Validity of exposure assessment methods has been a major concern in epidemiologic studies of magnetic field exposure and cancer. We conducted a study to evaluate the relative importance of distance to power lines and calculated historical magnetic fields when estimating past magnetic field exposure. Another goal was to compare results based on various estimates of magnetic field exposure, to assess the importance of short-term variability in magnetic fields, time between diagnosis and measurement, and sources of magnetic field exposure. We used data from a Swedish case-control study of residential exposure to magnetic fields and cancer. Childhood leukemia risk was associated with calculated historical annual average magnetic fields regardless of distance, and the association with distance disappeared when both variables were entered into the same logistic regression model. Relative risks for measurements at the time of the study (contemporary annual average fields, spot calculations, and spot measurements) were all close to or below unity. The results support the hypothesis that the difference between results using historical calculations and spot measurements is explained by the time interval between diagnosis and contemporary magnetic field estimates.