We report spatial separation of extracellular vesicle (EVs) populations based on particle size by using an approach that exploits Marangoni flow and the coffee-ring effect in microdroplets. Sequential transfer of a drying droplet progressively increases the mean size of EVs in the sample by repeated subsampling of a droplet during coffee-ring formation. This method allows size-based sorting, separation, and eventual retrieval of EVs for RNA and protein analysis. To demonstrate the biomedical relevance of this method, EVs from prostate cancer patients were analyzed; results revealed that the expression of cancer-associated genes and proteins was higher in small EVs than in large EVs. This ability to sort EVs using a combination of coffee ring with Marangoni flow and sequential droplet-transfer allows analysis of subpopulations of EVs, and will facilitate further studies of EVs.