Three-dimensional spherical shell models of mantle convection have been formulated to investigate the effects of ocean-continent contrast (i.e., different viscosity for continental and oceanic lithospheres) and continental keels on plate motion, net rotation of lithosphere (i.e., degree 1 toroidal plate motion), and the geoid. The models include relatively realistic plate rheology (i.e., strong plate interiors and weak plate margins) and continental keel structure based on seismic models. The mantle flow in the models is driven by slab buoyancy. The models demonstrate that the ocean-continent contrast and continental keels have minor effects on the long-wavelength geoid (for degrees 2--5). However, continental keels overriding subducted slabs, for example in North America, can produce large negative gravity anomalies at a regional scale. This may have important implications to interpreting the gravity anomalies in North America. Our models show that when plates bounded by weak plate margins have the same thickness, neither weak plate margins nor the ocean-continent contrast efficiently excite net rotation of lithosphere, although they excite toroidal motion at higher harmonic degrees. However, plate thickness variations like continental keels excite the net rotation. In spite of their ability to excite the net rotation, continental keels as thick as 300 km cannot provide the necessary coupling to the deep mantle to produce the observed net rotation of lithosphere and slow continental motion if there is a weak asthenosphere underlying them. If the thermomechanic structure of North American continental upper mantle derived from seismic and heat flow studies is representative, our models suggest that the temperature- and pressure-dependent mantle rheology may not produce sufficiently high viscosity below continental keels that is needed to explain the observed plate motion. Our study hints the necessity to include more realistic treatment of plate boundaries in future studies to assess possible effects of plate-plate coupling on plate motion. ¿ 2001 American Geophysical Union