Some theories of visuospatial attention propose that attention can be divided between separated zones of space that exclude the intervening region, whereas other theories state that the focus of attention must encompass a unitary, continuous zone. These contrasting views were evaluated in an experiment in which subjects were required to monitor two of four stimulus locations for targets; the two relevant locations were adjacent in one condition and were separated by an intervening irrelevant location in a second condition. To assess the distribution of attention across the relevant and irrelevant locations, event-related brain potentials (ERPs) were recorded to task-irrelevant "probe" stimuli that were occasionally presented at the individual stimulus locations. When the relevant locations were adjacent, probes presented at irrelevant locations elicited smaller sensory-evoked electrophysiological responses than probes presented at relevant locations, consistent with an attentional suppression of inputs from the unattended locations. When the relevant locations were separated by an irrelevant location, however, the sensory responses evoked by probes presented at this intervening irrelevant location were not suppressed, and target detection performance became slower and less accurate. These results suggest that attention forms a unitary zone that may expand to encompass multiple relevant locations but must also include the area between them; as a result, irrelevant information arising from intervening locations is not suppressed and perceptual processing is compromised.