A unified method of slowness analysis has been applied to the NORESS and ARCESS arrays and to the three-component stations within these arrays, with the following results: (1) there is a relatively large scatter in the three-component solutions for events from the same source region, and (2) there are significant differences between the solutions at the different three-component stations. Evidence is presented that a uniform half-space model is inadequate to derive slowness solutions from three-component data: layering of the crust and upper mantle affect the surface response at long periods, and surface topography and near-surface structure perturb the surface particle motion and slowness solutions especially at short periods. Topographic effects are evaluated by means of numerical simulations based on a newly developed multiple scattering method. The theoretical results explain up to about half of the observed anomalies in the three-component solutions. On the other hand, the array solutions based on vertical-component phase delays are relatively stable, since the nongeometrical phase perturbations are negligible.