There has been a substantial proliferation in the number of studies reporting endocrine effects as an endpoint. The vast majority have focused on oestrogenicity in vitro but, with recent recommendations by the USEPA Endocrine Disrupter Screening and Testing Advisory Committee, tests are now being developed for (anti)-androgenicity and effects on the thyroid, largely because of the potential for altering reproduction or development via these mechanisms. Despite being a vital organ and involved in reproduction and development, there is currently no provision for assessing adrenocortical function. Similarly, the entire process of steroidogenesis poses multiple molecular targets for toxic disruption that are not included in current test strategies and at present there is no clear position on the significance of the data being generated. This review provides a framework for approaching endocrine data: that all the glands, tissues, receptors, transporter proteins and enzymes that comprise the endocrine system are targets for toxicity. They should be considered in much the same way as other target organs, with appropriate provision for the special cases of carcinogenesis and teratogenesis, and a pragmatic weight of evidence approach should be adopted considering all available data and recognizing its limitations. In this approach, structure-activity relationships and in vitro and targeted in vivo screens provide useful data but repeat-dose regulatory studies with defined endpoints provide the most powerful tools for hazard assessment. Pragmatic consideration should be given to exposure issues (which may highlight the practical irrelevance, for example, of very low potency oestrogens) and subsequently whether endocrine disruption is the critical or most sensitive endpoint for a compound. Finally, endocrine disruption may be considered a mechanism and, as with other toxic endpoints, knowledge of effect and no-observable-effect levels and reversibility is as important as identifying the target tissue or any inherent hormone-like property.