The paper develops a novel learning model of clustering for Evoked Potential Single Trials, called Dynamic Grid Self- Organized Map (DG-SOM) designed according to the peculiarities of evoked potential data. The DG-SOM determines adaptively the number of clusters with a dynamic extension process which is able to exploit class information whenever exists. Specifically, it accepts available class information to control a dynamical extension process with an entropy criterion. In the case that there is no classification available, a similar dynamical extension is controlled with criteria based on the computation of local variances or resource counts. The results indicate that dynamic expansion can reveal (to a large extent) The many possible routes each of which leads from the input to the final "compulation". We employ these techniques in order to discriminate patterns from evoked potentials single trial data between alcoholic and non-alcoholic patients. From the classes provided, characteristic patterns for each class are extracted which can be valuable in studying the underlying brain dynamics.