Present bulk-sampling recommendations for the characterisation of fluvial sediments based upon ellipsoids of revolution do not take into account variations in particle shape and sorting. Disc-shaped particles with equivalent b -axes to spheres have lower relative masses and require more relaxed sampling criteria compared to existing recommendations, whereas heavier cube- and rod-shaped particles require more stringent criteria. Empirical data from the River Rede, Northumberland, UK, an upland gravel-bed channel, indicate that samples are dominated by discs. However, particle shape was not constant in every grain-size fraction; blocky material (cubes and rods) was more frequent in the > 63 mm fraction in comparison to the other finer fractions. Data based on the weight of the D 99 : 9 particle indicate that a larger sample weight is required when the coarsest size category of that sample is dominated by blocky material. Sorting also affects the size of the sample required to characterise a deposit; the more poorly sorted the deposit, the larger the sample has to be or the finer should be the truncation grain size. This is particularly evident in samples with a high matrix concentration. Two sample reference lines, developed from these data using the 0.1% by mass criteria, recommend larger minimum sample sizes (or finer truncation) in comparison to other published standards. Data from a second site support these results, indicating that they may be applicable to other poorly sorted upland gravel-bed streams. Analysis of the size category variance demonstrates that the new guidelines returned the lowest variance levels, suggesting that they should be used where particularly accurate sampling procedures are required.