# 英语高级 · 第 7 周 Day 1｜三源矩阵与利益立场

> 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**

同一个事件，三家媒体各写一篇，句子都挑不出错，可你一落笔就写成"三家都认为……"。今天要解决的不是读不懂，而是分不清：分不清哪句是测量出来的结果、哪句是记者的推断、哪句是带着利益立场的评价。这是本周论证主线的第一块地基——第 5 天你要写共识、分歧、空白三段，如果今天没把主张和证据拆开，那三段必然退化成三篇摘要的拼接。语域上今天练"归属但不站队"：让读者看得见每个说法出自谁口，同时看不出你替谁说话。学完你能交出一张可复查的来源矩阵，外加 6 句带距离的转述。

## 一、今日任务

| 环节 | 时长 | 具体动作 | 产出 |
| --- | --- | --- | --- |
| 1. 定题与预判 | 10 min | 选一个近半年被连续报道的题目（某类芯片出口、某城房租、某款减重药均可），先写下 3 句"我以为的共识"，读完再回来对照 | 预判 3 句 |
| 2. 三源通读 | 35 min | FT、The Economist、Nature 新闻各读 1 篇同题报道；每篇只做两件事：圈出主张句，标注证据类型（统计、访谈、文件、实验） | 3 篇标注稿 |
| 3. 填矩阵 | 20 min | 按"主张 / 证据一 / 证据二 / 证据类型 / 利益立场 / 未回答的问题"六列填写，每格写短句，不抄原句 | 矩阵 1 张 |
| 4. 立场审计 | 10 min | 回原文找 6 处评价性词语（形容词、情态、隐喻），逐处判强度：中性、有倾向、明确站队 | 6 处标记 |
| 5. 写转述 | 10 min | 用今日 10 个词中的 6 个，写 6 句带归属的转述，每句注明来源与证据类型 | 6 句 |

三篇报道的篇幅加起来约 2500 词，通读阶段不许查词、不许回读，遇到不懂的术语先画圈。矩阵填不出来的格子，恰恰是今天最值钱的信息——它说明这篇报道没有交代。

## 二、核心词汇与搭配

| 词 / 短语 | 音标 · 词性 | 释义与高频搭配 | 例句与翻译 | 语域 |
| --- | --- | --- | --- | --- |
| contend | /kənˈtend/（≈克恩泰恩德） v. | 主张、坚称，隐含争议；contend that, contend with | The report contends that the shortfall stems from delayed permits. 该报告坚称缺口源于许可审批延误。 | 学术 / 媒体 |
| attribute | /əˈtrɪbjuːt/（≈厄特里布尤特） v. | 把……归因于；attribute sth to sth, be attributed to | Much of the decline is attributed to tighter lending standards. 大部分下滑被归因于放贷标准收紧。 | 学术 / 商务 |
| vested interest | /ˌvestɪd ˈɪntrəst/ n. | 既得利益；have a vested interest in sth, a vested interest in the outcome | Equipment suppliers have a vested interest in prolonging the upgrade cycle. 设备供应商有既得利益去延长升级周期。 | 商务 / 媒体 |
| corroborate | /kəˈrɒbəreɪt/（≈克罗伯雷特） v. | 印证、佐证；corroborate a claim, corroborating evidence | Two independent audits corroborate the figures in the filing. 两份独立审计印证了申报文件中的数字。 | 学术 |
| anecdotal | /ˌænɪkˈdəʊtl/（≈艾尼克多特尔） adj. | 个案性的、传闻性的；anecdotal evidence, remain anecdotal | Evidence for the link is still largely anecdotal. 支持这一关联的证据在很大程度上仍是个案传闻。 | 学术 |
| disclose | /dɪsˈkləʊz/（≈迪斯克洛兹） v. | 披露；disclose data, disclose methodology, fail to disclose | Neither company disclosed how the sample was drawn. 两家公司都未披露样本的抽取方式。 | 商务 / 媒体 |
| stem from | /stem frəm/ phr. v. | 源于，用于追溯成因；stem from a shortage, stem largely from | The price spike stems from a shortage of refining capacity. 价格飙升源于炼油产能不足。 | 通用 / 学术 |
| partial | /ˈpɑːʃl/（≈帕希尔·长） adj. | 片面的、有偏向的（亦指"部分的"）；a partial account, a partial picture | What the article offers is a partial picture: accurate in its numbers, silent on who pays. 这篇文章给出的是片面的图景：数字准确，却对谁买单只字不提。 | 学术 |
| quantify | /ˈkwɒntɪfaɪ/（≈克沃恩提法） v. | 量化；quantify the impact, hard to quantify | The benefits are real but hard to quantify. 收益真实存在，但难以量化。 | 学术 / 商务 |
| incentive | /ɪnˈsentɪv/（≈伊恩塞恩提夫） n. | 激励、诱因；a perverse incentive, align incentives, create an incentive to | A subsidy tied to output creates an incentive to overproduce. 与产量挂钩的补贴会制造过度生产的诱因。 | 商务 / 学术 |

记词时不要只记中文对应，要连着介词记：attribute to、stem from、an incentive to do、a vested interest in doing。介词错了，词再高级也会被一眼看出不是母语者的搭配习惯。

## 三、语言点：事实、推断、评价三分

同一个意思可以写成三种句子，语法都成立，但确定性等级完全不同。中文靠"据说""可能""显然是"这类副词区分，英文更常靠句法区分：时态、情态、归属结构、主语选择。C1 写作成熟的标志，是让句子本身携带这个等级，而不是让读者去猜。

**规则**

1. **事实句**：给出可核对的量与具体来源，用一般过去时或现在完成时，不加情态，不加评价形容词。The trial enrolled 1,204 patients across 18 sites.
2. **推断句**：用 may、might、could 或 suggest、appear、point to。The timing suggests that the rule, not the subsidy, drove the shift.
3. **评价句**：必须显性归因，把评价放到某个主体名下，而不是伪装成对世界的描述。a figure the industry describes as conservative 而不是 the misleading figure。
4. **因果分级**：只有实验或准实验设计才配得上 cause。观察数据一律写 be associated with、be consistent with、coincide with。
5. **排列顺序**：先事实、再推断、后评价。读者按这个顺序进入，才不会把推断读成事实。

**强弱对比**

| 弱（等级混乱） | 强（等级清楚） | 差别在哪 |
| --- | --- | --- |
| Sales collapsed after the ad campaign. | Sales fell 31% in the two quarters after the campaign. | collapsed 是评价性动词，同时预设了因果；数字句只陈述幅度与时间。 |
| The policy failed because firms ignored it. | Compliance was low: only 38% of firms filed the required return. | 前者是判断，后者是任何人都能去核的观察。 |
| Critics say the data are unreliable. | The data have not been independently audited. | "critics say" 制造虚假平衡，仿佛两种意见等重；指出数据的实际状态更难反驳。 |

**常见误用**

1. suggest that 后面接 will：The data suggest that prices will fall。若这是研究结论，应写 may fall；will 把推断升级成预测，等于替作者背书。
2. attribute 后面接 because：~~was attributed because of poor management~~，应为 was attributed to poor management。attribute 的宾语是"被归因的事"，to 后才是原因。
3. 用 biased 形容人：~~a biased scientist~~。改为 a study funded by the manufacturer 或 an author who advises the industry。人身化指控会让整段失去说服力，而且你无法证明动机。
4. 需要提醒的是，做文献综述的方法描述时不要用评价句。审稿人会立刻问：你凭哪条标准判断它"弱"？方法部分只写设计与样本。

## 四、句型与范例

| 句式 | 功能 | 例句与翻译 |
| --- | --- | --- |
| X reports that …, citing … | 归属并点明证据类型 | The Economist reports that capacity will double by 2028, citing three unnamed plant operators. 《经济学人》称产能将在 2028 年前翻倍，援引三家匿名运营方。 |
| While X puts the figure at …, Y's estimate is closer to … | 并列两个数字，暂不裁决 | While the FT puts the shortfall at 12%, the industry estimate is closer to 20%. FT 给出的缺口是 12%，行业估计则接近 20%。 |
| The discrepancy is partly explained by … | 解释差异，避免指控 | The discrepancy is partly explained by different cut-off dates. 差异部分源于截止日期不同。 |
| X has a commercial stake in …, which may colour its reading of … | 点利益立场而不下道德判断 | The consultancy has a commercial stake in the forecast, which may colour its reading of demand. 该咨询公司在这份预测中有商业利益，这可能影响它对需求的解读。 |
| Neither account addresses … | 指出共同空白 | Neither account addresses who bears the cost of the upgrade. 两篇报道都没有回答谁承担升级成本。 |
| The claim rests on a single … | 指出证据强度，不否定结论 | The claim rests on a single plant visit in 2024. 这一主张仅建立在 2024 年的一次工厂走访上。 |
| What is presented as consensus is in fact … | 拆解伪共识 | What is presented as consensus is in fact one consultancy's model, repeated. 所谓共识，其实是一家咨询公司的模型被反复引用。 |
| X's figures are consistent with Y's, though the two part company on … | 一句内同时给共识与分歧 | The two sets of figures are consistent on volume, though they part company on price. 两组数字在销量上一致，在价格上则分道扬镳。 |
| That may be true of …, but it does not follow that … | 限定适用范围 | That may be true of the coastal plants, but it does not follow that inland capacity is idled. 这对沿海工厂或许成立，却不能推出内陆产能也已闲置。 |
| On the evidence available, … remains unresolved. | 保留判断，给后续段落留口子 | On the evidence available, whether the shortfall is cyclical remains unresolved. 就现有证据看，缺口是否周期性仍无定论。 |

连句示范（约 110 词，可直接改写成矩阵的叙述版）：

> On volume, the three accounts converge: all put 2025 output within 4% of each other. They part company on cause. The FT attributes the shortfall to permitting delays, citing two developers; The Economist prefers a demand story, resting on a single consultancy model; the Nature news piece stays with what can be measured, reporting that no independent audit of the figures has been published. The one claim none of them tests is who absorbs the cost. That gap, rather than any factual disagreement, is where a synthesis has something to add.

注意最后一句：它没有指控任何一家"错"，只是把"三家都没做"指出来，这比"三家都不够严谨"有力得多。

## 五、练习

**A 词汇与搭配（7 题）**

1. 用 corroborate、contend 填空：Three independent studies ______ the effect, but the authors ______ that the mechanism remains unclear.
2. 搭配配对：vested / anecdotal / perverse / partial 对 interest / evidence / incentive / picture。
3. 选词填空（quantify / disclose）：Regulators now require firms to ______ how the sample was drawn, though few attempt to ______ the resulting bias.
4. 语域改写：把 The company says it didn't do anything wrong 改成商务书面句。
5. 语域判断：下面哪一句不能出现在学术段落里，为什么？a) The sample was drawn from two provinces. b) This whole argument is a mess. c) The authors report no conflicts of interest.
6. 用 stem from 改写：The delay happened because the permit took eight months.
7. 名词化改写：把 The suppliers want the cycle to last longer 改成含 vested interest 的句子。

**B 句法、改写与压缩（6 题）**

1. 把 Prices went up a lot and it's the war's fault 改成合格的事实句（自拟数字）。
2. 把 Critics say the data are unreliable 改成可核查句。
3. 合并成一句含让步：FT 的数字高于行业估计；FT 未披露统计口径。
4. 压缩到 20 词以内，保留归属：According to a report published by the industry association last month, which surveyed 40 manufacturers, orders have fallen for three consecutive quarters.
5. 改错：The survey suggests that consumers will cut spending next year.
6. 把评价形容词去掉，改成可核查描述：The report's misleading cost estimate.

**C 输出（2 题）**

1. 矩阵 1 张（3 行 × 6 列）+ 6 句转述，合计不超过 140 词。6 句里至少 2 句含具体数字，至少 1 句指出空白。
2. 2 分钟口头陈述：向一位没读过这三篇的同事说明"三家在哪一点上其实完全一致，又在哪一点上根本没对话"。录音，回听时数一数自己有没有说"they all say"。

## 六、参考答案

**A**

1. corroborate；contend。前者接证据，后者接有争议的主张，两者常同时出现：证据印证事实，作者主张解释。
2. vested interest；anecdotal evidence；perverse incentive；partial picture。
3. disclose；quantify。先披露方法，才谈得上量化偏差，顺序不能反。
4. The company denies any wrongdoing. 或 The company said it had not breached the rules. 口语的 didn't do anything wrong 在书面语里要么换成 deny，要么给出被违反的具体规则。
5. b。a 与 c 都是可核查陈述，b 是纯评价且无标准，学术段落里出现会让读者质疑你的判断依据。
6. The delay stems from a permit that took eight months. 用 stem from 后，原因变成名词短语，句子重心落在因果链而非"发生"。
7. Suppliers have a vested interest in prolonging the upgrade cycle. 名词化把"想要"这种心理描述换成可观察的利益位置，语气更硬也更难反驳。

**B**

1. Prices rose 23% in the six months after the export ban, a period that also saw freight costs double. 给出幅度、时间窗，并把第二个变量并列出来，读者自己会判断因果是否成立。
2. The data have not been independently audited, and the authors decline to release the raw series.
3. The FT's figure is higher than the industry estimate, though the paper does not disclose how the series was compiled. 用 though 把"数字更高"和"口径不明"压进一句，避免两句并列显得像两条独立指控。
4. An industry survey of 40 manufacturers reports three consecutive quarters of falling orders. 从 34 词压到 13 词，保留来源、样本量、结论三要素，删掉发布时间因为不影响判断。
5. The survey suggests that consumers may cut spending next year. 若是预测而非结论，也可写 points to weaker spending；will 把可能说成必然。
6. The report's cost estimate, which excludes land, is roughly half the figure used by the regulator. 用"排除了什么"替代"误导"。

**C 评分维度**

| 维度 | 合格 | 优秀 |
| --- | --- | --- |
| 归属清楚 | 6 句都能看出是谁说的 | 归属方式有变化，不重复"according to" |
| 证据分级 | 区分了统计与访谈 | 明确指出某结论只有单一证据 |
| 利益立场 | 至少 1 处点到利益关系 | 点到利益关系但不下道德判断 |
| 空白识别 | 指出 1 个共同未答问题 | 空白具体到"缺哪类数据" |

参考转述（可对照，不要照抄）：The FT attributes the shortfall to permitting, citing two developers. The Economist, by contrast, rests its demand story on a single consultancy model. Both put 2025 output within 4% of each other, so the disagreement is about cause, not scale. Neither discloses who bears the upgrade cost.

## 七、今日输出与自查

| 自查项 | 合格线 | 我的结果 |
| --- | --- | --- |
| 矩阵完整度 | 3 源 × 6 列无空格；每源 ≥2 个证据并标类型 |  |
| 立场标注 | 6 处评价词，强度判断有依据 |  |
| 三分能力 | 6 句转述中，事实句与推断句可区分 |  |
| 词表落地 | 10 词中至少 6 个进入转述 |  |
| 预判对照 | 读完回看开头的 3 句，标出被推翻的那句 |  |

产出文件：`第07周/Day01_来源矩阵.md`，内含矩阵 1 张、6 句转述、预判对照 3 行。矩阵必须是表，不要写成段落。

## 八、明日预告

Day 2 转向图表语言：30 组趋势、比较、倍数、占比句式，每组各造一句，数字与方向必须全对。今天的矩阵解决"谁说"，明天的句式解决"怎么说清楚一个数在动"。请把今天三篇报道里出现的所有数字抄成一列带来，明天直接用它们造句，避免为造句编造数据。