We determine the burial dose in three known-age incompletely bleached fluvial samples using single grains of quartz. Estimation of burial dose in incompletely bleached samples requires that the characteristics of the well-bleached part of the distribution are known in order to distinguish between well-bleached and poorly bleached grains. It is especially important to investigate if the uncertainties assigned to individual estimates of dose adequately describe the observed variability in well-bleached dose distributions. We investigate this by quantifying the overdispersion in laboratory-bleached and gamma-irradiated samples. We show that there is a linear relationship between the over-dispersion and the mean dose. Knowing this uncertainty relationship enables us to estimate the burial dose by comparing the predicted uncertainty on the running mean with that calculated from actual dispersion. (c) 2007 Elsevier Ltd. All rights reserved.