An extensive remedial investigation conducted at the site of a former automobile assembly plant detected high soil levels of heavy metals and polycyclic aromatic hydrocarbons (PAHs). Future site plans included residential development and retail-store facilities. Prior to site development, a human health risk assessment was performed to assess the potential for health hazards to future residents of the site. A key input variable into the soil exposure equations for risk assessment is the chemical concentration term. Risk assessment guidelines recommend calculation of this term through use of upper 95 percent confidence limits on the mean of the soil concentration. Assuming a normal data distribution with resulting arithmetic means and confidence limits might not produce the most appropriate and defensible risk assessment. Selection by default of a lognormal distribution with calculation of geometric means and upper confidence limits based primarily on the poor fit to the normal distribution, a method frequently employed, is also not entirely defensible. Instead, normal and lognormal probability plots and a probability plot correlation coefficient test statistic were used here for selection of the data distribution. The final risk expression developed at this site using the method presented here gives a more representative hazard index and cancer risk estimate than either of the conventional methods.