Following a suggestion by Deardorff, Hess et al. have proposed that a necessary condition for the formation of dust devils is −h/L⩾50, where h is the convective boundary layer height and L the Obukhov length. A survey of pilots, air traffic controllers and meteorologists was conducted in Australia from 1 November 1987 to 31 January 1988 to test this hypothesis. The results of the survey showed that this criterion was satisfied for each dust devil event. Furthermore, the results were consistent with the idea that convective downdrafts play an important role in initiating dual devils. In strongly convective conditions there were two preferred heights attained by the dust—one at ≈0.09h(the height of the convergence wall between cells) and the other ≈0.51h(the height at which the vertical divergence of the updraft changes sign). The number density in these conditions is ≈4 per area of h2. The survey showed that dust devils on density currents have a different geometry; they have diameters five times ... Abstract Following a suggestion by Deardorff, Hess et al. have proposed that a necessary condition for the formation of dust devils is −h/L⩾50, where h is the convective boundary layer height and L the Obukhov length. A survey of pilots, air traffic controllers and meteorologists was conducted in Australia from 1 November 1987 to 31 January 1988 to test this hypothesis. The results of the survey showed that this criterion was satisfied for each dust devil event. Furthermore, the results were consistent with the idea that convective downdrafts play an important role in initiating dual devils. In strongly convective conditions there were two preferred heights attained by the dust—one at ≈0.09h(the height of the convergence wall between cells) and the other ≈0.51h(the height at which the vertical divergence of the updraft changes sign). The number density in these conditions is ≈4 per area of h2. The survey showed that dust devils on density currents have a different geometry; they have diameters five times ...