In biomedical research summary measures such as the curve maximum (Cmax)are frequently used to describe and analyze unimodal response curves. However, if the true curve is perturbed by autocorrelated noise the calculation of summary measures from raw data can be arbitrary and misleading due to high peaks produced by the correlated errors. A possible solution is to fit suitable nonlinear functions to the response curves and estimate the summary measures from these functions. In this paper formulas are derived providing a way to estimate important summary measures of unimodal curves by means of the parameters of the lognormal function. The method is illustrated by application to pharmacodynamic data.