Stacking methods are very important in overcoming the poor detection limits in capillary electrophoresis (CE). In this paper, the separation and determination of several tricyclic antidepressants by a stacking method is described. The inclusion of acetonitrile (ACN) in the sample causes stacking (transient pseudoisotachophoresis) especially in presence of sodium chloride. An experimental design (central composite design) together with the response surface methodology has been used to find the optimum composition of the separation buffer and the optimal stacking conditions in few experiments. The response functions used are the product of the total resolution by the number of peaks, for the optimization of the separation buffer, and the product of the total resolution by the mean of the peak heights, for the optimization of the stacking conditions. About 28% of the capillary volume is loaded with sample. The calibration curves are linear over the working range (50-300 ng/mL). With a bubble capillary, the limits of detection (LODs) are in the order of 5 ng/mL. For the analysis of serum samples, enrichment with sodium chloride and the protein precipitation with ACN are enough to avoid interferences and to get stacking. Recoveries between 91.6 and 104% and RSD between 0.6 and 12% are obtained in the analysis of samples of lyophilized human serum and non-lyophilized human serum, spiked with the drugs.