Monitoring sediment yields from catchments is important for assessing overall denudation rates and the impact of environmental change. One of the methods used to assess sediment yield is by quantifying sedimentation rates in reservoirs, lakes or small ponds. Before reliable sediment yield values (t ha−1 a−1) can be computed from such sedimentation records, the measured sediment volumes need to be converted to sediment masses using representative values of the dry sediment bulk density. In textbooks, simple relations predicting dry sediment bulk density from sediment texture, time since deposition and hydrologic condition are presented. In this study, 13 small flood retention ponds in central Belgium were sampled to reveal the variability in dry sediment bulk density and to test the commonly used relations to predict dry sediment bulk density. Dry sediment bulk density varies not only between the selected ponds (0·78–1·35 t m−3) but also within individual ponds (coefficient of variation at 95 per cent ranges from 7 to 80 per cent). The observed variability can be attributed primarily to the hydrologic condition of the retention pond and, also, to sediment texture. The existing relations are not a reliable predictor for the observed dry bulk densities, because they are primarily based on sediment texture. Thus, when using volumetric sedimentation data from small ponds with varying hydrologic condition to predict sediment yield, existing relations predicting dry sediment bulk density cannot be applied. Instead, frequent and dense sampling of sediments is necessary to calculate a representative value of the dry sediment bulk density. Copyright © 2001 John Wiley & Sons, Ltd.