Coherence analysis techniques are used to compare Southern Ocean wind forcing with Antarctic Circumpolar Current transport. Winds are derived from five different products: ERS winds that have been bin-averaged, weekly gridded ERS winds produced by the Centre ERS d'Archivage et de Traitement, 5 day winds from the Special Sensor Microwave Imager, analysis winds from the European Centre for Medium-Range Weather Forecasts, and reanalysis winds from the National Centers for Environmental Prediction. Barotropic transport is estimated from the pressure difference between bottom pressure gauges deployed on either side of Drake Passage by Proudman Oceanographic Laboratory as part of the World Ocean Circulation Experiment. Surface transport is estimated from TOPEX altimetry. Results indicate that transport and wind forcing are coherent over a broad range of frequencies, corresponding to time periods of roughly 10–256 days. Highest coherences occur for winds at latitudes on the south side of Drake Passage. Barotropic ocean transport lags wind forcing not by a constant time interval but by a constant phase lag of about one eighteenth of a cycle at a broad range of frequencies, suggesting that the oceanic response to wind is controlled by both the tendency term and a frequency-dependent viscous process. Surface transport lags by a longer phase interval. Wind stress curl north of Drake Passage is more coherent with transport than is wind stress curl in the latitudes of Drake Passage. Ocean transport lags wind stress curl, suggesting that transport fluctuations are not governed by a simple Sverdrupian vorticity balance. Like the observations, general circulation model transports from the Parallel Ocean Program and from the Parallel Ocean Climate Model are coherent with wind stress from the south side of Drake Passage and with wind stress curl from latitudes north of Drake Passage. Unlike the observations, model transport and bottom pressure vary almost simultaneously with the wind and do not replicate the observed phase lags, implying that the effective model viscosity may be too large.