For seventy years the mind has been imagined as a grammar machine that builds sentences like branching trees; a new study argues that much of language rests instead on flat, familiar chunks — and that could narrow the gulf between human speech and animal calls
七十多年来，语言学家习惯把人的心智想象成一台“语法机器”：先由零散的词组成短语，再层层组合，像树枝一样分叉，最终长成一整句完整的话。这套“层级结构”理论相当优雅，也能解释我们为何能说出从未说过、从未听过的新句子。但有一个细节始终让这幅图景不太安生：英语里最常用的三四个词的组合，往往根本算不上语法单位。像“in the middle of the”或“can I have a”这样的串，被反复使用，却偏偏横切过语法树本该画出的边界；因为无处归类，它们长期被忽视。2026 年 1 月，期刊《自然·人类行为》（Nature Human Behaviour）刊出的一篇论文提出：这些串并非杂音，而是语言在大脑中存储方式的一项基本成分。作者 Yngwie Nielsen 与 Morten Christiansen（奥胡斯大学 / 康奈尔大学）认为，心智表征所依赖的单位，比标准模型所允许的更扁平、更短。为验证这一想法，团队做了四项预注册实验（近 500 名参与者），并分析了两份真实语料：68 人的眼动阅读记录，以及 358 段电话对话录音。结果显示，即使某个串在语法上不是“成分”，只要刚刚见过，人对它的反应就会更快——这种“启动效应”说明它确实进入了心智表征。研究由此挑战了乔姆斯基一脉的生成语法与较新的构式语法；作者更大胆地指出：如果人的语言并不必依赖层级句法的复杂机器，那么人类语言与动物交流之间的鸿沟，也许比想象的要窄得多。本文循“背景—疑难—研究—证据—挑战—启示—边界—收束”的链条，追问一个朴素的问题：我们开口说话时，大脑究竟在搭一棵树，还是在拼一堆积木？
【1】For most of the past seventy years, linguists have pictured the human mind as a machine that builds sentences from nested, tree-like structures.On this view, words are grouped first into small units, such as noun phrases, then combined step by step into larger ones until a full sentence emerges.The theory is elegant, and it explains why we can produce sentences we have never spoken or heard before.
【2】Yet one awkward detail has always resisted this picture: the most common three- or four-word sequences in everyday English are often not units of grammar at all.Expressions such as “in the middle of the” or “can I have a” are used constantly, and they cut across the very boundaries the tree is meant to draw; because they fit no neat category, such strings have largely been ignored.
【3】A study published in January in Nature Human Behaviour suggests that these sequences are not mere noise but a basic ingredient of how language is stored in the mind.Its authors, Yngwie Nielsen and Morten Christiansen, of Aarhus University and Cornell University, argue that mental representationrests on flatter, shorter units than the standard model allows.
【4】To test the idea, the team ran four pre-registered experiments in which nearly five hundred participants judged, as fast as they could, whether a string of words formed a natural phrase.The trick lay in priming, a well-studied phenomenon: having just met a sequence, people respond to it faster the next time, even when the sequence was not a grammatical constituent at all.
【5】Two further analyses carried the finding into ordinary life.One tracked the eye movements of sixty-eight people as they read; the other combed through three hundred and fifty-eight recorded telephone conversations.Both showed the same pattern, which suggests the effect is not an artefact of artificial tasks.
【6】The results challenge the generative tradition associated with Noam Chomsky, and newer constructionist accounts too, since neither explains why strings that break the rules of grammar are handled like familiar patterns.If the mind really relies on flat, linear chunks, the elaborate machinery of hierarchical syntax may be doing less work than linguists have long assumed.
【7】The implications reach well beyond academic debate: language teachers, for instance, might do better to drill whole useful chunks, such as “it turns out that”, rather than dissecting every sentence into its parts.The work also bears on how children acquire their first language, and on how machines that process speech are designed.
【8】Most provocatively, the authors suggest that if human language does not need the complex machinery of hierarchical syntax, the gulf between our language and animal communication may be far narrower than once believed.The work is confined to English, however, so the claim is a starting point rather than a verdict.
【9】What the study ultimately offers is a humbler picture of the mind, in which fluent speech grows less from an inner rulebook than from the patient accumulation of well-worn phrases.We may speak in trees on paper, but in the mind we seem to build with bricks.
原文出处：本文根据 Nature Human Behaviour 论文《Evidence for the representation of non-hierarchical structures in language》（Yngwie A. Nielsen & Morten H. Christiansen；Aarhus University / Cornell University；Nature Human Behaviour, 10(3): 579–588, 2026；DOI 10.1038/s41562-025-02387-z）改写整理，供学习使用。
三、长难句分析（三步拆解法）
Yet one awkward detail has always resisted this picture: the most common three- or four-word sequences in everyday English are often not units of grammar at all.
主干：one awkward detail has always resisted this picture（主语 + 谓语 + 宾语）
冒号后的解释：the most common three- or four-word sequences ... are often not units of grammar at all（说明“这个别扭的细节”究竟是什么）
程度状语：at all（与否定连用，意为“根本不、一点也不”，加重语气）
考点：Yet 的转折；resist 的引申义“使难以成立、与……相抵触”；冒号引出同位解释；not ... at all 的强调否定
Expressions such as “in the middle of the” or “can I have a” are used constantly, and they cut across the very boundaries the tree is meant to draw; because they fit no neat category, such strings have largely been ignored.
并列分句一：Expressions ... are used constantly（被动语态；constantly 作方式状语）
并列分句二：they cut across the very boundaries the tree is meant to draw（cut across 意为“横切、穿过”；the tree is meant to draw 省略关系词的定语从句，修饰 boundaries）
very 的强调：the very boundaries（“恰恰是那些边界”，very 作形容词加强限定）
分号后的原因状语从句：because they fit no neat category, such strings have largely been ignored
考点：分号连接的并列分句；省略关系代词的定语从句；cut across 的搭配；because 引导原因状语从句
译文：像“in the middle of the”或“can I have a”这样的表达被反复使用，它们恰恰横切过语法树本应划出的那些边界；正因为不属于任何整齐的类别，这类词串在很大程度上一直被忽视。
The trick lay in priming, a well-studied phenomenon: having just met a sequence, people respond to it faster the next time, even when the sequence was not a grammatical constituent at all.
主干：The trick lay in priming（lay in 意为“在于”；the trick lay in ... 是“诀窍在于……”的常用句式）
同位语：a well-studied phenomenon（对 priming 作补充说明）
现在分词作状语：having just met a sequence（先完成、后发生的逻辑，说明前提）
让步状语从句：even when the sequence was not a grammatical constituent at all
考点：the trick lay in 的固定表达；同位语；完成式分词作状语；even when 引导让步状语从句
The results challenge the generative tradition associated with Noam Chomsky, and newer constructionist accounts too, since neither explains why strings that break the rules of grammar are handled like familiar patterns.
主干：The results challenge the generative tradition ... and newer constructionist accounts too（challenge 后接两个并列宾语）
过去分词作定语：associated with Noam Chomsky（修饰 the generative tradition）
Most provocatively, the authors suggest that if human language does not need the complex machinery of hierarchical syntax, the gulf between our language and animal communication may be far narrower than once believed.
仿写题：仿照第 4 段 "The trick lay in X, a well-studied phenomenon: having just done A, people do B faster the next time, even when C."（the trick lay in + 同位语 + 完成式分词状语 + even when 让步）的句式，围绕「某项看似神奇的效果，其实源于一个已被研究过的心理机制」写一个句子。
