Plotting the movement of air through typical tsetse habitats in Zimbabwe revealed that it is liable to change direction by over 90° within a few metres. In winds of < 1 m/sec, there was a negative correlation between windspeed and the wind's tendency to meander, whether in thick bush or out in the open. And at the mean windspeed of wet season mopane woodland (0.3 m/sec), the wind changed direction by c. 15°/sec. Accordingly, host odour does not move downwind in simple trajectories that tsetse flies could easily follow. Even 5 m from a source, odour (modelled with smoke) approached a notional tsetse fly for a quarter of the time from >90° away from the true source direction. Also, air turbulence at common tsetse resting sites generates much nonsense information about the “true” direction of any odour-bearing wind. The suggested answer to how tsetse manage to find distant, invisible hosts, in spite of this confusing information from the wind, is that they progress by a biased random walk which is the outcome of their upwind anemotactic responses to odour-bearing wind and the proportion of the time for which the wind does blow in the “correct” direction. Computer simulation shows that this might work in principle.