A number of recent papers have suggested basing the statistical analysis of Salmonella (Ames) mutagenicity test results on a mathematical model of the complete dose-response curve. For most mutagens at low doses the curve increases linearly; then, as the dose increases, the curve may flatten and finally turn downwards due primarily to effects of toxicity. The exact mechanism underlying this shape is, however, not well understood and is likely to vary for different chemicals. A different approach is to assume that the initial part of the curve is linear and to base the statistical analysis solely on this region, reasoning that it contains most of the interpretable information about the mutagenesis dose response. In this paper a formal method of deciding which points are on the initial linear part of the curve is described, and a statistical method is proposed for analyzing these points. Computer simulations are used to examine the properties of the procedure and comparisons are made with a previously proposed mathematical model of the whole curve. It is concluded that the method suggested here provides a very satisfactory, robust method for the standard analysis of Salmonella data.