We studied the global climatology of mid- and low-latitude F region daytime neutral winds using extensive measurements by the Wind Imaging Interferometer (WINDII) instrument on board the Upper Atmosphere Research Satellite (UARS). Quiet-time winds are mostly poleward and westward during the day, and are generally 5-20 m/s smaller in the longitudinal sector closest to the magnetic pole, compared to longitudinally averaged winds. The pre-noon zonal winds are less westward with increasing solar flux, while the post-noon meridional winds are less poleward . Our quiet-time results are in good agreement with the NCAR ThermosphereIonosphere- Electrodynamics General Circulation Model (TIEGCM). We computed residual winds by subtracting quiet-time values calculated along the satellite orbit, which effectively removes average measurement bias. Using these residuals, we studied the average change in the winds under disturbed conditions. The zonal disturbance winds are mostly westward, increase with latitude, and have largest values in the late afternoon sector. In general, the meridional perturbation winds are equatorward, increase linearly with latitude, and decrease from morning to afternoon hours. The zonal and meridional perturbations increase roughly linearly with Kp. We developed empirical analytical models for the disturbance winds from 60° to the equator; these model winds are in poor agreement with results from the empirical Horizontal Wind Model. There are also important discrepancies between the average perturbations winds from WINDII and TIEGCM. We studied the average time-dependent development of disturbance winds during geomagnetic storms. The onset of a storm is characterized by equatorward surges, mostly in the morning sector, that reach the equator in about 2 h. These surges lessen 5-6 h after the onset of a storm, but subsequently increase, reaching their largest values about 15 h after the start of the storm before leveling off or diminishing. Following the end of typical storms, the disturbance winds decrease quickly but oscillate for at least one 8-9 h cycle. We developed time-dependent analytical models of the disturbance winds as a function of the polar cap index at key storm time lags. Our results are consistent with predictions from theoretical models. (146 pages)