Language switching has been one of the main tasks to investigate language control, a process that restricts bilingual language processing to the target language. In the current review, we discuss the How (i.e., mechanisms) and Where (i.e., locus of these mechanisms) of language control in language switching. As regards the mechanisms of language control, we describe several empirical markers of language switching and their relation to inhibition, as a potentially important mechanism of language control. From this overview it becomes apparent that some, but not all, markers indicate the occurrence of inhibition during language switching and, thus, language control. In a second part, we turn to the potential locus of language control and the role of different processing stages (concept level, lemma level, phonology, orthography, and outside language processing). Previous studies provide evidence for the employment of several of these processing stages during language control so that either a complex control mechanism involving different processing stages and/or multiple loci of language control have to be assumed. Based on the discussed results, several established and new theoretical avenues are considered.

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.

The current study systematically examined the influence of sequential predictability of languages and concepts on language switching. To this end, 2 language switching paradigms were combined. To measure language switching with a random sequence of languages and/or concepts, we used a language switching paradigm that implements visual cues and stimuli. The other paradigm implements a fixed sequence of languages and/or concepts to measure predictable language switching. Four experiments that used these 2 paradigms showed that switch costs were smaller when both the language and concept were predictably known, whereas no overall switch cost reduction was found when just the language or concept was predictable. These results indicate that knowing both language and concept (i.e., response) can resolve language interference. However, interference resolution does not start solely based on the knowledge of which concept or language one has to produce. We discuss how existent models should be revised to accommodate these results.