Are alternation and co-occurrence of stimuli of different sensory modalities conspicuous? In a novel audio-visual oddball paradigm, the P300 was used as an index of the allocation of attention to investigate stimulus- and task-related interactions between modalities. Specifically, we assessed effects of modality alternation and the salience of conjunct oddball stimuli that were defined by the co-occurrence of both modalities. We presented (a) crossmodal audio-visual oddball sequences, where both oddballs and standards were unimodal, but of a different modality (i.e., visual oddball with auditory standard, or vice versa), and (b) oddball sequences where standards were randomly of either modality while the oddballs were a combination of both modalities (conjunct stimuli). Subjects were instructed to attend to one of the modalities (whether part of a conjunct stimulus or not). In addition, we also tested specific attention to the conjunct stimuli. P300-like responses occurred even when the oddball was of the unattended modality. The pattern of event-related potential (ERP) responses obtained with the two crossmodal oddball sequences switched symmetrically between stimulus modalities when the task modality was switched. Conjunct oddballs elicited no oddball response if only one modality was attended. However, when conjunctness was specifically attended, an oddball response was obtained. Crossmodal oddballs capture sufficient attention even when not attended. Conjunct oddballs, however, are not sufficiently salient to attract attention when the task is unimodal. Even when specifically attended, the processing of conjunctness appears to involve additional steps that delay the oddball response.