The effects of intermodal selective attention on event-related brain potentials (ERPs) were examined in 2 experiments. In experiment 1, auditory ERPs were compared (1) when subjects responded to easy and difficult-to-detect target tones in sequences of tone bursts; and (2) when they ignored the same auditory sequences and played a demanding video game. In experiment 2, auditory ERPs to tone bursts and visual ERPs to vertical line gratings were compared as subjects responded to difficult-to-detect targets in one modality or the other. Attention to auditory stimuli resulted in biphasic enhancements in auditory ERPs, the Nda (negative auditory difference wave, latency 120-160 msec) and the Pda (positive auditory difference wave, latency 200-240 msec) waves. These had longer latencies and somewhat different scalp distributions than N1 and P2 components evoked by non-attended tones. The Nda and Pda could be contrasted with the monophasic processing negativities typically found in dichotic selective attention tasks. Nda amplitudes were larger for difficult-to-detect targets (closely resembling standards) than for standards themselves, but no Ndas were recorded to highly deviant targets. Deviant auditory stimuli evoked mismatch negativities (MMNs) that persisted during visual attention. MMN amplitudes to difficult-to-detect deviants were enlarged with attention, but no change was found in MMN amplitudes to easy-to-detect deviants. In experiment 2 intermodal attention was associated with biphasic changes in visual ERPs over the posterior scalp: the occipital Pdv (100-130 msec), and contralateral-temporal Ndv (120-320 msec) deflections. Deviant visual stimuli also elicited mismatch negativity/N2b components, largest over the inferotemporal cortex contralateral to the stimulated visual field. Like the auditory MMN, the MMN increased in amplitude with attention, but it was also evident during attend auditory conditions. The results suggest that sustained, intermodal attention depends primarily in processing modulations in modality-specific cortex. We found no evidence of the participation of modality non-specific cortex. This excludes the possibility that intermodal attention depends on a single, supramodal attention system. The relatively long latency of intermodal effects suggests that they may depend on the reafferent (top down) modulation, and do not index "template matching" operations.