The spreading of smoke puffs in the atmosphere should be governed by the laws of relative diffusion that have been advanced by Brier and by Batchelor, and not (contrary to what has been previously assumed) by Taylor's fixed-source diffusion law. The predicted mean-square puff spreading should proceed as time cubed, according to Batchelor's similarity theory of relative diffusion. Recent detailed smoke-puff measurements reported by Frenkiel and Katz, and Kellogg, make it possible to test the similarity theory of relative diffusion. Examination of these data indicates the existence of a t3 spreading regime. Values of the rate of eddy-energy dissipation are also inferred from the smoke-puff data.