Concern with health effects of extremely low frequency magnetic fields has been raised by epidemiologic studies of childhood cancer in relation to proximity to electric power distribution lines. This case-control study was designed to assess the relation between residential exposure to magnetic fields and the development of childhood cancer. Eligible cases consisted of all 356 residents of the five-county 1970 Denver, Colorado Standard Metropolitan Statistical Area aged 0-14 years who were diagnosed with any form of cancer between 1976 and 1983. Controls were selected by random digit dialing to approximate the case distribution by age, sex, and telephone exchange area. Exposure was characterized through in-home electric and magnetic field measurements under low and high power use conditions and wire configuration codes, a surrogate measure of long-term magnetic field levels. Measured magnetic fields under low power use conditions had a modest association with cancer incidence; a cutoff score of 2.0 milligauss resulted in an odds ratio of 1.4 (95% confidence interval (CI) = 0.6-2.9) for total cancers and somewhat larger odds ratios (ORs) for leukemias (OR = 1.9), lymphomas (OR = 2.2), and soft tissue sarcomas (OR = 3.3). Neither magnetic fields (OR = 1.0) nor electric fields (OR = 0.9) under high power use conditions were related to total cancers. Wire codes associated with higher magnetic fields were more common among case than control homes. The odds ratio to contrast very high and high to very low, low, and buried wire codes was 1.5 (95% CI = 1.0-2.3) for total cases, with consistency across cancer subgroups except for brain cancer (OR = 2.0) and lymphomas (OR = 0.8). Contrasts of very high to buried wire code homes produced larger, less precise odds ratios of 2.3 for total cases, 2.9 for leukemias, and 3.3 for lymphomas. Adjusted estimates for measured fields and wire codes did not differ from crude results, indicating an absence of confounding. Limitations to the study are nonresponse (especially for field measurements), differential mobility of cases and controls, and a presumably nondifferential exposure misclassification from the use of imperfect surrogates for long-term magnetic field exposure history. In spite of these concerns, the results encourage further examination of the carcinogenic potential from this form of nonionizing radiation.