The Neurocognitive Underpinnings of Second Language Processing: Knowledge Gains From the Past and Future Outlook: A Response to Open Peer Commentaries

Abstract

Writing a review of the neural underpinnings of second language (L2) learning and processing, with a serious eye to future avenues for research, is among the most fun writing invitations that I have ever received. If not curtailed by Language Learning’s word limit, this article would have become a full issue, or even a book! I am thrilled that my passion for this field and enthusiasm for the future of neurocognitive inquiries into L2 learning and processing is shared by eminent and highly esteemed colleagues in the field who read and commented on this keynote article. These commentators lauded the field's amazing achievements, offered their praise and thoughtful insights on future promises and avenues outlined in my review, and extended several of these ideas in interesting and engaging directions.

In my review paper, I started with two lines of classical studies that set the research stage and sparked highly productive lines of research. I then illustrated the field's impressive achievements by selectively reviewing electrophysiological and neuroimaging research on L2 processing and bilingual brain organization and outlined major insights acquired over the past 25 years. I also discussed changing perspectives (including individual variability and experience-based perspectives, neural network approaches, neuroplasticity and L2-learning related neural changes) and identified challenges, promises and future directions in order to better understand the neurocognitive underpinnings of L2 learning and processing. Such future directions include revisiting the native-speaker benchmark for L2 attainment and related methodological implications, applying advanced electrophysiological and neuroimaging techniques to better capture newer perspectives in the field, increasing linguistic diversity in neurocognitive research on L2 processing, enhancing the ecological validity of neurocognitive experimentation, intensifying research on child L2 learners’ brain, and adopting a lifelong approach to L2 learning.

One theme that emerged from the commentaries is the overall agreement on the critical importance of incorporating individual differences perspectives and approaches in future research on L2 learning and processing to push knowledge forward (as explicitly voiced by Martin and Stoehr, Wong, Rossi and Nakamura, Birdsong, and Marian). As I had concluded in my article, future research should move beyond studying the roles of age of acquisition and L2 proficiency and embrace a wider focus on learner-internal and learner-external variables that shape L2 learning trajectories and L2 learners’ neurocognitive profiles. We need to better capture how L2 learners’ experiences (including age of acquisition but also current language uses and environmental context; see, e.g., DeLuca et al., 2019; Gullifer et al., 2018) and variability in cognitive functions (e.g., cognitive control, working memory, declarative and procedural memory abilities), language learning aptitude, and motivation impact the neural correlates of L2 learning and processing. Moreover, an experience-based perspective also encompasses the notion that, building on Grosjean's language modes (e.g., Grosjean, 2001), bilingualism is not a static but a dynamic phenomenon that varies along a continuum of how bilinguals utilize their languages in various sociolinguistic contexts and that changes across the life-span for most bilingual speakers.

In their commentary, Clara Martin and Antje Stoehr elaborated on the critical importance of studying individual variability in neural correlates of L2 learning and processing by highlighting several variables that so far have received relatively little empirical attention. One of these variables is auditory processing precision (“having a good ear”), an individual's lower-order abilities in precisely perceiving domain-general acoustic information (i.e., pitch, formants, duration, and intensity). Auditory processing has been associated with L2 speech learning success (for review, see Saito, in press). Martin and Stoehr convincingly argued that, because auditory processing is critical in identifying word and phrase boundaries, morphosyntactic markers, and syntactic structures, the assessment of L2 learners’ auditory processing precision is important to better understand individual variability in L2 learning and processing (Martin and Stoehr also pointed at open-source tools [Mora-Plaza et al., 2022]) to measure auditory processing precision). A related point was offered by Patrick Wong, from a neurocognitive perspective. Highlighting research on individual differences in neural speech tracking and research from his lab demonstrating that pretraining differences in learners’ cortical functional networks were associated with their future success in learning words of an artificial spoken language (Sheppard et al., 2012), Wong proposed that future work may explore how individual variation in neural speech tracking of different chunk sizes (cf. Ding et al., 2015) may result in variability in L2 learning outcomes. To further advance research on how individual differences impact L2 learning and processing, Wong made the valuable suggestion to adopt machine learning techniques to make predictions about individual learners’ learning outcomes as has been successfully done in research on neural speech encoding in native language acquisition (Wong et al., 2021).

Martin and Stoehr also highlighted that variability in L2 processing may be partially explained by variability in first language (L1) processing, and I concur with the importance of measuring L2 learners’ variability in L1 processing. I add the caveat here that, as recently evidenced by Vermeiren and Brysbaert (2023), researchers should be cautious using vocabulary tests developed for native speakers of that language, even when testing advanced L2 speakers. The critical importance of studying bilinguals’ L1 processing was also highlighted in Jorge Valdés Kroff and Keng-Yu Lin's commentary, yet for a different but somewhat related reason: L1 processing can change because of L2 learning. Valdés Kroff and Lin postulated that, in fact, successful L2 learning and real-time processing require adaptive changes to the L1, and the recruitment of domain-general processes to regulate the language systems. A comprehensive understanding of L2 learning and processing should therefore also entail a close inspection of the learner's L1 processing and L2-learning-induced changes therein. Valdés Kroff and Lin exemplified this by the observation that Spanish–English bilinguals’ frequent exposure to specific patterns of codeswitched determiner–noun phrases induced changes in how L1 Spanish was processed in monolingual contexts and that the Spanish–English bilinguals showed adaptive changes that differed from monolingual speakers (Valdés Kroff & Dussias, 2023).

Variability in native language processing was also highlighted by David Birdsong. In addition to further historically contextualizing ongoing debates on the critical period hypothesis and the native-speaker benchmark for L2 learning outcomes, Birdsong advocated studying patterns of dispersion in native speaker data as well as in L2 learner data. He specifically encouraged researchers to conduct analyses of signal dispersion in their behavioral, electrophysiological, and neuroimaging data and to study patterns of signal dispersion (e.g., via the coefficient of variation [CV], to quantify the signal's variability) within and across participant groups, measures, tasks, and stimulus types. I concur that signal dispersion analyses add value to the researchers’ toolbox to further quantify how signal variability across languages shapes the cognitive and neural correlates of L1 and L2 processing, and L2-learning-induced changes in language processing, in L2 learners and bilinguals.

Two sources of (inter)individual variability highlighted by Martin and Stoehr, namely the speakers’ experience with target and nontarget languages and their exposure to native- and nonnative-accented input, align with the key point made by Eleonora Rossi and Megan Nakamura. Rossi and Nakamura expanded on the importance of better capturing variability in the L2 learner and bilingual experiences and ways to optimally model this variability in order to better understand how it shapes neural indices of L2 processing. While acknowledging the value of the language entropy measure (that estimates the social diversity of language use and has been used to characterize individual differences in bilingual/multilingual language experience related to the social diversity of language use [Gullifer et al., 2018; Gullifer & Titone, 2020]), Rossi and Nakamura illustrated how personal social network (PSN) analysis can further advance our understanding of how bilingual experience may affect the neurocognitive correlates of L2 processing. Social network analysis identifies patterns of relationships, behaviors, or experiences among social actors, enabling researchers to explore how variability in individuals’ social environment predicts or affects particular outcomes. PSN analysis (or egocentric network analysis) is concerned with social networks around specific individuals (i.e., egos), the members of their networks (i.e., alters), and the relationships among alters. Cuartero, Rossi, and colleagues (2023) unpacked how PSN analysis can be used to better understand the complex language-related dynamics and heterogeneity that characterize heritage speaker bilingualism. In their thoughtful commentary, Rossi and Nakamura proposed to extend the use of PSN analysis to understand how variability in language use affects the behavioral and neural correlates of L2 learning and processing. Rossi and Nakamura highlighted a particularly valuable aspect of PSN analysis, namely that it captures variability in language use beyond the individual (i.e., ego). Specifically, PSN analysis not only measures variability in language at the level of the L2 learner (ego)—as do language experience questionnaires (such as Language Experience and Proficiency Questionnaire [LEAP-Q; Marian et al., 2007] and the Language History Questionnaire [LHQ; Li et al., 2020, the language entropy measure [Gullifer & Titone, 2020], and the bilingualism quotient [Marian & Hayakawa, 2021])—it also collects information on communicative behaviors of the L2 learner (ego) and members of their network (i.e., alters), as well as communication behaviors among the members of the network. I agree with Rossi and Nakamura that these unique indices of structural and compositional features of communication patterns in L2 speakers’ networks (such as codeswitching patters among network members; Navarro et al., 2022) have strong potential to further advance our understanding of how complex language-related dynamics and sources of sociolinguistic variation can shape L2 learners’ language use and neurocognitive profiles. I will add that integrating PSN analysis into neurocognitive research on L2 learning and processing also aligns with recent calls to incorporate sociolinguistic and sociocultural approaches to better understand the cognitive and neural bases of L2 learning and processing (as also voiced in Titone and Tiv's [2023] “Systems Framework of Bilingualism”; Tiv et al. [2022]), as well as neural network science approaches that use a data-driven quantitative approach to model language structure.

Viorica Marian explicitly related language experience to neural networks and agreed that neural network approaches to L2 processing are a valuable newer research direction. In her commentary, she cited research from her lab and others evidencing that (bi/multi)language experience changes the neural signatures associated with a multitude of language processes (including speech, language learning, and competition within and across languages) and cognitive functions (such as attentional and executive control); language experience can even impact the subcortical encoding of sounds and otoacoustic emissions (sounds generated from within the inner ear). Marian further concurred that the neural network approach is a promising research avenue in the neuroscience of L2 learning and bilingualism, particularly in light of the fast developments in generative AI and large language models that utilize deep learning in natural language processing and natural language generation tasks. As these large language models are pretrained on vast amounts of data and potentially challenge long-held beliefs and established empirical knowledge in language science, Marian is exactly right that our field needs to make sure that questions and insights on the neurocognitive underpinnings of L2 learning, bi/multilingual experience, and linguistic diversity take a central stage in research and discussions in generative AI and large language models. In fact, with my colleagues at Penn State, I lead an NSF-funded research training program for graduate students in the language sciences, psychology, communication sciences and disorders, information sciences and technology, learning design and technology, and computer science and engineering (entitled “Linguistic diversity across the lifespan: Transforming training to advance human–technology interaction”) in which these discussions are integral to the students’ research training and research design projects. This also illustrates another parallel in the many ways Viorica Marian's and my professional and personal lives overlap—as she so elegantly portrayed in her commentary.

Taomei Guo, Cristina Sanz, Jorge Valdés Kroff and Keng-Yu Lin, and Patrick Wong also commemorated the strides that the field has made, reinforced and acknowledged the value of the directions of future research that I had outlined in my target article, and picked up on several themes and extended them in interesting and engaging directions. Guo highlighted the value of noninvasive brain stimulation techniques, such as transcranial magnetic stimulation and transcranial direct current stimulation, to examine the causal relations between specific brain regions and L2 learning and processing. These neurocognitive intervention techniques carry a strong promise to push the field forward, as they allow the field to leverage current insights largely based on observational neurocognitive methods (electroencephalography, functional magnetic resonance imaging) to make causal inferences about specific brain regions and language functions (for a detailed review of using noninvasive brain stimulation in L2 learning and bilingualism research, see Pandža, in press).

Patrick Wong reinforced my future research recommendation to make an effort to enhance the ecological validity of neurocognitive research on L2 learning and processing. I fully endorse his suggestion to examine how the brains of learners and teachers interact by studying brain synchronies during conversations involving L2 learners and interactions in the L2 classroom, using, for example, hyperscanning techniques. Indeed, interbrain coupling during face-to-face interactions and (electroencephalography-based) hyperscanning techniques have been successfully used in public spaces, such as museums and festivals (Dikker et al., 2021), and in classrooms (e.g., Davidesco et al., 2023; Dikker et al., 2017); the technical know-how and insights obtained in “real-world neuroscience” (Matusz et al., 2019) can be readily applied to L2 classroom contexts. Jorge Valdés Kroff and Keng-Yu Lin also voiced a belief that enhancing ecological validity is imperative for future research endeavors and highlighted the importance of understanding speakers’ more nuanced use of their L2 and their processing of discourse and pragmatic expressions beyond the level of morphosyntactic processing. I fully agree with their statement that much more work is needed to better understand the neural and cognitive mechanisms associated with L2 learners “high-end” L2 language use, such as figurative language (including idioms and metaphors), irony, politeness, humor, narrative and expository discourse, and emotional expressions (for a review on the neuropragmatics of L2 processing, see Citron, 2023).

In her particularly creative commentary, Cristina Sanz took several topics that I had identified as issues that need to be resolved, gaps in current knowledge, and promising avenues of future research as the starting point for designing an empirical study (“thought experiment”) that overcomes these limitations and that models a key step forward in understanding the neurocognitive underpinnings of L2. This exemplary experiment elegantly incorporated many of my and others’ recommendations, including using rigorous research designs, moving beyond the native speaker benchmark and acknowledging that L2 learning trajectories are complex and multifaceted, incorporating individual variability and dynamic changes in both L1 and L2 processing, as well as enriching linguistic diversity and ecological validity and considering the translational implications of research outcomes. Sanz's thought experiment is an example of how we can optimally move the field forward. So let us do it! And let us leverage the insights and discussions on open science practices in Marsden and Morgan-Short's (in press) keynote article in Language Learning’s 75^th Jubilee volume.

To conclude, the neurocognition of L2 learning and processing is a relatively young field that has yielded tremendously rich insights and has made significant strides forward in the past decades. The peer commentators to my keynote article have each made, and continue to make, foundational contributions to this field, and I thank all of the commentators for their thoughtful engagement with my ideas and their invaluable insights. As part of the celebration of Language Learning’s 75^th Jubilee edition, I hope that my keynote article and the commentators’ insights will inspire readers, contribute to shaping and paving the way for new discoveries, and nudge knowledge forward to new levels of fully understanding the complexity and enchantment of learning and processing multiple languages.