An ecologic study was conducted to determine whether leukemia rates among children born to mothers residing in the vicinity of Ontario (Canada) nuclear facilities differed from the provincial average. Childhood leukemia mortality and incidence ratios for the period 1950 to 1987 were examined for five regions within 25 km of a nuclear facility. The nuclear facilities included a research development facility, a uranium refinery, a uranium mining and milling facility, and two nuclear-power generating stations. Overall, the observed number of leukemia deaths (O = 54) was slightly greater than expected (E = 46.1) during the period when the facilities operated, but the difference was not statistically significant (O/E = 1.17, 95 percent confidence interval [CI] = 0.88-1.53). There was no indication of a birth cohort effect, as the mortality ratios based on place of birth were not significantly greater than the mortality ratios based on place of death. In the analyses of individual facilities, CIs included the null value and were generally wide because of the small observed and expected numbers; however, in the vicinity of the nuclear generating stations, the observed relative excess of leukemia deaths (O = 36, O/E = 1.40) had a lower confidence limit that was close to the null value.