Contemporary deformation of the Cascadia forearc consists of an elastic interseismic strain build-up as part of the subduction earthquake deformation “cycle” anda secular deformation primarily in the form ofarc-parallel translation and clockwise rotationofforearc blocks. Athree-dimensional (3-D) elastic dislocation model, constrainedby vertical deformation data, was developed previously to study the interseismic deformation. In this study, we develop a 3-D viscoelastic finite element model for the Cascadia subduction zone to study the temporal and spatial variations of interseismic deformation, and we compare the model results primarily with horizontal geodetic deformation observations. The model has an elastic lithosphere/slab and a viscoelastic mantle which has a viscosity of 1019 Pa s as constrained by recent postglacial rebound analyses. For comparison, we adopt a seismogenic zone geometry that was used in the previous elastic dislocation model, and we test the effects of different estimates of relative plate motion on the model predictions. Interseismic deformation is simulated by assigning a backslip rate to the locked zone of the subduction fault, preceded by an earthquake rupture of the same zone. Based on preliminary model results, we draw the following conclusions: (1) The deformation rate decreases through the interseismic period. A seaward motion is predicted for inland sites early in the interseismic period, an effect of postseismic creep of the mantle. (2) Model strain rates 300 years after the earthquake are consistent with the observed values, regardless of the plate motion models used. The horizontal velocities in northern Cascadia decrease landward at a slower rate than predicted by the elastic dislocation model, providing a better fit to observations. (3) Oblique subduction causes strain partitioning. As a result, the direction of local maximum contraction is much less oblique than plate convergence. The northerly direction of the GPS velocities in southern Cascadia represent a northward translation of the forearc. The secular deformation of the forearc may be partially accommodated through earthquake deformation cycles, but it may be better modeled as a process independent of the earthquake cycle.