In their target article, Branigan & Pickering (B&P) argue for structural priming as a key implicit methodology to probe the nature of linguistic representations. They provide extensive data supporting their model, ultimately arguing that structural priming provides a tool to understand the nature of language.
The authors also discuss the implications of their proposal for language representation in bilinguals, focusing primarily on cross-language structural priming between single-language sentences. This research has led to vital insights on cross-language activation at the syntactic processing level (cf., e.g., Hartsuiker & Pickering Reference Hartsuiker and Pickering2008), adding to the accumulation of evidence that language use in bilinguals involves ubiquitous cross-language activation at multiple levels of processing (cf. e.g., de Bot et al. Reference De Bot, Broersma, Isurin, Isurin, Winford and de Bot2009; Kootstra et al. Reference Kootstra, van Hell, Dijkstra, Isurin, Winford and de Bot2009; Kroll et al. Reference Kroll, Bobb and Wodniecka2006). Importantly, however, everyday bilingual communication does not normally happen according to a scripted cross-language priming paradigm with primes in one language and targets in the other (cf. Fricke & Kootstra Reference Fricke and Kootstra2016). To truly understand the nature of language in all its respects, as is B&P's ambition, we propose that their model and approach should be further developed to explain a larger number of bilingual language scenarios as they occur in real life.
One such bilingual language scenario – and a true hallmark of bilingualism – is code-switching, i.e., the use of multiple languages within one single sentence. Code-switching is susceptible to exactly the same structural priming mechanisms as the production of unilingual sentences, in the sense that bilinguals' syntactic choices in the production of code-switched sentences are primed by those of their dialogue partner (Kootstra et al. Reference Kootstra, van Hell and Dijkstra2010). But priming also occurs with dependent variables that are specific to code-switching, namely priming of the sentence position of code-switching (Kootstra et al. Reference Kootstra, van Hell and Dijkstra2012), priming of the language of the inflected verb (i.e., matrix language; Fricke & Kootstra Reference Fricke and Kootstra2016; Kootstra et al. Reference Kootstra, van Hell and Dijkstra2010), and priming of the actual choice to code-switch or not (Fricke & Kootstra Reference Fricke and Kootstra2016; Kootstra et al., Reference Kootstra, van Hell and Dijkstrain revision). Importantly, these findings are based on both experimental and corpus research. This indicates that structural priming is more than a method to investigate linguistic representations; it is a core mechanism of language use that, together with adaptive control processes (cf., Green & Abutalebi Reference Green and Abutalebi2013), appears to guide bilinguals' linguistic behavior in real-life language use.
The critical implication of this code-switching evidence is that B&P's model should be extended with a representation of language membership. After all, for priming of linguistic elements from multiple languages to take place, these multiple languages must somehow be encoded within the representational system. In most models of bilingual language processing, this is implemented by assuming a language node that is linked to linguistic representations (e.g., Hartsuiker & Pickering Reference Hartsuiker and Pickering2008; Kroll et al. Reference Kroll, Bobb and Wodniecka2006). Based on the omnipresence of cross-language activation at all levels of processing, we assume this language node is linked to linguistic representations at all levels of processing (de Bot Reference De Bot2004; de Bot et al. Reference De Bot, Broersma, Isurin, Isurin, Winford and de Bot2009; Kootstra et al. Reference Kootstra, van Hell, Dijkstra, Isurin, Winford and de Bot2009; Reference Kootstra, van Hell and Dijkstra2010). Primed code-switching can then be explained in the form of persisting co-activation of language nodes from the recently experienced discourse (see Fricke & Kootstra Reference Fricke and Kootstra2016; Kootstra et al. Reference Kootstra, van Hell and Dijkstra2010).
In addition to code-switching, the just-described extension of B&P's structural priming account also may serve to explain another fascinating bilingual language scenario: first language (L1) attrition (i.e., loss of or decreased access to L1 representations, mostly due to immersion in a second-language environment, leading to infrequent use of the first language [e.g., Schmid Reference Schmid2011]). Recently, a number of psycholinguistic paradigms have been used to study first-language attrition using offline, online, and neural measures of language comprehension and production (Rossi et al., Reference Rossi, Prystauka, Diaz, Köpke and Keijzerin revision), but the mechanism of priming so far has not been used to study first-language attrition. Based on B&P's point that priming can be seen as evidence of access to linguistic representations, it can be predicted that, if L1 representations are completely inaccessible as a consequence of attrition, L1 structural priming should be nonexistent, whereas if L1 representations merely become less accessible as a consequence of attrition, rates of L1 structural priming may well be relatively strong. This would be consistent with inverse-frequency and surprisal effects found in structural priming studies (e.g., Bernolet & Hartsuiker Reference Bernolet and Hartsuiker2010; Bock Reference Bock1986; Ferreira Reference Ferreira2003; Jaeger & Snider Reference Jaeger and Snider2007; Reference Jaeger and Snider2013). Another prediction that can be made is that structural priming can serve as a very sensitive measure of changing levels of access to L1 linguistic representations, thus making it possible to boost L1 activation for speakers who are undergoing L1 attrition, much along the lines of what has been proposed for aphasic speakers (Rossi Reference Rossi2013). To continue, under the assumption that structural priming boosts access to linguistic representations by easing the demands on cognitive abilities such as memory, structures that are difficult and/or cognitively taxing should benefit the most from structural priming. Interestingly, these predictions not only show how B&P's model and its bilingual extension can be utilized to further test existing issues in L1 attrition, but also showcase the intricate relation between structural priming and implicit language learning (e.g., Chang et al. Reference Chang, Dell and Bock2006; Dell & Chang Reference Dell and Chang2014; Ferreira & Bock Reference Ferreira and Bock2006).
In sum, we propose to extend B&P's account with the notion of a language node connected to linguistic representations at all levels of processing. This extended account makes it possible to capture the dynamics of real-life bilingual language use beyond cross-language priming of unilingual sentences, explaining the processes of both cross-language interactivity (e.g., code-switching) and language accessibility (e.g., L1 attrition). Given that more than half of the world's population is bilingual (e.g., Grosjean Reference Grosjean2010), this extension is by no means trivial; it is relevant and necessary, and strengthens the generalizability of B&P's account.
In their target article, Branigan & Pickering (B&P) argue for structural priming as a key implicit methodology to probe the nature of linguistic representations. They provide extensive data supporting their model, ultimately arguing that structural priming provides a tool to understand the nature of language.
The authors also discuss the implications of their proposal for language representation in bilinguals, focusing primarily on cross-language structural priming between single-language sentences. This research has led to vital insights on cross-language activation at the syntactic processing level (cf., e.g., Hartsuiker & Pickering Reference Hartsuiker and Pickering2008), adding to the accumulation of evidence that language use in bilinguals involves ubiquitous cross-language activation at multiple levels of processing (cf. e.g., de Bot et al. Reference De Bot, Broersma, Isurin, Isurin, Winford and de Bot2009; Kootstra et al. Reference Kootstra, van Hell, Dijkstra, Isurin, Winford and de Bot2009; Kroll et al. Reference Kroll, Bobb and Wodniecka2006). Importantly, however, everyday bilingual communication does not normally happen according to a scripted cross-language priming paradigm with primes in one language and targets in the other (cf. Fricke & Kootstra Reference Fricke and Kootstra2016). To truly understand the nature of language in all its respects, as is B&P's ambition, we propose that their model and approach should be further developed to explain a larger number of bilingual language scenarios as they occur in real life.
One such bilingual language scenario – and a true hallmark of bilingualism – is code-switching, i.e., the use of multiple languages within one single sentence. Code-switching is susceptible to exactly the same structural priming mechanisms as the production of unilingual sentences, in the sense that bilinguals' syntactic choices in the production of code-switched sentences are primed by those of their dialogue partner (Kootstra et al. Reference Kootstra, van Hell and Dijkstra2010). But priming also occurs with dependent variables that are specific to code-switching, namely priming of the sentence position of code-switching (Kootstra et al. Reference Kootstra, van Hell and Dijkstra2012), priming of the language of the inflected verb (i.e., matrix language; Fricke & Kootstra Reference Fricke and Kootstra2016; Kootstra et al. Reference Kootstra, van Hell and Dijkstra2010), and priming of the actual choice to code-switch or not (Fricke & Kootstra Reference Fricke and Kootstra2016; Kootstra et al., Reference Kootstra, van Hell and Dijkstrain revision). Importantly, these findings are based on both experimental and corpus research. This indicates that structural priming is more than a method to investigate linguistic representations; it is a core mechanism of language use that, together with adaptive control processes (cf., Green & Abutalebi Reference Green and Abutalebi2013), appears to guide bilinguals' linguistic behavior in real-life language use.
The critical implication of this code-switching evidence is that B&P's model should be extended with a representation of language membership. After all, for priming of linguistic elements from multiple languages to take place, these multiple languages must somehow be encoded within the representational system. In most models of bilingual language processing, this is implemented by assuming a language node that is linked to linguistic representations (e.g., Hartsuiker & Pickering Reference Hartsuiker and Pickering2008; Kroll et al. Reference Kroll, Bobb and Wodniecka2006). Based on the omnipresence of cross-language activation at all levels of processing, we assume this language node is linked to linguistic representations at all levels of processing (de Bot Reference De Bot2004; de Bot et al. Reference De Bot, Broersma, Isurin, Isurin, Winford and de Bot2009; Kootstra et al. Reference Kootstra, van Hell, Dijkstra, Isurin, Winford and de Bot2009; Reference Kootstra, van Hell and Dijkstra2010). Primed code-switching can then be explained in the form of persisting co-activation of language nodes from the recently experienced discourse (see Fricke & Kootstra Reference Fricke and Kootstra2016; Kootstra et al. Reference Kootstra, van Hell and Dijkstra2010).
In addition to code-switching, the just-described extension of B&P's structural priming account also may serve to explain another fascinating bilingual language scenario: first language (L1) attrition (i.e., loss of or decreased access to L1 representations, mostly due to immersion in a second-language environment, leading to infrequent use of the first language [e.g., Schmid Reference Schmid2011]). Recently, a number of psycholinguistic paradigms have been used to study first-language attrition using offline, online, and neural measures of language comprehension and production (Rossi et al., Reference Rossi, Prystauka, Diaz, Köpke and Keijzerin revision), but the mechanism of priming so far has not been used to study first-language attrition. Based on B&P's point that priming can be seen as evidence of access to linguistic representations, it can be predicted that, if L1 representations are completely inaccessible as a consequence of attrition, L1 structural priming should be nonexistent, whereas if L1 representations merely become less accessible as a consequence of attrition, rates of L1 structural priming may well be relatively strong. This would be consistent with inverse-frequency and surprisal effects found in structural priming studies (e.g., Bernolet & Hartsuiker Reference Bernolet and Hartsuiker2010; Bock Reference Bock1986; Ferreira Reference Ferreira2003; Jaeger & Snider Reference Jaeger and Snider2007; Reference Jaeger and Snider2013). Another prediction that can be made is that structural priming can serve as a very sensitive measure of changing levels of access to L1 linguistic representations, thus making it possible to boost L1 activation for speakers who are undergoing L1 attrition, much along the lines of what has been proposed for aphasic speakers (Rossi Reference Rossi2013). To continue, under the assumption that structural priming boosts access to linguistic representations by easing the demands on cognitive abilities such as memory, structures that are difficult and/or cognitively taxing should benefit the most from structural priming. Interestingly, these predictions not only show how B&P's model and its bilingual extension can be utilized to further test existing issues in L1 attrition, but also showcase the intricate relation between structural priming and implicit language learning (e.g., Chang et al. Reference Chang, Dell and Bock2006; Dell & Chang Reference Dell and Chang2014; Ferreira & Bock Reference Ferreira and Bock2006).
In sum, we propose to extend B&P's account with the notion of a language node connected to linguistic representations at all levels of processing. This extended account makes it possible to capture the dynamics of real-life bilingual language use beyond cross-language priming of unilingual sentences, explaining the processes of both cross-language interactivity (e.g., code-switching) and language accessibility (e.g., L1 attrition). Given that more than half of the world's population is bilingual (e.g., Grosjean Reference Grosjean2010), this extension is by no means trivial; it is relevant and necessary, and strengthens the generalizability of B&P's account.