# 英语高级 · 第 5 周 Day 3｜四种句法压缩手段

> 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**

今天解决"压缩就是删句子"这个误解。真正的压缩是换句法：把两个短句并进一个分词短语，把定语从句降成同位语，把并列分句收进 with 结构，把三项罗列交给一个冒号。改完之后字数少四成，信息一个不少。学完你能把两百词的方法与结果段稳定压到一百二十词，并且说得出每一处删减换来了什么——这套动作是本周一百五十词摘要的技术底座。四种手段不是炫技，而是让关系直接进入句法，把连接词省掉。

## 一、今日任务

| 环节 | 时长 | 具体动作 | 产出 |
| --- | --- | --- | --- |
| 1. 规则对照 | 10 min | 读第三部分五条规则，把四种手段各抄一个例句并标出被删掉的成分 | 4 句对照 |
| 2. 单句改写 | 25 min | 每种手段改 3 句，共 12 句，每句写出原字数与改后字数 | 12 句改写 |
| 3. 整段压缩 | 25 min | 把练习 C 的 200 词原文压到 110—130 词，先列数据清单再动笔 | 压缩稿 1 份 |
| 4. 信息核对 | 12 min | 拿数据清单逐项打勾，缺一项就回补，不许用"大约"糊过去 | 核对表 1 张 |
| 5. 决策记录 | 8 min | 写下三处最主要的删减及理由，每处一句 | 压缩说明 |

压缩时先做减法再做加法：第一遍只删重复主语和解释性从句，第二遍才用句法手段合并。反过来做，你会为了套句式而丢掉数字。

## 二、核心词汇与搭配

这十个词全部来自方法与结果部分，是压缩时最容易被保留、也最容易被写错的成分。它们多半是名词化形式，正好和今天的句法手段互为表里。

| 词/短语 | 音标 | 词性 | 释义 | 语域 |
| --- | --- | --- | --- | --- |
| deploy | /dɪˈplɔɪ/（≈迪普洛伊·短） | v. | 运用（方法、资源） | 学术/商务 |
| draw on | /drɔː ɒn/ | phr. v. | 依托，借用（数据、文献） | 学术 |
| comprise | /kəmˈpraɪz/（≈克姆普赖兹） | v. | 由……组成，包含 | 学术 |
| yield | /jiːld/（≈伊尔德·长） | v. | 得出（结果、估计） | 学术 |
| aggregate | /ˈæɡrɪɡət/（≈艾格里厄特） | adj. | 汇总的，合计的 | 学术 |
| subset | /ˈsʌbset/（≈萨布塞特·短） | n. | 子集，部分样本 | 学术 |
| adjusted for | /əˈdʒʌstɪd fɔː/ | phr. | 校正了（混杂因素） | 学术 |
| attrition | /əˈtrɪʃn/（≈厄特里希恩） | n. | 样本流失 | 学术 |
| impute | /ɪmˈpjuːt/（≈伊姆普尤特） | v. | 插补（缺失值） | 学术 |
| sensitivity analysis | /ˌsensəˈtɪvəti əˈnæləsɪs/ | n. | 敏感性分析 | 学术 |

| 高频搭配 | 英文例句 | 中文 |
| --- | --- | --- |
| deploy a mixed-methods design | The team deployed a mixed-methods design across four sites. | 团队在四个点采用了混合方法设计。 |
| draw on archival records | The analysis draws on archival records from 1998 onward. | 该分析依托 1998 年以来的档案记录。 |
| the sample comprised 412 adults | The sample comprised 412 adults from three clinics. | 样本由三家诊所的 412 名成人组成。 |
| the model yields an estimate | The model yields an estimate of 0.3 standard deviations. | 该模型给出的估计为 0.3 个标准差。 |
| aggregate estimates mask variation | Aggregate estimates mask wide variation across districts. | 汇总估计掩盖了各区之间的巨大差异。 |
| a subset of respondents | A subset of respondents, 96 in all, were re-interviewed. | 一部分受访者共 96 人接受了二次访谈。 |
| adjusted for age and income | Estimates were adjusted for age, income and housing type. | 估计值校正了年龄、收入与住房类型。 |
| differential attrition | Differential attrition between renters and owners biases the gap. | 租户与业主之间的差别流失使差距产生偏差。 |
| impute missing values | Missing income values were imputed from district medians. | 缺失的收入值按各区中位数插补。 |
| a sensitivity analysis excluding X | A sensitivity analysis excluding imputed cases left the result unchanged. | 剔除插补样本的敏感性分析显示结果未变。 |

四个词最值得先练：comprise、yield、adjusted for、attrition。它们能替掉一整句解释。注意 comprise 直接接宾语，不写 comprised of；yield 的主语通常是模型、分析、试验，不写人；adjusted for 是过去分词短语，放句末或句中都要跟逗号；attrition 的重音在第二音节 /əˈtrɪʃn/。

## 三、语言点：四种句法压缩手段

压缩的本质是把"两句之间的逻辑关系"编码进句法，而不是靠连接词说出来。关系一旦进了句法，字数自然下来，语域也随之上移。四种手段各管一类关系，用错就会产生歧义或悬垂。

**规则**

1. 分词短语：删掉关系代词和 be，主动关系用 -ing，被动或完成用 -ed，放在句首或句末。分词的逻辑主语必须与主句主语一致。
2. 同位语：把定语从句或整句解释降成名词短语，紧贴被解释的名词。同位语只保留身份、数量、地点这类可核查信息。
3. with 结构：with 加名词加分词或介词短语，替代一个并列分句，表示伴随状态或属性。它适合放"限定条件"，不适合放主要结果。
4. 冒号列举：冒号前必须是完整句，冒号后是三个以上平行成分。用它可以替掉 there are three factors 这类空句。
5. 压缩红线：数字、样本量、效应方向、显著性、局限条件一律不压。能压的是重复主语、解释性从句、评价性形容词和过程性叙述。

| 手段 | 压缩对象 | 改前 | 改后 |
| --- | --- | --- | --- |
| 分词短语 | 两个同主语短句 | The survey was conducted in 2022. It covered 1,240 households. | Conducted in 2022, the survey covered 1,240 households. |
| 同位语 | 定语从句 | Li and colleagues, who work at a public health institute in Wuhan, tracked 400 families. | Li and colleagues, a public health team in Wuhan, tracked 400 families. |
| with 结构 | 并列分句表条件 | Response rates were below 40% in two sites. This limits the site-level estimates. | With response rates below 40% in two sites, the site-level estimates are limited. |
| 冒号列举 | 空句加罗列 | There were three factors that mattered. They were price, distance and timing. | Three factors mattered: price, distance and timing. |

**强弱对比**

| 偏弱或偏误 | 更稳妥 | 差别在哪 |
| --- | --- | --- |
| There were 412 participants, and 38 of them dropped out. | Of 412 participants, 38 dropped out. | 后者把样本量与流失放在同一句，读者一眼能算比例。 |
| The sample was small. This means the estimates are unstable. | The small sample leaves the estimates unstable. | 名词化加使役动词，省掉一句解释，语域更学术。 |
| The study was done in three provinces and it found that prices rose. | Conducted in three provinces, the study found that prices rose. | 分词短语把两次陈述并成一次，主语只出现一次。 |

**常见误用**

1. 悬垂分词：Having collected the data, the results were analysed. 分词的主语是研究者，主句主语却是 results，逻辑上成了"结果收集了数据"。改法有两种：把主句改成 We analysed the results，或把分词改成 After data collection。
2. 同位语堆叠：Wang, a professor, the lead author, at Fudan University, argues that… 四个成分挤在一起，读者找不到主语。同位语最多两个，且要按"身份加机构"的顺序。
3. with 结构歧义：With response rates falling, the team extended the survey. 是响应率在下降，还是团队在下降？当 with 后面的名词可能被读成主语时，改用 Because response rates were falling。

**什么时候用，什么时候不能用**

四种手段都用于"关系明确、主语清楚"的句子。关系不明时，宁可多写一个从句，也不要用分词硬并：读者读不懂，压缩就失败了。冒号列举只适合三项以上的平行成分，两项用 and 就够，用冒号会显得刻意。with 结构适合限定条件，不适合承载主要发现——把结果塞进 with 短语，读者会把它当背景。用了会显得怎样？手段精准，文字紧、密度高、语域稳；手段滥用，句子短了，但每句都要读两遍。

## 四、句型与范例

| 句式 | 功能 | 中文 |
| --- | --- | --- |
| Conducted across 24 cities, the survey covers 1,240 households. | 分词前置交代范围 | 该调查覆盖 24 座城市的 1,240 户家庭。 |
| Using a difference-in-differences design, we estimate the policy effect. | 分词前置交代方法 | 我们采用双重差分设计估计政策效应。 |
| The sample, 412 adults from three clinics, was re-interviewed. | 同位语补样本信息 | 样本为三家诊所的 412 名成人，接受了二次访谈。 |
| With attrition at 27% among renters, the gap is likely overstated. | with 结构限定条件 | 由于租户流失率达 27%，这一差距可能被高估。 |
| Three mechanisms are plausible: price, distance and timing. | 冒号列举 | 三种机制说得通：价格、距离与时机。 |
| Of 1,240 households approached, 986 completed the survey. | 介词前置给比例 | 接触的 1,240 户中，986 户完成了问卷。 |
| After adjusting for age and income, the gap narrows to 4 points. | 过去分词短语表校正 | 校正年龄与收入后，差距缩小到 4 个百分点。 |
| The model yields an effect of 0.3 SD, robust to specification. | 结果加稳健性 | 模型给出的效应为 0.3 个标准差，且对设定稳健。 |
| Missing values were imputed; a sensitivity analysis leaves the estimate unchanged. | 分号并置处理与检验 | 缺失值已插补；敏感性分析显示估计未变。 |
| Aggregating across sites masks variation: the effect is positive in two and null in the third. | 冒号引出对比细节 | 汇总各点会掩盖差异：两点为正，第三点为零。 |

这十句合起来就是一段标准的方法与结果压缩段。注意它们的分工：前四句交代设计与样本，中间三句给数字，后三句处理缺失与稳健性。写一百五十词摘要时，方法部分用前三句的骨架，结果部分用中间三句，局限部分把 with 结构和冒号句调换主语即可，不必重新造句。

## 五、练习

先列数据清单再动笔，是今天唯一的硬要求。清单没列完就压缩，一定会丢数字。

**A 词汇与搭配**

1. 填空：deploy a ______ design；draw ______ archival records；the sample ______ 412 adults；the model ______ an estimate；adjusted ______ income；a subset ______ respondents。
2. 改错：The sample was comprised of 412 adults.
3. 语域判断：哪一句更适合放进方法段？(a) We left out the missing values. (b) Missing values were excluded from the analysis.
4. 写出 aggregate 作形容词和动词时的重音差别，并各造一句。
5. 中译英：校正年龄与收入后，两区差距缩小到 3.1 摄氏度。
6. 用 attrition 写一句说明流失如何影响结论。
7. 用 yield 和 subset 各写一句，主语必须是研究、模型或样本。

**B 句法与压缩**

1. 用分词短语合并：The survey was conducted in 2022. It covered 1,240 households.
2. 用同位语压缩：Li and colleagues, who are based at a public health institute in Wuhan, tracked 400 families.
3. 用 with 结构压缩：Response rates were below 40% in two sites. This limits the site-level estimates.
4. 用冒号列举压缩：There were three factors that mattered. They were price, distance and timing.
5. 把下面 62 词压到 40 词以内，保留全部数字：The team interviewed 320 residents in two districts. The interviews took place in June 2023. Each interview lasted about 40 minutes. Of the 320 residents, 54 had moved within the previous year. The team found that movers reported lower satisfaction, with a mean score of 3.1 against 4.2 for stayers.
6. 找出并改正悬垂分词：Having adjusted for income, the gap remained at 4 points.

**C 输出**

1. 把下面 200 词压到 110—130 词，保留全部数字与局限条件。先列数据清单（样本量、响应率、温度、比例、差距、缺失比例、敏感性结果），压完逐项打勾。
2. 写 120—150 词压缩说明：三处主要删减各删了什么、换成了哪种句法、为什么这样换不丢信息。

原文：We surveyed 1,240 households in four districts of a mid-sized city between March and August 2023. Households were selected through stratified random sampling, with strata defined by district and housing type. Of the 1,240 households approached, 986 completed the full questionnaire, a response rate of 79.5%. Attrition was higher among renters (27%) than among owners (12%). Interviewers recorded indoor temperature at three points during the day and asked respondents about cooling behaviour, income and household size. We used linear mixed models with district-level random intercepts, adjusting for outdoor temperature, housing type and household income. Indoor temperatures averaged 29.4°C in the afternoon, and 41% of respondents reported difficulty sleeping at least three nights a week. The gap between the coolest and warmest districts was 3.1°C after adjustment. Missing income data, 8% of cases, were imputed using district medians; a sensitivity analysis excluding imputed cases changed the estimate by less than 0.2°C. The design cannot separate the effect of housing quality from that of neighbourhood green space, because the two are strongly correlated in all four districts.

## 六、参考答案

**A**

1. deploy a mixed-methods design；draw on archival records；the sample comprised 412 adults；the model yields an estimate；adjusted for income；a subset of respondents。
2. 改为 The sample comprised 412 adults. comprise 直接接宾语；若要用 of，应改成 was composed of。两种写法不要混用。
3. (b)。(a) 的 left out 是口语短语动词，(b) 用被动加名词化，语域与方法段一致。
4. 形容词重音在第一音节 /ˈæɡrɪɡət/，动词在第三音节 /ˈæɡrɪɡeɪt/。例：Aggregate estimates hide local variation. / We aggregate the data by district.
5. After adjusting for age and income, the gap between the two districts narrows to 3.1°C.
6. 示例：Attrition was 27% among renters, which probably overstates the gap between tenures. 流失句要写出偏差方向，不只是报数字。
7. 示例：The model yields an effect of 0.3 SD. / A subset of 96 respondents was re-interviewed.

**B**

1. Conducted in 2022, the survey covered 1,240 households. 分词的主语是 survey，与主句一致。
2. Li and colleagues, a public health team in Wuhan, tracked 400 families. 同位语只留身份与地点，删掉 who are based at。
3. With response rates below 40% in two sites, the site-level estimates are limited. with 短语放句首，主句保留结论。
4. Three factors mattered: price, distance and timing. 冒号前是完整句，冒号后三项平行。
5. 示例：In June 2023, we interviewed 320 residents across two districts; of these, 54 had moved within the previous year, and movers scored 3.1 on satisfaction against 4.2 for stayers. 数字全保留，过程性信息（每次 40 分钟）可删。
6. 改为 Having adjusted for income, we found the gap remained at 4 points. 或者 After adjustment for income, the gap remained at 4 points.

**C**

1. 参考压缩稿（114 词）：We surveyed 986 of 1,240 households across four districts of one mid-sized city (79.5% response), recording indoor temperature three times daily and asking about cooling behaviour and income. Using mixed models adjusted for outdoor temperature, housing type and income, we found afternoon indoor temperatures averaging 29.4°C, a 3.1°C gap between the coolest and warmest districts, and 41% of respondents reporting poor sleep on three or more nights a week. Attrition was higher among renters (27%) than owners (12%). Imputing or excluding the 8% of missing income cases moved estimates by under 0.2°C. Housing quality and neighbourhood green space are too closely correlated here to be separated, so the estimates describe conditions rather than causes.
   数据清单：1,240 接触户、986 完成户、79.5%、27% 与 12%、29.4°C、41%、3.1°C、8%、0.2°C，共九项，缺一即不合格。
2. 评分维度：三处删减是否都是过程性信息、句法手段是否用对、数字是否齐全、改写后是否仍能读出局限。参考说法：删掉抽样分层细节，因为设计名称不能替代样本量；删掉"每天三次"的时间点，保留测量频率；把两句局限并成 because 从句，省掉重复主语。

## 七、今日输出与自查

- 产出 Day03_压缩对照.md 与 Day03_数据清单.md：左栏原文 200 词，右栏压缩稿 110—130 词，九项数字逐项打勾。
- 产出 Day03_压缩说明.md，120—150 词，写清三处删减与三种句法手段。
- 自查一：压缩稿有没有出现"several""about"这类模糊词替代原数字。
- 自查二：每个分词短语的逻辑主语是否与主句一致。
- 自查三：局限条件是否完整保留，尤其"相关不等于因果"这一层。

## 八、明日预告

Day 4 换到耳朵：听十二分钟学术讲座，记演讲者的论证顺序，再做两分钟口头摘要录音。今天的压缩是在纸上做减法，明天的压缩要在时间里做选择——你不能回看，只能靠结构记忆。请先选好一段你熟悉的讲座音频，明天只听一遍就画结构图。