# 英语高级 · 第 2 周 Day 6|摘要复盘与信息核对 > 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** 写摘要时最危险的不是写错语法,而是写对语法却说错事实。今天做一件慢活:把昨天的 150 词拆回句子,逐句对照原文,判断每一句是保留、遗漏、添加还是失真。四类判定里,失真最难发现——它往往不是编造,而是名词化悄悄抬高了断言强度,或删掉一个限定词让相关性变成了因果。今天的目标是漏掉的关键信息不超过 1 条,并整理 8 条错句。这一天的产出不是新文章,是一双能看见自己偏差的眼睛。 ## 一、今日任务 | 环节 | 时长 | 具体动作 | 产出 | | --- | --- | --- | --- | | 逐句对齐 | 25 min | 把摘要切成 8—10 句,每句编号,在原文里找对应段落,填对照表四列 | 对照表 | | 判定与归因 | 20 min | 每句判为保留/遗漏/添加/失真;失真的句子写出"强度或范围被改成了什么" | 判定结果 + 失真说明 | | 词卡复习 | 15 min | 过本周 50 张卡,只留三次都没说顺的 10 张,其余归档 | 10 张顽固卡 | | 错句整理 | 15 min | 从摘要与练习中挑 8 条,每条写错在哪类(搭配/冠词/介词/强度/冗余);连同对照表归档,标注明日要带的 3 个问题 | 8 条错句本 + 归档 | 计划给这一天 60 分钟;本篇排到 75 分钟,多出的时间全部投在逐句对照上(从抽查 8 句改为全句过一遍),产出标准不变。当天状态差就按 60 分钟压缩,先保证对照表完整,再补错句本。 对照表必须用四列,不要用自由笔记。自由笔记会让你只记下感觉不对的地方,而漏掉那些读起来很顺、其实已经改了强度的句子——后者才是失真的主体。四列强制你逐句走,走完一遍通常会发现两三处自己完全没意识到的问题。 对照时不要在原文里找「我写的意思」,要找「原文到底说了什么」。这两件事看起来一样,实际差很远:前者是确认偏误,后者才是核对。 ## 二、核心词汇与搭配 今天的词都服务于"判断信息是否忠实",写复盘笔记和修改说明时会反复用到。 | 词/短语 | 音标 | 词性 | 释义 | 语域 | | --- | --- | --- | --- | --- | | overstate | /ˌəʊvəˈsteɪt/(≈欧弗斯泰特) | v. | 夸大 | 学术/媒体 | | understate | /ˌʌndəˈsteɪt/(≈阿恩德斯泰特) | v. | 轻描淡写、低估 | 学术/媒体 | | misrepresent | /ˌmɪsˌreprɪˈzent/(≈米斯雷普里泽恩特) | v. | 歪曲、误述 | 学术/媒体 | | faithful to | /ˈfeɪθfl tu/ | 形容词短语 | 忠实于(原意) | 学术 | | at the expense of | /ət ði ɪkˈspens əv/ | 介词短语 | 以……为代价 | 通用 | | in proportion to | /ɪn prəˈpɔːʃn tu/ | 介词短语 | 与……成比例 | 学术 | | trace X to Y | /treɪs/(≈特雷斯) | v. | 把 X 追溯至 Y | 学术 | | check X against Y | /tʃek əˈɡenst/ | 动词短语 | 用 Y 核对 X | 通用/学术 | | discrepancy | /dɪsˈkrepənsi/(≈迪斯克雷珀恩西) | n. | 不一致之处 | 学术/商务 | | reconcile | /ˈrekənsaɪl/(≈雷克恩赛尔) | v. | 使一致、调和 | 学术/商务 | **搭配与例句** 1. **overstate** — 搭配:overstate the case / overstate the effect。The summary overstates the case by turning an association into a cause.(摘要把关联说成因果,夸大了结论。)反义词 understate,一对词就能说清失真方向。 2. **understate** — 搭配:understate the scale of / understate the risk。Aggregate figures understate the scale of the shortfall in rural areas.(汇总数字低估了农村短缺的规模。)注意它是"低估",不是"少说"。 3. **misrepresent** — 搭配:misrepresent the findings / misrepresent what was said。Removing the caveat misrepresents the authors' position.(删掉限定条件就歪曲了作者的立场。)比 wrong 更准确,因为它指向"表述"而非"事实"。 4. **faithful to** — 搭配:be faithful to the original / to the evidence。A 150-word summary can be faithful to the argument without covering every result.(150 词摘要可以不覆盖每项结果,却仍忠实于论证。)这是复盘的验收语言。 5. **at the expense of** — 搭配:at the expense of accuracy / clarity。Compression came at the expense of precision.(压缩以精确为代价。)写复盘时用它指出取舍的代价。 6. **in proportion to** — 搭配:in proportion to its importance / to the evidence。Space was allocated in proportion to the strength of the evidence.(篇幅按证据强度分配。)这是判断摘要比例是否合理的一把尺。 7. **trace X to Y** — 搭配:trace the error to / trace the difference to。The distortion can be traced to one deleted modal.(失真可追溯到一个被删掉的情态动词。)追溯比"大概是"有用。 8. **check X against Y** — 搭配:check the summary against the source。Check each claim against the results section, not the abstract.(把每条陈述对照结果段,而不是摘要。)这条建议本身就是方法。 9. **discrepancy** — 搭配:a discrepancy between X and Y。The discrepancy between the two counts is unexplained.(两个计数之间的不一致未获解释。)可数,常与 between 连用。 10. **reconcile** — 搭配:reconcile the figures / reconcile A with B。The two estimates cannot be reconciled without assuming a different base year.(除非假设不同的基年,两个估计无法调和。)用于"差异无法消除"的场合。 ## 三、语言点:四类信息误差 核对的本质是把"摘要句"与"原文句"放在一起,看差异落在哪一类。四类里只有一类可以接受。 | 类型 | 定义 | 例 | 判定 | | --- | --- | --- | --- | | 保留 | 信息、强度、范围都与原文一致 | 原文 may reduce → 摘要 may reduce | 通过 | | 遗漏 | 原文有、摘要无,且属于关键信息 | 原文给出置信区间,摘要只给点估计 | 最多 1 条 | | 添加 | 原文无、摘要有,通常是作者补的解释 | 原文只报关联,摘要写"说明政策有效" | 不可接受 | | 失真 | 信息在,但强度、范围或因果被改 | may → does;in urban samples → 泛指 | 不可接受 | | 规则 | 说明 | 例 | | --- | --- | --- | | 1. 先查情态,再查数字 | 数字错容易被发现,情态错不会 | may / suggests / is consistent with 最易被删 | | 2. 查限定语是否随行 | 样本、时间、地区限定必须跟着结论走 | in 42 cities 不能省成 globally | | 3. 查因果词 | 原文的 associated with 不能改成 leads to | 一词之差,结论性质全变 | | 4. 关键信息的判定标准 | 影响结论方向的才算关键;修饰性细节不算 | 置信区间算,作者单位不算 | | 5. 名词化会偷偷改强度 | 动词带情态,名词不带 | may fail → the possible failure(保住了);→ the failure(丢了) | 第四条最容易被跳过,因为它要求你承认判断:什么算关键信息。一个可操作的标准是看它是否改变结论的方向或适用范围——样本从 42 城变成全球,结论就换了适用域,这算关键;作者来自哪所大学,不改变任何判断,这算细节。按这个标准删,摘要会变短而不变浅;按「看起来重要」删,删掉的往往是限定语,留下的往往是背景套话。 **强弱对比** | 原文 | 失真版 | 修回版 | | --- | --- | --- | | Canopy was associated with cooler tracts. | Canopy makes tracts cooler. | Canopy was associated with cooler tracts, though the design cannot establish direction. | | The effect may fade after three years. | The effect fades after three years. | The effect may fade after three years. | | The sample covers 42 cities, all in temperate zones. | The pattern holds globally. | The pattern holds across the 42 temperate-zone cities studied. | **常见误用** - **把"删掉限定"当成"精简"**:为了省 5 个词删掉 may,读者会把推测读成结论。省词要从不承载强度的形容词开始,不是从情态开始。 - **用名词化修补字数**:字数不够时把 because 改成 due to the fact that,把 may 改成 the possibility of。前者是废话,后者虽然保住了强度却拉长了句子,两样都不划算。 - **只对照摘要段**:论文摘要本身已经是二次压缩,且常偏乐观。核对应以结果段与局限段为准,否则你只是复制了别人的偏差。 修正有固定顺序,按这个顺序改能省一半时间:先补被删掉的情态与限定语,这一步不加字数反而可能减字;再删添加的解释句,通常整句拿掉;最后才调词序与搭配。多数人反着来,先改搭配,结果改完发现整句结论都是错的,白改。改完再数一遍词数,如果因为补限定超出预算,从背景段删一句通用描述,不要动结果。 ## 四、句型与范例 | 句式骨架 | 功能 | 例句(中译) | | --- | --- | --- | | The summary traces the discrepancy to X. | 复盘笔记,定位问题 | The summary traces the discrepancy to a single deleted modal.(复盘把不一致追溯到一个被删掉的情态动词。) | | This claim is faithful to X but not to Y. | 区分忠实范围 | The claim is faithful to the abstract but not to the results table.(这一说法忠实于摘要,却不忠实于结果表。) | | Compression came at the expense of X. | 说明取舍代价 | Compression came at the expense of the confidence interval.(压缩的代价是丢掉了置信区间。) | | The original reports X; the summary reports Y. | 并列呈现差异 | The original reports a 1.2 °C difference; the summary reports a "marked" cooling.(原文报告 1.2 度差异,摘要写成"显著"降温。) | | Nothing in the source supports X. | 判定"添加" | Nothing in the source supports the claim that planting is cost-effective.(原文没有任何内容支持种植具有成本效益这一说法。) | | The caveat was dropped in revision. | 记录修改动作 | The caveat was dropped in revision, which raised the claim by one notch.(修订时删掉了限定,断言强度升了一档。) | | Read against X, the sentence overstates Y. | 对照后的判断 | Read against the methods section, the sentence overstates the sample's reach.(对照方法段,这句话夸大了样本的覆盖范围。) | | Once X is restored, the two versions agree. | 修复确认 | Once the modal is restored, the two versions agree.(情态补回后,两版一致。) | | The gap is small enough to leave as is. | 接受轻微差异 | The gap is small enough to leave as is.(差异小到可以保留。) | | In proportion to the evidence, X receives too much space. | 评价篇幅分配 | In proportion to the evidence, the policy implication receives too much space.(相对证据强度,政策含义占的篇幅过多。) | 这十句可以直接抄进复盘笔记。它们的作用不是显得专业,而是让"我觉得这里不太对"变成"这里删了一个情态,强度升了一档"——后者才能改。 ## 五、练习 **A 词汇与搭配(7 题)** 1. 填介词:The error can be traced ___ a single assumption. 2. 填介词:The summary is faithful ___ the original argument. 3. 填介词:Space was allocated in proportion ___ the evidence. 4. 改错:The report overstates about the size of the effect. 5. 改错:There is a discrepancy of the two counts. 6. 选词(understate / misrepresent):By calling it a "minor adjustment", the memo ___ the scale of the change. 7. 选词(reconcile / check):The two estimates cannot be ___ without a different base year. **B 句法改写(5 题)** 8. 把失真句修回:The programme cuts dropout rates by half. 9. 把失真句修回:Higher fees deter all applicants. 10. 把"添加"删掉:The results show a correlation, which proves that funding matters most. 11. 给下面的句子补回范围限定:The effect holds for older workers.(原文:in manufacturing firms with more than 500 employees) 12. 把复盘判断写成一句话(用 trace ... to):摘要说"显著下降",原文是"小幅下降但统计显著"。 **C 输出(2 题)** 13. 完成 Day 5 摘要的对照表:每句一行,四列(摘要句 / 原文对应 / 判定 / 处理),至少覆盖 8 句。判定为遗漏的,写清是否关键。 14. 整理 8 条错句:每条给原句、改句、错因(搭配/冠词/介词/强度/冗余),错因不许写"不熟"。 ## 六、参考答案 1. **to**。trace X to Y,表示"把 X 追溯到 Y";trace back to 也可,但 to 不能省。 2. **to**。faithful to 是固定搭配,不用 with。 3. **to**。in proportion to 表"与……成比例",of 不成立。 4. **The report overstates the size of the effect.** overstate 是及物动词,直接接宾语,不加 about。 5. **There is a discrepancy between the two counts.** discrepancy 接 between,不接 of。 6. **understates**。memo 把大改动说成"小幅调整",是低估而非歪曲;misrepresent 指表述与事实不符,此处重点在尺度。 7. **reconciled**。两个估计无法调和,用 reconcile;check 是核对动作,不表示"使一致"。 8. **The programme was associated with a halving of dropout rates.** 原文若为观察性研究,cuts 是因果断言;改为 associated with + 名词化 halving,强度回到相关。 9. **Higher fees were followed by a fall in applications among price-sensitive applicants.** 原句 all 是范围失真;改写把范围收回到原文支持的子群,并用 were followed by 避免因果。 10. **The results show a correlation between funding and outcomes.** 删掉 which proves 与 most;proves 是添加的因果,most 是添加的比较级。 11. **The effect holds for older workers in manufacturing firms with more than 500 employees.** 限定语必须贴在名词后,不能放句末另起一句,否则容易被读成补充说明而非限制。 12. **The discrepancy can be traced to the word "significant", which the summary reads as "large".** 一句话点明:差异来自统计显著被读成幅度显著。 13. **对照表评分**:① 8 句全覆盖;② 每句有原文位置(段名即可);③ 判定不出现"好像""大概";④ 遗漏项 ≤1 且能说明为何关键。参考行:*摘要句 "Canopy cut temperatures by 1.2 °C" / 原文 "associated with a 1.2 °C difference" / 判定:失真(因果)/ 处理:改为 was associated with*。 14. **错句本评分**:8 条中至少 5 条属于搭配或介词类(本周高发),强度类至少 1 条。示例:*原句 The study makes a research on heat. → conducted research on heat. 错因:搭配(research 不可数,动词用 conduct)*。 ## 七、今日输出与自查 | 检查项 | 合格线 | 我的结果 | | --- | --- | --- | | 对照表句数 | ≥8 句,四列齐全 | | | 遗漏的关键信息 | ≤1 条 | | | 失真句 | 全部修回,且写明强度或范围的变化 | | | 错句本 | 8 条,每条有错因分类 | | | 顽固词卡 | 10 张,能说出为什么记不住 | | 产出文件:`第02周_Day06_对照表与错句.md`。对照表左栏贴摘要原句,不要改写后再对照——改写过的句子已经不是你要检查的对象。 ## 八、明日预告 Day 7 高压:60 分钟读 1200 词论文摘要与引言,写 200 词摘要;40 分钟口头复述并录音。今天查出的失真类型,明天要在没有对照时间的情况下避开——这才是真正的检验。