Wind measurements from 39 chemical release rockets launched from auroral belt and polar cap sites between 1967 and 1979 are summarized and interpreted. Multipoint measurements between 200 and 320 km altitude were obtained from all flights by tracking the neutral strontium clouds produced as by-products of Ba-CuO releases. These releases, producing Ba+ clouds, also provided the simultaneous measurements of ion drift that are needed for wind analysis. Trail releases of trimethylaluminum and lithium provided measurements between 90 and 200 km from 19 rockets. The wind measurements at different local times and latitudes are disorganized when presented in geographic and solar local time coordinates. An organized wind system at altitudes, above 160 km emerges when wind vectors are presented in magnetic local time and invariant latitude coordinates. The dominant role of electric field ion drag as a driving force and the inertial characteristics of the wind system are clearly illustrated when the wind vectors are superposed on model patterns of the global high-latitude ion convection. At altitudes above the 110 km shear region and below a transitional zone in the 140--160 km range, relationships between the observed winds and the causative mechanisms become less apparent as a consequence of the slower response to changing ion drag conditions. A notable exception occurs in the polar cap where high velocity winds are observed at all altitudes under the influence of continuous antisolar ion flow. A key factor affecting the wind pattern is the formation of a pressure ridge along the interface zone between the polar cap and the auroral belt in the evening hours. Back pressures from this ridge suppress the Coriolis forces in the adjacent regions and contribute to the development of a dusk to dawn component in the antisolar polar cap flow. The observations clearly indicate that the antisolar polar cap wind loses momentum and deposits energy in the early morning auroral belt. At high altitudes (i.e., 160--320 km) there is directional continuity but a velocity loss as the polar cap wind inertially penetrates the morning auroral region. At lower altitudes (i.e.,<140 km) the observations are less definitive but indicate that the wind is discontinuous along the morning auroral belt interface. In general, the observations do not reveal wind characteristics that can be attributed to pressure gardients from joule heating or heating by precipitating particles. A possible, but speculative, exception is that the perturbing effects of heating may explain why observed velocities tend to be lower than velocities calculated for model ion drag conditions. Other wind effects from high-latitude heating apparently have less influence than the time-space variability of the electric field and are thus obscured by these variations. The existence of the classical day-night flow from dayside EUV heating is demonstrated by measurements from three rockets launched at subauroral latitudes on the dayside. The influence of this flow at higher latitudes is similarly not identifiable in the presence of the ion drag winds. A number of the gross characteristics of the observed high latitude wind pattern are simulated reasonably well by Fuller--Rowell and Rees' recent model.