Assuming the human head to bf a spherical conductor, we used a numerical st
udy to obtain the localization error of the helmet-shaped 150-channel SQUID
magnetoencephalography system. 95 % confidence limits for the dipole param
eters were estimated using the semi-analytical formula. The error was compa
red with the results of a Monte Carlo simulation obtained by solving the in
verse problem. Using the methods, we found that the magnetometer system was
superior to the gradiometer system in source localization ability.