The necessary condition for the seismic refraction method to succeed i
s that the refracted first arrivals from each layer in a multilayered
earth system should be detected on a seismogram as first arrivals, and
this is possible only when velocities of all underlying layers are su
ccessively greater. The usual procedure to interpret the refraction tr
avel times is to fit such a data set with several intersecting straigh
t lines by employing a visual technique which may lead to errors of su
bjective judgment, as the velocity model depends on the selection of v
arious line segments through the data. To remove the visual fit we pro
pose here a layer stripping method based on minimum intercept time, ap
parent velocity, rms residual, and maximum data points by least-square
s fitting to yield several intersecting straight lines. Once data are
segmented out, the conventional equations can be used to determine the
velocity structure.