Résumé : Objective: Studies exploring neurophysiological correlates of main psychiatric disorders have commonly used event-related potentials (ERP) during a visual or an auditory oddball task. The main results concern modulations of the P3b amplitude and/or latency. The present study aims to increase the clinical sensitivity of these P3b modulations by using a more ecological oddball design, using synchronized pairs of audio-visual emotional stimuli. Method: Two groups of healthy participants, one composed of controls and the other of students displaying anxious and depressive tendencies completed visual, auditory and audio-visual (cross-modal) oddball tasks, in which they had to detect deviant happy and sad stimuli among neutral ones as quickly as possible. Behavioral performance and P3b ERP data were analyzed. Results: Subjects displaying anxious and depressive tendencies exhibited lower P3b amplitude than the controls, but only in the cross-modal oddball task. Conclusions: Although the two groups of subjects differed in their levels of co-morbid anxiety and depression, unimodal visual and auditory oddball tasks did not allow us to detect this difference by P3b amplitude modulations, but the cross-modal task did. Significance: These results suggest that a cross-modal oddball design should be used in future studies to increase the sensitivity of the P300 amplitude differences between healthy participants and those with clinical symptoms. © 2010 International Federation of Clinical Neurophysiology.