When acting intentionally, humans respond to stimuli based on their goals, which sometimes requires responding differently to the same stimulation or performing the same response to a different stimulus. Such recombining of responses and stimuli often hinders performance, yielding partial repetition costs (PRCs). PRCs are considered an after-effect of feature-binding processes, but the underlying process remains unclear. PRCs were alternatively attributed to time-consuming unbinding, or to solving the conflict between the current and the previous episode. Crucially, the role of control processes remains unexplored. If PRCs imply conflict, control might be recruited to handle PRCs. Hence, control level might be adjusted following a partial repetition, as in the well-known conflict adaption phenomenon. We re-analysed seven task-switching datasets (Ntotal = 591) seeking adaptation to partial repetitions, i.e. trials where the task repeats and the response switches, or vice versa. In every dataset, reaction times in partial repetitions shrank when trial N-1 was also a partial repetition, compared to a full switch. This could reflect reduced feature binding or retrieval, or a carryover of a heightened control state from the previous trial. Interestingly, in 3 out of 7 datasets, full repetitions were faster when trial N-1 was a partial repetition, compared to a full switch (and only numerically in two additional datasets). This contrasts with weaker binding or retrieval and it favours more control being deployed after a partial repetition compared to a full switch. These findings suggest that cognitive control is recruited to deal with partial repetitions in task switching.