The above reviews on language switching and mixing indicated a measurable processing cost occurred when individuals switch or mix languages, but did not specify in which directions.
Thus, this section discusses processing cost in terms of the direction of language switching and mixing.
Many studies considered mostly unidirectional switching (mostly switches from L1 into L2), and not the opposite (from L2 to L1). However, numerous studies have shown that the cost of processing was asymmetric in these two switching directions and that such asymmetry might be able to be accounted for with the idea of cognitive control (Meuter & Allport, 1999;
Alvarez et al., 2003; Liao & Chan, 2016), as demonstrated in the two models that will be described below: the Inhibitory Control model (IC) and the Bilingual Interactive Activation Plus model (BIA+).
From a psycholinguistic perspective, Green (1998) proposed a production model, the Inhibitory Control Model (IC), to explain the modulation of the selective activation of the target language and inhibition of the non-target language (see Figure 2 for the diagram). In this model, the communication goal motivates the conceptualizer to construct the conceptual information during a linguistic task. The lexico-semantic system consists of lemmas (entries in the lexicon that contain information on the morphology, syntax, and phonology for each lexical item (Levelt, 1989) that were tagged for specific language. Thus, it was expected to find more highly
activated lemmas for a more dominant language. The language task schemas are networks that individuals may construct or alter in order to achieve a specific task. In addition, the SAS (Supervisory Attentional System) controlled the language task schema through inhibiting or activating the lemmas of unintended or intended language in the bilingual lexico-semantic system. According to Green, while executing a task in L2, stronger inhibition for the L1 was required as L1 lexical nodes were usually more activated. As a consequence, a longer reaction time and more efforts were required to switch into a more dominant and more suppressed language.
Figure 2: The Inhibitory Control Model (IC), proposed by Green (1998)
Bilingual Interactive Activation Plus model (BIA+) (Dijkstra & Van Heuven, 2002), on the other hand, was built to explain the process of bilingual language comprehension and contained two interactive subsystems, respectively a word identification system and a
task/decision system (see Figure 3 for the diagram). The processing of the information flow will be directed from a bottom-up manner starting from the word identification system (the linguistic context). The input will first activate the orthographic, phonological and semantics representations and consequently builds an interactive network with the language nodes, which reflect the membership to particular language. All the information collected at the word identification system will thus be transmitted to the task/decision system to carry out further executive actions, which include all the non-linguistic information. In addition, this model proposes that the resting-level activation of words in a language depends on their recency of use. In most cases of bilinguals, L1 attains higher frequency of usage as compared to L2, and so higher L1 representation is expected, and the cross-linguistic influence will be greater from L1 to L2. Therefore, L1 to L2 language switches will be more energy consuming than the other way round.
Figure 3: The Bilingual Interactive Activation Plus model (BIA+), proposed by Dijkstra and Van Heuven (2002)
The language control in the IC model was achieved through the implementation of language task schemas. It was hypothesized that the language selection was expected to occur at an early stage and cognitive control was required to inhibit non-target language (Ye & Zhou, 2009). Whereas, in the language comprehension model, BIA+, the processing of language selection occurred at a late stage and the function of its cognitive control is to reanalyze and resolve ambiguous sentences (Ye & Zhou, 2009). In terms of processing costs in language switching, IC stressed the effect of inhibition. It was proposed that the dominant language will attain greater lemma activation and thus stronger efforts were required to inhibit the switching to L1 while performing a task in L2. However, BIA+ provided an alternative proposal. It
suggested that the frequently used language (L1) will have higher resting-level activation and greater cross-linguistic influence from L1 to L2.
Since the current study required bilinguals to comprehend switching and mixing sentences, studies concerned with the comprehension model, BIA+ and the language switching/mixing direction will be reviewed below.
Numerous studies have shown temporal delay of the infrequently used language in processing(in most cases is the L2), which supports the hypothesis that the frequency use of a language would determine the activation of the language representation and thus affect the reaction time asserted (Ardal et al., 1990; Moreno & Kutas, 2005).
Proverbio et al.’s (2004) behavioral data provided support towards the above studies. It was found that switching from L1 to L2 required longer response time, but not the other way around. Thus, it is possible to infer that the representations in the more dominant language were more active and thus easier to retrieve from the bilinguals’ mind. Liao and Chan’s (2016) study has also shown that switching from a more dominant language (Mandarin) to less dominant language (Taiwanese) required extra time, which indicates more effort and strength were exerted to access a word in Taiwanese, or to incorporate a word in the Mandarin context. This result supported the BIA+ model (Dijkstra & Van Heuven, 2002), which hypothesized that switching from a more dominant language to a weaker language would be more difficult. The
elicitation of N400 effect in their ERP’s experiment further supported this model, as the N400 effect reflected the difficulty in lexical integration and lexical access.
An ERP experiment was conducted by Jackson et al. (2001) to investigate the executive control while the language used switch from one to the other repeatedly. Participants were instructed to name visually cued digits in the targeted language (L1 or L2). It was found that the N2 component, which is functionally linked to response inhibition, was more negative when switching from a more dominant language (L1) to a weaker language (L2). This finding is in consistent with Verhoef et al. (2006), suggesting that a more active suppression is required to inhibit the more dominant language. Moreover, an increased N2 component in the parietal and frontal cortices were found when the demands on response selection increased, which implies that these regions were involved in the inhibition process of the non-target language and in selecting a relevant response.
Wang et al. (2007) adopted an event related-fMRI (ER-fMRI) technique to examine the underlying neural substrates of second-language learners in language mixing. Twelve Chinese-English bilinguals were recruited to perform the picture-naming tasks. The result showed that mixing conditions did elicit greater activation in the right superior prefrontal cortex, left middle and superior frontal cortex, and also the right middle cingulum and caudate. In addition, the switching direction was found to be one of the dependent factors that will affect the involvement of the neural substrates. In comparison with the backward control (L2 to L1), the
forward switching (L1 to L2) activated brain regions in bilateral frontal cortices and ACC, in which link to the function of executive control. Therefore, it was suggested that the direction of language switch might activate different neural mechanisms and should be controlled while designing our materials.
Neuroimaging studies have found that, while subjects were undergoing a language switching task, the ACC and the dorsolateral prefrontal areas, which were responsible for general executive function, were activated, showing that they might be where the task/decision system in the BIA+ model was located (Price et al., 1999; Hernandez, 2009). In addition, Abutalebi and Green (2007) and Wang et al. (2007) show that switching from a more dominant language into a less dominant language required more cognitive efforts, as revealed by the activation of bilateral frontal cortices and the ACC. In sum, these studies all points to one direction that switching from L1 to L2 required more cognitive efforts.
Taken together, the current study intended to explore the brain regions that are responsible for the processing of language switching and mixing sentences in healthy individuals as few studies were conducted to differentiate between these two phenomena. In addition, the current study also aimed to inspect if similar brain regions are activated for the healthy bilinguals and brain-damaged bilinguals as reviewed in previous studies in processing switching and mixing sentences since not many studies seek out to compare this between the healthy and pathological individuals. Furthermore, by using the fMRI technique, we can examine if the
switching/mixing direction in processing language switching or mixing sentences is a crucial factor in affecting the processing of cognitive control.