The beta-particle emission from a sediment or rock sample can be measured very precisely using beta-counting instruments. The observed count rate is largely a function of the radionuclide concentration in the sample, so has the potential to provide a precise estimate of the natural radiation dose rate. However, the count rate is also sensitive to the attenuation of beta particles in the sample, and the relative proportions of the different radionuclide sources. Here we devise a correction for the self-attenuation effect using dilution analysis, and show that imprecise prior knowledge of radionuclide activity is sufficient for calculation of an accurate combined beta-plus-gamma dry dose rate. The method is tested on a selection of archive samples, and compared with results from high-resolution gamma-spectrometry. We show that with counting uncertainty ~2 %, and calibration uncertainty ~2 %, the total random uncertainty of the beta-plus-gamma dry dose rate is less than 3 %. For most natural sediments, this level of precision equal to, or better than, that obtainable with other methods.