Research on brain mechanisms of deviance detection and sensory memory trace formation, best indexed by the mismatch negativity, mainly relied on the investigation of responses elicited by auditory stimuli. However, comparable less research reported the mismatch negativity elicited by somatosensory stimuli. More importantly, little is known on the functional features of mismatch deviant and standard responses across different sensory modalities. To directly compare different sensory modalities, we adopted a crossmodal roving paradigm and collected event-related potentials elicited by auditory, non-nociceptive somatosensory, and nociceptive trains of stimuli, during Active and Passive attentional conditions. We applied a topographical segmentation analysis to cluster successive scalp topographies with quasi-stable landscape of significant differences to extract crossmodal mismatch responses. We obtained three main findings. First, across different sensory modalities and attentional conditions, the formation of a standard sensory trace became robust mainly after the second stimulus repetition. Second, the neural representation of a modality deviant stimulus was influenced by the preceding sensory modality. Third, the mismatch negativity significantly covaried between Active and Passive attentional conditions within the same sensory modality, but not between different sensory modalities. These findings provide robust evidence that, while different modalities share a similar process of standard trace formation, the process of deviance detection is largely modality dependent.