The U. S. Environmental Protection Agency recommended protocol for the fathead minnow life-cycle test specifies the measurement of numerous basic, but fragmentary, response parameters. Various researchers have recognized the need for combining those basic responses into global responses, which are more indicative of species survivability and ecological impact. One method of satisfying this need is to construct a hierarchy of responses. We illustrate a hierarchy of responses, currently being used in SRI experiments on munitions wastewaters, which consists of 31 basic responses (including new cumulative survivability responses) and 4 global responses.

The use of a hierarchy of responses including global responses necessitates the development of statistical tests for these responses and highlights the problems associated with potentially large overall significance levels. The use of Dunnett's test for assessing the statistical significance of the global responses is recommended along with the subdivision of aquaria to provide adequate degrees of freedom. In addition, we discuss methods for accurately estimating and reducing the generally unacceptably large overall significance level associated with the policy of declaring a concentration to be above the no-observed-effect level (NOEL) whenever any response is statistically significant.

The importance of incorporating biological judgment and expertise into the NOEL decision is gaining recognition, and a schema for this incorporation is presented. Methods for modifying the statistical tests to detect only biologically important response differences and statistical and graphical techniques for assisting the application of biological judgment are examined. The application of biological judgment in the NOEL decision is found to lower the overall significance level, and a method (postrandomization) for estimating that reduction is presented.