A practical method is proposed for determining human dose-response curves based on reasonable assumptions and simplifications. The epidemiological data needed are the fractions of the population suffering an adverse response from exposures to two or more patterns of fluctuating concentrations of a pollutant, and the statistical parameters of each pattern. The method calculates the two parameters of the threshold type dose-response curve of the pollutant, represented by a cumulative lognormal distribution. This distribution was derived from a reasonable statistical model. The calculation does not require any arbitrary safety factors and yields central values. The dose-response parameters then may be used to calculate the health risk rate of exposure to any other fluctuating concentration pattern. Another method is proposed to select appropriate threshold limit values (TLVs) using calculations involving these parameters. Examples are given to illustrate the calculations. Results with hypothetical data gave apparently reasonable results. They showed the importance not only of the geometric mean concentration but also of the geometric standard deviations of both the concentrations and of the dose-response curve, that greatly influence the results. It is believed that results of useful accuracy should be obtained. Health risk rates are readily understood. They are useful for cost-benefit calculations. Relative rates can be used to compare the hazards of different operations and different plants. The method may make possible the development of standards specifying maximum allowable risk rates.