Moored thermistor chains at 2 ¿N and 2 ¿S and current-temperature moorings at 0¿ are used to examine the accuracy of geostrophically estimated zonal velocity on the equator in the eastern (110 ¿W) and western (165 ¿E) Pacific. The meridionally differentiated form of the geostrophic balance is used to eliminate large errors due to wind-balanced cross-equatorial pressure gradients. Statistical analyses indicate that for time scales longer than 30--50 days, the observed and geostrophically estimated zonal velocities are similar (correlation coefficients of 0.6--0.9 and comparable amplitudes). Thus low-frequency equatorial current oscillations are reasonably well represented by the geostrophic approximation. However, the mean currents are poorly resolved with the available array. In the eastern Pacific the mean zonal speed difference over the 10-month comparison period is 25 cm s-1 at 25 m and increases to 60 cm s-1 at 125 m. At 165 ¿E mean difference in the upper 250 m are typically 50 cm s-1 over a 4-month record. The principal reason for these large mean differences is that the meridional scale of the mean currents is smaller than the spacing of the moorings. Comparison of observed and geostrophic velocity profiles obtained from shipboard sampling indicates that meridional spacing of about 1¿ latitude would be optimum for estimating the zonal velocity. ¿ American Geophysical Union 1989