What causes dual-task costs?

Why are dual-task costs reduced with ideomotor (IM) compatible tasks (see e.g. Greenwald & Shulman, 1973)? In a series of experiments, I tested the way in which task structure affects dual-task performance (Halvorson et al., 2012). The results suggest that in some cases, typical dual-task costs arise from task structure rather than response limitations. Further examination of this question has shown that dual-task costs cannot be predicted solely on the basis of the relationship between the stimuli and the responses; the relationship between the tasks, or the task pairing, plays a critical role in whether the tasks overlap and performance is impaired. A series of experiments using novel task pairings showed that when one task uses a spatial central code and the other uses a verbal central code, dual-task costs are eliminated.