In this paper we describe a novel data driven spatial filtering technique that can be applied to the ERP analysis in order to find statistically significant hidden differential activations in the EEG data. The technique is based on the known morphological characteristics of the response. Underlying optimization problem is formulated as a generalized Rayleigh quotient maximization problem. We supply our technique with a relevant randomization-based statistical test to assess the significance of the discovered phenomenon. Furthermore, we describe an application of the proposed method to the EEG data acquired in the study devoted to the analysis of the auditory neuroplasticity. We show how the mismatch negativity component, a tiny and short-lasting negative response that hallmarks the novel stimuli activating primary error-detection mechanisms, can be detected after filtration.