In several countries, epidemiological studies are being planned, or are in progress, to test the hypothesis that a causal relation exists between exposure to extremely low frequency magnetic fields and cancer incidence. One of the major difficulties in these studies is the development of valid and efficient protocols to assess magnetic field exposure. In studies focusing on residential magnetic fields, many researchers are turning to recently developed stationary automated magnetic field monitors to characterize exposure. We argue that a relatively small number of manually collected 'spot' measurements may be an adequate alternative which has several advantages. We compared a dichotomous exposure classification based on continuous magnetic field monitoring of 40 houses with that obtained through 'spot' measurements randomly sampled from the continuous records. We found that a single spot measurement had at least an 80% chance of classifying houses correctly and that this probability did not increase significantly as the number of readings was increased. We also calculated the sensitivity and specificity of various simulated measurement protocols and, from these, the effect of misclassification on estimates of relative risk. Since relatively large spatial variations in background magnetic field exist in many homes, we suggest that a small number of readings collected manually at several points within a residence may characterize the magnetic field better than continuous monitoring at one fixed location.