Tensor-based signal processing has been regarded as a promising tool for MEG (Magnetoencephalogram) and EEG (Electroencephalogram) data analysis by preserving and exploiting the multi-dimensional nature of the data. This paper addresses the application of a tensor decomposition called PARAFAC2 in the analysis of somatosensory evoked magnetic fields (SEFs) and somatosensory evoked electrical potentials (SEPs). This new perspective of studying such MEG and EEG signals was first motivated by the seemingly good match between the PARAFAC2 model and the nature of these data. Based on this observation, we propose to compute the PARAFAC2 decomposition on three-dimensional tensors constructed from the time-frequency representation of the data. Results show that this PARAFAC2-based component analysis of SEFs and SEPs is able to extract the spectral, temporal and time-varying spatial signatures and to corroborate previous findings concerning the multiplicity in the signals evoked by peripheral nerve stimulations.