The reflection of interannual Rossby waves over a 2.7-year period (i.e., November 1986 to August 1989) is examined in the western tropical Pacific utilizing altimetric seal level observations from the Geosat Exact Repeat Mission. In 1987 this reflection process generates upwelling Kelvin waves that tended to limit the growth of 1986--1987 El Ni¿o in the eastern equatorial Pacific. Two-dimensional autocorrelation and spectral analyses verify the existence of westward traveling Rossby waves in both the equatorial and off-equatorial domains, and eastward propagating Kelvin waves in the equatorial domain. Extended empirical orthogonal function (EEOF) analysis characterizes the time-space evolutionary behavior of the Rossby wave reflection process, extending from the El Ni¿o in 1986--1987 toward the anti-El Ni¿o in 1988--1989. Both the Philippines Archipelago in the northern hemisphere and the Solomon Archipelago/Bismarck Archipelago/New Guinea complex in the southern hemisphere appear to reflect incident Rossby waves, with maximum amplitude near 8¿ N and 8¿ S, generating Kelvin waves in the equatorial wave guide that conduct the anomalous signal eastward along the equator. This apparent reflection process is examined in terms of the linear theory, with sea level phase information in the northern hemisphere at the western boundary (130¿ W) expanded in terms of the Kelvin and symmetric Rossby meridional wave modes. These modes are formed as weighted sums of parabolic cylinder functions. This procedure, operating under the constraint that the reflection process occurred according to linear theory, determines the percent contribution that each incident Rossby wave mode amplitude has upon the amplitude of the reflected Kelvin wave. This percentage changed little over the evolution of the EEOF pattern. The amplitude of the first-mode Rossby wave (i.e., mode 2) accounts for most of the Kelvin wave amplitude (i.e., 70--80%), with the higher-mode Rossby wave amplitudes (i.e., 4, 6, and 8) contributing less (i.e., 20%, -10%, and 10%, respectively), interacting both constructively and destructively with the first. ¿ American Geophysical Union 1992