Whether attention can be split between multiple regions in space simultaneously is an ongoing controversy in attention research. We argue that the debate could be resolved if the distribution of target locations over hemifields and task difficulty are both considered. This premise was tested in five experiments in which 48 subjects compared the identity of two out of four stimuli. In an easy task, within each hemifield, performance (reaction times and error rates) was better for adjacent targets than for separated ones, but across hemifields, performance for separated and adjacent stimuli was similar. In difficult tasks, performance was always better when the stimuli were presented across the hemifields indicating a bilateral field advantage. Moreover, the difference between adjacent and separate conditions within one hemifield diminished with increasing task difficulty. We propose a modified model of visuo-spatial attention, which permits the hemispheres to maintain and control simultaneous attentional foci.