The California Institute of Technology (CIT) operates a dense network of short-period vertical seismometers which we have used as a large seismic array to measure the azimuth of approach &zgr; and ray parameter dT/dΔ of teleseismic P waves. analysis of over 145 globally distributed events indicates that the &zgr; and dT/dΔ values measured by the array are in close agreement with the values predicted by the U.S. Geological Survey event locations. Most &zgr; anomalies (measured minus predicted values) do not exceed 1 ¿. The small magnitude of the anomalies suggests that the measured dT/dΔ values can be used to construct a radial velocity model for the lower mantle. The curve of dT/dΔ versus Δ is in agreement with the curbe determined for the Tonto Forest Seismological (TFSO) by Johnson (1969). The two curves differ in that there is not compelling evidence in the CIT data for first- or second-order discontinuities in the velocity gradient. It is shown that discontinuities in the dT/dΔ data proposed by Johnson near distances of 40.5, 49.5, 59.5, and 70.5 ¿ are due to a strong dependence upon azimuth to source. The same least squares straight line can be fit through the CIT and TESO dT/dΔ values. Residuals from this straight line display the same azimuthal dependence at both arrays. A velocity-depth curve has been constructed by Weichert-Herglotz inversion of the CIT dT/dΔ data. A current upper mantle model for western North America was used as the top of the velocity profile. It is also shown that if fine structure does exist in the lower mantle velocity profile, the best way to find it may be through a combined amplitude-dT/dΔ study.