Ocean tidal effects on universal time and polar motion (UTPM) are investigated at four nearly diurnal (K1, P1, O1, and Q1) and four nearly semidiurnal (K2, S2, M2, and N2) frequencies by analyzing very long baseline interferometry (VLBI) data extending from 1978 to 1992. We discuss limitations of comparisons between experiment and theory for the retrograde nearly diurnal polar motion components due to their degeneracy with prograde components of the nutation model. Estimating amplitudes of contributions to the modeled VLBI observables at these eight frequencies produces a statistically highly significant improvement of 7 mm to the residuals of a fit to the observed delays. Use of such an improved UTPM model also reduces the 14--30 mm scatter of baseline lengths about a time-linear model of tectonic motion by 3--14 mm, also with high significance levels. A total of 28 UTPM ocean tidal amplitudes can be unambiguously estimated from the data, with resulting UT1 and PM magnitudes as large as 21 μs and 270 microarc seconds (μas) and formal uncertainties of the order of 0.3 μs and 5 μas for UTI and PM, respectively. Empirically determined UTPM amplitudes and phases are compared to values calculated theoretically by Gross from Seiler's global ocean tide model. The discrepancy between theory and experiment is larger by a factor of 3 for UT1 amplitudes (9 μs) than for prograde PM amplitudes (42 μas). The 14-year VLBI data span strongly attenuates the influence of mismodeled effects on estimated UTPM amplitudes and phases that are not coherent with the eight frequencies of interest. Magnitudes of coherent and quasi-coherent systematic errors are quantified by means of internal consistency tests. We conclude that coherent systematic effects are many times larger than the formal uncertainties and can be as large as 4 μs for UT1 and 60 μas for polar motion. On the basis of such realistic error estimates, 22 of the 31 fitted UTPM ocean tidal amplitudes differ from zero by more than 2&sgr;. ¿ American Geophysical Union 1993