In this paper, two different visions of the relationship between science and policy are contrasted with one another: the "modern" vision and the "precautionary" vision. Conditions which must apply in order to invoke the Precautionary Principle are presented, as are some of the main challenges posed by the principle. The following central question remains: If scientific certainty cannot be provided, what may then justify regulatory interventions, and what degree of intervention is justifiable? The notion of "quality of information" is explored, and it is emphasized that there can be no absolute definition of good or bad quality. Collective judgments of quality are only possible through deliberation on the characteristics of the information, and on the relevance of the information to the policy context. Reference to a relative criterion therefore seems inevitable and legal complexities are to be expected. Uncertainty is presented as a multidimensional concept, reaching far beyond the conventional statistical interpretation of the concept. Of critical importance is the development of methods for assessing qualitative categories of uncertainty. Model quality assessment should observe the following rationale: identify a model that is suited to the purpose, yet bears some reasonable resemblance to the "real" phenomena. In this context, "purpose" relates to the policy and societal contexts in which the assessment results are to be used. It is therefore increasingly agreed that judgment of the quality of assessments necessarily involves the participation of non-modellers and nonscientists. A challenging final question is: How to use uncertainty information in policy contexts? More research is required in order to answer this question.