As part of the International Workshop on Standardization of Genotoxicity Test Procedures, in Melbourne, 27-28 February 1993, various international guidelines were examined with respect to protocol issues in the area of mammalian cell gene mutation assays. The working group on mammalian cell gene mutation assays discussed a wide range of protocol issues related to study design; in most cases the recommendations are reasonably consistent with existing guidelines. Agreement was reached on several issues as follows. The upper limit of concentration for testing non-toxic substances should be 10 mM or 5 mg/ml, whichever is lower. For testing toxic substances the criteria of an acceptable upper limit of concentration should yield 10-20% survival. Any of several established mammalian cell mutation assays (L5178Y TK+/-, CHO/HPRT, AS52/XPRT, V79/HPRT) can be used to evaluate mutagenesis in mammalian cells; the ouabain (Na/K-ATPase) system is not an acceptable mutation assay for routine evaluation of mutagenesis in mammalian cells. Ability to recover small colonies must be convincingly demonstrated when using the L5178Y TK+/- mouse lymphoma assay. In the mouse lymphoma assay (L5178Y TK+/-), colonies in positive controls and at least two (if available) representative positive doses of the test compound should be sized if a positive response is seen; in the event of a negative response due to the test compound, colony sizing of the positive control is necessary to validate the conduct of the assay. Testing both in the presence and absence of S9 metabolic activation is necessary. It was not possible to come to a firm conclusion about the length of treatment. There was a general agreement that extended treatment times (> 2 cell cycles) often bear more disadvantages than advantages and should only be used with adequate justification. It is not necessary to repeat clear positive or clear negative tests when the assay has been adequately performed; this recommendation differs significantly from the UK guidelines. If treatment groups are not replicated, the numbers of doses tested should be increased; this recommendation differs significantly from the UK guidelines. Each laboratory should establish a historical database for the performance of a given assay in that laboratory.