When people alternate between two tasks, substantial switch and mixing costs arise, yet these costs can be reduced through training. Previous work has shown that auditory task-switching training can induce cross-modal transfer to an analogous visual task-switching situation, suggesting that training-related improvements can operate at an amodal process level. However, because earlier studies used identical semantic stimulus content and responses on the auditory training and visual transfer tasks, it remains unclear whether the amodal training-related improvements arise at the stimulus level or at a more abstract category level.

To address this question, participants completed three sessions of auditory task-switching training involving two semantic-categorization tasks with spoken words (experimental group). Two control groups received either single-task or no training. In pre- and post-training sessions participants performed both, the trained auditory and untrained visual task-switching situation. The visual tasks included the same semantic-categorization tasks as the auditory condition but with new additionally presented category members.

The results revealed reduced auditory switch (but not mixing) costs for participants in the experimental group relative to both control groups, indicating successful training-related improvement of task-switching. However, we found no group-specific improvements for the visual transfer tasks when new category members were used. This lack of transfer suggests that previously reported cross-modal training benefits in task-switching arose from the reuse of identical semantic content. Consequently, this pattern supports the assumption that cross-modal transfer is limited to situations in which trained and transfer tasks overlap at the stimulus level, thereby reflecting stimulus-specific rather than category-based learning.