Understanding of the role of radiation as a cause of kidney cancer remains limited. The most common types of kidney cancer are renal cell carcinoma and renal pelvis carcinoma. It has been posited that these entities differ in their degree of radiogenicity. Recent analyses of cancer incidence and mortality in the Life Span Study (LSS) of Japanese atomic bomb survivors have examined associations between ionizing radiation and renal cell carcinoma, but these analyses have not reported results for cancer of the renal pelvis and ureters. This paper reports the results of analyses of kidney cancer incidence during the period 1958-1998 among 105,427 atomic bomb survivors. Poisson regression methods were used to derive estimates of associations between radiation dose (in sievert, Sv) and cancer of the renal parenchyma (n = 167), and cancer of the renal pelvis and ureter (n = 80). Heterogeneity by cancer site was tested by joint modeling of cancer risks. Radiation dose was positively associated with cancers of the renal pelvis and ureter [excess relative rate (ERR)/Sv = 1.65; 90% confidence interval (CI): 0.37, 3.78]. The magnitude of this association was larger than the estimated association between radiation dose and cancer of the renal parenchyma (ERR/Sv = 0.27; 90% CI = -0.19, 0.98). While the association between radiation and cancer of the renal parenchyma was of greater magnitude at ages <55 years (ERR/Sv = 2.82; 90% CI = 0.45, 8.89) than at older attained ages (ERR/Sv = -0.11; 90% CI = nd, 0.53), the association between radiation and cancers of the renal pelvis and ureter varied minimally across these categories of attained age. A test of heterogeneity of type-specific risks provides modest support for the conclusion that risks vary by kidney cancer site (LRT = 2.34, 1 d.f., P = 0.13). Since some studies of radiation-exposed populations examine these sites in aggregate, results were also derived for the combined category of cancer of the renal parenchyma, renal pelvis and ureters. Overall, there was a positive association between radiation and the combined category of cancer of the renal parenchyma, renal pelvis and ureters (ERR/Sv = 0.60, 90% CI: 0.09, 1.30). Updated follow-up of the LSS cohort provides substantial additional information on the association between radiation and cancer of the renal pelvis and ureter, a site not examined in recent reports on analyses of these data. The results are suggestive of differences between the different regions of the kidney in sensitivity to the carcinogenic effects of ionizing radiation.