There is a debate regarding the processes contributing to n-2 repetition costs in task switching, which are thought to reflect an inhibition of task sets. N-2 repetition costs are assessed when switching between three tasks and are defined as a worse performance if the task was repeated from trial n-2 to trial n (sequence ABA) compared to when the task switched from trial n-2 to trial n (sequence CBA). Research showed that besides inhibitory mechanisms, also episodic interference effects acting on the level of a task or stimulus episode contribute to n-2 repetition costs. By manipulating whether participants have to process repeated vs. unrepeated stimuli, with the latter one being never encountered before, we aimed to further investigate the contribution of interference effects to n-2 repetition costs. As interference effects are likely to increase with prolonged experimental duration, i.e., when more episodes of the same stimulus have been processed, we also examined how n-2 repetition costs changed with practice. In this experiment – in contrast to previous task switching studies – n-2 repetition costs increased after prolonged practice. Furthermore, they were more pronounced when stimuli were repeated. This highlights the importance of interference effects between stimulus episodes for n-2 repetition costs, with this interference increasing with the number of previously encountered stimulus episodes. This suggests that the mechanisms contributing to n-2 repetition costs depend on the experimental context and practice.