Abstract: A computer implemented method and apparatus is disclosed for automatically picking faults in a recorded 3D seismic trace data volume. The method employs test planes, which are mathematically inserted into the seismic data volume to approximate dip and azimuth of a potential fault plane surface. A large number of data points, which are selected points on the seismic traces, are defined within the seismic volume, such that each test plane positioned in the seismic volume contains data points corresponding to at least a significant portion of a trace. The method then determines a factor for each data point which is representative of the probability that the data point resides on a fault plane. This probability is based on planar discontinuity and average amplitude difference between corresponding traces in adjacent parallel test planes, and the method selects locations, in an x, y grid, of a strip of locations having high probability of residing on a fault surface.