The Doppler Wind Experiment on the Galileo probe provided the first in situ data on wind speeds in Jupiter's atmosphere. Initial analysis of the results indicated that wind speeds increase with depth, rather than decaying to zero below the cloud tops or remaining relatively constant as had previously been assumed. But this earlier analysis was subject to several potential sources of error, as highlighted by the fact that wind speeds measured at the cloud tops did not seem to match those inferred from tracking clouds in images obtained by the Voyager spacecraft. Here we report new analyses of the probe data that use a corrected treatment of the timing errors, adopt the measured (rather than predicted) descent trajectory, and incorporate a new calibration of the instrumentation that takes into account the unexpectedly high temperatures encountered by the probe. We determine wind speeds at the cloud tops (700-mbar level) in the range 80–100 m s⁻¹, in agreement with the results of cloud tracking; the speed increases dramatically between 1 and 4 bar, and then remains nearly constant at ∼170 m s⁻¹down to the 21-bar level. The increase in wind speed implies a latitudinal density gradient of 0.5% per degree in the 1–2 bar altitude range, but whether these winds are driven by internal heat or absorbed sunlight remains uncertain.