A variety of regulatory requirements for ecotoxicity test data exist in the US and each of these is able to specify what test protocols and endpoints are used. General practice is to calculate regression curve endpoints, usually EC50 values, for acute tests, and hypothesis test endpoints (e.g. NOEC, LOEC) for chronic studies. However, tests of wastewater effluents often use a hypothesis test endpoint which has been derived using site-specific information to represent a pass-fail standard for compliance. Field work, such as site assessments, tends to use hypothesis tests, but such work does not seek the doseresponse curve sought in standard laboratory tests. The risk-based approaches being developed use cither type of endpoint, but this seems to be an accommodation to existing data; preference is for dose-response curves, not just a single ECx value. Endpoints are only one component of the conventional paradigm of environmental protection. Experience with various tests and endpoints suggest several perspectives: quality is critical, test species must be reliable and relevant, extrapolations will dominate decisions, and basing environmental decisions on the most sensitive endpoint of the most sensitive species may not remain a feasible paradigm for protecting ecological systems.