A new all-sky imaging, wavelength scanning Fabry-Perot spectrometer was used to record high-resolution (R≃200,000) spectra of the &lgr;630 nm thermospheric optical emission above Poker Flat, Alaska. These spectra were used to derive spatially resolved maps of the horizontal wind vector at approximately 250 km altitude. We describe the procedure used to infer vector winds from line-of-sight Doppler shifts, along with its limitations. We present the time evolution of the vector wind fields obtained from this method for 6 nights of observation. Five of the 6 nights contained periods when we inferred the existence of significant curvature, divergence or shear in the thermospheric wind across our instrument's ~1000 km diameter field of view. The sixth night exhibited little spatial structure and is shown for comparison. We compare these results with a generic solar minimum winter time run of the National Center for Atmospheric Research's Thermosphere, Ionosphere, and Electrodynamics General Circulation Model. While agreement was good at the start and end of the night, considerable differences were found in the late evening and midnight sectors. Some possible origins for these discrepancies are proposed. In particular, we suggest that the F region horizontal wind may be deflected by upwelling vertical winds, which are in turn driven by E region heating in the auroral zone. We note that both the instrument used and our high-latitude implementation of the analysis procedure are new experimental techniques. Thus the data presented here should be regarded as preliminary and, if possible, be validated by comparison with results from other techniques. ¿ 1998 American Geophysical Union