Research suggests that dual-task interference is caused by a central bottleneck (together with response grouping and impaired preparation). The emphasis placed on the 1st response in these experiments, however, may have discouraged the sharing of processing resources between tasks. In the present experiment, instructions placed equal emphasis on 2 choice reaction-time tasks in which stimuli were presented simultaneously on 20% of the trials. In contrast to a graded trade-off of resources, a bottleneck predicts bimodality in the distribution of interresponse intervals for the 2 tasks, reflecting the 2 possible orders in which their respective central stages might be performed. Most subjects showed such a bimodality, along with other signs of a bottleneck; the remainder showed evidence of response grouping. The data suggest that the bottleneck is structural rather than strategic and make the graded sharing of resources less plausible.