The identification of seismically active fault planes is important for assessment of earthquake potential as well as development of neotectonic models and analysis of stress systems for a region. An automated set of computer programs, SEISPLN, has been developed for the identification of three-dimensional (3-D) alignments of seismic hypocenters. The SEISPLN technique is based on rotating a 3-D elongate grid system through the hypocenter distribution. The number of events in each cell is tabulated, and those cells which contain statistically significant numbers of events are considered to be potential fault surfaces. A multiple linear regression is applied to the set of hypocenters in each delineated cell to model the 3-D geometry of each of the inferred fault surfaces. We applied the technique to earthquakes located by the Washington Regional Seismic Network to separate the seismic events associated with the subducted Juan de Fuca (JDF) plate from events in the overlying crust. A total of 259 JDF-associated events were delineated, and the resulting 3-D models of the JDF slab beneath Washington reveal a complex combination of arching and flattening. Also, a steep orientation for the slab is observed beneath the southern portion of the Washington Cascades, while a much shallower orientation is inferred beneath the northern Cascades.