The center of Hurricane Felix passed 85 km to the southwest of the Bermuda Testbed Mooring (BTM; 31¿44'N, 64¿10'W) site on 15 August 1995. Data collected in the upper ocean from the BTM during this encounter provide a rare opportunity to investigate the physical processes that occur in a hurricane's wake. Data analyses indicate that the storm caused a large increase in kinetic energy at near-inertial frequencies, internal gravity waves in the thermocline, and inertial pumping, mixed layer deepening, and significant vertical redistribution of heat, with cooling of the upper 30 m and warming at depths of 30--70 m. The temperature evolution was simulated using four one-dimensional mixed layer models: Price-Weller-Pinkel (PWP), K Profile Parameterization (KPP), Mellor-Yamada 2.5 (MY), and a modified version of MY2.5 (MY2). The primary differences in the model results were in their simulations of temperature evolution. In particular, when forced using a drag coefficient that had a linear dependence on wind speed, the KPP model predicted sea surface cooling, mixed layer currents, and the maximum depth of cooling closer to the observations than any of the other models. This was shown to be partly because of a special parameterization for gradient Richardson number (RgKPP) shear instability mixing in response to resolved shear in the interior. The MY2 model predicted more sea surface cooling and greater depth penetration of kinetic energy than the MY model. In the MY2 model the dissipation rate of turbulent kinetic energy is parameterized as a function of a locally defined Richardson number (RgMY2) allowing for a reduction in dissipation rate for stable Richardson numbers (RgMY2) when internal gravity waves are likely to be present. Sensitivity simulations with the PWP model, which has specifically defined mixing procedures, show that most of the heat lost from the upper layer was due to entrainment (parameterized as a function of bulk Richardson number RbPWP), with the remainder due to local Richardson number (RgPWP) instabilities. With the exception of the MY model the models predicted reasonable estimates of the north and east current components during and after the hurricane passage at 25 and 45 m. Although the results emphasize differences between the modeled responses to a given wind stress, current controversy over the formulation of wind stress from wind speed measurements (including possible sea state and wave age and sheltering effects) cautions against using our results for assessing model skill. In particular, sensitivity studies show that MY2 simulations of the temperature evolution are excellent when the wind stress is increased, albeit with currents that are larger than observed. Sensitivity experiments also indicate that preexisting inertialmotion modulated the amplitude of poststorm currents, but that there was probably not a significant resonant response because of clockwise wind rotation for our study site.