Many natural deposits contain grains that have different burial histories, but reliable procedures to extract the component doses from mixed-dose samples have not been developed in optical dating. Here we present results for synthetic two- and three-component mixtures of data derived from laboratory-dosed sedimentary quartz using a single-aliquot regenerative-dose protocol and statistical models for finite mixtures. Composite sets of data were created from the doses measured for individual grains that had been bleached by sunlight and then given a beta dose of 5, 10 or 20 Gy. We found that the correct number of dose components, corresponding doses, and relative proportions of each component may be estimated if the within-component dispersion is small or is known. We also found, however, that recuperation of the OSL signal in sun-bleached quartz may produce artificially high dose estimates for some grains, an effect that should be taken into account when modelling natural mixtures.