Global Positioning System (GPS) measurements of a geodetic network in southern and central California have been used to investigate the errors introduced by adopting different sets of stations as fixed. Such fixed points, called fiducial stations, are necessary to eliminate the errors of imprecise satellites orbits, which otherwise would dominate the error budget for distances greater than tens of kilometers. These fiducial stations aslo define the reference frame of the crustal deformation network. Establishing the magnitude of the effect of changing the fiducial network is essential for crustal deformation studies, so that these artifacts of the differences between fiducial networks used for the data analyses are not interpreted as geophysical signals. Solutions for a crustal deformation network spanning distances up to 350 km were computed with a variety of fiducial networks. We use fiducial coordinates determined from very long baseline interferometry (VLBI). We compare these solutions by computing the equivalent uniform strain and rotation that best maps one solution into another. If we use a continental-scale fiducial network with good geometry, the distortions between the solutions are about 10-8, largely independent of the exact choice of stations. The one case of large- scale fiducial network where the distortions are larger is when the three fiducial stations chosen all lie close to a great circle. Use of a fiducial network no larger than the crustal deformation network can produce apparent strains of up to 10-7. Our work suggests that fiducial corordinates determined from GPS data analysis may be used, although they should be determined using a consistent reference frame, such as provided by VLBL and satellite laser ranging. ¿American Geophysical Union 1991