Geological models of buried thrust faults indicate that fault propagation folds (FPF) form and grow with a geometry that depends on that of the fault (Suppe, 1985; Suppe and Medwedeff, 1990). The displacement gradient fields for faults and kink folds are very similar, and both can be modeled using dislocations. In this paper we test a geological model of the FPF associated with the January 17, 1994 Northridge, California, earthquake to determine whether folding along axial planes inferred from geologic models accounts in part for the coseismic surface displacements measured with Global Positioning System (GPS). We test for coseismic deformation on both the main rupture plane and active axial planes of related folds by inverting for the displacements on dislocation planes in an elastic half-space. A model incorporating two axial planes is preferred to a model with a single rupture plane in a normalized root mean square (NRMS) sense; however, the distribution of axial plane displacements does not correlate with the displacements on the main rupture plane in the way expected for a fault propagation fold. Our results indicate that the deformation associated with folding is too distributed to be resolved on a discrete plane, that the deformation occurs interseismically, or that one or both of the kink bands does not exist. A model of a single elevated plane, which is parallel to, but not coplanar with, the aftershocks, is better in a NRMS sense than the FPF model, indicating that anelastic deformation in the hanging wall may be distributed.¿ 1997 American Geophysical Union