We examine simulations of the mean seasonal cycle in the tropical Pacific using a multiple vertical model linear numerical model forced with three different surface wind stress products averaged over the period 1979--1981. The model is run to equilibrium for each of four vertical modes, and results are summed. Simulated mean seasonal cycles in dynamic height and sea level are then compared with observed variations based on expendable bathythermograph and island tide gauge data averaged over the same 1979--1981 period. All simulations show characteristic features of the mean meridional ridgetrough structure in surface topography. However, north and south equatorial ridges at 20¿N and 20¿S are much higher than those observed, only weak equatorial ridges are generated near 4¿N, and none of the simulations exhibits a significant equatorial trough. These discrepancies are due principally to limitations in model physics and in the wind forcing. Observed and modeled mean seasonal variations in surface height are of the order of a few centimeters. Coherence estimates of 0.5--0.7 are found between the model simulations and the observations for the 1 cycle per year harmonic, which dominates the seasonal cycle over most of the tropical Pacific. This suggests that about 25--50% of the variance in the observed annual surface height is accounted for by the linear model, given current estimates of the surface wind field. Harmonics higher than the annual are less well modeled because modedled because of their weaker signal levels. Regional patterns are observed in coherence levels between modeled and observed variability; i.e., longitudinally, the eastern Pacific is most poorly modeled, while latitudinally, the equatorial band (5¿N to 5¿S) is best modeled. However, no wind stress product is clearly superior to the others for simulating the mean seasonal cycle. Thus uncertainty in the surface stress field remains a fundamental obstacle to more accurate modeling of the variability in tropical. Pacific sea surface topography.