The use of antimicrobial agents in aquaculture represents a potential risk for the efficacy of the use of these agents in the therapy of human infectious diseases. This potential was first identified in 1971 but, despite twenty-eight years of research, little or no information relevant to estimating the size of the risk currently exists. This paper argues that risk analysis will be possible only after the implementation of a systematic, planned data acquisition programme. The movement of pathogenic bacteria and the movement of resistance genes between aquaculture enterprises and humans are the two routes by which any hazard could be mediated. An analysis of each mechanism is presented and the key data which would be required for risk analysis of each are also identified. The paper stresses the fact that aquaculture is a diverse activity and that separate data collection would have to be performed for each major type of enterprise. In planning any systematic data collection it would be necessary, therefore, to prioritise the enterprises to be investigated. The author argues that, as a general principle, small subsistence-farming systems, which constitute the majority of aquaculture production in the world but which are not involved in international trade, should not be investigated.