微精选

人类理解语言本质的一个早期里程碑，是认识到语言形式无需与其所指对象相似。巧合的是，古希腊的柏拉图与中国的荀子遥相呼应，相距大半个地球的两位哲人几乎在同一时期提出了这一观点。

另一里程碑是陈第（1606）基于典籍考据重构古汉语音系，揭示了语言的动态属性：口语和书面语都在时空维度上不断演变。1786年，威廉·琼斯（William Jones）在加尔各答发表的演说不仅深化了这一认知，更为印欧语系的系统性研究开启了百年探索之门。

二十世纪人类学及文化相对主义思潮的兴起，进一步推动语言研究向全球范围拓展。弗朗茨·博厄斯（Franz Boas，1858-1942 ）开创性地开展美洲原住民语言田野调查。他在哥伦比亚大学的讲座深深影响了当时尚在求学的约瑟夫·格林伯格（Joseph Greenberg, 1915-2001）。格林伯格随后在非洲大陆进行了深入调查，探索世界各语系间的普遍性与近普遍性规律。

我在哥伦比亚大学读本科时，有幸选修格林伯格的课程，也首次接触到他的多边比较法。数十年后，格林伯格任职于斯坦福，而我在伯克利任教。两所学校相距仅一小时车程，因此我们经常互相拜访，我也得以继续向他请教。

图：格林伯格与作者在加州大学伯克利分校

与他的早期思想碰撞，使我心中萌发这样的思考——语言是生物与文化双重演化的独特产物。1977年夏，在夏威夷举办的美国暑期语言学院，威廉·拉波夫（William Labov）和我共同开设的课程正是这一理念的实践：拉波夫教授聚焦语言结构与语言演变的社会维度，我则侧重阐释其生物学基础。

图：拉波夫与作者在香港中文大学

继由陈第到琼斯，由博厄斯到格林伯格所奠定的里程碑后，语言科学迎来又一重大突破。1861年，保罗·布洛卡（Paul Broca）在巴黎向解剖学和人类学界展示首例语言障碍的研究。布洛卡与十九世纪其他先驱开辟了崭新的研究路径，他们开始探索语言（包含口语和书面语）在大脑中的加工与感知机制。与许多其他领域类似，研究功能障碍为理解语言的常规机制和寻求治疗方案提供了一扇独特的窗口。

然而，这一研究路径的真正实现得益于二十世纪神经科学的进展，从神经元、突触、神经递质的微观发现，到眼动追踪与脑成像技术（ 如脑电图EEG、脑磁图MEG、磁共振MRI与功能性磁共振fMRI）的成熟。 这些设备面世之后，测量大脑加工过程反应和反应时的传统方法得到了更新，测量的精确性大幅提升。头皮电极与颅内电极可精准捕捉毫秒级的脑电信号，磁共振成像则可通过更高分辨率精准定位静态或动态下的大脑活动。

2007年，著名失语症学家尼娜·德龙克尔斯（Nina Dronkers）及其团队发表的研究，集中体现了自布洛卡时代以来的新进展。布洛卡只能在尸检时通过肉眼观察患者的大脑皮层，而德龙克尔斯借助三维成像技术重新分析布洛卡具有前瞻性地保存的两位患者的脑标本。其磁共振图像清晰显示：语言障碍不仅源于后来被命名为“布洛卡区”的局部损伤， 更与全脑多个区域的病变相关。

图：改编自Dronkers, N. et al. 2007. Brain 130.1437.

当代研究证实，大脑的神经网络始终以复杂的方式整体协同运作。语言作为核心认知功能更是如此，依赖全脑神经网络的高度协作。最新的发现 不仅深化了19世纪先驱奠定的失语症研究，更关乎全球日益增多的神经退行性疾病患者（如阿尔茨海默症）的认知衰退问题。

当前，我们正置身于语言科学史上的最新里程碑——21世纪人工智能的革命性进展。机器不依赖人类真实经验，仅凭海量数据（文本、语音频、视觉影像）就能成功学习多门语言。这一技术突破为我们理解语言本质和语言障碍提供了新的视角。同时，也以前所未有的方式向人类发起挑战：面对日益智能化的机器，我们应当如何加以规范并与其协作， 以维护我们脆弱星球的共同福祉。

（译者：唐仪  校对：王旭）

Landmarks in Language Research

Wang Shiyuan, 2025.

An early landmark toward understanding the nature of language, was the recognition that a linguistic form need not resemble the object the form represents.  By coincidence, this insight was expressed at nearly the same time by philosophers half a world apart: Plato in Greece and Xunzi in China.

Another landmark was CHEN Di’s work (1606) in reconstructing Old Chinese based on textual evidence, which highlighted the dynamic nature of language: that both spoken and written language are constantly changing in space and time.  This insight was broadened in the famous lecture (1786) by William Jones in Kolkata, which launched a century of systematic research on the Indo-European family of languages.

The generalization of linguistic studies to the world-at-large took place in the 20 ^th century, under the influence of anthropology and its attention to cultural relativism.  It was Franz Boas (1858-1942) who first pioneered fieldwork on the languages of the Americas.  His lectures at Columbia University inspired Joseph Greenberg (1915-2001), then an undergraduate there, who did extensive fieldwork in Africa, and went on to explore universals and near-universals across all the language families of the world.

When I was an undergraduate at Columbia, I had the good fortune of enrolling in one of Greenberg’s classes, and was first exposed to his method of multilateral comparison.  Decades later, Greenberg had moved to Stanford and I was teaching at Berkeley.  Since the two universities were only an hour’s drive apart, we visited each other often and I was able to continue to learn from him.

These early interactions with him first planted in my mind the seeds for thinking about language as a unique product of biological and cultural evolution in humans.  It was not until summer 1977, at the Linguistic Institute held in Hawaii, that William Labov and I offered a course on this basis, in which he discussed the social aspects of language structure and language change, and I discussed the biological aspects.

Following the landmarks set by CHEN-Jones, and by Boas-Greenberg, another important landmark in research on the language sciences was the first report (1861) on language disorders Paul Broca made to anatomists and anthropologists in Paris.  The approach that Broca and other 19 ^th century pioneers took began by asking where language, both spoken and written, is produced and perceived in our brain.  As in many other fields, studying disorders is an excellent way toward understanding how language normally works, as well as toward finding therapy for the disorders.

However, for this scientific approach to become feasible, we had to wait till the 20 ^th century for neuroscience to discover neurons, synapses, and neurotransmitters, etc. as well as for the technology to develop, of eye-tracking and brain imaging, from EEG, MEG, MRI, to fMRI, etc.  Methods for measuring brain processes by responses and reaction times became updated with instruments that yield much greater precision.  Electrodes on the scalp or inserted into the brain now can time oscillations in the brain in milliseconds.  Magnetic resonance images now can place brain activities, either static or dynamic, by ever smaller pixels.

A vivid demonstration of the progress that has been made since Broca is the 2007 report by the well-known aphasiologist, Nina Dronkers, and her colleagues.  Whereas Broca was able to examine only the cortex of his patients at autopsy by eye, Dronkers could subject two of Broca’s patients by detailed 3-dimensional imaging of their brains, which Broca himself with amazing foresight had preserved.  The MRI photos she published show conclusively that their disorder was not caused by lesion localized merely at the site that has become known as Broca’s Area, but caused by many other sites as well distributed across the entire brain.

Current knowledge tells us that the numerous networks of the brain almost always work globally together in complex synchronized ways. This is especially true for language; as the major form of cognition, it is integral to all internal and external behavior.  Importantly, the new knowledge applies not only to the various aphasias that the 19 ^th century pioneers laid the foundation for studying, but even more to the ever increasing population of old people in the world who have succumbed to neurodegeneration and cognitive decline, exemplified by Alzheimer’s Disease.

We are now in the midst of the last landmark for studying language sciences, which began with the remarkable advances of Artificial Intelligence in the 21 ^st century.  Machines have been unexpectedly successful at learning multiple languages from mere artificial data (texts, audio and visual images), instead of from real world experiences that humans do. This historic breakthrough in technology opens up completely new vistas for viewing the nature of language and language disorders.  Even more importantly, it challenges mankind as never before: how we might regulate and synchronize with machines that are ever growing in intelligence for the collective benefit of our already fragile planet.