The available studies on upper-mantle structure in North America can be broadly divided into two categories: delineation of one-dimensional models, that is, the determination of P- and S-velocities as a function of depth; and computation of two- and three-dimensional models to take into account lateral heterogeneities in structure. About 50 one-dimensional models based on traveltimes, synthetic seismograms, and surface-wave velocities are currently available for continental North America. The gross features of these models are sharp velocity increases at depths near 400 and 650 km in the upper mantle beneath the whole continent and the presence of a low-velocity layer in the uppermost part of the upper mantle in the western half of the continent. A few other velocity discontinuities and velocity-gradient changes have also been documented. The most important finding from the available studies is a quantitative confirmation of what was suspected even in the early 1950s, namely, that the upper-mantle structure, particularly the structure related to the low-velocity layer, is drastically different in the tectonically active Cordillera from the stable central and eastern shield of North America. In western North America, in general, the upper-mantle velocities are low, and the low-velocity zone is well developed and occurs at shallow depths. On the other hand, in the central and eastern parts of the continent the upper-mantle velocities are higher than in the west, and a low-velocity layer—if present at all—tends to be deep and to have a smaller velocity contrast than in the west. Available data and modeling techniques are inadequate to unambiguously resolve spatial variation in the depths to the 400-km and 650-km discontinuities and the boundaries of the low-velocity layer in North America. Apart from the broad division of the one-dimensional models into tectonically active and shield structures, any further finer scale quantitative division of the models within each tectonic unit is not warranted by the available data. Qualitatively, however, it is clear that such finer scale differences do exist, particularly in the upper 250 to 300 km of the upper mantle. The laterally heterogeneous structure of the upper mantle in North America has been studied by three-dimensional modeling using teleseismic P- and S-wave residuals. Three-dimensional inversion of P- and S-wave residual data collected over a substantial part of the North American continent show the existence of long-wavelength heterogeneous structure extending throughout the upper mantle. In addition, short-wavelength lateral heterogeneities are revealed by regional investigations. These include heterogeneous velocity structures associated with: (1) the ongoing subduction of the Juan de Fuca plate beneath Washington and Oregon, (2) the cessation of subduction of the Farallon plate beneath California during early Tertiary time, (3) rifting in Imperial Valley, (4) hot-spot magmatism in the Yellowstone Plateau, (5) large-scale asthenospheric upwelling in the region of the Rio Grande Rift, and (6) the orogenic belts in the northeastern United States.