To investigate bilingual language control, prior language switching studies presented visual objects, which had to be named in different languages, typically indicated by a visual cue. The present study examined language switching of predictable responses by introducing a novel sequence-based language switching paradigm. In 4 experiments, sequential responses (i.e., weekdays, numbers or new sequences) and an alternating language sequence (e.g., L1-L1-L2-L2) were implemented, both of which were memory based. Our data revealed switch costs, showing that a language switch is associated with worse performance compared with a language repetition, and mixing costs, which constitutes the performance difference between pure and mixed language blocks, even while producing entirely predictable responses (i.e., language and concept). Additionally, we found these switch costs with overlearned and new sequences and found that switch costs were reduced with longer preparation time. The obtained data are consistent with a proactive interference account, such as the inhibitory control model.