# 英语高级 · 第 5 周 Day 5｜结构化摘要四段配比

> 系列：语种学习 / 英语 / 高级 · 每日学习｜AI生成｜2026-10-27｜codebuddy

**今天学什么**

今天把前四天的动作合成一份成品：一百五十词，背景三十、方法四十、结果五十、局限三十。字数配比不是形式主义，它逼你决定哪条信息值得占十个词。今天真正要解决的是"写不短"和"写过头"两个毛病：背景铺陈太多，结果只剩一句；或者结果写得很足，局限一句"样本有限"糊过去。学完你手上会有一份能直接交出去的结构化摘要，以及一套可复用的配比检查法。配比本身就是一次取舍训练，删哪句、留哪句，都在这三十词里见分晓。

## 一、今日任务

| 环节 | 时长 | 具体动作 | 产出 |
| --- | --- | --- | --- |
| 1. 选文与拆解 | 15 min | 选一篇 800 词左右的实证论文，用四色标出四要素，抄下所有数字 | 四要素表 1 张 |
| 2. 分段起草 | 40 min | 按 30/40/50/30 分四段写，每段先写超，再删到配额 | 四段初稿 |
| 3. 查重与改写 | 15 min | 逐句与原文比对，凡连续五词重合就重构句子 | 查重记录 |
| 4. 朗读校对 | 10 min | 出声读一遍，删掉读不顺的从句，核对数字与单位 | 二稿 |
| 5. 配比复核 | 10 min | 数每段词数，超出配比 ±5 词就调，写一句调整理由 | 终稿与词数表 |

配比是硬约束，不是建议。写的时候允许某段先超二十词，但删的时候只能从这一段删，不能从局限段借词给结果段——那等于用结论掩盖边界。

## 二、核心词汇与搭配

这十个词专治"结果写不具体、局限写不具体"。它们都带程度或边界含义，用对了，一句话能顶三句解释。

| 词/短语 | 音标 | 词性 | 释义 | 语域 |
| --- | --- | --- | --- | --- |
| overarching | /ˌəʊvərˈɑːtʃɪŋ/（≈欧弗拉奇英） | adj. | 总体的，统摄性的 | 学术 |
| twofold | /ˈtuːfəʊld/（≈图福尔德·长） | adj./adv. | 双重的；两倍地 | 学术 |
| magnitude | /ˈmæɡnɪtjuːd/（≈麦格尼特尤德） | n. | 幅度，量级 | 学术 |
| on the order of | /ɒn ði ˈɔːdə əv/ | phr. | 大约（数量级） | 学术 |
| marginal | /ˈmɑːdʒɪnl/（≈马知恩尔） | adj. | 边际的；微小的 | 学术/商务 |
| pronounced | /prəˈnaʊnst/（≈普若瑙恩斯特） | adj. | 显著的，明显的 | 学术 |
| hold for | /həʊld fɔː/ | phr. v. | 对……成立，适用于 | 学术 |
| underpowered | /ˌʌndəˈpaʊəd/（≈阿恩德保厄德） | adj. | 统计功效不足的 | 学术 |
| cross-sectional | /ˌkrɒs senˈseʃənl/ | adj. | 横断面的 | 学术 |
| generalizability | /ˌdʒenrəlaɪzəˈbɪləti/（≈真若莱泽比勒提） | n. | 可推广性，外推能力 | 学术 |

| 高频搭配 | 英文例句 | 中文 |
| --- | --- | --- |
| the overarching question | The overarching question is whether shorter hours change behaviour. | 总体问题是更短工时是否改变行为。 |
| a twofold increase | Reports of burnout showed a twofold increase over three years. | 三年内职业倦怠报告增长了一倍。 |
| the magnitude of the effect | The magnitude of the effect surprised the authors. | 效应幅度让作者意外。 |
| on the order of 40 minutes | Workers gained sleep on the order of 40 minutes a night. | 劳动者每晚多睡约四十分钟。 |
| a marginal gain | The second intervention produced a marginal gain. | 第二项干预只带来微小增益。 |
| a pronounced difference | The difference is most pronounced among parents. | 这一差异在有孩子的群体中最明显。 |
| the pattern holds for | The pattern holds for salaried staff but not for shift workers. | 该模式适用于受薪员工，不适用于轮班工人。 |
| underpowered to detect | The trial was underpowered to detect effects below 0.2 SD. | 该试验功效不足，无法检出 0.2 个标准差以下的效应。 |
| cross-sectional evidence | Cross-sectional evidence cannot establish direction. | 横断面证据无法确定方向。 |
| limit generalizability | Volunteer firms limit generalizability. | 自愿参与的企业限制了可推广性。 |

先练 magnitude 与 on the order of，它们能替掉"很大""差不多"这类模糊表达。发音注意 magnitude 的重音在第一音节，第二个音节是 /nɪ/ 不是 /naɪ/；pronounced 作形容词时词尾是 /st/，不要读成 pronounce 的 /s/（≈斯·清）；generalizability 有七个音节，重音在 /bɪlə/ 前。

## 三、语言点：局限句的具体化与 hedge 梯度

一百五十词摘要里，局限段只有三十词，却最见功力。写虚了，读者会怀疑你不敢面对设计缺陷；写重了，前面五十词的结果被自己推翻。核心是两件事：局限要落到具体机制上，断言要落在合适的强度上。

**规则**

1. 局限只从四类里选：设计（横断面、观察性、无对照）、样本（规模、代表性、流失）、测量（自报、单一来源、替代指标）、时间（随访长度、单一时点）。一段写两类即可，不要罗列。
2. 一句合格局限等于限制加后果加边界：Because the design is cross-sectional, we cannot tell whether X preceded Y。只说限制不说后果，读者不知道影响什么。
3. hedge 有梯度：may / might 最弱，could 次之，suggest / point to 中等，indicate 较强，show 很强，demonstrate 最强。选词只看证据，不看你想让结论多漂亮。
4. 样本问题必须给数字或功效。With 84 participants, the study was underpowered for effects below 0.3 SD 比 the sample was small 有用十倍。
5. 不写 future research is needed 这类空话。要写就写具体：a panel design following the same households would settle the direction。

| 局限类型 | 该问的问题 | 合格写法 |
| --- | --- | --- |
| 设计 | 能不能排除反向因果 | The cross-sectional design cannot establish whether X preceded Y. |
| 样本 | 覆盖了谁，漏了谁 | Volunteer firms over-represent white-collar work. |
| 测量 | 指标是否替代了真实变量 | Actigraphy cannot separate time in bed from sleep. |
| 时间 | 随访够不够长 | Six months is too short to capture seasonal variation. |

写局限前先问一句：如果审稿人只能提一个问题，他会提哪个？答案通常就是最该写的那一条。三十词放不下两条以上，所以要选"一旦不写就会让结论失效"的那条，其余放进方法段的限定语里，用 with、among、in 这类介词悄悄交代。比如样本只覆盖城市租户，可以写成 among urban renters 嵌进结果句，把宝贵的三十词留给设计层面的方向问题。

**强弱对比**

| 偏弱或偏误 | 更稳妥 | 差别在哪 |
| --- | --- | --- |
| The sample was small. | With 84 participants, the study was underpowered for effects below 0.3 SD. | 给了规模与后果，读者能判断结论能推到哪。 |
| This may be limited. | The estimates cannot be generalised beyond urban renters. | 前者没有对象，后者指明了边界。 |
| The results prove that shorter hours improve sleep. | The results indicate that shorter hours improve sleep among parents of young children. | prove 超出单组证据，限定人群反而让结论更可信。 |

indicate 与 show 之间没有硬边界，差别在读者预期：indicate 允许例外，show 几乎不允许。选词时先问"如果出现反例，我这个句子还成立吗"，成立就用 indicate。

**常见误用**

1. 通段 may、might、could：读者读完不知道你到底发现了什么。hedge 是分级工具，不是安全毯，一段最多两处。
2. 把方法复述当局限：We used a survey 不是局限；We relied on self-reported sleep, which tends to overstate duration 才是。
3. 用 prove、confirm、establish 描述单一致结果。这三个词要求证据链完整，单篇摘要几乎达不到。

**什么时候用，什么时候不能用**

结果段用 show 或 indicate，前提是效应量大、设计有对照、样本足够；只要缺一项，就退回 suggest。局限段必须用 can not 或 is limited to 这类明确的能力表述，不要用 may not 把话说软——局限本来就是确定的。反过来，解释机制时不要用 demonstrate，因为你没有直接测机制。用了会显得怎样？hedge 精准，读者觉得你清楚自己证据的边界；hedge 泛滥，读者觉得你在回避；hedge 缺失，审稿人第一句就会问"你凭什么"。

## 四、句型与范例

| 句式 | 功能 | 中文 |
| --- | --- | --- |
| Debate over X has centred on Y, while Z remains unexamined. | 背景：指出空白 | 关于 X 的争论集中在 Y，而 Z 尚未被考察。 |
| Whether the pattern holds outside X is unknown. | 背景：限定范围 | 该模式在 X 之外是否成立尚不清楚。 |
| We ran a six-month trial in 12 firms, randomising 2,400 employees. | 方法：设计与样本一句给全 | 我们在 12 家企业开展为期六个月的试验，随机分配 2,400 名员工。 |
| Sleep was measured with wrist actigraphy and daily diaries. | 方法：测量工具 | 睡眠通过腕部活动记录仪与每日日记测量。 |
| Employees slept 41 minutes longer, a gain of roughly 9%. | 结果：绝对量加相对量 | 员工多睡 41 分钟，增幅约 9%。 |
| The effect was concentrated among parents of young children. | 结果：异质性 | 效应集中在幼儿父母身上。 |
| The association was robust to controls for workload. | 结果：稳健性 | 控制工作量后该关联依然稳健。 |
| With 2,400 employees, the trial was not powered for subgroup analysis. | 局限：功效 | 该试验有 2,400 人，但不足以支撑亚组分析。 |
| Because firms volunteered, the sample over-represents white-collar work. | 局限：选择偏差 | 企业自愿参与，样本高估了白领工作。 |
| The estimates describe motivated employers, not the labour market. | 局限：边界收束 | 这些估计描述的是有积极性的雇主，而非整个劳动力市场。 |

这十句按四段排列，正好是一份一百五十词摘要的骨架。写的时候把第三、四句的方法细节按你的论文替换，第五到七句至少保留两个数字，最后三句只选两句——三十词放不下三句局限，硬塞会让每句都残缺。

## 五、练习

A 部分动笔，B 部分改完要数字数，C 部分是今天的正式产出。写之前先把原文的数字抄成一张清单，写完逐项核对。整个过程计时：前十五分钟只做拆解，不写句子；四十分钟起草时允许超词，最后二十分钟只做删除和查重。删除的顺序是先砍形容词、再砍过程性细节、最后才动从句，因为从句往往承载因果或条件。

**A 词汇与搭配**

1. 按断言强度从弱到强排序：indicate / may / suggest / demonstrate / show。
2. 填空：the ______ question；a ______ increase；the ______ of the effect；______ the order of 40 minutes；a ______ gain；the pattern ______ for renters。
3. 语域判断：哪一句能进摘要？(a) The effect was kind of big. (b) The effect was pronounced among parents.
4. 改错：The results prove that the policy works.
5. 中译英：该试验功效不足，无法检出 0.2 个标准差以下的效应。
6. 用 generalizability 写一句，说明自愿样本的问题。
7. 把 "The sample was small" 改成含数字与后果的句子（自定 n）。

**B 句法与压缩**

1. 把 45 词背景压到 30 词以内：There has been a long debate among economists and policymakers about whether reducing working hours can raise productivity, and a large number of studies have examined output per worker, but very little attention has been paid to what happens to sleep.
2. 把两句结果合并，保留两个数字：Employees slept 41 minutes longer. That is a gain of about 9%.
3. 给下面这句加后果：The design is cross-sectional.
4. 把 60 词结果压到 50 词，数字一个不丢：Employees on the shorter week slept 41 minutes longer on workdays than those on usual hours, which is equivalent to a 9% gain, and they also reported 23% fewer nights of insufficient sleep; the effect was concentrated among employees with children under five, who gained 58 minutes, and it was absent in firms where workload was not reduced.
5. 改掉过度 hedge：The results may possibly suggest that there might be an association.
6. 检查照抄：下面哪一句与原文连续五词重合？原文为 "We ran a six-month stepped-wedge trial in 12 firms, randomising 2,400 employees to a 32-hour week." (a) We conducted a six-month stepped-wedge trial in 12 firms. (b) A six-month trial in 12 firms randomised 2,400 employees.

**C 输出**

1. 写 150 词结构化摘要，允许 135—165 词，四段配比 30/40/50/30，每段 ±5 词。四段之间不空行，用一句过渡或直接并置。
2. 交出词数表与查重记录：每段实际词数、与原文重合的最长片段（不得超过四词）、三处改写方法。

## 六、参考答案

**A**

1. may < suggest < indicate < show < demonstrate。排序依据是证据要求的完整度，不是语气强弱的感觉。
2. the overarching question；a twofold increase；the magnitude of the effect；on the order of 40 minutes；a marginal gain；the pattern holds for renters。
3. (b)。kind of big 是口语，且没有说明在谁身上明显。
4. 改为 The results indicate that the policy works. 单组前后比较不足以用 prove。
5. The trial was underpowered to detect effects below 0.2 SD.
6. 示例：Volunteer firms limit the generalizability of the findings to the wider labour market.
7. 示例：With 84 participants, the study was underpowered to detect effects smaller than 0.3 SD.

**B**

1. 示例：Debate over shorter hours has focused on output, while effects on sleep remain largely unexamined.（17 词）压缩靠删"经济学家与政策制定者""大量研究"这类不承载信息的成分。
2. Employees slept 41 minutes longer, a gain of roughly 9%. 同位语把两句并成一句。
3. Because the design is cross-sectional, we cannot tell whether shorter hours preceded better sleep.
4. 示例：Employees on the shorter week slept 41 minutes longer, a 9% gain, and reported 23% fewer nights of insufficient sleep. The effect was concentrated among parents of children under five, who gained 58 minutes, and was absent where workload was not reduced.（45 词）删除重复主语与比较从句即可，数字全留。
5. 改为 The results suggest an association. 三层 hedge 叠加等于没说。
6. (a)。conducted a six-month stepped-wedge trial in 12 firms 与原句连续六词重合，必须改写，如 (b) 的语序调整。

**C**

参考摘要（148 词）：Debate over the four-day week has focused on output, yet whether shorter hours improve sleep, and for whom, remains largely unexamined. Policymakers cite productivity gains, while sleep is treated as a private matter. We ran a six-month stepped-wedge trial in 12 firms, randomising 2,400 employees to a 32-hour week at full pay or to usual hours. Sleep was measured with wrist actigraphy and daily diaries; adherence was tracked through payroll records. Employees on the shorter week slept 41 minutes longer on workdays, a gain of roughly 9%, and reported 23% fewer nights of insufficient sleep. The effect was concentrated among parents of children under five, who gained 58 minutes, and was absent in firms where workload was not reduced. Firms volunteered, so the sample over-represents white-collar work; actigraphy also cannot separate time in bed from sleep. The estimates describe motivated employers, not the labour market as a whole.

## 七、今日输出与自查

- 产出 Day05_结构化摘要.md，135—165 词，四段配比 30/40/50/30，各段 ±5 词。
- 产出 Day05_词数与查重.md，逐段词数、最长重合片段、三处改写方法。
- 自查一：背景段有没有出现"since ancient times""with the development of"这类空转开头。
- 自查二：结果段是否至少有两个具体数字，且都带单位或比例；局限段是否写到"因此不能判断什么"。
- 自查三：全文有没有连续五词与原文相同；有就重写那一句，不是换同义词。

## 八、明日预告

Day 6 进入复盘：把你的摘要与原文逐段对照，做一张增删表，标出漏掉的信息与多加的信息，再把三十张词卡过一遍。写完之后隔一天再看，你会发现自己最常犯的不是语法错，而是把作者的推测写成了结果。请把今天的终稿冷置一晚，明天带着陌生感去读。