This paper questions the usefulness of the no-observed-effect level (NOEL) as a summary statistic for ecotoxicological experiments. Quantification of the NOEL depends critically on size and variability of an experiment: smaller and less precise experiments lead to higher values for the NOEL, which gives the wrong signal to suppliers of a NOEL. The NOEL will generally lie in a dose range where possible effects cannot be excluded, even within the context of the experiment. Hence it does not form a suitable starting point for extrapolation to field situations. Alternative methods of finding a dose with a negligible effect are discussed. A two-step procedure is proposed that involves finding the dose whose effect is at most 25%, followed by linear extrapolation to a dose whose effect is acceptably small. The procedure leads to higher values if efforts are made to increase the power of the experiment.