All participants agreed that there are now great possibilities for improvements in the rationale regarding the use of experimental animals in safety studies. The discussion centred around the three Rs-the replacement by alternative methods, the reduction in the number of animals used, and the refinement of the techniques applied. Although it is not possible to replace the whole animal or human in pivotal safety studies, the question is, are all the test systems widely used today really suitable for indicating or assessing hazards for human safety? It was emphasized that we have to take greater advantage of the possibility of studying humans, or blood or other samples from humans, who have voluntarily been exposed, who are occupationally exposed, or who are victims of an accident. This seems to be a sadly neglected area. The possibility of using human cells or tissues for in vitro experiments, and particularly for comparison of the results with those obtained with animal cells and tissues, is another neglected area. Improvements could also be made today with regard to the replacement of animal models that require death or severe disease as the endpoint (e.g. by replacing host-resistance assays with biochemical or immunological markers or microbe counts). Another way of reducing the number of animals required is to design experiments with multiple endpoints, instead of performing a large set of separate animal experiments. In the area of immunotoxicology, a tiered approach has been used in many laboratories for several years. Useful information can be obtained from regular 28-day general toxicity tests if increased attention is paid to the study of the histopathology of a large variety of lymphoid tissues, coupled with immunohistochemical measurements and the determination of classes of antibodies. Fluorescence Activated Cell Analyser (FACS) analysis has also been shown to be of great value. A proposal should be made for revision of the OECD guidelines. A greater understanding of the immunobiological mechanisms that result in chemical allergy provides new opportunities for designing predictive test methods for sensitization that do not require that allergic reactions are provoked in animals. Many participants questioned the ethics of continuing to perform the Magnusson and Kligman guinea pig maximization test. The use of complete Freund's adjuvant was also questioned, from both the scientific and the ethical point of view. With regard to the validation of in vitro methods one important point that was. made was that the validation should be against information gained from humans, and not against results from one of the common laboratory animal species.