# 英语高级 · 第 12 周 Day 5|模拟面试与追问应答 > 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** 面试真正的分水岭在第一层追问。第一遍回答可以靠准备,追问只能靠结构:你有没有把"我做了什么"和"结果为什么发生"分开说,决定了对方能不能继续往下挖。今天做 30 分钟连续模拟,每题至少追问一层,目标是 6 题里 4 题扛得住。语言上只练一个动作:**把模糊的追问变成可回答的问题**——先用一句话复述对方真正在问什么,再给 15 词以内的直接答案,最后给边界。产出是一条 30 分钟不剪辑的录音和一张追问预判表。 ## 一、今日任务 | 环节 | 时长 | 具体动作 | 产出 | | --- | --- | --- | --- | | 1 写追问预判 | 15 min | 六张卡各预写两层追问:机制层(为什么会这样)、替代层(换一种做法会不会更好) | 追问表 1 张(12 问) | | 2 墙钟模拟 | 30 min | 30 分钟连续进行,每题 2 分钟回答 + 至少 1 层追问 + 30 秒补充;全程录音,中途不看笔记 | 录音 1 条(30 min) | | 3 回听标崩点 | 25 min | 只标三类:答非所问、防御性反驳、换词重复原答案;每类记时间戳 | 崩点清单 | | 4 重录四题 | 20 min | 只重录扛不住的那几题(至少 2 题,最多 4 题),每题仍带 1 层追问 | 重录 2—4 条 | **为什么用墙钟而不是逐题练**:逐题练会把每题的 2 分钟切成可以重来的小段,而真实面试不给你重来。连续 30 分钟会暴露真正的问题——第 4 题开始语速变快、第 5 题开始重复用词、第 6 题开始答得比第 1 题短。这些衰减只能在全真时长里看见。 **30 分钟的分配**:6 题 × 2 分钟 = 12 分钟;6 层追问 × 1 分钟 = 6 分钟;补充与过渡 4 分钟;剩下 8 分钟是你说错、停顿、重来的真实损耗。不要为了凑满而加速。 ## 二、核心词汇与搭配 | 词/短语 | 音标 | 词性 | 中文释义 | 高频搭配 | 语域 | | --- | --- | --- | --- | --- | --- | | probe | prəʊb | v. | 追问、深挖 | probe the assumption; probe further | 面试、媒体 | | follow-up | ˈfɒləʊ ʌp | n. | 追问;后续问题 | a natural follow-up; take a follow-up | 面试、商务 | | push back | pʊʃ bæk | phr. v. | 提出异议、顶回去 | push back on the number; push back gently | 商务 | | quantify | ˈkwɒntɪfaɪ | v. | 量化 | quantify the impact; hard to quantify | 学术、商务 | | attribute | əˈtrɪbjuːt | v. | 归因于 | attribute the gain to X | 学术、商务 | | caveat | ˈkæviæt | n. | 限定条件、保留意见 | one caveat; with the caveat that | 学术、法律 | | contingent | kənˈtɪndʒənt | adj. | 取决于、视……而定 | contingent on approval; contingent on demand | 商务、法律 | | concede | kənˈsiːd | v. | 承认(对方某点成立) | concede the point; concede that | 通用、谈判 | | defensible | dɪˈfensəbl | adj. | 站得住、经得起质疑 | a defensible number; defensible in hindsight | 学术、商务 | | stress-test | stres test | v. | 压力测试(把结论推向极端) | stress-test the assumption | 商务、学术 | **例句与用法** 1. **probe** — They will probe the assumption behind the number, not the number itself. 他们会追问数字背后的假设,而不是数字。(probe 是及物动词,直接接对象;面试里最常被 probe 的是因果链。) 2. **follow-up** — The follow-up is where prepared answers run out. 追问才是准备好的答案用尽的地方。(follow-up 可数,a follow-up 指一次追问。) 3. **push back** — I pushed back on the target, not on the person who set it. 我质疑的是目标,不是定目标的人。(push back on + 对象,必须指明对象,否则听起来像对抗。) 4. **quantify** — I could not quantify the retention gain, so I did not claim one. 我无法量化留存收益,所以没有声称有。(quantify 的诚实用法:说不能量化,比编一个数字更可信。) 5. **attribute** — We attributed the drop to one change, and tested it by reverting. 我们把下降归因于一处改动,并通过回滚做了验证。(attribute X to Y 是固定搭配,介词不能换。) 6. **caveat** — One caveat: the numbers cover nine months, not a full year. 一点保留:数据是九个月,不是整年。(caveat 后接冒号给具体限制,比说 it depends 精确。) 7. **contingent** — The whole plan is contingent on finance releasing the budget in Q1. 整个计划取决于财务在一季度放预算。(contingent on + 名词,正式但清楚。) 8. **concede** — I conceded the point about sample size and moved to what the data could support. 我承认样本量的问题,然后转到数据能支持的部分。(concede 只让一点,不让整段。) 9. **defensible** — A conservative number is more defensible than an impressive one. 保守的数字比漂亮的数字更站得住。(defensible 是面试里对"你这数字可靠吗"的最佳回应词。) 10. **stress-test** — Stress-test the claim: what happens if the effect is half of what you say? 把结论推向极端:如果效果只有你说的一半呢?(stress-test 用于主动示弱,显示你想过边界。) ## 三、语言点:追问的三步闭环 追问之所以难,是因为它常常是**模糊的**:Why did that work? 可能问机制、可能问归因、可能问"换个人做会不会也成"。直接回答,八成答偏;先问回去,又显得躲。可用的解法是把追问重写成一句可回答的问题,再回答。 **规则** 1. **第一句先定焦,用 If you mean X, then ...**。If you mean why the queue moved, then it was the approval change. 这一句不是拖延,它把"你为什么觉得是你的功劳"翻译成"机制是什么",同时向对方展示你听懂了。 2. **直接答案放在第二句,≤15 词**。The short answer is no. / It was one change, not five. 面试官在等这一句;先给结论,再给支撑,顺序反了会被当成绕。 3. **hedge 放在边界,不放在结论**。结论用直陈(It cut the delay by half),边界用 as long as / unless / provided(as long as volume stayed flat)。反过来最糟:结论全是 might、possibly,边界却说得斩钉截铁。 4. **数字被追问时补口径,不补新数字**。回答分三层:怎么量的、覆盖多长时间、谁核对的。改口径可以,改数字不行——一改,前面所有数字都不可信。 5. **两轮答不上来就收口**:That I don't know. Here is how I would find out, and by when. 现编一个答案的代价远大于承认不知道。 **什么时候用**:面试追问、答辩质询、客户异议、上级的"为什么"。凡是对方的问题带着怀疑但还没说清怀疑什么的时候。 **什么时候不能用**:对方的问题本身很清楚(What was the churn rate?),此时先复述反而显得啰嗦,直接给数字即可。时间极紧的场合(30 秒电梯问答)也只用第二句。 **用了会显得怎样**:定焦 + 直答 + 边界,显得你想过反面且有分寸;只定焦不直答,显得圆滑;只直答不给边界,显得绝对,一被 stress-test 就崩。 **强弱对比** | 弱版 | 强版 | 差别 | | --- | --- | --- | | That's a great question. Well, there were many factors, and I think teamwork was really important. | If you mean why the queue moved, it was one change: we collapsed two approvals. Teamwork helped, but it was not the cause. | 弱版是拖延 + 归因给氛围;强版先定焦,再给单一机制,最后排除干扰项 | | Yes, absolutely, we could have done more. | Up to a point. With the budget we had, the alternative was to delay the launch — and that was worse. | 弱版是无边界让步;强版承认一半、给出替代方案的代价,保住立场 | | I already said that in my previous answer, but to repeat... | Let me put it differently: the constraint was not speed but sign-offs. | 弱版是换词重复并带情绪;强版换角度重述,提供新信息 | **常见误用** 1. **用 So, great question 换时间**。这句话零信息,且暴露你没准备。要争取两秒,用 Let me take the second part first,因为它同时交代了结构。 2. **所有追问都答成"是的,而且我们还……"**。从不设边界的人不可信,因为听起来什么都能做到。至少在一次追问里明确说 This is where it stops working。 3. **换词重复原答案**。追问方听到同一结构第二次,会判断你在回避。正确做法是换层级:从"做了什么"降到"机制是什么",或升到"如果重来会怎么改"。 ## 四、句型与范例 **10 个追问应答句式** 1. **If you mean X, then Y.** 如果你问的是 X,那么 Y。(定焦句,最通用。) 2. **The short answer is no — and the reason is one line: ...** 简短回答是不——理由一句话:……(直答 + 一句支撑,控制在 20 词内。) 3. **Let me separate two things.** 我把两件事分开说。(追问把两件事混在一起时,先拆再答。) 4. **That number is measured over nine months, not a year.** 这个数字是九个月的,不是整年。(补口径的标准句。) 5. **I would not claim that.** 我不会这么讲。(拒绝被塞进一个过强的结论,比 Yes, exactly 安全。) 6. **It holds as long as volume stays flat.** 只要量持平,它就成立。(边界句,hedge 放在这里。) 7. **The trade-off was real: we shipped later to ship once.** 代价是真的:晚上线,但只上线一次。(承认代价,换可信度。) 8. **Partly. Here is the part I would not defend.** 部分成立。这部分我不辩护。(半让步,比全让步或全反驳都稳。) 9. **That I don't know. I would find out by X, and I would know by Friday.** 这个我不知道。我会用 X 去查,周五能有答案。(承认 + 路径 + 时限,三件套。) 10. **Which is why I changed how I ran the next project.** 所以下一个项目我改了做法。(把追问转成学习证据,收尾用。) **六题 × 两层追问预判表** | 题 | 第一层(机制) | 第二层(替代/边界) | 应答支点 | | --- | --- | --- | --- | | 冲突 | 双方真正分歧在范围还是资源? | 如果对方坚持不让,你会怎么做? | 用范围文档把争点外化 | | 失败 | 你什么时候意识到判断错了? | 早两周发现,结果会不同吗? | 承认信号被忽略,给出新机制 | | 领导 | 你没有正式授权,怎么让人配合? | 如果对方就是不配合呢? | 用指标与可见度换配合 | | 优先级 | 你砍掉的那个需求,谁承担后果? | 如果被砍的那个更重要呢? | 用影响 ÷ 成本排序并留复核点 | | 跨部门 | 对方为什么一开始不同意? | 你的方案对他们有什么损失? | 先算对方的成本,再谈交换 | | 超预期 | 标准是谁定的? | 如果没有你,会不会也发生? | 区分"我推动"与"时势推动" | **追问实录示范(约 170 词,带批注)** > **Q:** You said attrition fell from 24 to 11 per cent. How much of that was you? > **A:** If you mean what caused the drop, one change accounts for most of it: we collapsed two approvals into one. If you mean whether it would have happened without me, probably — but not in nine months. The queue had been visible for two years and nobody owned it. > **Q:** So you are saying the idea was obvious? > **A:** The idea was obvious. The decision was not. Two directors had to give up a signature, and that is the part I did. > **Q:** What if finance had refused the audit condition? > **A:** Then I would have run it as a pilot on one team for a quarter and brought back the numbers. That was the fallback, and I had it ready before the meeting. 批注:① If you mean ... 拆开了一个含混的 how much was you,把"功劳比例"转成"机制"与"必要性"两个可答问题;② probably — but not in nine months 是精准让步:承认时势,保住时限贡献;③ The idea was obvious. The decision was not. 六词对照,承担了第二层的全部说服力;④ 最后一段给 fallback,说明追问方担心的风险你早有准备;⑤ 全段没有一个 great question,也没有一个 I think。 ## 五、练习 **A 词汇与搭配(7 题)** 1. 填空:They will ______ the assumption behind the number. 2. 搭配:用 attribute 写一句归因,并说明你如何验证。 3. 改错:We attributed the drop on the new onboarding flow. 4. 用 caveat 写一句,必须含具体限制(时间或样本)。 5. concede 与 admit 在面试语境下的差别,各造一句。 6. 用 contingent 写一句,说明计划依赖的条件。 7. defensible 与 impressive 在描述数字时有何取舍?各造一句。 **B 句法与改写(5 题)** 1. 把拖延句改成结构句:That's a really great question, let me think about how to answer it. 2. 给结论加边界(不改结论):The change cut approval time by half. 3. 半让步改写:Yes, you're absolutely right, we should have done that differently. 4. 承认不知道并给路径:I'm not sure about that figure. 5. 把重复原答案改成换层级:As I mentioned, the constraint was sign-offs. **C 输出(2 题)** 1. 交 30 分钟模拟录音(1 条,不剪辑),六题各含至少 1 层追问;附崩点清单(含时间戳)。 2. 写 400 词复盘:哪两题扛住了、哪题崩了、崩在定焦还是边界;每类给一句下次要用的句式。 ## 六、参考答案 **A** 1. probe(probe the assumption 是固定搭配。) 2. 示例:We attributed the gain to the approval change, and tested it by reverting for two weeks.(attribute X to Y + 验证方式。) 3. 错。介词是 to:We attributed the drop to the new onboarding flow. 4. 示例:One caveat: the figure covers nine months and one region.(时间 + 范围双限制。) 5. concede 指承认对方某一点成立,仍保留立场;admit 常含"承认错误",语气更重。例:I concede the sample is small.(让一点)/I admit I missed the signal.(认错) 6. 示例:The rollout is contingent on legal sign-off in March.(contingent on + 条件。) 7. defensible 强调经得起追问,可能不亮眼;impressive 强调好看,可能经不起追问。例:Eleven days to four is defensible.(可核查)/A 70 per cent cut would be impressive but untrue.(好看但假) **B** 1. 示例:Let me take the second part first: the number, then the cause.(交代结构,替代拖延。) 2. 示例:The change cut approval time by half, as long as volume stayed flat.(as long as 引出边界,结论不动。) 3. 示例:Partly. With the budget we had, the alternative was to delay the launch, which was worse.(承认一半,给出替代方案的代价。) 4. 示例:That figure I don't have. I would pull it from the finance dashboard and confirm by Friday.(承认 + 来源 + 时限。) 5. 示例:Let me put it at the level of mechanism: four signatures, one decision, so the delay sat in the queue, not in the work.(换层级:从"我说过"降到机制。) **C 评分维度** 录音看四项:① 六题是否都有明显追问(无追问的题不计入"扛住");② 扛住的标准是追问后给出新信息,而不是重复;③ 至少 4 题达标;④ 无"So, great question"类填充开场。复盘看三点:① 崩点必须带时间戳,写"整体还行"不得分;② 每题给出可复用的句式,不写"多练";③ 承认不知道的题是否给了路径与时限。 ## 七、今日输出与自查 产出文件:第12周_Day05_模拟录音(1 条,30 分钟)、第12周_Day05_追问表.md、第12周_Day05_复盘400词.md。 | 自查项 | 合格线 | 怎么查 | | --- | --- | --- | | 追问覆盖 | 6 题全部有至少 1 层追问 | 听录音数问句,无追问的题补录 | | 扛住率 | ≥4/6 题在追问后给出新信息 | 逐题判定,重复原答案记为未扛住 | | 定焦 | 至少 3 题用 If you mean 类句式 | 搜录音转写中的 if you mean / let me separate | | 边界 | 至少 2 题出现 as long as / unless / up to a point | 无边界 = 过度承诺 | | 承认 | 至少 1 处明确说不知道并给路径 | 全程无一处承认,说明在硬撑 | **一个反直觉的判断**:模拟面试里表现最好的往往不是答得最全的那次,而是承认得最准的那次。追问方真正在测的是"你会不会为了好看而编",不是"你知不知道"。所以今天允许自己说三次 That I don't know,但每次后面必须跟路径和时限。 ## 八、明日预告 Day 6 做 90 天总复盘:把第 1 天的录音和今天的放在一起对比,列 5 个仍弱项与对应练法。今天标出的崩点清单,就是明天第一栏的素材——复盘表不写感受,只写可观测的差异,比如填充词从每分钟 9 次降到 3 次,而不是"口语更流利了"。