The performance of two methods, used to localize brain activity from e
voked potential fields measured on the scalp, was assessed in a tank m
odel of the human head. This physical model contained a human skull en
cased in a polymer simulating the resistivity and geometry of brain an
d scalp. The dipole localization method mislocalized the positions of
known dipole sources by several centimeters. The mislocalization was s
ystematic. The dipoles were localized too deeply in the head. The Lapl
acian method yielded a field resembling the brain surface field (epico
rtical potential field) provided that the iso-potential contours of th
e scalp field closed within the measurement range. Clipping resulted i
n a serious mislocalization of the position of the peak of the epicort
ical potential field.