restricted to Hiroshima or Nagasaki. Although restriction of honseki to these two cities would have materially simplified the follow-up, the loss of data which this step would have entailed seemed prohibitive.

3. Multiple births were excluded on the grounds that the mortality rates are appreciably different from those of single births, and they are too infrequent to warrant the added analytic complication they introduce.

4. The samples were to include all infants born to parents one or both of whom were known to the ABCC to have been within 2000 meters of the hypocenter at the time of the bombing. In addition, from among births occurring in the two cities to parents whose exposure status was on file with ABCC, equal numbers of children were to be randomly drawn from among births to parents (a) one or both of whom were exposed in the interval 2500–9999 meters (neither parent was, of course, to be nearer) and (b) neither of whom were in the cities at the time of the atomic bombings. These latter two samples were to be matched with the former group by sex and year of birth. This procedure yields three major or, if joint parental exposure is considered, nine minor exposure groups, as shown in Fig. 1. In view of some findings to be brought out later, it should be emphasized that the matching by sex and year of birth was by major but not by minor exposure group.

Fact of death was established by routine, periodic inspection of the appropriate koseki. A home visit by a trained “contactor” was utilized to establish honseki in those cases where the koseki could not be located by routine methods. The first cycle of koseki checking was begun in January, 1961, and completed in January, 1964, and it is the results of this cycle which are here reported. Since this cycle embraced some three years of record checking, we shall restrict our attention to deaths prior to December 31, 1961, and the initiation of the first cycle, to obviate the possibility, remote but nonetheless real, that exposure class may have been correlated with the ease or difficulty of follow-up and hence the number of years at risk of death. We present in Table 1 a summary of the results of this cycle of record reviews. It will be noted that among some 54,243 cases, the predominant reason for failure to determine survival status was non-Japanese parentage (i.e., no koseki record); in all except 53 instances where the child was Japanese, the survival status of the child in 1961 could be determined. Failure of follow-up because of lack of koseki can scarcely be regarded as a source of bias; the 53 children of Japanese parents whose survival status is unknown are too few to influence the data significantly, even if all the cases involved children no longer alive. In subsequent tables, the 810, largely non-Japanese, children in Hiroshima and Nagasaki for whom koseki records were not available have been excluded, as have 14 births in the Sex Ratio Program for whom the information was incomplete. The resulting distribution of children on the basis of the radiation histories of their parents is shown in Table 2. Where death had occurred, a record of the