In the present study, we combined the psychological refractory period (PRP) paradigm with a novel task-pair switching logic which enabled us to isolate performance costs occurring at the global level of task-pairs. In Experiment 1, in which we used conceptually overlapping responses for Task 1 (T1) and Task 2 (T2), we generated 3 task-pairs by combining 1 of 3 visual tasks (T1) with an auditory task (T2). In addition to worse performance after a short SOA than a long SOA (i.e., PRP effect), we found impaired performance in n − 1 task-pair switches as compared to n − 1 task-pair repetitions (i.e., n − 1 task-pair switch costs), suggesting that task-pairs were activated during dual-task processing. In Experiment 2, we increased the interference between T1 and T2 by using physically overlapping responses and we again observed n − 1 task-pair switch costs. To investigate whether the activation of task-pairs is adjusted by inhibitory control, we looked at the n − 2 task-pair sequence and found performance to be better in n − 2 task-pair repetitions than in n − 2 task-pair switches in both experiments. This n − 2 task-pair repetition benefit was replicated in Experiment 3 in which no immediate task-pair repetitions were included. Hence, the evidence suggests enhanced activation rather than inhibition as a crucial selection mechanism at the global level of dual-task processing.