We propose a framework for an unsupervised analysis of electroencephalography (EEG) data based on possibilistic clustering, including a preliminary noise and artefact rejection. The proposed data flow identifies the existing similarities in a set of segments of EEG signals and their grouping according to relevant experimental conditions. The analysis is applied to a set of event-related potentials (ERPs) recorded during the performance of an emotional Go/NoGo task. We show that the clusterization rate of trials in two experimental conditions is able to characterize the participants. The extension of the method and its generalization is discussed.