pith. sign in

arxiv: 2606.19170 · v1 · pith:HAMKX2B3new · submitted 2026-06-17 · 💻 cs.CL

Dango: A Strictly L1-Only Large Language Model for Studying Second Language Acquisition

Shiho Matta , Yin Jou Huang , Fei Cheng , Takashi Kodama , Hirokazu Kiyomaru , Yugo Murawaki This is my paper

Pith reviewed 2026-06-26 21:09 UTC · model grok-4.3

classification 💻 cs.CL
keywords second language acquisitionL1-to-L2 transferlarge language modelsJapanese-Englishdata contaminationcomputational SLAmodel filteringdecoder-only models
0
0 comments X

The pith

A 1.8B model pretrained only on filtered Japanese data acquires human-like English production after targeted lessons.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper builds Dango to study how Japanese speakers acquire English inside a large decoder model without the usual data leaks. It first removes most English from the Japanese pretraining corpus while leaving a small realistic trace, then adds LLM-generated lessons that mimic classroom input. After this sequence the model produces English that matches patterns seen in human learners and beats both unfiltered and ordinary multilingual baselines. The released model and data are meant to let researchers run controlled experiments on L1-to-L2 transfer at scale.

Core claim

Dango is created by filtering a Japanese corpus to limit premature English exposure, pretraining the 1.8B decoder on the cleaned data, and then fine-tuning on generated L2 lessons; the resulting model exhibits human-like L2 production patterns that outperform unfiltered and standard multilingual baselines.

What carries the argument

The filtering method that reduces L2 contamination in the monolingual pretraining corpus while preserving realistic minimal exposure.

If this is right

  • Dango supplies a controllable simulator for testing specific L1-to-L2 transfer predictions at decoder scale.
  • The same filtering-plus-lesson pipeline can be applied to other language pairs to isolate exposure effects.
  • Releasing the model, data, and code enables direct replication of SLA experiments that were previously limited to smaller models.
  • Human-like error patterns in the fine-tuned model can be compared against learner corpora to validate computational SLA claims.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the filtering step proves robust, similar cleaning could improve other multilingual models that currently suffer from unintended cross-lingual leakage.
  • The approach opens the possibility of running ablation studies that vary only the amount of minimal L2 exposure while holding model size fixed.
  • Learner-facing tools could eventually be built by swapping the lesson generator for real classroom materials and measuring alignment with actual student output.

Load-bearing premise

The filtering method successfully reduces premature L2 exposure in the monolingual pretraining corpus while still preserving realistic minimal exposure.

What would settle it

A direct measurement of English token probability or translation accuracy on the filtered model before any L2 fine-tuning that shows no reduction compared with the unfiltered baseline.

Figures

Figures reproduced from arXiv: 2606.19170 by Fei Cheng, Hirokazu Kiyomaru, Shiho Matta, Takashi Kodama, Yin Jou Huang, Yugo Murawaki.

Figure 1
Figure 1. Figure 1: English contamination examples: severe case [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: An overview of our proposed methodology. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Linguistic proficiency assessments across the L1-pretraining stage. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Dango and llm-jp-3 Japanese-to-English translation performance when trained and evaluated on different [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Error Rates (raw values) of human and LLMs. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Contamination: document-level parallel data. [PITH_FULL_IMAGE:figures/full_fig_p011_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Contamination: word/phrase-level parallel [PITH_FULL_IMAGE:figures/full_fig_p011_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: English appearances that are allowed by the [PITH_FULL_IMAGE:figures/full_fig_p012_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Task category breakdown of llm-jp-eval per [PITH_FULL_IMAGE:figures/full_fig_p012_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Case study of English production after com [PITH_FULL_IMAGE:figures/full_fig_p013_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Prompt used to generate textbook-style L2 learning data. [PITH_FULL_IMAGE:figures/full_fig_p014_11.png] view at source ↗
Figure 13
Figure 13. Figure 13: Prompt format used to evaluate the models [PITH_FULL_IMAGE:figures/full_fig_p014_13.png] view at source ↗
Figure 12
Figure 12. Figure 12: Prompt format used to train Dango and llm [PITH_FULL_IMAGE:figures/full_fig_p014_12.png] view at source ↗
Figure 14
Figure 14. Figure 14: Sentence length distribution of the LLM [PITH_FULL_IMAGE:figures/full_fig_p015_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: System prompt for linguistic feature annotation. We build on the framework of [PITH_FULL_IMAGE:figures/full_fig_p016_15.png] view at source ↗
Figure 17
Figure 17. Figure 17: Case study of Dango’s English production [PITH_FULL_IMAGE:figures/full_fig_p017_17.png] view at source ↗
Figure 16
Figure 16. Figure 16: User prompt for annotating Numbers Agree￾ment. Human L1 JSD of UF JSD of ER BCD of ER Japanese 0.0100 0.0214 0.1491 Korean 0.0136 0.0180 0.1496 Urdu 0.0123 0.0257 0.1778 Mandarin 0.0142 0.0253 0.1430 Cantonese 0.0131 0.0436 0.2140 Thai 0.0162 0.0303 0.1818 Malay 0.0112 0.0404 0.1863 [PITH_FULL_IMAGE:figures/full_fig_p017_16.png] view at source ↗
Figure 19
Figure 19. Figure 19: Error Rate distribution of human and LLMs [PITH_FULL_IMAGE:figures/full_fig_p017_19.png] view at source ↗
Figure 21
Figure 21. Figure 21: L1 knowledge injection user prompt for L2 [PITH_FULL_IMAGE:figures/full_fig_p018_21.png] view at source ↗
Figure 22
Figure 22. Figure 22: Dango and llm-jp-3 when trained and evaluated on different levels of Japanese-English data, showing L2 [PITH_FULL_IMAGE:figures/full_fig_p019_22.png] view at source ↗
Figure 23
Figure 23. Figure 23: L1 knowledge injection system prompt for L2 role-playing. The original prompt is from [PITH_FULL_IMAGE:figures/full_fig_p019_23.png] view at source ↗
Figure 24
Figure 24. Figure 24: Our simplified L1 knowledge-injection prompt for L2 role-playing, designed for llm-jp-3.1-instruct. [PITH_FULL_IMAGE:figures/full_fig_p020_24.png] view at source ↗
read the original abstract

We introduce Dango, a 1.8B-parameter large language model designed for controlled studies of L1-to-L2 (Japanese-to-English) transfer in second language acquisition (SLA). While previous studies have explored SLA in language models, they have predominantly relied on smaller or non-decoder models, limiting their ability to generate open-ended text and reducing their suitability as practical L2 simulators. We identify a key challenge when scaling models to this size: L2 contamination within the "monolingual" pretraining corpus used for L1 acquisition. To address this, we propose a filtering method to reduce premature exposure to English while preserving realistic, minimal exposure. We then fine-tune the model on LLM-generated L2-learning lessons to simulate the L2 acquisition process. Our evaluations confirm that Dango develops human-like L2 production patterns, outperforming both unfiltered and standard multilingual baselines. We release the model, data, and code to facilitate reproducible computational SLA research and learner-facing applications.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper introduces Dango, a 1.8B-parameter decoder-only LLM for controlled studies of Japanese-to-English second language acquisition. It identifies L2 contamination in monolingual pretraining corpora as a scaling challenge, proposes a filtering method to enforce realistic minimal English exposure during L1 pretraining, and then fine-tunes the model on LLM-generated L2 lessons. The central claim is that evaluations demonstrate human-like L2 production patterns, with Dango outperforming both unfiltered and standard multilingual baselines; the model, data, and code are released for reproducibility.

Significance. If the filtering successfully produces an L1-dominant corpus and the performance claims are substantiated with detailed metrics, this work would supply a scalable, open decoder-only model for computational SLA research, enabling controlled transfer experiments that smaller or non-decoder models cannot support. The explicit release of model, data, and code is a concrete strength that directly aids reproducibility and downstream learner-facing applications.

major comments (2)
  1. [Abstract] Abstract and evaluation sections: the claim that 'evaluations confirm that Dango develops human-like L2 production patterns, outperforming both unfiltered and standard multilingual baselines' is asserted without any reported metrics, test sets, baselines, statistical tests, or human-judgment protocols, rendering the central empirical result unverifiable from the provided description.
  2. [Methods (filtering procedure)] Methods section on corpus filtering: the proposed filtering procedure is described as the sole mechanism for reducing premature L2 exposure while preserving 'realistic, minimal exposure,' yet no quantitative validation (residual English n-gram rates, ablation on filter aggressiveness, or comparison against actual Japanese child-directed corpora) is supplied; this leaves the load-bearing precondition for attributing later L2 transfer to controlled SLA unanchored.
minor comments (1)
  1. [Introduction] Notation for the 1.8B model size and the distinction between 'unfiltered' and 'standard multilingual' baselines should be defined explicitly on first use to avoid ambiguity for readers outside the immediate SLA subfield.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their thoughtful review and constructive comments. We address each major comment below and plan to incorporate revisions to strengthen the paper.

read point-by-point responses

  1. Referee: [Abstract] Abstract and evaluation sections: the claim that 'evaluations confirm that Dango develops human-like L2 production patterns, outperforming both unfiltered and standard multilingual baselines' is asserted without any reported metrics, test sets, baselines, statistical tests, or human-judgment protocols, rendering the central empirical result unverifiable from the provided description.

    Authors: We agree with the referee that the abstract and evaluation sections should provide the supporting metrics, test sets, baselines, statistical tests, and human-judgment protocols to substantiate the central claim. We will revise the manuscript to include these details, ensuring the empirical results are fully verifiable. revision: yes

  2. Referee: [Methods (filtering procedure)] Methods section on corpus filtering: the proposed filtering procedure is described as the sole mechanism for reducing premature L2 exposure while preserving 'realistic, minimal exposure,' yet no quantitative validation (residual English n-gram rates, ablation on filter aggressiveness, or comparison against actual Japanese child-directed corpora) is supplied; this leaves the load-bearing precondition for attributing later L2 transfer to controlled SLA unanchored.

    Authors: We agree that quantitative validation is essential for the filtering procedure. We will supplement the methods section with residual English n-gram rates, ablations on filter aggressiveness, and comparisons against actual Japanese child-directed corpora. revision: yes

Circularity Check

0 steps flagged

No circularity; sequential training and evaluation steps remain independent

full rationale

The paper describes a linear pipeline: propose and apply an L2 filter to a Japanese corpus, pretrain the 1.8B model, fine-tune on generated L2 lessons, then run separate evaluations against unfiltered and multilingual baselines. No equations, fitted parameters renamed as predictions, or self-citations are invoked to derive the human-like L2 patterns; the outcome is presented as an empirical result of the procedure rather than a definitional or load-bearing reduction to the filter itself. The filtering step is an assumption whose effectiveness is not independently verified in the provided text, but that is a question of evidence strength, not circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract provides no explicit free parameters, axioms, or invented entities; all details on training objectives, filtering thresholds, or evaluation metrics are absent.

pith-pipeline@v0.9.1-grok · 5721 in / 1090 out tokens · 32743 ms · 2026-06-26T21:09:07.981806+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

44 extracted references · 16 canonical work pages

  1. [1]

    Proceedings of the Thirtieth Annual Meeting of the Association for Natural Language Processing (NLP2024) , year =

    Uzushio: A Distributed Huge Corpus Processor for the LLM Era , author =. Proceedings of the Thirtieth Annual Meeting of the Association for Natural Language Processing (NLP2024) , year =

  2. [2]

    Proceedings of the Thirtieth Annual Meeting of the Association for Natural Language Processing (NLP2024) , year =

    Han, Namgi and. Proceedings of the Thirtieth Annual Meeting of the Association for Natural Language Processing (NLP2024) , year =

  3. [3]

    2024 , eprint=

    LLM-jp: A Cross-organizational Project for the Research and Development of Fully Open Japanese LLMs , author=. 2024 , eprint=

    2024

  4. [4]

    Can LLM s Simulate L 2- E nglish Dialogue? An Information-Theoretic Analysis of L 1-Dependent Biases

    Gao, Rena and Wu, Xuetong and Kuribayashi, Tatsuki and Ye, Mingrui and Qi, Siya and Roever, Carsten and Liu, Yuanxing and Yuan, Zheng and Lau, Jey Han. Can LLM s Simulate L 2- E nglish Dialogue? An Information-Theoretic Analysis of L 1-Dependent Biases. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long...

    work page doi:10.18653/v1/2025.acl-long.219 2025

  5. [5]

    Second Language Acquisition of Neural Language Models

    Oba, Miyu and Kuribayashi, Tatsuki and Ouchi, Hiroki and Watanabe, Taro. Second Language Acquisition of Neural Language Models. Findings of the Association for Computational Linguistics: ACL 2023. 2023. doi:10.18653/v1/2023.findings-acl.856

    work page doi:10.18653/v1/2023.findings-acl.856 2023

  6. [6]

    Modeling Nonnative Sentence Processing with L 2 Language Models

    Aoyama, Tatsuya and Schneider, Nathan. Modeling Nonnative Sentence Processing with L 2 Language Models. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing. 2024. doi:10.18653/v1/2024.emnlp-main.283

    work page doi:10.18653/v1/2024.emnlp-main.283 2024

  7. [7]

    BL i MP : The Benchmark of Linguistic Minimal Pairs for E nglish

    Warstadt, Alex and Parrish, Alicia and Liu, Haokun and Mohananey, Anhad and Peng, Wei and Wang, Sheng-Fu and Bowman, Samuel R. BL i MP : The Benchmark of Linguistic Minimal Pairs for E nglish. Transactions of the Association for Computational Linguistics. 2020. doi:10.1162/tacl_a_00321

    work page doi:10.1162/tacl_a_00321 2020

  8. [8]

    2025 , eprint=

    MultiBLiMP 1.0: A Massively Multilingual Benchmark of Linguistic Minimal Pairs , author=. 2025 , eprint=

    2025

  9. [9]

    Proceedings of the AAAI Conference on Artificial Intelligence (AAAI-26) , year =

    HSKBenchmark: Modeling and Benchmarking Chinese Second Language Acquisition in Large Language Models through Curriculum Tuning , author =. Proceedings of the AAAI Conference on Artificial Intelligence (AAAI-26) , year =. 2511.15574 , archivePrefix =

    arXiv

  10. [10]

    SLABERT Talk Pretty One Day: Modeling Second Language Acquisition with BERT

    Yadavalli, Aditya and Yadavalli, Alekhya and Tobin, Vera. SLABERT Talk Pretty One Day: Modeling Second Language Acquisition with BERT. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2023. doi:10.18653/v1/2023.acl-long.657

    work page doi:10.18653/v1/2023.acl-long.657 2023

  11. [11]

    Unsupervised

    Conneau, Alexis and Khandelwal, Kartikay and Goyal, Naman and Chaudhary, Vishrav and Wenzek, Guillaume and Guzm \'a n, Francisco and Grave, Edouard and Ott, Myle and Zettlemoyer, Luke and Stoyanov, Veselin. Unsupervised Cross-lingual Representation Learning at Scale. Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics. ...

    work page doi:10.18653/v1/2020.acl-main.747 2020

  12. [12]

    JBL i MP : J apanese Benchmark of Linguistic Minimal Pairs

    Someya, Taiga and Oseki, Yohei. JBL i MP : J apanese Benchmark of Linguistic Minimal Pairs. Findings of the Association for Computational Linguistics: EACL 2023. 2023. doi:10.18653/v1/2023.findings-eacl.117

    work page doi:10.18653/v1/2023.findings-eacl.117 2023

  13. [13]

    Second Language Acquisition Modeling

    Settles, Burr and Brust, Chris and Gustafson, Erin and Hagiwara, Masato and Madnani, Nitin. Second Language Acquisition Modeling. Proceedings of the Thirteenth Workshop on Innovative Use of NLP for Building Educational Applications. 2018. doi:10.18653/v1/W18-0506

    work page doi:10.18653/v1/w18-0506 2018

  14. [14]

    CALL for Widening Participation: Short Papers from EUROCALL 2020 , editor =

    The Development of an Online Game-based Simulation for the Training of English Language Teachers in Virtual Environments , author =. CALL for Widening Participation: Short Papers from EUROCALL 2020 , editor =. 2020 , month = dec, pages =. doi:10.14705/rpnet.2020.48.1210 , url =

    work page doi:10.14705/rpnet.2020.48.1210 2020

  15. [15]

    Effectiveness of Chatbots in Improving Language Learning: A Meta‐Analysis of Comparative Studies , doi =

    Lyu, Boning and Lai, Chun and Guo, Jianing , month =. Effectiveness of Chatbots in Improving Language Learning: A Meta‐Analysis of Comparative Studies , doi =. 2024 , journal =

    2024

  16. [16]

    Japanese English: Language and Culture Contact , urldate =

    James Stanlaw , publisher =. Japanese English: Language and Culture Contact , urldate =

  17. [17]

    The ICNALE Spoken Dialogue: A New Dataset for the Study of Asian Learners’ Performance in L2 English Interviews , doi =

    Ishikawa, Shin’ichiro , month =. The ICNALE Spoken Dialogue: A New Dataset for the Study of Asian Learners’ Performance in L2 English Interviews , doi =. 2019 , journal =

    2019

  18. [18]

    2022 , eprint=

    Training Compute-Optimal Large Language Models , author=. 2022 , eprint=

    2022

  19. [19]

    言語処理学会 第30回年次大会 発表論文集 , pages =

    ichikara-instruction: LLM のための日本語インストラクションデータの作成 , author =. 言語処理学会 第30回年次大会 発表論文集 , pages =. 2024 , url =

    2024

  20. [20]

    2020 , eprint=

    Language Models are Few-Shot Learners , author=. 2020 , eprint=

    2020

  21. [21]

    Council of Europe , title =

  22. [22]

    Common European Framework of Reference for Languages: Learning, Teaching, Assessment , year =

  23. [23]

    Tono, Yukio , title =

  24. [24]

    Syahid, A

    Tono, Yukio , title =. CEFR Journal---Research and Practice , year =. doi:10.37546/JALTSIG.CEFR1-1 , url =

    work page doi:10.37546/jaltsig.cefr1-1

  25. [25]

    C hat GPT Beyond E nglish: Towards a Comprehensive Evaluation of Large Language Models in Multilingual Learning

    Lai, Viet Dac and Ngo, Nghia and Pouran Ben Veyseh, Amir and Man, Hieu and Dernoncourt, Franck and Bui, Trung and Nguyen, Thien Huu. C hat GPT Beyond E nglish: Towards a Comprehensive Evaluation of Large Language Models in Multilingual Learning. Findings of the Association for Computational Linguistics: EMNLP 2023. 2023. doi:10.18653/v1/2023.findings-emnlp.878

    work page doi:10.18653/v1/2023.findings-emnlp.878 2023

  26. [26]

    1957 , publisher =

    Robert Lado , title =. 1957 , publisher =

    1957

  27. [27]

    1989 , publisher =

    Terence Odlin , title =. 1989 , publisher =

    1989

  28. [28]

    2023 , eprint=

    BLOOM: A 176B-Parameter Open-Access Multilingual Language Model , author=. 2023 , eprint=

    2023

  29. [29]

    2023 , eprint=

    YAYI 2: Multilingual Open-Source Large Language Models , author=. 2023 , eprint=

    2023

  30. [30]

    Investigating

    Constantinescu, Ionut and Pimentel, Tiago and Cotterell, Ryan and Warstadt, Alex. Investigating Critical Period Effects in Language Acquisition through Neural Language Models. Transactions of the Association for Computational Linguistics. 2025. doi:10.1162/tacl_a_00725

    work page doi:10.1162/tacl_a_00725 2025

  31. [31]

    Issues in the Assessment and Evaluation of English Language Education at the Elementary School Level: Implications for Policies in South Korea, Taiwan, and Japan , volume =

    Butler, Yuko , year =. Issues in the Assessment and Evaluation of English Language Education at the Elementary School Level: Implications for Policies in South Korea, Taiwan, and Japan , volume =

  32. [32]

    Japan's emblematic English , doi =

    Hyde, Barbara , month =. Japan's emblematic English , doi =. 2002 , journal =

    2002

  33. [33]

    Learning from the Linguistic Landscape: A Project-Based Learning Approach to Investigating English in Japan , doi =

    Barrs, Keith , month =. Learning from the Linguistic Landscape: A Project-Based Learning Approach to Investigating English in Japan , doi =. 2020 , journal =

    2020

  34. [34]

    JALT2021 Postconference Publication: Reflections and New Perspectives , editor =

    Nakayama, Shusaku , title =. JALT2021 Postconference Publication: Reflections and New Perspectives , editor =. 2022 , publisher =. doi:10.37546/JALTPCP2021-24 , url =

    work page doi:10.37546/jaltpcp2021-24 2022

  35. [35]

    CCN et: Extracting High Quality Monolingual Datasets from Web Crawl Data

    Wenzek, Guillaume and Lachaux, Marie-Anne and Conneau, Alexis and Chaudhary, Vishrav and Guzm \'a n, Francisco and Joulin, Armand and Grave, Edouard. CCN et: Extracting High Quality Monolingual Datasets from Web Crawl Data. Proceedings of the Twelfth Language Resources and Evaluation Conference. 2020

    2020

  36. [36]

    Language ID in the Wild: Unexpected Challenges on the Path to a Thousand-Language Web Text Corpus

    Caswell, Isaac and Breiner, Theresa and van Esch, Daan and Bapna, Ankur. Language ID in the Wild: Unexpected Challenges on the Path to a Thousand-Language Web Text Corpus. Proceedings of the 28th International Conference on Computational Linguistics. 2020. doi:10.18653/v1/2020.coling-main.579

    work page doi:10.18653/v1/2020.coling-main.579 2020

  37. [37]

    Language Contamination Helps Explains the Cross-lingual Capabilities of E nglish Pretrained Models

    Blevins, Terra and Zettlemoyer, Luke. Language Contamination Helps Explains the Cross-lingual Capabilities of E nglish Pretrained Models. Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing. 2022. doi:10.18653/v1/2022.emnlp-main.233

    work page doi:10.18653/v1/2022.emnlp-main.233 2022

  38. [38]

    L1 INFLUENCE ON THE ACQUISITION ORDER OF ENGLISH GRAMMATICAL MORPHEMES , doi =

    Murakami, Akira and Alexopoulou, Theodora , month =. L1 INFLUENCE ON THE ACQUISITION ORDER OF ENGLISH GRAMMATICAL MORPHEMES , doi =. 2015 , journal =

    2015

  39. [39]

    and Miller, Paul W

    Chiswick, Barry R. and Miller, Paul W. , month =. Linguistic Distance: A Quantitative Measure of the Distance Between English and Other Languages , doi =. 2005 , journal =

    2005

  40. [40]

    The ICNALE and Sophisticated Contrastive Interlanguage Analysis of Asian Learners of English , doi =

    Ishikawa, Shin'ichiro , month =. The ICNALE and Sophisticated Contrastive Interlanguage Analysis of Asian Learners of English , doi =. 2013 , journal =

    2013

  41. [41]

    2023 , publisher =

    Ishikawa, Shin'ichiro , title =. 2023 , publisher =. doi:10.4324/9781003252528 , isbn =

    work page doi:10.4324/9781003252528 2023

  42. [42]

    2025 , eprint=

    A Systematic Assessment of Language Models with Linguistic Minimal Pairs in Chinese , author=. 2025 , eprint=

    2025

  43. [43]

    Jarvis, Scott and Pavlenko, Aneta , title =

  44. [44]

    1983 , isbn =

    Strategies in Interlanguage Communication , publisher =. 1983 , isbn =

    1983