Transformer stations in apartment buildings may offer a possibility to conduct epidemiological studies that involve high exposure to extremely low frequency magnetic fields (MF), avoid selection bias and minimize confounding factors. To validate exposure assessment based on transformer stations, measurements were performed in thirty buildings in three Finnish cities. In each building, spot measurements in all rooms and a 24-h recording in a bedroom were performed in one apartment above a transformer station (AAT), in one first floor (FF) reference apartment, and one reference apartment on upper floors (UF). The apartment mean of spot measurements was 0.62 microT in the AATs, 0.21 microT in the FF and 0.11 microT in the UF reference apartments The 24-h apartment mean (estimated from the spot measurements and the bedroom 24-h recording) was 0.2 microT or higher in 29 (97%) AATs, in 7 (25%) FF and in 3 (10 %) UF reference apartments. The corresponding numbers for the 0.4 microT cut-off point were 19 (63%), 4 (14%), and 1 (3.3%). The higher MF level in the FF reference apartments indicates that they should not be considered "unexposed" in epidemiological studies. If such apartments are excluded, a transformer station under the floor predicts 24-h apartment mean MF with a sensitivity of 0.41 (or 0.58) and a specificity of 0.997 (or 0.97), depending on the MF cut-off point (0.2 or 0.4 microT). The results indicate that apartments can be reliably classified as high and low MF field categories based on the known location of transformer stations.