Electroencephalographic Event-Related Potentials (ERPs) present special interest for investigating cognitive processes. ERP curves contain peaks called components, such as the P600 component. P600 may be present in the averaged ERP waveform as a not-easily discernible secondary peak. In order to extract the P600 component, Independent Component Analysis (ICA) is used. ICA is a method used for solving the Blind Source Separation (BSS) problem, based on the assumption that the sources are statisti cally independent. The focus of the present study was to select a subset of independent components (ICs) for the reconstruction of the averaged ERPs in the original electrode recording positions and in the time frame of the P600 component. Two empirical criteria are used for the selection of the IC’s; the first criterion is based on the temporal proximity of the original P600 and the IC projection that reconstructs the P600 component and the second criterion is maximizing the reconstructed P600 amplitude based on the selection of a combination of IC projections using up to a preselected maximum number of them. The techniques are tested on ERPs recorded from healthy subjects and heroin addicts.