The effectiveness of different functions of statistical distribution (Gaussian, log-normal, gamma and Weibull) is examined for the deconvolution of chromatograms with overlapped peaks. A modified Gaussian function is also considered for comparison. The results obtained in the curve fitting of individual peaks, corresponding to diverse solutes run in polar, semi-polar and apolar columns, show that the log-normal function is generally the best for description of the chromatographic shapes. This function proved to be the most useful in the resolution of a test set of 32 overlapped chromatographic profiles prepared from mixtures of standards of known quantitative composition. © 1987.