In the Deese–Roediger–McDermott (DRM) paradigm, participants falsely recall or recognize a nonpresented word (critical lure), highly associated with previously studied words. As numerous DRM studies have found a robust false memory effect at the behavioural level, eventrelated potentials (ERPs) studies have searched for possible overlapping in brain electrical activity between true and false memory. Using the DRM paradigm, the present experiment manipulated the sensory modality of stimulus presentation (auditory vs. visual) in the study phase to analyse the effect of modality match between study and test on true and false recognition. Words were therefore presented either visually or auditorily at study and always visually at test. True recognition was found to be significantly higher in the modality ‘match’ condition (visual–visual) than in the ‘mismatch’ condition (auditory–visual), whereas there was no modality-match effect on false recognition of critical lures. A general, overlapping was found between ERP correlates of true and false recognition: FN400 (300–500 ms), left-parietal (400–800 ms) and late right-frontal (1000–1500 ms) old/new effects were similar for both studied words and critical lures. No sensory modality-match effect was associated with FN400 or left-parietal old/new effects. Only the late right-frontal activity was modulated by modality manipulation, with significantly more positive ERPs in the modality-match condition. Sensory modality match of stimulus presentation, therefore, dissociated true and false recognition memory only at the behavioural level but not at the ERP level. Overall, true and false recognition memories seem to share common underlying processes.