Following a decision taken by the National Co-ordinators of the OECD Test Guideline Programme and the OECD Risk Assessment Advisory Body (RAAB) at their joint session in December 1995, an OECD Workshop on Statistical Analysis of Aquatic Ecotoxicity Data was held in Braunschweig, Germany on 15-17 October 1996. The workshop was hosted by the Biologische Bundesanstalt für Land-und Forstwirtschaft (BBA) in Braunschweig and was chaired by Dr Arno Lange of the German Umweltbundesamt (UBA). The objectives of the workshop were: • to review the options available for the analysis of data from ecotoxicity tests; • to compare their advantages and disadvantages; • to recommend (a) the most appropriate approach for deriving a summary parameter(s) which has (have) scientific validity, and (b) further work to be undertaken by the OECD and/or others, as appropriate. In plenary and working group sessions participants discussed statistical data analysis appropriate for single-species chronic/subchronic studies using a number of test concentrations, i.e. dose-response tests. Aquatic tests served as a basis for these discussions, however the issues addressed may be similar for toxicity tests in general. Background documents had been prepared on the following main existing data analysis approaches for such tests: analysis of variance/hypothesis testing (“ANOVA/NOEC approach”); regression analysis (based on empirical models); and mechanistic modelling (theory-based). It was concluded that the NOEC, as the main summary parameter of aquatic ecotoxicity tests, is inappropriate for a number of reasons (see detailed discussion in the report) and should therefore be phased out. It was recommended that the OECD should move towards a regression-based estimation procedure. The time course of effects should be incorporated in the analytical procedures, and the OECD should initiate a study of the available time-dependent regression models (both mechanistic and empirical) in order to select those which best meet its needs. The study should also address the issue of appropriate values of x for ECx and the optimal experimental designs. A steering group should be set up to direct the mathematical, statistical and biological work required to take the workshop recommendations forward. This group should include representatives from the appropriate scientific and regulatory communities.