A procedure was proposed for the speciation of chromium by differential pulse adsorptive stripping voltammetry (DPAdSV) using pyrocatechol violet (PCV) as a complexing agent. In this procedure a partial least squares regression (PLS) was used for the resolution of the strongly overlapping voltammetric signals from mixtures of Cr III and Cr VI in the presence of PCV. The application of a genetic algorithm to select some of the predictors variables (potentials of the voltammogram) allows one to considerably reduce the number of experimental variables, but does not diminish the prediction capacity of the PLS model constructed with these selected variables. The relative error in absolute value is less than 5 % when concentrations of several mixtures are calculated. The analysis of the possible effect of the presence of foreign ions in the solution was performed. The procedure was successfully applied to the speciation of chromium in the waste water from a tanning factory.