Despite a growing body of research into task switching phenomena, little is known of the interactions between the performance decrements that result from shifting task sets and shifting stimulus modalities. Seven experiments are presented which replicate a previously documented underadditivity of task and modality switch costs, and extend the known limits and behavior of this underadditivity by showing that the cost of switching tasks can be reversed under cross-modal stimulus conditions. Cross-modal task switch costs and benefits are also shown to vary widely as a function of experimental protocols, the tasks subjects perform, and the degree of cross-modal overlap in imperative stimuli. Active preparation in response to modality predictive cues was not observed to result in differential reductions in cross -modal task switch costs, contrary to suggestions made by other researchers.