1. The accuracy of death-rate estimates provided by the egg-ratio method was assessed by directly measuring the real death rates of six large laboratory Daphnia populations for 40 days.2. Of six models tested the Kawecki iterative model gave the best relationship to real death rates in terms of variance explained, bias, and fit at both 4-and 8-day sampling intervals. Although considered theoretically correct by persons familiar with the mathematics of egg-ratio models, the Edmondson-Paloheimo model gave slightly worse results in terms of bias and fit. 3. Using raw data, as opposed to smoothed data, gave the best fit between estimated and observed rates for both the Kawecki iterative and Edmondson-Paloheimo models. 4. Size-selective predation did not increase error in estimates of the death rate; however, non-steady-state dynamics were correlated with error in model estimates. 5. In general, the models gave a relatively good description of real dynamics, but they* regularly produced negative death rates, and were grossly wrong for some individual estimates. The present study also gives practical suggestions on application of this technique to field populations.