This paper describes the study design, measurement protocols, and results of a project examining residential magnetic -field exposures at eight sites across the contiguous United States. The goal of the project was to investigate surrogates that have been used in epidemiologic studies to characterize residential magnetic -field exposure. These surrogates include: personal -exposure ( PE ), fixed -location long -term ( LT ), and outside and inside point -in -time ( PIT ) magnetic -field measurements; net -service ( or ground current ) measurements; and the ''wire -code category'' of the residence. ( The latter is a surrogate for magnetic -field exposure based on the nature and proximity of electric power lines outside the house. ) Measurements were conducted on four visits to each of eight sites between January 1994 and June 1997 for a study population of 218 single -unit detached dwellings. Information on the residence, residents, and neighborhood was collected. A simple random sample of 392 single -unit detached dwellings at the sites was used to create a weighted sample of houses representative of the population of single -unit residences. The correlations among the various types of 60 -Hz magnetic -field measures were relatively strong ( Pearson r > 0.74, Spearman > 0.78 ). Variability of PE and LT measurements, as measured by the standard deviations during a visit, was independent of wire -code category. Visit means for PE, LT, and outside and inside PIT were well correlated over periods between visits of from 1.5 to 20 months ( r > 0.62, > 0.76 ). These results support the use of survey measurements ( less demanding than personal monitoring ) to represent exposure that occurred up to 20 months in the past. The principal component of the total variance in PE measurements was the between -house variance; betweenvisit and between -site variances were generally less important. This supports the sampling of many houses with relatively few visits in residential exposure characterization studies. There was a trend for presumably higher wire -code categories to be associated with higher field summary measures for all summary measures related to magnetic -field magnitude, including PE and LT resultant, and inside and outside resultant ( 60 Hz ) and harmonics. However, because of the overlap in field levels between categories, wire code was not a good predictor of magnetic -field levels, accounting for less than 21% of the variance in magnetic -field measurements.