In 1998, while at a symposium in New York City organized by applied linguist Professor Aneta Pavlenko, I was introduced to one of the other panelists whose last name started with “Van” and who turned out to be from the same small village in the Netherlands as my Dutch in-laws. That day, a professional camaraderie was born, one that saw us reconnecting at conferences and meetings in many countries and universities over the years. It saw us start faculty positions, establish and run laboratories, design experiments, write papers, teach and mentor students, earn tenure, then promotion, then even more service. It also saw us raise children, cope with aging parents, and navigate life. Decades summed in one paragraph, because academic journals leave out the intertwined work and life challenges of the professoriate. And yet, academia is punctuated by precisely this type of professional relationships built over decades of conference conversations in the pursuit of knowledge—the short-term ontogeny of individual lives evolving alongside the long-term phylogeny of science.

Science can seem relentless at times—there are always more questions to be answered, and asked, new things to be discovered, knowledge to be gained. Especially when it comes to a field as young as neuroscience, where new tools and methodologies drive discoveries every day. It can be exciting to figure out the answers to questions that nobody in the world knows yet. In the case of second language (L2) learning and processing, these are questions about how the brain learns and manages two or more languages.

Professor Janet van Hell's contributions to understanding the neurocognitive correlates of L2 learning and processing have helped shape the field over the years. Her keynote article has offered readers a comprehensive review of the neuroscience of L2 processing, starting with the basics of electroencephalography and functional magnetic resonance imaging technologies through the many variables that impact L2 neural, structural, and functional changes, all the way to the neural networks that support language processing in multilingual speakers, as well as current challenges and future directions. It has provided a concise overview of the literature and is an excellent introduction to the field for newcomers who want to understand its evolution.

Although in science at large the monolingual prism has continued to dominate, those who study language are acutely aware of the diversity of language experiences around the world and know that language experience shapes the brain. The studies discussed in this review have clearly shown that as language experiences change, so do the neural networks subserving them. The networks activated by language are not identical across speakers with different language backgrounds and indeed not even within speakers as their language experiences evolve.

Moment by moment, individual experiences, including language experiences, continually rewrite one's neural networks (Marian, 2023). The variability in language use is tied to variability in neural networks. As language knowledge advances, as new words and grammars and functions are acquired, the connectivity of the neural network is modified. It has been proposed that the constant juggling of two or more languages creates a more interconnected neural network that is responsible for the cognitive and neural reserve observed in speakers of multiple languages.

Experimental evidence has suggested that using multiple languages changes the neural processing of speech in children (Krizman et al., 2015) and the neural consistency and attentional control of adolescents (Krizman et al., 2014). In adults, experience with multiple languages changes the neural signatures of language learning (Bartolotti et al., 2017), the cortical control of between- and within-language competition (Marian et al., 2017), and the recruitment of executive control regions during phonological competition (Marian et al., 2014). Beyond cortical changes, experience with L2 processing also modifies the subcortical encoding of sound (Krizman et al., 2012). Language experience even impacts biological phenomena such as otoacoustic emissions, which are sounds generated from within the inner ear (Marian et al., 2018).

Indeed, there is a rich body of evidence from empirical research conducted in laboratories around the world that has yielded support for Dr. van Hell's review of the neurocognitive underpinnings of L2 processing. I particularly appreciated that the article highlighted recent findings that the degree of functional connectivity within the language network is shaped by the developmental timeline of L2 learning. Also notable, the review did not miss the need for increased diversity in neurocognitive research, the limitations of averaging across individual differences, and the acknowledgment of the many other yet-unanswered puzzles waiting to be solved. The science of L2 acquisition has evolved from a history of case studies and descriptive analyses to empirical cognitive and behavioral experiments, to current neuroscientific investigations, to a future of large data and human-machine interaction.

I believe that, at the present time, the neural network approach is a particularly productive direction to pursue in the neuroscience of bilingualism, despite the admittedly still-limited tools at researchers’ disposal. The review mentioned recent work on the language network, the default-mode network, and resting-state functional connectivity in those who know more than one language. The neural network approach to studying L2 processing is particularly relevant in today's emerging world of large language models, with generative pretrained transformers taking center stage. These models challenge many long-held assumptions in the field of language learning, from innate grammar to critical age, and raise questions about similarities and differences between how humans and machines learn language and even about the definition of language itself.

The current discussions about large language models and Dr. van Hell's review of the neurocognition of L2 processing suggest that the field of L2 learning and processing is entering an era in which the neurocognitive underpinnings of language will gain increased relevance. As the AI race heats up, researchers must make sure that questions about L2 processing, bilingual and multilingual experience, and linguistic diversity are an integral part of those conversations (Marian, 2023).

1998 年，在应用语言学家 Aneta Pavlenko 教授在纽约举办的一次研讨会上，我被介绍给了另一位小组成员，他的姓氏以“Van”开头，结果证明他来自荷兰的同一个小村庄作为我的荷兰姻亲。那天，一种专业的友情诞生了，多年来我们在许多国家和大学的会议上重新建立了联系。我们开始担任教职，建立和运行实验室，设计实验，撰写论文，教学和指导学生，获得终身教职，然后晋升，然后甚至提供更多服务。它还见证了我们抚养孩子、应对年迈的父母以及驾驭生活。几十年来在一个段落中进行了总结，因为学术期刊忽略了教授职位相互交织的工作和生活挑战。但是，

科学有时似乎是无情的——总是有更多的问题需要回答和提出，新的事物需要发现，知识需要获得。尤其是对于像神经科学这样年轻的领域，新的工具和方法每天都会推动新的发现。找出世界上尚无人知道的问题的答案可能会令人兴奋。就第二语言 (L2) 学习和处理而言，这些问题涉及大脑如何学习和管理两种或多种语言。

Janet van Hell 教授多年来在理解第二语言学习和处理的神经认知相关性方面做出的贡献帮助塑造了该领域。她的主题文章为读者提供了对 L2 处理的神经科学的全面回顾，从脑电图和功能磁共振成像技术的基础知识开始，通过影响 L2 神经、结构和功能变化的许多变量，一直到神经网络支持多语言使用者的语言处理，以及当前的挑战和未来的方向。它提供了对文献的简明概述，对于想要了解其演变的新手来说是对该领域的极好的介绍。

尽管在整个科学领域，单语棱镜仍然占据主导地位，但研究语言的人敏锐地意识到世界各地语言体验的多样性，并且知道语言体验塑造了大脑。本综述中讨论的研究清楚地表明，随着语言体验的变化，支持它们的神经网络也会发生变化。不同语言背景的说话者之间由语言激活的网络并不相同，甚至随着语言体验的发展，说话者内部也不尽相同。

每时每刻，个人经历，包括语言经历，都会不断重写一个人的神经网络（Marian，2023）。语言使用的可变性与神经网络的可变性有关。随着语言知识的进步，随着新单词、语法和功能的获得，神经网络的连接性也会发生变化。有人提出，两种或多种语言的不断混合会产生一个更加相互关联的神经网络，负责在多种语言的使用者中观察到的认知和神经储备。

实验证据表明，使用多种语言会改变儿童言语的神经处理（Krizman 等人，2015）以及青少年的神经一致性和注意力控制（Krizman 等人，2014）。在成年人中，使用多种语言的经验会改变语言学习的神经特征（Bartolotti et al., 2017）、语言间和语言内竞争的皮层控制（Marian et al., 2017）以及执行控制区域的招募在语音竞争期间（Marian et al., 2014）。除了皮质变化之外，L2 处理的经验也会修改声音的皮质下编码（Krizman 等，2012 ））。语言体验甚至会影响生物现象，例如耳声发射，这是内耳内产生的声音（Marian et al., 2018）。

事实上，世界各地实验室进行的实证研究提供了大量证据，为 van Hell 博士对第二语言处理的神经认知基础的回顾提供了支持。我特别欣赏这篇文章强调了最近的发现，即语言网络内的功能连接程度是由第二语言学习的发展时间表决定的。同样值得注意的是，该评论并没有忽略增加神经认知研究多样性的必要性、对个体差异进行平均的局限性，以及承认许多其他尚未解答的难题有待解决。第二语言习得的科学已经从案例研究和描述性分析的历史发展到实证认知和行为实验，再到当前的神经科学研究，

我相信，目前，神经网络方法是双语神经科学中一个特别富有成效的方向，尽管研究人员可以使用的工具仍然有限。该评论提到了最近在语言网络、默认模式网络和了解一种以上语言的人的静息状态功能连接方面的研究。研究 L2 处理的神经网络方法在当今新兴的大型语言模型领域尤其重要，其中生成式预训练 Transformer 占据了中心舞台。这些模型挑战了语言学习领域中许多长期存在的假设，从先天语法到关键年龄，并提出了关于人类和机器如何学习语言之间的异同，甚至语言本身的定义的问题。

当前关于大型语言模型的讨论以及 van Hell 博士对 L2 处理的神经认知的回顾表明，L2 学习和处理领域正在进入一个时代，在这个时代中，语言的神经认知基础将获得越来越多的相关性。随着人工智能竞赛的升温，研究人员必须确保有关 L2 处理、双语和多语言体验以及语言多样性的问题成为这些对话中不可或缺的一部分（Marian，2023 ）。