To examine sequential adjustments of dual-task performance, we re-analyzed the data of four published task-pair switching experiments. In task-pair switching experiments, three tasks (e.g., A, B, C) are combined in two task-pairs (e.g., task-pair 1: A as Task 1 [T1] and C as Task 2 [T2]; task-pair 2: B as T1 and C as T2). Switching between task-pairs typically results in task-pair switch costs (i.e., worse performance in task-pair switch trials than in task-pair repetition trials). These performance costs indicate that the identity of the two component tasks performed in a dual task is jointly represented in a single mental representation, termed task-pair set. The data re-analysis revealed that task-pair switch costs were reduced after a task-pair switch relative to after a task-pair repetition. To explain this sequential modulation of task-pair switch costs by the task-pair sequence in the previous trial, we refer to top-down cognitive control theories and bottom-up automatic priming theories.