How toddlers' attention is distributed in the visual field during a magic trick was examined using three expectation conditions. Results showed that 2.5-year-old children assigned to the condition with major task-relevant information (i.e., a verbal cue to attend to one of the visual targets) (i) outperformed those who were assigned to the condition with no task-relevant information, (ii) displayed more attentional switches between visual targets than those who were assigned to the condition with no task-relevant information, and (iii) did not look significantly longer at one of the visual targets in contrast to children assigned to the condition with no task-relevant information. The findings of an additional control condition suggest that the performance by children in the condition with major task-relevant information cannot merely be the consequence of the larger quantity of auditory information that was given during the interaction prior to the magic trick. In our task, verbal cue affected the switching of attention, not the prioritization of a specific region of space. These results are discussed in terms of advantage conferred by a diffuse mode of attention.