Between February and April 2003, a total of 80 single-dwelling households close (< 70 m) to high-tension (161 or 345 kV) power lines in a northern community of Taiwan received measurements of indoor extremely low-frequency (ELF) magnetic field for 72 h. Measurements were performed with EMDEX II meter at a sampling rate of every 300 s, yielding some 860 readings of ELF magnetic field for each household. In addition to the 72-h mean ELF magnetic field, we also calculated arithmetic means of the first 2, 6, 12, and 288 readings taken in each household to represent the information on spot, 30-min, 60-min, and 24-h exposures, respectively. The mean 72-h exposure to ELF magnetic field for the 80 study households was estimated at 0.80 micro-Tesla (microT) with a standard deviation (SD) of 1.13 microT. The mean for pot, 30-min, 60-min, and 24-h exposure was 0.88 (SD 1.38), 0.90 (SD 1.40), 0.9 (SD 1.17), and 0.83 (SD 1.17) microT, respectively. There were high agreements, indicated by a nearly perfect intra-class correlation coefficient, between 72-h mean exposure and those short-term exposure measures. Additionally, the sensitivity and specificity of various short-term exposures in the prediction of 72-h exposure greater than 0.4 microT were similar at values of 0.82-0.87 and 0.93-0.95, respectively. This study indicates that short-term measurements of indoor ELF magnetic field seem adequate to represent the mean 72-h exposure, but tended to overpredict 72-h exposure greater than 0.4 microT. Further investigation is needed to assess whether these findings can be replicated in households far away from high-tension power lines.