Early results from two transgenic mouse mutation assays, Big Blue [Kohler SW et al. (1991): Proc Natl Acad Sci USA 88:7958-7962] and Muta Mouse [Myhr BC (1991): Environ Mol Mutagen 18:308-315], raised questions about appropriate study design and methods for statistical analysis. First, there were a number of potential sources of variability in the technical aspects of the assay. These are "how we do it in our laboratory" differences, which, if not controlled, ultimately make inter-laboratory comparison of data difficult. Second, separate from the technical sources of variability were a number of study design issues, e.g., how many animals are needed in each treatment group, how many times should the animal be dosed, what is the appropriate expression period for a particular tissue, how many plaques need to be collected from each animal, and so on. To address these questions and to identify and understand the sources of variability in mutation data from these systems, a workshop was held in June, 1992 in Cincinnati, Ohio, USA. A core group of biologists and statisticians discussed possible sources of bias (tissue sampling, phage recovery and transgene recovery), possible sources of variability in mutant frequency (between animals, protocol-based sources, and design-based sources) and assay sensitivity. Following two days of discussion on protocol design and assay procedures, three action steps were recommended: (1) compile a data base of existing mutation data in transgenic mice to study its statistical features, (2) develop standard protocols for the mutation assays; and (3) use the standard protocol to generate a large data base of mutant frequencies in liver DNA from untreated mice for statistical study and analysis. This report summarizes the proceedings and recommendations of the workshop. The progress made toward these recommendations was reviewed in a second workshop, held in April, 1993, in Norfolk, Virginia, part of which is the subject of the accompanying paper by Piegorsch et al. To date, a standard protocol has been developed for the Big Blue mutagenesis assay and a data base of over 90 million plaques from seven labs using either the Big Blue or Muta Mouse system has been assembled, including a large data set of spontaneous liver mutant frequencies in the Big Blue system.