Recent dual-task studies suggest that a bottleneck prevents central mental operations from working on more than one task at a time, especially at relatively low practice levels. It remains highly controversial, however, whether this bottleneck is structural (inherent to human cognitive architecture) or merely a strategic choice. If the strategic hypothesis is testable, it ought to predict that, under sufficiently strong incentives, people could choose to bypass the bottleneck and perform both tasks in parallel. Because the incentives for parallel processing in previous studies have been modest, the authors introduced a novel dual-task paradigm with much greater incentives, induced by strict time deadlines for each task. With this paradigm, bottleneck delays would cause participants to frequently miss the time deadline or make errors, triggering immediate negative consequences (failure feedback). Nevertheless, participants had little success performing central operations in parallel; severe dual-task performance costs were observed, even with relatively easy tasks. These results greatly strengthen the case that the central bottleneck reflects a structural limitation that, at least at modest practice levels, cannot be avoided merely by trying harder. (Contains 2 tables, 2 footnotes, and 8 figures.)