Observations of semidiurnal currents from high-frequency radio Doppler current meters and moored acoustic Doppler current profilers (ADCPs) in the Kauai Channel, Hawaii, are described and compared with two primitive equation numerical models of the tides. The Kauai Channel, separating the islands of Oahu and Kauai, is a site of strong internal tide generation by the barotropic tides flowing over Kaena Ridge, the subsurface extension of Oahu. The nature and impacts of internal tide generation in the Kauai Channel were intensively studied during the 2002–03 near-field component of the Hawaii Ocean Mixing Experiment. Comparisons of observed coherent (i.e., phase locked to the astronomical forcing) M2 and S2 surface currents with model predictions show good agreement for the phases, indicating propagation of internal tides away from the ridge. Although the predicted M2 and S2 surface currents are similar (except for their magnitudes), as expected for internal waves at periods closer to each other (12.4 and 12 h, respectively) than to the inertial period (33 h), the observed M2 and S2 surface currents differ significantly. The S2 kinetic energy pattern resembles the predicted pattern. In contrast, the observed structure and magnitude of the more important M2 kinetic energy pattern differs significantly from the model predictions. The models predict a band of enhanced M2 surface kinetic energy 30–40 km from the ridge axis, corresponding to the first surface reflection of internal tide beams generated on the ridge flanks. The beams are clearly observed by the moored ADCPs, albeit with weaker amplitudes than predicted. Observations at the surface show an area of enhanced kinetic energy that is 10–20 km farther away from the ridge than predicted, with weaker magnitude. Observed M2 surface currents also exhibit apparent seasonal variability, with magnitudes weaker in spring 2003 than in fall 2002. Complex-demodulated semidiurnal currents exhibit significant temporal variability in amplitude and phase, not only because of the interference between semidiurnal constituents (e.g., the spring–neap cycle) but also on shorter and irregular time scales. The result is that ~20% of semidiurnal energy is incoherent with astronomical forcing. Furthermore, the temporal variability is not spatially coherent; the spatial patterns of semidiurnal kinetic energy resemble those predicted by the numerical models during the strongest spring tides but differ from them at other times. As a result, M2 and S2 kinetic energy patterns phase locked to the astronomical forcing differ from each other. Some features of the observed spatial pattern and amplitude modulations can be qualitatively reproduced by a simple analytical model of the effects of homogeneous barotropic background currents on internal tide beams.