Language Attrition and Reactivation in a Simultaneous Japanese–English Bilingual Teenager: A 6-Year Longitudinal Psycholinguistic and Neurolinguistic Case Study

Abstract

Aims and objectives/purpose/research questions:

This study examines long-term bilingual (attrition/reactivation) trajectories in a simultaneous Japanese–English bilingual adolescent during two consecutive cross-national moves. It asks the following: (1) To what extent do the data support the regression hypothesis on a multiyear timescale? (2) When patterns deviate from monotonic loss/recovery, do they align with complex dynamic systems theory (CDST)? (3) Does change in brain activation precede subsequent psycholinguistic attrition or reactivation? and (4) How susceptible is a simultaneous bilingual who moves postpuberty to environment-driven attrition relative to children versus adults?

Design/methodology/approach:

A 6-year longitudinal single-case design tracked a simultaneous Japanese–English bilingual who moved from the United States (birth–16) to Japan for 3 years of local schooling and then returned to the United States for 3 years of tertiary education. Annual psycholinguistic and neurolinguistic assessments were administered across both relocations.

Data and analysis:

Yearly behavioral (psycholinguistic) measures and brain-based indices (neurolinguistic activation) were analyzed to characterize language-specific change, cross-linguistic coupling, non-linear trends, and lead–lag relations between neural and behavioral indices across attrition and reactivation phases.

Findings/conclusion:

(1) A subset of outcomes supports the regression hypothesis, but over a considerably longer horizon than typically reported, indicating multiyear rather than monthscale dynamics. (2) The remaining outcomes exhibit non-linear, context-sensitive fluctuations consistent with CDST. (3) Shifts in brain activation preceded and predicted later behavioral attrition/reactivation, suggesting neurofunctional reconfiguration as a precursor of performance change. (4) Despite simultaneous bilingualism, the adolescent was not impervious to postpubertal, environment-driven attrition; susceptibility appeared attenuated relative to younger children and closer to adult-like resilience.

Originality:

The study integrates annual psycholinguistic and neurolinguistic evidence across two transnational moves, demonstrating that neural change can foreshadow behavioral attrition/reactivation in a simultaneous bilingual adolescent.

Significance/implications:

Findings refine models of bilingual maintenance and loss by linking long-horizon trajectories to both regression hypothesis and CDST accounts, and by positioning neural reconfiguration as an early marker for forthcoming behavioral change—informing timing and targets for support during educational transitions.

Keywords

Language attrition regression hypothesis complex dynamic systems theory fNIRS

Introduction

There are a number of theoretical and methodological problems inherent in previous attrition studies, including (1) hypotheses and theories on language attrition, (2) attributes of attriters or extralinguistic factors (onset age of bilingualism and attrition), and (3) research design (receptive vs. productive, longitudinal vs. cross-sectional, and behavioral vs. neurolinguistic). These issues are first reviewed in an attempt to disclose areas that have seldom been addressed in attrition studies, before formulating the research questions (RQs) in this study.

This study addresses the following overarching questions early to foreground the aims: (RQ1) To what extent does first-language attrition emerge in a highly dominant language under reduced use, and how does it unfold over time? (RQ2) How do behavioral and neurolinguistic indices converge—or diverge—when tracking attrition trajectories longitudinally? (RQ3) To what degree do the findings adjudicate among leading accounts (e.g., regression hypothesis and complex dynamic systems theory [CDST]). We also make explicit that the present work is exploratory in scope.

Hypotheses on Language Attrition

We distinguish incomplete acquisition from attrition as follows. Conceptually, incomplete acquisition refers to structures that never reached target-like representations; attrition denotes subsequent decline in previously established knowledge or performance following reduced use. Operationally, we index incomplete acquisition via pre-reduction patterns or age-/education-matched baselines, and attrition via within-individual longitudinal change from an earlier, better-established level.

To explain the phenomenon of language attrition, a number of hypotheses have been put forward, including (1) the regression hypothesis (the order of language attrition is a mirror image or reverse order of language acquisition, which was tested by Schmid in 2002 with conflicting results, (2) the critical threshold theory (Neisser, 1984 proposed a critical threshold level which guaranteed immunity to loss once an item reached that level, which gained support from Hansen, 1999), (3) the interface hypothesis (Sorace, 2011 posited that external interfaces such as pragmatics were more susceptible to loss than internal language modules such as morpho-syntax, which was supported by a limited number of L1 or heritage attrition studies such as Van Osch et al., 2014), and (4) the 4-M Model (Myers-Scotton and Jake proposed in 2017 that conceptually elected morphemes are more prone to attrition than structurally activated system morphemes, which gained both supportive (Bolonyai, 2002) and contradictory (Taura, 2008) results.

Recently, particular attention has been given to the CDST. Chomsky’s Universal Grammar takes the nativist stance of language acquisition as taking place in a particular language faculty module of the brain, whereas emergentism regards language structures as emerging across a number of modules through meaning-making processes (Mitchell et al., 2019; Ortega, 2009). A rebuttal against UG comes both from linguists (Wulff & Ellis, 2018) and from neurolinguists (Lieven, 2016; Schwieter, 2019; Tomasello, 2003) who claim that different brain networks may come into play, depending on which aspects or levels of language processing are examined. One of the prominent emergentism theories is the CDST (henceforth), which is a combined transdisciplinary label for the complexity theory (Larsen-Freeman, 2017) and the dynamic systems theory (de Bot et al., 2007), sharing the same view that language progression and regression is adaptive, complex, dynamic, systematic, but chaotic, non-linear, and variable in nature (Mitchell et al., 2019). In comparing the CDST to a static model of language development, Dornyei (2014) emphasizes the time scope which makes straightforward outcomes among variables in the static model rather complex because each variable changes over time, impacting the outcome and resulting in an endless chain reaction. Linck and Kroll (2019) claim that the CDST accords well with most bilingualism studies in that bilingual language acquisition and attrition take place by dynamically and constantly adapting to changing demands imposed by the linguistic environments surrounding bilinguals. Bialystok and Kroll (2018) also argue that language development is far from static from infancy through to adolescence, which Knowland and Donlan (2014, p. 155) echo “with change evident at the levels of behaviour, brain, and brain–behaviour relationships.” Although the CDST seems promising to explain language processing, it may also be pointed out that attrition should be treated as a part of the whole language system development by longitudinally integrating the positive development of the other (acquired) language(s), and that cross-sectional group statistical analyses should be replaced by longitudinal intra-personal accounts (Opitz, 2019).

Schmitt (2019) and Mertzen et al. (2021) claim that these theories and hypotheses need to undergo rigorous scrutiny because the number of studies that test them is still sparse. With regard to hypothesis testing, Mehotcheva and Kopke (2019) state that these hypotheses have been tested in L1 attrition studies at least to some extent, but not in research on L2 or FL.

Attributes of Attriters

There has been a call to control the types of target attriters in order for research results to be comparable across studies (Schmid, 2011b, 2019; Taura, 2008). In the literature, attriters vary from first-generation immigrants (Karayayla & Schmid, 2019) to international adoptees (Rakhlin et al., 2015), returnees (Flores, 2020; Taura, 2008), heritage language speakers (see Brehmer & Treffers-Daller, 2020 for overview), and language learners who have terminated their L2 learning (Li, 2018). Mehotcheva and Mytara (2019) argue that it is crucial to identify which category the attriters in question belong to. Thus, Schmid et al. (2004) strongly recommend a minimal requirement that attrition studies take into account the following variables—the age at onset of L2 acquisition (simultaneous, early, or late bilingualism), age at onset of L1 attrition (pre-puberty or postpuberty), incubation period (the defined span of reduced use preceding observed change), literacy education, use (frequency, amount, and settings) of the attriting languages, and the combination of languages being attrited and acquired.

Among these, the age of onset of bilingualism is one of the most crucial variables that make research results difficult to generalize (Karayayla & Schmid, 2019). When adult immigrants who have reached a stable state in their L1 development are compared with pre-puberty immigrants or international adoptees who have experienced a limited amount of exposure to their L1 before being relocated to a totally different language environment, the degree of L1 erosion is more evident in the pre-puberty immigrants and adoptees (Pierce et al., 2019). In these studies, the onset age of abruptly discontinuing to use their L1 ranges from a few months old to those well into adulthood. However, hardly any research has been conducted on simultaneous bilinguals whose acquisition of their two L1s (Lα being the socially dominant language and Lß being the home language, for instance) begins at birth, then continues on when they move to the Lα-dominant environment or change over to the Lß dominant environment. Although the returnee participants in Flores’ (2020) and Taura’s (2008) attrition studies included such simultaneous bilinguals, their cross-sectional design failed to disentangle the onset age zero group from the rest. Studies on heritage speakers traditionally focus on the end state (attrition or incomplete language acquisition) while research on simultaneous bilinguals typically focuses on development (Kupisch, 2019), not attrition, indicating the necessity of attrition studies on simultaneous bilinguals. In addition, claims from Bylund (2019) and Nicoladis and Montanari (2016) show under-researched attrition studies on teenagers whose L2 exposure and use only begins after puberty but before adulthood—between the ages of 12 and 18. The significance of the age of onset of bilingualism is also pointed out by neurolinguistic studies on bilinguals, such as Burgaleta et al. (2016), who claim the possibility of a restructuring in the language network in the bilingual brain due to the cumulative effects of controlling two languages.

Second, the combination of the languages in attrition studies is limited mostly to that of alphabetical languages (Borragan et al., 2021), and there is “virtually no research that involves languages of other typological groups” (Schmitt, 2019: 240), apart from several Japanese–English returnee studies (see Taura, 2019 for summary) and several Chinese attrition studies (Aalberse et al., 2020; Li, 2018). In conducting attrition studies with typologically varied language combinations, Larson-Hall (2019) states that such studies could benefit more from a longitudinal design because it allows researchers to observe precisely what happens after a complete absence or drastic reduction in a once dominant language input.

A third issue involves the types of circumstances where bilingual language acquisition takes place—whether it is (1) in a single language context where one language is dominantly used and the other is used only on limited occasions or (2) in a dual language context where two languages are equally used in the same environment. This distinction, made in the adaptive control hypothesis (Abutalebi & Green, 2016) is said to be one of the predictor variables for a bilingual outcome—complete or incomplete acquisition or attrition.

These issues signify the need for a new line of attrition research to examine simultaneous bilinguals who have acquired two typologically distant languages from birth in a dual language context as they move from an Lα dominant environment to an Lß dominant environment in their teens.

Methodological Considerations

Attrition studies vary in the types of data collected from attriters: (1) receptive versus productive data, (2) cross-sectional versus longitudinal design, (3) behavioral (psycholinguistic) versus neurolinguistic data, and (4) attrition data versus (re)acquisition data or both.

First, the literature indicates that on-line production data are more desirable to detect subtle changes, as opposed to receptive data, based on the fact that productive skills precede receptive skills when language erosion manifests itself (Schmid, 2010; Schwieter, 2019). In addition, Andrews (2019) and Schmid (2011b) claim that it is desirable to collect multiple types of linguistic data such as free speech production and written data.

Second, longitudinal case studies on language attrition have been repeatedly recommended over cross-sectional studies for a number of reasons, admitting such studies lack in generalizability. The first reason involves a long list of variables inherent in attrition studies, which are intertwined and practically impossible to control to form a homogeneous attrition group (Mehotcheva & Kopke, 2019). In case studies, meanwhile, an attriter’s attributes such as socioeconomic status, age, and gender can be all controlled, except for the change in language environment (Baker & Wright, 2021). The second reason pertains to a greater number of identifiable individual differences among bilinguals than monolinguals (Bice & Kroll, 2021; Marian & Hayakawa, 2021; Nicoladis & Montanari, 2016). The third reason is that an attriter’s baseline data are obtained for a more accurate picture of the attrition (Flores, 2020; Montrul, 2020; Schmid & Kopke, 2017; Tomiyama, 2009). The fourth merit of longitudinal case studies lies in detecting which aspect or level of language deteriorates first intra-personally (Schmid & de Leeuw, 2019). The fifth reason is that cross-sectional group averages do not represent any particular individual (de Bot et al., 2007; Schmid, 2011a), which also gains support from neuro-cognitive (Mishra, 2018) and neurolinguistic studies (Rossi et al., 2019). More recently when summarizing the indeterminate results of neurolinguistic studies examining if there is a bilingual cognitive advantage or not, Leivada et al. (2021) regard such an advantage as phantom-like possibly due to cross-sectional research design, therefore calling for more longitudinal case studies. In the same vein, Andrews (2019) calls for such studies when intra-subject variances were found among homogeneous bilinguals, depending upon what aspect or level of language was tapped into. Green (2013, p. 910) even states that “longitudinal studies must represent a gold standard for any effort at creating a causal account of the relationship between bilingual experience and neuroplastic changes.” Thus, it seems desirable for a longitudinal case study design to be used in the future both linguistically or neurolinguistically.

Third, attrition research began with the collection of linguistic data to describe what takes place in the dominant language (usually the L1) when the linguistic environment drastically changes. With the advent of neuro-imaging data technology, collecting both neurolinguistic and behavioral/linguistic attrition data has been called for. From a cognitive neurolinguistic point of view, Andrews (2019) and Steinhauer and Kasparian (2020) reiterate the importance of including behavioral/linguistic data in combination with neurolinguistic data. From a psycholinguistic perspective on L1 attrition, Kopke and Keijzer (2019) and Schmid (2011b) emphasize the inclusion of neurophysiological data in addition to behavioral data. In comparison with abundant neurolinguistic research on bilinguals versus monolinguals, there is a scarcity of neuro-imaging studies on attrition (Paz-Alonso et al., 2019).

fMRI (functional magnetic resonance imaging) is a neuro-imaging technique to observe the hemodynamic response of brain activity. While the EEG/ERP technique has a high temporal resolution, fMRI has a high spatial resolution and is capable of identifying exact brain regions that are activated. Up to date, only a few fMRI attrition studies have been conducted, which all target international adoptees. Among such fMRI studies, the results are contradictory: Pallier et al. (2003) failed to detect any L1 memory traces in Korean adoptees adopted by French families in childhood, whereas Pierce et al. (2014) revealed traces of neural sensitivity in Chinese adoptees. Rossi et al. (2019) point out that these two studies focus on highly specialized target groups and thus suggest future fMRI studies should be conducted on attriters who have completed their L1 acquisition and moved to other linguistic environments.

Such discrepant results are also seen among ERP studies on L1 attrition: (1) no attrition effects in grammatical gender usage in German attriters (Bergmann et al., 2015) and (2) evident attrition in verb and modifier positions in Italian immigrants (Kasparian & Steinhauer, 2017). This is despite the fact that both target immigrant groups had acquired and established their L1 before they emigrated and were now predominantly immersed in their L2. Steinhauer and Kasparian (2020) ascribe such discrepancies to the different approaches to define attrition (1) as a change in competence using the Chomskyan term, where performance errors are not regarded as such in Bergmann et al.’s study, and (2) as a change in the L1 neuro-cognitive process that is revealed as a qualitative effect in Kasparian and Steinhauser’s research.

Rossi et al. (2019) proposed that neuro-imaging approaches to attrition studies should take a longitudinal design, based on the assumption that L1 language attrition is “a flexible, continuous adaptation of the language system that is characterized by modulations in the linguistic, cognitive, and neural functioning of the L1 in the face of bilingualism.” If a higher level of scrutiny is called for concerning the CDST, which suggests the constant and ever-changing nature of the languages spoken by bilinguals, Kasparian and Steinhauser’s definition of attrition needs to be tested even further than Bergmann et al. (2015).

The final issue concerning attrition research design relates to the question of whether attrition data alone are sufficient. The CDST approach to language attrition takes the stance that language development is dynamic and both attrited and acquired languages need to be equally taken into account in that the former is regarded as negative development while the latter is positive (Montrul, 2020). Opitz also (2019, p. 53) raises this precise issue by proposing “a need to consider language attrition vis-a-vis the development of the overall system.” Except for a handful of studies, including Opitz (2011), who examined adult migrants’ L1 and L2 proficiency after having been extensively immersed in L2 environments, most attrition studies focus solely on the attrited languages, not the interplay of two languages in interplay.

There is also scant research to date which longitudinally tracks those who undergo language attrition and then re-learn or re-acquire the attrited language. They include research focusing on (1) international adoptees whose exposure to L1 is limited to early childhood and then residual knowledge is tested in adulthood (Pierce et al., 2019; Sorokina, 2019), (2) heritage speakers who reside in an L2 dominant environment with very limited exposure to their heritage language during childhood, and then, they learn it later in life (see Oh et al., 2019 for overview), (3) missionaries who learn an L2 but attrite it back in their home country (Hansen, 2011), and (4) school age returnees who spend a lengthy period of time in L2 communities and then return to their home country (Flores, 2020). The first three are not strictly longitudinal because they failed to obtain baseline data before attrition set in. This leaves Flores (2020) to be the only longitudinal study, where Flores tracked a girl who grew up in a first-generation Portugal immigrant family in Germany, where she acquired German at school and Portuguese at home until her family moved back to Portugal when she was 9. The family then stayed in Portugal for 4 years, when her Portuguese became dominant until they went back to Germany when she was 13. The first data were collected 13 months after her return to Portugal and showed inaccurate use of German case, gender, and plural marking. These errors had significantly decreased in the second data collected 11 months after re-entry into Germany. In this way, Flores’ behavioral data successfully showed the dynamic state of the girl’s L1 regression and progression, lending support to the CDST and regression hypothesis.

Among the longitudinal studies that examine both progression and regression, a study by Osterhout et al. (2019) alone collected both behavioral and neurolinguistic data. They tracked 16 native speakers of English enrolled in Finnish courses at an American university over 9 months of instruction. Their acquisition data had been collected three times at the mean instruction hours of 36, 71, and 129 hours before the attrition phase, followed for an identical amount of time. Both behavioral (receptive phonological sensitivity) and ERP data supported the regression hypothesis—a linear progression in the acquisition phase and regression in the attrition phase. Apart from this one study on a classroom-learned foreign language, no longitudinal exploration has ever been made on L1 or L2 acquisition and the subsequent attrition (or attrition followed by subsequent reactivation) using a combination of both behavioral and neurolinguistic approaches.

Research Questions

By synthesizing the literature, some gaps are revealed that have not been examined as yet. One such gap is the need to investigate attrition in simultaneous bilinguals who have acquired two typologically distant languages from birth as they move from an Lα dominant environment to an Lß dominant environment in their teens. In addition, it is desirable to design such research to (1) be longitudinal, (2) employ both psycholinguistic and neurolinguistic approaches, and (3) test existing hypotheses on language attrition. Thus, the following research questions have been formulated.

Research Question 1 (RQ1): When simultaneous bilinguals whose Lα and Lß are typologically distant, move between those two language environments, do their longitudinal data show a dynamic regression and progression as seen in Flores (2020)?

Research Question 2 (RQ2): Are such dynamic changes equally observable in their neurolinguistic data, as seen in the work by Osterhout et al. (2019)?

Research Question 3 (RQ3): Do such data support any existing hypotheses on language attrition?

Method

Participants and Soundscape

A right-handed Japanese–English simultaneous bilingual was tracked for 6 years. Hana (a pseudonym) was born to a Japanese family and spent the first 16 years of her life in the United States where she attended local nursery, primary, and junior high schools. Her parents’ language policy had her use only Japanese among family members at home and attend a Saturday Japanese school, which helped her to acquire Japanese literacy and become a balanced bilingual. Hana’s family moved back to Japan when she was 16 years old, where she enrolled in a local senior high school, and her social and school-medium language totally shifted from English to Japanese, resulting in a drastic reduction in her exposure to and use of English. However, she was fortunate enough to receive a daily 60-minute English lesson by joining a mainstream (not EFL) class offered to her at the appropriate grade level by the international school that shared the same school site as hers. This enabled her to maintain a minimum daily contact in English.

Baseline data were collected 2 months after Hana’s return to Japan (incubation period INC .02) with subsequent yearly data collection at INC 1.02, 2.03, and 3.01. Upon graduating from high school, she pursued a tertiary education in the United States, but yearly data collection continued for another 3 years at the end of each academic year. Thus, two fully symmetrical periods of data collection took place, covering a 3-year-long English attrition in Japan and a subsequent 3 years of English reactivation in the United States.

Yearly data collection included a self-assessment of her own receptive and productive skills in Japanese and English. The data showed it took her 2 years to catch up in Japanese literacy skills, whereas her Japanese oral skills were always equal to those of her friends. Meanwhile, her self-assessment of her English skills was on par with her grade level. Interviews with her high school teachers in Japan confirmed her self-assessment. Retrospective interview data with Hana revealed that her contact with the Japanese language was virtually reduced to nil once back in the United States where she lived in a university dormitory.

A signed consent form was obtained yearly from Hana and her parents.

Procedures

Both psycholinguistic and neurolinguistic data were collected yearly. Spontaneous productive (oral and written) data were collected for psycholinguistic analyses. An oral interview lasting a few minutes was audio-recorded for fluency analyses. For writing, The Test of Written Language 3rd edition (TOWL-3, Hammill & Larsen, 1996) was used to elicit English writing for 15 minutes, which was then analyzed for fluency, accuracy (based on the 4-M model), and vocabulary, along with the writing scores provided by the TOWL-3. The semi-structured interview focused on asking about the participant’s experiences over the past year, including academic life, social interactions, and changes since the previous session. Approximate duration per language per session: ~10–15 min. Outcome measures included speech rate (syllables/s), mean length of utterance, and error profiles.

Second, neurolinguistic data were collected while Hana undertook verbal fluency tasks (VFT) in English and Japanese, using a 42-channel Shimadzu OMM-3000 fNIRS (functional near-infrared spectroscopy) apparatus, which detects oxygenated (Oxy-Hb) every 130 ms. fNIRS was used in preference to fMRI because of its superiority in temporal resolutions, quiet (suitable for language tasks), and above all its “relatively resistant to movement artifacts which makes it ideal for use with vulnerable and/or mobile population” (Minagawa & Cristia, 2019: 154). Out of the 42 channels, the main focus is placed on the left inferior frontal gyrus (BA 44 and 45, IFG, hereafter) which is identified as the region controlling language, in particular generating verbal words (Abutalebi & Green, 2008; Minagawa & Cristia, 2019). The number of words she produced for each VFT was counted as behavioral data while the fNIRS raw data (mM-mm) underwent standard analysis for a statistical comparison possible (between-tasks and intra-person). In this study’s VFT, we administered letter (phonemic) and semantic/category fluency (such as color names) in each language, with fixed-duration production epochs (e.g., 30 seconds) and standard exclusions (no proper nouns, no inflected variants). Task order was counterbalanced across sessions; instructions were scripted; a brief practice preceded each block.

Results

Psycholinguistic Data

Lα (English) Spontaneous Oral Interview Data

Pauses lasting more than 200 ms were identified as not being caused by physiological reasons (Howell et al., 1994). Table 1 summarizes the total interview duration and the average pause duration over the 6 years. A severe fluency attrition (slower speech) is evident at IINC 2.03 and 3.01, from under 300 ms before INC 1.02 to more than 600 ms after INC 203 (Figure 1). This disfluent speech rate, however, reverted back to a previous time of 300 ms (more fluent speech) as soon as Hana had returned to the English dominant milieu as seen at USA 1.00. Thus, Hana’s Lα fluency suffered from severe attrition in the second and third years in the Lß dominant environment, while her Lα fluency picked up in the very first year back in the Lα dominant environment.

Table 1.

Oral Interview Data.

Incubation period (years)	0.02	1.02	2.03	3.01	USA1.00	USA2.00	USA3.00
Total interview duration (milliseconds)	122,986	162,226	149,291	226,471	93,731	102,746	214,647
Average pause duration (milliseconds)	280.3	278.1	664.4	729.4	317.6	331.4	319.8

Figure 1.

Average pause duration (milliseconds).

Lα (English) Spontaneous Written Data

The TOWL-3 uses a picture prompt to elicit a test taker’s English writing data. Two writing excerpts are provided below:

Excerpt @ INC 0.02 (two months after returning to Japan)

December 21, 3010; It’s been exactly ten years since my space team and I left on a journey to outer space. Our mission was to go to Mars and look for life, when we discovered there was a civilization already developing in one of the many craters . . .

Excerpt @ USA 3 years (three years back in the USA)

It’s 2050 and I finally made it on the moon. Here I am, living here. I think of my family back home on Earth, but life on the moon is also quite interesting. I would not change it for anything else. 20 years ago, I took a spaceship to the moon with two other . . .

All the collected texts were analyzed for vocabulary, fluency, accuracy, and writing skills (Table 2).

Table 2.

Written Data.

Incubation period (years)	0.02	1.02	2.03	3.01	USA1.00	USA2.00	USA3.00
Vocabulary (#token)	166	230	197	196	206	168	239
Vocabulary (#type)	112	130	104	103	105	119	129
Type–token ratio (TTR)	0.67	0.57	0.53	0.53	0.51	0.71	0.54
Fluency (#tokens/minute)	11.1	15.3	13.1	13.1	13.7	11.2	15.9
Accuracy (content)	100.0	99.4	99.3	100.0	100.0	100.0	100.0
Accuracy (early)	100.0	100.0	100.0	100.0	93.8	100.0	100.0
Accuracy (late)	100.0	97.9	98.2	93.3	100.0	96.8	100.0
Accuracy (total)	100.0	99.3	99.2	99.1	98.9	99.5	100.0
Writing (CC)	13	13	14	15	14	14	17
Writing (CL)	15	16	18	18	19	17	18
Writing (StC)	15	14	14	15	14	15	13
Writing (quotient)	128	128	134	139	136	134	139

Jarvis (2019) emphasizes the importance of analyzing writing data in terms of lexical diversity and density in attrition studies. Such an analysis on Hana’s data showed a non-linear fluctuation in the number of types and tokens, but a gradual and incremental attrition in Japan and re-acquisition in the United States in type–token ratio (TTR; unique types / total tokens) (Table 1).

Writing fluency, the number of tokens produced per minute, underwent fluctuation in the range of 11.1 to 15.9 words, showing neither attrition nor re-acquisition.

An accuracy analysis was conducted, employing the 4-M model (Myers-Scotton & Jake, 2017), which postulates that content morphemes should be the first acquired and attrited, followed by early system morphemes and finally by two late system morphemes. This 4-M based analysis managed to detect such late system errors as “he was left” instead of “he had left” at IINC 2.03 and “every day passed by peaceful” instead of “peacefully” at INC 3.01, although such grammatically erroneous expressions are sometimes uttered even by an English native speaker. However, Hana maintained extremely high accuracy, showing no attrition in particular types of morphemes as seen in Table 2 which indicates her accuracy fluctuated in a very narrow range from 100.0% at INC .02 (baseline) to 99.3% at INC 1.02, 98.2% at INC 2.03, and 99.1% at INC 3.01.

Hana’s TOWL-3 scores (Table 2) show more than 13 in the sub-components of CC (contextual conventions—basic writing rules), CL (contextual language—grammar and vocabulary), and StC (story construction) with the range of 8 to 12 being set as the NS average, and more than 128 in quotient (overall writing skill) with the range of 90 to 110 being set as the NS average. This signifies that her Lα writing was always in the above-average range compared with her English native speaker counterparts of the same age throughout the observed 6 years.

Behavioral Outcomes During Verbal Fluency (Moved From 3.2.1)

The number of words generated in the VFT is summarized in Table 3. Hana’s linguistic environment, whether in Japan or the United States, failed to determine the number of words produced for each task, that is (1) the Japanese letter and English category tasks showed an overall constant increase in volume, and (2) the Japanese category and English letter tasks underwent fluctuations in number.

Table 3.

VFT Behavioral Data.

Task/incubation	0.02	1.02	2.03	3.01	USA1.00	USA2.00	USA3.00
E-letter	26	16	32	20	24	26	30
J-letter	16	20	26	22	22	26	30
E-category	25	20	28	36	34	40	38
J-category	28	34	26	30	26	32	32

[MINOR: Reassigned under Behavioral data; concurrent with fNIRS recording]

[MINOR: Content moved to 3.1.3]

Neurolinguistic Data

Behavioral Data

fNIRS Data

fNIRS studies on hemodynamic behavior in the brain are “based on the assumption that an increase in blood flow signals an increase in brain activity” (Minagawa & Cristia, 2019, p. 155), which leads to the general interpretation that greater oxygenated hemoglobin (Oxy-Hb) values signify more of an effort needed for the task. This accords well with the notion of Green’s (1998) Inhibitory Control Model and Paradis’s (2007) Activation Threshold Hypothesis in that the dominant and more proficient language of the two in a bilingual speaker means less brain activation due to more frequent and recent access, but more brain activation is required once the dominant language use has been discontinued and attrition sets in.

The Oxy-Hb (mM-mm) levels in the left IFG during both tasks in both languages are summarized in Table 4. The results show both a straightforward (E-letter and J-category tasks) and an indeterminate (J-letter and E-category tasks) trend in brain activation.

Table 4.

fNIRS Data (mM-mm) in the Left IFG.

Task/incubation		0.02	1.02	2.03	3.01	USA 1.00	USA 2.00	USA 3.00	ANOVA	Eta squared
E-letter	mean	−0.1738	−0.0619	0.4010	0.6578	1.0797	0.1420	−0.2138	F(6,838) = 100.062**	.4174
E-letter	SD	0.5276	0.4494	0.6932	0.2600	0.5841	0.2594	0.3824
J-letter	mean	0.1125	0.5698	0.3762	0.0105	0.3828	0.5871	0.3022	F(6,837) = 27,954**	.1669
J-letter	SD	0.4716	0.4980	0.5130	0.4048	0.5607	0.3345	0.4409
E-category	mean	−0.287	0.1825	0.0871	0.7339	0.9422	0.8578	0.9546	F(6,837) = 39.539**	.2210
E-category	SD	0.52552	0.71591	0.99469	0.90292	1.12517	0.75727	0.72768
J-category	mean	0.2998	0.1292	−0.0524	−0.0546	−0.0489	0.268	0.7414	F(6,838) = 20.133**	.1260
J-category	SD	0.75553	0.63396	0.65813	0.69477	0.61807	0.56164	0.98253

E(nglish) and J(apanese).

p < .001.

Two sets of analysis of variance (ANOVA) and post hoc Bonferroni tests reveal that the Oxy-Hb in the IFG on the English (Lα) letter task gradually rose during Hana’s stay in Japan but dropped when she was in the United States, F(6,838) = 100.62, p < .001, EtaSquared = .4172 as shown in Figures 2 and 3, while the activation level in the Japanese (Lß) category task gradually went down in Japan but picked up again in the United States, F(6,838) = 20.133, p < .001, EtaSquared = .1260 as shown in Figures 4 and 5. In the English VFT, Hana did not need to activate the brain much at all in the first 14 months in Japan (INC 0.02 and 1.02) and also back in the United States from the second year onwards (USA 2.00 and 3.00). However, she needed significantly more energy to retrieve English lexical items in her second year back in Japan (IINC 2.03 and 3.01) and in the first year back in the United States (USA 1.00). This seems to present a straightforward picture of her English activation, depending on the linguistic environment where she lived and received education (except for the Oxy-Hb data in the United States 1.00 which were expected to be significantly lower than at INC 3.01 in Japan). This unexpectedly high brain activation at USA 1.00 was apparently triggered by her uneasy and nervous psychological state as revealed in a retrospective interview where she expressed her concern over the researcher’s high expectations of her English after having studied for a year in the United States. This emotional impact on language processing within neural networks is reported by Kopke and Keijzer (2019).

Figure 2.

fNIRS data in English VFT (letter).

Figure 3.

Brain imaging in English letter task.

Figure 4.

fNIRS data in Japanese VFT (category).

Figure 5.

Brain imaging in Japanese category task.

When examining Hana’s Japanese (Lß), using VFT (see Figures 4 and 5), her brain activation was the least during her second and third year in Japan (INC 2.03 and 3.01) and the first year in the United States (USA 1.00). Significantly more brain activation was observed in the INC periods of 0.02 and 1.02 during the first 14 months in Japan and USA INC 2.00 (the second year in the United States), until the third year in the United States (USA 3.00) when the most brain activation occurred. This seems to signify a distinct but linear brain activation pattern in the following way: (1) as Hana’s stay in Japan lengthened, she needed increasingly less energy to retrieve Japanese lexical items, but (2) this effortless state with her Japanese lasted for only 1 year after she returned to the United States, (3) in the second year in the United States her Japanese activation level reverted back to that seen at INC 0.02, and (4) back in the United States 3.00 the effort required to access her Japanese was the highest.

Thus, as seen in the English letter task and the Japanese category task, brain activation patterns followed the expected trend. They showed a linear change where the socially dominant language progressively induced less brain activation whereas the socially less-dominant language gradually resulted in higher brain activation for lexical retrieval. However, this was not the case with the English category and the Japanese letter tasks.

Data Triangulation

Schmid (2011b) points out the discrepancy often observed between measured decline and experienced decline by attriters themselves. This was observed in Hana’s self-assessment of her Lα English speaking skills that had showed no decline (always 3 out of 5-point Likert-type scale) in Japan whereas her English oral fluency deteriorated acutely in the third year in Japan (Table 1). On the contrary, looking at her Lß Japanese, in line with her own retrospective remark, it took her 2 years to catch up with her friends in Japanese language skills, which was supported by her fNIRS data. Thus, the importance of data triangulation seems upheld in attrition studies which should integrate neurolinguistic data.

Discussion

To avoid a diffuse engagement with many models, we foreground two complementary frameworks—regression hypothesis and CDST—and examine how behavioral and fNIRS findings converge/diverge with these accounts, reserving other models for brief consideration.

Psycholinguistic analyses of Hana’s Lα (English) data show the selective nature of attrition and re-acquisition. First, linear attrition in lexical density (TTR) of her writing and oral fluency was observed in both the second and third years in the Lß (Japanese) environment, although such erosion quickly disappeared in the very first year of her re-immersion in the Lα milieu. Second, Hana’s high Lα accuracy and writing skills were maintained throughout the 6 years. Third, a non-linear fluctuation was observed in her writing fluency and in the number of types and tokens produced in the spontaneous written tasks. Similarly, Hana’s neurolinguistic data in the left IFG showed the selective nature of attrition and re-acquisition. First, when she was engaged in the English letter and Japanese category tasks, the fNIRS data revealed a linear change where it was progressively more difficult for Hana to retrieve Lα lexical items in the Lß (Japanese) environment and also for her to retrieve Lß lexical items in the Lα (English) environment. Second, the brain oscillatory signals from the English category and Japanese letter tasks exhibited a totally non-linear fluctuation, which was also seen in the behavioral data.

Thus, both the psycholinguistic and neurolinguistic data analyses share the same features of either linear development or non-linear fluctuation. The feature of linear development appears to lend strong support to the regression hypothesis. This psycholinguistic attrition and subsequent reactivation due to the change of socially dominant language shows an identical pattern to Flores’ (2020) longitudinal study, where an early but not simultaneous Portuguese–German bilingual suffered from severe attrition in the accurate use of German. The morphological errors observed in the 13th month of her time in Portugal at age 9 were no longer observable at age 13 in the 11th month after her family had returned to Germany.

Next, neurolinguistically, the linear changes in Hana’s brain activation when she was engaging in English letter tasks exhibited a mirror image of the progressively greater activation of her Lα English when she was in Japan but a gradually smaller activation occurred while she was in the United States. Furthermore, Hana’s Lß Japanese brain activation data when she engaged herself in Japanese letter tasks showed the striking symmetry of an acquisitional and attritional pattern identical to that of Osterhout et al.’s (2019) longitudinal study on adult Anglophone speakers, where ERP showed a mirror image between a 9-month-long FL (Finnish) learning phase and a post-instruction phase in discerning phonological violations. To date, data supporting the regression hypothesis have been reported in adult immigrants or children who move to a linguistically new environment before their L1 has solidified. Thus, this study seems to fill a missing gap by providing supportive evidence that comes from a simultaneous teenage bilingual who has established a firm language base in both Lα English and Lß Japanese when she was engaged in an on-line productive task.

In addition, a closer look at Figures 1 and 2 discloses that brain activation appears to be more sensitive to the linguistic environment and could possibly be a precursor of the psycholinguistic attrition that follows. The Oxy-Hb values in the left IFG showed an increase (though not significant yet with p > .05) in the 14th month compared with the second month (Figure 2), only to reach significantly greater brain activation in the 27th month (p < .05) due to incremental effects over the first 2 years in Japan. The non-alignment of behavioral and neurolinguistic data is pointed out by Paap et al. (2015) and Beatty-Martinez and Dussias (2019). Green and Kroll (2019) take it a step further in mentioning possible brain sensitivity to the everyday language environment or even to classroom language learning before this becomes evident in the behavioral form. Rossi et al. (2019) claim that neurolinguistic changes derived from L1 attrition can be manifested even in the absence of linguistic changes. This study seems to exemplify these arguments in the longitudinal attrition data showing that neurolinguistic changes in lexical retrieval precede psycholinguistic changes.

Apart from these data, which exhibit an adaptive, dynamic, and linear change, Hana’s remaining data indicate either a partial or a lack of sensitivity to changes in her linguistic environment. Throughout the observed period, her English (Lα) accuracy and writing proficiency remained high, whereas her productive vocabulary in terms of the number of types and tokens underwent non-linear fluctuation. Another non-linear fluctuation was detected in her brain activation when she was engaging in the Japanese (Lß) letter and English category tasks. When examining the psycholinguistic data, initial attrition in local sub-skills and subsequent erosion in global skills to synthesize sub-skills are reported by Kopke and Keijzer (2019) and Yoshitomi (1999). This seems to adequately explain the discrepancy of Hana’s high maintenance of accurate morpho-syntax and writing skills versus lexical fluctuation, which indicates a possible attrition in writing and accuracy with a longer span than 3 years. Though showing a fluctuating nature during the 6 years, the number of types produced in the spontaneous writing tasks did not differ greatly when the range was compared. The number of types ranged between 103 and 130 in Japan and between 105 and 129 in the United States. With a closer look at the types, it was found that (1) more than 90% of the words used in each session belong to the list of the first 2,000 words (K1 and K2) most frequently used by English native speakers, according to the (Complete Lexical Tutor v.8.3, n.d.) based on the British National Corpus, but (2) it does not seem to be a simple re-emergence of forgotten words in the re-acquisition stage as observed in Tomiyama’s (2009) attriter. K3 and K4 level words such as civilization, immense, and crater were used in the earlier sessions in Japan, but words at the same level, such as architect, devastate, and simultaneous appeared in later sessions in the United States. This implies that Hana’s acquisition of new lexical items was possibly due to her total immersion in English in the United States where she was undertaking a tertiary education rather than remembering forgotten words. Jarvis (2019) mentions possible factors in lexical attrition, namely the recency and frequency of words appearing in the attriting language and the onset age of language attrition. Hana continued to be exposed to and use English throughout the 6 years, and when she moved to the Lß environment, she was 16 years of age by which time her English was well established. The complex interplay of these factors in Hana’s data could mean no clear-cut linear lexical attrition or re-acquisition. On the contrary, the fNIRS data in the IFG resulting from English letter and Japanese category tasks proved sensitive to her linguistic environments. However, this was not the case with the English category and Japanese letter tasks, which appeared to indicate varied time scales for different variables that could have interacted with one another as seen in language acquisition (Bylund, 2019). This state of non-linear and somehow chaotic language development over time due to the complex interplay of different variables, with linear positive and negative language development sensitive to the linguistic environment, appears to lend strong support to the CDST. However, no conclusion can be made about the cause of these two distinctively differing patterns (linear versus fluctuating). Indeed, there is no guarantee that this pattern is replicated in other language combinations, as Garcia-Penton et al. (2016: 321) argue about bilingual language combinations that “languages of different typologies shape the human brain and its functions in different ways.”

Kopke and Keijzer (2019) and Schmitt (2019) discuss existing hypotheses related to attrition and conclude that there is no single hypothesis or theory that can explain the complexity of the attrition processes. Taking this point of view, this study in analyzing Hana’s psycholinguistic and neurolinguistic data seems to support both the regression hypothesis and the CDST. In contrast, while the 4-M Model gained support from Schmitt (2010), Hana’s data using this model were indeterminate. Partial support for the critical threshold theory and the interface hypothesis was gained in terms of the robustness to loss of the high level of acquired morpho-syntax and writing skills. The dynamic change in brain activation levels in both Lα and Lß when Hana was engaged in lexical retrieval tasks is possibly supportive of neuroemergentist approaches such as the adaptive control hypothesis (Green & Abutalebi, 2013) and The Dynamic Restructuring Model (Pliatsikas, 2020) that posit that structural adaptations take place in the brain, induced by the linguistic environments that bilinguals live in over time.

Conclusion and Future Scope

Hana’s data exhibited a dynamic state of language progression and regression from both psycholinguistic and neurolinguistic perspectives over a much longer time span than in the studies by Flores (2020) and Osterhout et al. (2019), lending strong support to the regression hypothesis.

At the same time, the remaining psycholinguistic and neurolinguistic data also revealed non-linear and fluctuating language development, which can be interpreted as supportive of the CDST. The data appeared to lend partial support to the critical threshold theory and the interface hypothesis, but showed indeterminate support toward the 4-M Model. The neurolinguistic data are possibly indicative of supporting neuroemergentist approaches.

Triangulation of the two sets of data revealed that brain activation appeared to be more sensitive to the linguistic environment and could be a precursor of psycholinguistic attrition or the re-acquisition that followed. Meanwhile, both psycholinguistic and neurolinguistic data underwent a similar linear regression/progression and non-linear fluctuation, depending on which aspects of the language were examined. Even a simultaneous bilingual whose Lα and Lß had been firmly established by age 16 experienced both positive and negative language development as she moved between the Lα and Lß dominant environments. This signifies that bilingual teenagers are not impervious to attrition when they move to a different language environment after puberty, although the extent may not be as evident as seen in younger children, being situated somewhat closer to adults on the attrition-resistant continuum.

This study was undertaken to fill in the missing gap in the literature by formulating three research questions. Answers were suggested above, including several new findings. However, the current research is not without flaws. First, Hana’s Lß (Japanese) data were not sufficient to enable a valid comparison with her Lα (English). Second, the present neurolinguistic enquiry focused on the brain activation in the left IFG alone, but how activation networks need to be examined further. Third, phonological accounts could have been included to draw a more persuasive conclusion as to whether or not Hana’s psycholinguistic data support the existing hypotheses on attrition. When a preliminary version of this paper was presented at a conference, a phonological shift was pointed out—Hana’s initial Californian accent became neutralized (typically heard at international schools worldwide) during her sojourn in Japan, and then, she had a Boston accent in the end. Incorporating these additional data, a more decisive conclusion could have been drawn about the three research questions posed in this study.

Footnotes

Acknowledgements

The authors are grateful for the participation of the participant in this study.

ORCID iD

Hideyuki Taura

Ethical Considerations

This study was approved by the Institutional Ethics Committee of Ritsumeikan University, Japan.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported in part by the research grants awarded to the author by the Japan Society for the Promotion of Science (B#25284111 and B#19H01291).

Declaration of Conflicting Interests

The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Author Biography

Hideyuki Taura is Professor of Applied Linguistics at the Graduate School of Language Education and Information Science, Ritsumeikan University, Japan. He obtained his PhD from Macquarie University, Sydney. His research interests include (l) bilingual language acquisition and attrition from both linguistic and neuro-linguistic perspectives, (2) early bilingual cognitive development, and (3) bilingual narrative development.

References

Aalberse

Andringa

Faber

Lippe

(2020). Definiteness in Wenzhounese Chinese in the Netherlands and in China evidence for generational change in two locations. In Brehmer

Treffers-Daller

(Eds.), Lost in transmission: The role of attrition and input in heritage language development (pp. 16–31). John Benjamins.

Abutalebi

Green

D. W.

(2008). Control mechanisms in bilingual language production: Neural evidence from language switching studies. Language and Cognitive Process, 23, 557–582.

Abutalebi

Green

D. W.

(2016). Neuroimaging of language control in bilinguals: Neural adaptation and reserve. Bilingualism: Language and Cognition, 19(4), 689–698.

Andrews

(2019). Cognitive neuroscience and multilingualism. In Schwieter

J. M.

(Ed.), The handbook of the neuroscience of multilingualism (pp. 19–47). Wiley Blackwell.

Baker

Wright

W. E.

(2021). Foundations of bilingual education and bilingualism. Multilingual Matters.

Beatty-Martinez

A. L.

Dussias

P. E.

(2019). Bilingualism, executive function, and beyond: Questions and insights. John Benjamins.

Bergmann

Meulman

Stowe

L. A.

Sprenger

S. A.

Schmid

M. S.

(2015). Prolonged L2 immersion engenders little change in morphosyntactic processing of bilingual natives. NeuroReport, 26, 1065–1070.

Bialystok

Kroll

(2018). Can the critical period be saved? A bilingual perspective. Bilingualism: Language and Cognition, 21, 908–910.

Bice

Kroll

J. F.

(2021). Grammatical processing in two languages: How individual differences in language experience and cognitive abilities shape comprehension in heritage bilinguals. Journal of Neurolinguistics, 58, 100963.

10.

Bolonyai

(2002). Case systems in contact: Syntactic and lexical case in bilingual child language. Southwest Journal of Linguistics, 21, 21–36.

11.

Borragan

Bruin

Hava

Diego-Balaguer

Vulchanova

M. D.

Vulchanov

Dunabeitia

J. A.

(2021). Differences in word learning in children: Bilingualism or linguistic experience? Applied Psycholinguistics, 42, 345–366.

12.

Brehmer

Treffers-Daller

(2020). Lost in transmission: The role of attrition and input in heritage language development. John Benjamins.

13.

Burgaleta

Sanjua

Ventura-Campos

Sebastian-Galles

Avial

(2016). Bilingualism at the core of the brain. Structural differences between bilinguals and monolinguals revealed by subcortical shape analysis. NeuroImage, 125, 437–445.

14.

Bylund

(2019). Age effects in language attrition. In Schmid

M. S.

Kopke

(Eds.), The Oxford handbook of language attrition (pp. 277–287). Oxford University Press.

15.

Complete Lexical Tutor v.8.3. (n.d.). https://www.lextutor.ca/vp/comp/

16.

de Bot

Lowie

Verspoor

M. H

. (2007). A dynamic systems theory approach to second language acquisition. Bilingualism: Language and Cognition, 10, 7–21.

17.

Dornyei

(2014). Researching complex dynamic systems: “Qualitative retrodictive modeling” in the language classroom. Language Teaching, 47, 80–91.

18.

Flores

(2020). Attrition and reactivation of a childhood language: The case of returnee heritage speakers. Language Learning, 70(Suppl. 1), 85–121.

19.

Garcia-Penton

Garcia

Y. F.

Costella

Dunabeitia

J. A.

Carreiras

(2016). The neuroanatomy of bilingualism: How to turn a hazy view into the full picture. Language, Cognition and Neuroscience, 31, 3030–3327.

20.

Green

D. W.

(1998). Mental control of the bilingual lexico-semantic system. Bilingualism: Language and Cognition, 1(2), 67–81.

21.

Green

D. W.

(2013). Individual variability and neuroplastic changes. Applied Psycholinguistics, 35, 910–912.

22.

Green

D. W.

Abutalebi

(2013). Language control in bilinguals: The adaptive control hypothesis. Journal of Cognitive Psychology, 25, 515–530.

23.

Green

D. W.

Kroll

(2019). The neurolinguistics of bilingualism: Plasticity and control. In de Zubicary

G. I.

Schiller

N. O.

(Eds.), The Oxford handbook of neurolinguistics (pp. 261–294). Oxford University Press.

24.

Hammill

D. D.

Larsen

S. C.

(1996). Test of written language (3rd ed.). pro-ed.

25.

Hansen

(1999). Second language attrition in Japanese contexts. Oxford University Press.

26.

Hansen

(2011). The acquisition, attrition, and relearning of mission vocabulary. In Schmid

M. S.

Lowi

(Eds.), Modeling bilingualism: From structure to chaos. In honor of Kees de Bot (pp. 115–134). John Benjamins.

27.

Howell

R. A.

Saling

M. M.

Bradley

D. C.

Berkovic

S. F.

(1994). Inerictal language fluency in temporal lobe epilepsy. Cortex, 30, 469–478.

28.

Jarvis

(2019). Lexical attrition. In Schmid

M. S.

Kopke

(Eds.), The Oxford handbook of language attrition (pp. 241–250). Oxford University Press.

29.

Karayayla

Schmid

M. S.

(2019). First language attrition as a function of age at onset of bilingualism: First language attainment of Turkish-English bilinguals in the United States. Language Learning, 69, 106–142.

30.

Kasparian

Steinhauer

(2017). When the second language takes the lead: Neurocognitive processing changes in the first language of adult atrriters. Frontiers in Psychology, 8, 1–22.

31.

Knowland

Donlan

(2014). Language development. In Mareschal

Butterworth

Tolmie

(Eds.), Educational neuroscience (pp. 134–171). Wiley Blackwell.

32.

Kopke

Keijzer

(2019). Introduction to psycholinguistic and neurolinguistic approaches to language attrition. In Schmid

M. S.

Kopke

(Eds.), The Oxford handbook of language attrition (pp. 61–72). Oxford University Press.

33.

Kupisch

(2019). 2L1 simultaneous bilinguals as heritage speakers. In Schmid

M. S.

Kopke

(Eds.), The Oxford handbook of language attrition (pp. 458–469). Oxford University Press.

34.

Larsen-Freeman

(2017). Complexity theory: The lessons continue. In Ortega

Han

Z. H.

(Eds.), Complexity theory and language development: In celebration of Diane Larsen-Freeman (pp. 11–50). John Benjamins.

35.

Larson-Hall

(2019). Lexical attrition. In Schmid

M. S.

Kopke

(Eds.), The Oxford handbook of language attrition (pp. 377–390). Oxford University Press.

36.

Leivada

Westergaard

Dunabeitia

J. A.

Rothman

(2021). On the phantom-like appearance of bilingualism effects on neurocognition: (How) should we proceed? Bilingualism: Language and Cognition, 24, 197–210.

37.

(2018). Empirical research on English language attrition of students studying non-English majors in a Chinese University. Journal of the World Federation of Associations for Teacher Education, 3a, 40–55.

38.

Lieven

(2016). Usage-based approaches to language development: Where do we go from here? Language and Cognition, 8, 346–368.

39.

Linck

J. A.

Kroll

J. F.

(2019). Memory retrieval and language attrition: Language loss or manifestations of a dynamic system? In Schmid

M. S.

Kopke

(Eds.), The Oxford handbook of language attrition (pp. 87–97). Oxford University Press.

40.

Marian

Hayakawa

(2021). Measuring bilingualism: The quest for a “bilingualism quotient”. Applied Psycholinguistics, 42, 527–548.

41.

Mehotcheva

T. H.

Kopke

(2019). Introduction to L2 attrition. In Schmid

M. S.

Kopke

(Eds.), The Oxford handbook of language attrition (pp. 330–348). Oxford: Oxford University Press.

42.

Mehotcheva

T. H.

Mytara

(2019). Exploring the impact of extralinguistic factors on L2/FL attrition. In Schmid

M. S.

Kopke

(Eds.), The Oxford handbook of language attrition (pp. 349–363). Oxford University Press.

43.

Mertzen

Lago

Vasishth

(2021). The benefits of preregistration for hypothesis-driven bilingualism research. Bilingualism: Language and Cognition, 24, 344–357.

44.

Minagawa

Cristia

(2019). Shedding light on language function and its development with optical brain imaging. In de Zubicary

G. I.

Schiller

N. O.

(Eds.), The Oxford handbook of neurolinguistics (pp. 154–185). Oxford University Press.

45.

Mishra

R. K.

(2018). Bilingualism and cognitive control. Springer.

46.

Mitchell

Myles

Marsden

(2019). Second language learning theories. Routledge.

47.

Montrul

(2020). Attrition, addition, and age. Second Language Research, 36, 213–217.

48.

Myers-Scotton

Jake

(2017). Revisiting the 4-M model: Codeswitching and morpheme election at the abstract level. International Journal of Bilingualism, 21, 40–366.

49.

Neisser

(1984). Interpreting Harry Bahrick’s discovery: What confers immunity against forgetting? Journal of Experimental Psychology, General, 113, 32–35.

50.

Nicoladis

Montanari

(2016). Bilingualism across the lifespan: Factors moderating language proficiency. De Gruyter Mouton.

51.

J. S.

T. K.

Jun

S. A.

Lee

R. M.

(2019). Childhood language memory in adult heritage (re)learners. In Schmid

M. S.

Kopke

(Eds.), The Oxford handbook of language attrition (pp. 481–492). Oxford University Press.

52.

Opitz

(2011). First language attrition and second language acquisition in a second language environment [Unpublished PhD thesis].

53.

Opitz

(2019). A complete dynamic systems perspective on personal background variables in L1 attrition. In Schmid

M. S.

Kopke

(Eds.), The Oxford handbook of language attrition (pp. 49–60). Oxford University Press.

54.

Ortega

(2009). Understanding second language acquisition. Routledge.

55.

Osterhout

Pitkanen

McLaughlin

Zeitlin

(2019). Event-related potentials as metrics of foreign language learning and loss. In Schmid

M. S.

Kopke

(Eds.), The Oxford handbook of language attrition (pp. 403–416). Oxford University Press.

56.

Paap

R. K.

Johnson

H. A.

Sawi

(2015). Bilingual advantages in executive functioning either do not exist or are restricted to very specific and undermined circumstances. Cortex, 69, 265–278.

57.

Pallier

Dehaene

Poline

J. B.

LeBihan

Argenti

A. M.

Dupoux

Mehler

(2003). Brain imaging of language plasticity in adopted adults: Can a second language replace the first? Cerebral Cortex, 13, 155–161.

58.

Paradis

(2007). L1 attrition features predicted by a neurolinguistic theory of bilingualism. In Kopke

Schmid

M. S.

Keijzer

Dostert

(Eds.), Language attrition: Amsterdam: A theoretical perspective (pp. 121–133). John Benjamins.

59.

Paz-Alonso

P. M.

Oliver

Quinones

Carreiras

(2019). Neural basis of monolingual and bilingual reading. In Greig

Niels

(Eds.), The Oxford handbook of neurolinguistics (pp. 603–625). Oxford University Press.

60.

Pierce

Klein

Chen

J. K.

Delcenserie

Genesee

(2014). Mapping the unconscious maintenance of a lost first language. Proceeding of the National Academy of Sciences of the United States of America, 111, 17314–17319.

61.

Pierce

Genesee

J. F

Klein

(2019). Language loss and language learning in internationally adopted children. In Schmid

M. S.

Kopke

(Eds.), The Oxford handbook of language attrition (pp. 470–480). Oxford University Press.

62.

Pliatsikas

(2020). Understanding structural plasticity in the bilingual brain: The dynamic restructuring model. Bilingualism: Language and Cognition, 23, 459–471.

63.

Rakhlin

Hein

Doyle

Hart

Macomber

Ruchkin

Tan

Grigorenko

(2015). Language development of internationally adopted children: Adverse early experiences outweigh the age of acquisition effect. Journal of Communication Disorders, 57, 66–80.

64.

Rossi

Prystauka

Diaz

(2019). Neuroimaging perspectives on L1 attrition and language change. In Schmid

M. S.

Kopke

(Eds.), The Oxford handbook of language attrition (pp. 166–177). Oxford University Press.

65.

Schmid

M. S.

(2010). Languages at play: The relevance of L1 attrition to the study of bilingualism. Bilingualism: Language and Cognition, 13, 1–7.

66.

Schmid

M. S.

(2011a). Contact x time: External factors and variability in L1 attrition. In Schmid

M. S.

Lowie

(Eds.), Modeling bilingualism: From structure to chaos. In honor of Kees de Bot (pp. 155–176). John Benjamins.

67.

Schmid

M. S.

(2011b). Language attrition. Cambridge University Press.

68.

Schmid

M. S.

de Leeuw

(2019). Introduction to linguistic factors in language attrition. In Schmid

M. S.

Kopke

(Eds.), The Oxford handbook of language attrition (pp. 180–190). Oxford University Press.

69.

Schmid

M. S.

Kopke

(2017). The relevance of first language attrition to theories of bilingual development. Linguistic Approaches to Bilingualism, 7, 637–667.

70.

Schmid

M. S.

Kopke

Keijzer

Weilemar

(2004). First language attrition: Interdisciplinary perspectives on methodological issues. John Benjamins.

71.

Schmitt

(2010). When boundaries are crossed: Evaluating language attrition data from two perspectives. Bilingualism: Language and Cognition, 13, 63–72.

72.

Schmitt

(2019). Morphological attrition. In Schmid

M. S.

Kopke

(Eds.), The Oxford handbook of language attrition (pp. 228–240). Oxford University Press.

73.

Schwieter

J. W.

(2019). The Handbook of the neuroscience of multilingualism. Wiley Blackwell.

74.

Sorace

(2011). Pinning down the concept of ‘interface’ in bilingualism. Linguistic Approaches to Bilingualism, 1(1), 1–33.

75.

Sorokina

(2019). When do memories go away? L1 attrition effect on bilingual autobiographical memory [Unpublished doctoral dissertation, Temple University].

76.

Steinhauer

Kasparian

(2020). Brain plasticity in adulthood—ERP evidence for L1-attrition in Lexicon and morphosyntax after predominant l2 use. Language Learning, 70(Suppl. 2), 171–193.

77.

Taura

(2008). Bilinguality and bilingualism in Japanese school-aged children. Akashi Shoten.

78.

Taura

(2019). Attrition studies on Japanese returnees. In Schmid

M. S.

Kopke

(Eds.), The Oxford handbook of language attrition (pp. 391–402). Oxford University Press.

79.

Tomasello

(2003). Constructing a language: A usage-based theory of language acquisition. Harvard University Press.

80.

Tomiyama

(2009). Age and proficiency in L1 attrition: Data from two siblings. Applied Linguistics, 30, 253–275.

81.

Van Osch

Hulk

Sleeman

Van Suchtelen

P. I

. (2014). Gender agreement in interface contexts in the oral production of heritage speakers of Spanish in the Netherlands. Linguistics in the Netherlands, 31, 93–106.

82.

Wulff

Ellis

N. C.

(2018). Usage-based approaches to SLA. In Mille

Bayram

Rothman

Serratrice

(Eds.), Bilingual cognition and language: The state of the science across its subfields (pp. 37–56). John Benjamins.

83.

Yoshitomi

(1999). On the loss of English as an L2 by Japanese returnee children. In Hansen

(Ed.), Second language attrition in Japanese contexts (pp. 21–58). Oxford University Press.