Interplanetary scintillation observations are used to build up a picture of the distribution of solar wind velocities in latitude and longitude. Such a picture is derived for periods of several rotations during the years 1973-1977. The observations are interpreted as if they were point observations and mapped to the sun with the constant radial velocity assumption. Since only eight scintillating raido sources are available, a useful instantaneous picture cannot be obtained, so some integrating in time is required to build up coverage. This integration is restricted to a few solar rotations by both the changing spatial distribution of the radio sources and possible temporal variation of the corona and solar wind. The observed average distributions of velocity are found to be not isotropic or even symmetric about the rotation axis of the sun. The averaged distributions are found to be similar to and to evolve as the distribution of low intensities in white light coronameter data taken at 1.5 R⊙ and suggest that high-speed solar wind originates in regions of low coronal density. This extends to high latitudes the evidence for coronal holes as the source of high-speed streams, including the polar holes as the source of high speed at polar latitudes.