Series ID: EDKS-ADV-ZH-0014 · How to Learn Advanced English (Chinese Edition) · Lesson No.014 · C1 → C2

Calibrate Certainty Without Becoming Vague | 精准控制学术英语中的确定程度

Advanced academic English does not sound certain all the time. It sounds exactly as certain as the evidence allows.

高级学术英语，不是永远“说得很肯定”。真正高级，是让语言的确定程度刚好匹配证据。

Writers often make one of two mistakes. Some write too strongly:

This proves that X causes Y.

Others become so cautious that the sentence loses intellectual content:

It may perhaps possibly be the case that X could maybe have some relationship with Y.

两种都不是 precision。第一种把 evidence 说强了；第二种把意思稀释到几乎无法判断。

The advanced task is calibration:

evidence state → claim strength → linguistic form.

证据状态 → 主张强度 → 语言形式。

---

Before we begin: this lesson does not replace the modality owner | 开始之前：本课不取代 modality 主页面

eduKate Sengkang already has a canonical general mechanism page for modality and certainty. Lesson 014 does not reteach the entire grammar system. It owns the C1–C2 academic calibration layer: deciding exactly how strong a research or professional claim should sound after considering evidence quality, scope, precision, alternative explanations and source distance.

• How Modality and Certainty Calibrate English Claims — general modality/certainty mechanism owner.
• Lesson 009 | Evidence-Based Reports and Recommendations — evidence-to-action reporting.
• Lesson 011 | Multi-Source Synthesis — evidence comparison across sources.
• Lesson 012 | Literature Review as Scholarly Conversation — field-level academic synthesis.
• Lesson 013 | Academic Argument That Survives Counterevidence — argument robustness.

1. CEFR treats precision of certainty as advanced language control | CEFR 把 certainty calibration 视为高级 precision

The CEFR Companion Volume describes propositional precision as the ability to formulate what one wishes to express accurately and to qualify likelihood, commitment, belief and related shades of meaning. At C1, learners can qualify opinions and statements precisely in relation to degrees of certainty, uncertainty, belief and doubt; at C2, they can convey finer shades of meaning through a wide range of qualifying devices.

Council of Europe | CEFR Companion Volume

所以，本课不是“学几个 hedge”。它训练的是 propositional precision。

2. Manchester calls this epistemological strength | Academic writing 在控制“知识强度”

The University of Manchester Academic Phrasebank explains that academic writers often mitigate the epistemological strength of a claim where uncertainty remains, especially in discussion sections where writers interpret findings.

University of Manchester Academic Phrasebank | Being Cautious

3. Certainty is not personality | 确定程度不是“作者性格”

Weak calibration:

I am very confident that X is true.

Better academic calibration:

Across three independent replications, X is consistently associated with Y under comparable conditions.

第二句告诉读者 confidence 从哪里来。

4. Separate psychological confidence from evidential confidence | 主观自信 ≠ 证据置信度

You can feel certain and still have weak evidence.

You can feel cautious and still possess strong evidence.

Academic wording should follow the evidence state, not your mood.

5. Certainty is multidimensional | Certainty 不是单一刻度

A claim can be:

• highly certain about direction;
• uncertain about magnitude;
• certain in one population;
• uncertain in another;
• strong for immediate outcomes;
• weak for long-term outcomes;
• strongly associated;
• causally uncertain.

不要把所有 uncertainty 压成一个 “maybe”。

6. Build an evidence-state profile | 建立 evidence-state profile

Dimension	Question
Existence	Does the effect/pattern appear at all?
Direction	Positive, negative or mixed?
Magnitude	How large?
Precision	How tightly estimated?
Causation	Can alternatives be ruled out?
Scope	For whom/where?
Persistence	How long?
Mechanism	Why?
Replicability	Does it recur independently?

7. The language should match the dimension under discussion | 语言要匹配 uncertainty 的具体位置

Bad:

The result is uncertain.

Better:

The direction of the effect is consistent across studies, but its long-term magnitude remains uncertain.

8. Build a calibration ladder—but do not turn it into fake mathematics | 建立语言阶梯，但不要假装每个词有固定百分比

Rough linguistic strength	Examples
Open possibility	may, might, could, possibly
Plausible interpretation	may reflect, is consistent with, could indicate
Moderate support	suggests, indicates, appears to
Strong support	strongly suggests, provides strong evidence for
Direct establishment within scope	shows, demonstrates, establishes — only where justified

这些不是 universal probability bands。不同 discipline、genre 和 evidence type 对词强度的习惯不同。

9. “May” does many jobs | may 不只表示“概率”

X may explain Y. — epistemic possibility.

Participants may leave at any time. — permission.

The treatment may cause nausea. — possible outcome.

先判断 function，再判断 certainty。

10. “Should” is especially dangerous for Mandarin speakers | should 有多种意义

The result should be available tomorrow. — expectation.

The team should repeat the test. — recommendation.

Participants should meet the criteria. — obligation/requirement depending context.

中文“应该”不能机械翻成 should。

11. “Must” can signal logical inference or obligation | must 也有双重功能

You must submit the form. — obligation.

There must be another explanation. — strong inference.

Academic writing usually needs to distinguish these clearly.

12. “Can” is often overused as evidence language | can 容易被滥用

This method can improve performance.

Does this mean:

• it is theoretically possible?
• one study observed improvement?
• it reliably improves performance?

Choose a more exact form if the distinction matters.

13. Modal verbs are only one calibration system | Hedge 不只是 modal verb

Other systems include:

• lexical verbs;
• adverbs;
• adjectives;
• nouns;
• reporting structures;
• conditional clauses;
• scope phrases;
• source attribution.

14. Lexical verbs carry epistemic force | 动词本身带 certainty

Verb	Typical commitment
speculates	low
suggests	cautious support
indicates	evidence points toward
supports	evidence is compatible and strengthens claim
shows	directly establishes stated observation within scope
demonstrates	strong; use carefully

15. “Shows” is not automatically wrong | shows 不是禁词

Appropriate:

The table shows that response time fell from 12 to 9 minutes.

Risky:

The observational study shows that the intervention caused the reduction.

第一句是直接 observation；第二句跨到了 causation。

16. “Demonstrates” carries a high burden | demonstrates 需要高证据负担

Use it when the design and the exact proposition justify the strength—not because it sounds academic.

17. “Suggests” is useful but can become lazy | suggests 也会被滥用

If every paragraph says suggests, you have stopped calibrating.

Ask what kind of support exists.

18. “Is consistent with” has a special job | is consistent with 不是 suggests 的花式替代

The pattern is consistent with a motivation effect.

This means the evidence fits that explanation.

It does not mean the explanation is uniquely established.

19. “Cannot rule out” also has a special job | cannot rule out 的意思很窄

Residual confounding cannot be ruled out.

This does not mean confounding is likely.

It means the evidence/design cannot exclude it.

20. “Appears” can mark provisional interpretation | appears 表示目前看起来

The effect appears stronger among novice learners.

Good when the pattern is visible but uncertainty remains.

21. “Tends to” describes pattern, not law | tends to 表趋势

Participants with lower baseline scores tend to improve more.

It implies exceptions are expected.

22. Adverbs adjust strength | 副词也可以精细调节

• possibly;
• perhaps;
• apparently;
• likely;
• probably;
• generally;
• typically;
• clearly;
• strongly.

23. Avoid stacking adverbs without adding precision | 不要堆 hedge

Weak:

This perhaps may possibly indicate…

Better:

This may indicate…

or explain the exact source of uncertainty.

24. “Clearly” is a dangerous booster | clearly 常常是危险 booster

Ask:

Clear to whom, from what evidence?

Replace rhetorical force with evidence description when possible.

25. “Obviously” is rarely an evidence argument | obviously 往往没有证据功能

If a point is genuinely obvious, you may not need to say so.

26. “Undoubtedly” creates a very high burden | undoubtedly 极强

Do not use it merely to sound authoritative.

27. Nouns can carry uncertainty | 名词也可以表达 uncertainty

• possibility;
• likelihood;
• probability;
• uncertainty;
• evidence;
• indication;
• hypothesis;
• interpretation.

28. Adjectives calibrate propositions | 形容词也可以调节 claim

• possible;
• plausible;
• likely;
• probable;
• consistent;
• uncertain;
• tentative;
• robust.

29. “Robust” should describe robustness, not confidence theatre | robust 要有明确对象

Better:

The association is robust to three alternative model specifications.

Not:

The result is robust.

without explaining what test it survived.

30. Clauses can carry conditions | Clause 是 advanced calibration 的核心

The effect is strongest when prior knowledge is low.

The conclusion holds provided that the measurement remains comparable across groups.

Although the direction is stable, the magnitude remains imprecise.

这类句子比单个 hedge 更精确。

31. Scope phrases are calibration devices | Scope phrase 也是 hedge

• in this sample;
• under these conditions;
• during the six-week trial;
• among novice learners;
• for the measured outcome;
• within the current evidence base.

高级写作经常通过限制范围提高精确度，而不是通过更多 maybe。

32. Source attribution controls commitment | Attribution 可以改变作者承诺

Smith argues that X.

is not the same as:

X is established.

Attribution tells the reader whose proposition it is.

33. “According to” does not make the claim true | according to 只标来源

Source distance is not evidence quality.

34. Report source confidence separately from your confidence | 来源立场与自己立场分开

Study A concludes that X causes Y; however, the observational design supports a causal interpretation less strongly than the authors’ wording suggests.

这就是 advanced critical writing。

35. Academic Phrasebank distinguishes distancing devices | Distance 可以通过结构表达

Examples include impersonal structures such as it has been suggested that… or attribution to a source. Use these when source ownership matters, not as automatic academic decoration.

36. Impersonal passive can hide responsibility | Impersonal passive 也有风险

It is widely believed that…

Who believes it?

What evidence?

If these matter, name the community or evidence.

37. “Research shows” is often too broad | research shows 常常太粗

Better:

Three recent controlled studies report…

or:

Across the reviewed literature…

when you truly synthesised the field.

38. “Studies have shown” can create citation fog | studies have shown 容易制造 fog

Tell the reader which studies and what exactly they show.

39. Match verbs to evidence type | Verb 要匹配 evidence type

Evidence	Safer language
survey preference	respondents reported / preferred
observational association	was associated with / correlated with
controlled causal design	increased / reduced, within design limits
qualitative interview	participants described / interviews suggest mechanism
case study	illustrates / documents a possible process

40. Causal language needs a causal gate | Causal wording 需要 gate

Before writing caused, led to, resulted in, produced, ask:

• Was temporal order established?
• Were major alternatives controlled or addressed?
• Was assignment/exposure design strong enough?
• Is the outcome measured appropriately?

41. Temporal sequence is weaker than causation | happened after ≠ caused by

Performance improved after the intervention.

does not automatically mean:

The intervention improved performance.

42. “Contributed to” can be useful for multi-causal systems | contributed to 适合多因系统

But it still makes a causal claim. Use evidence accordingly.

43. “May have contributed to” separates causal plausibility from proof | may have contributed to

Useful when the mechanism is plausible but alternative causes remain.

44. “Associated with” is not timid if causation is not established | association wording 不是“弱英语”

It may be exactly the strongest honest claim.

45. Generalisation has its own certainty | Generalisation 不是从 result 自动延伸

A result can be internally strong but externally narrow.

The intervention caused improvement in this trial may still not establish:

the intervention works for all learners.

46. Separate internal conclusion from external conclusion | 内部结论与外部推广分开

Within the study…

Whether the effect generalises to… remains uncertain.

47. Replication changes certainty differently from sample size | Replication 与 sample size 作用不同

A very large single sample improves precision.

Independent replication adds evidence that the result is not unique to one study context or analytic choice.

48. Consistency across sources strengthens some claims | 多来源一致可以增强 confidence

But only if sources are genuinely independent and comparable.

49. Five papers using the same dataset are not five replications | 同一数据多篇 paper ≠ 多次独立验证

50. Contradictory evidence should change the language | Counterevidence 应该改变措辞

Before:

The intervention improves retention.

After strong null replication:

Evidence for a retention benefit is mixed and may depend on implementation conditions.

51. Do not hide disagreement behind “some studies” | 不要用 some studies 模糊冲突

State where disagreement occurs.

52. Certainty can differ by outcome | 不同 outcome 可以不同 certainty

Evidence strongly supports faster task completion, while evidence for improved learning remains limited.

53. Certainty can differ by subgroup | 不同 subgroup 可以不同 certainty

The benefit is consistently observed among novice learners but remains uncertain among advanced learners.

54. Certainty can differ by timeframe | 时间会改变 certainty

Immediate effects are well supported; persistence beyond three months is not yet established.

55. Certainty can differ by mechanism | Mechanism 也可能比 effect 更不确定

The effect is replicated, but the proposed mechanism remains contested.

56. Do not make one weak dimension contaminate all dimensions | 一处不确定，不等于全部不确定

Bad:

Everything is uncertain.

Better:

The effect exists reliably, but its mechanism and long-term magnitude remain uncertain.

57. Statistical significance is not certainty | statistical significance 不是“确定”按钮

A statistically significant result can still be:

• small;
• imprecise;
• biased;
• non-generalizable;
• practically trivial.

58. Non-significance is not proof of no effect | non-significant ≠ no effect

It may reflect:

• small effect;
• low precision;
• small sample;
• high variability;
• true absence.

59. Equivalence requires appropriate evidence | “差不多”也需要设计

Do not rewrite:

No significant difference was detected.

as:

The two methods are equivalent.

60. Confidence intervals communicate precision, not rhetorical confidence | Confidence interval 不是作者“有多自信”

Use intervals to show a range of values compatible with the statistical model and data, according to the method used.

Do not turn the interval into vague prose if the numerical range matters.

61. Wide interval should weaken magnitude claims | 区间很宽，magnitude wording 要谨慎

The estimate is positive but imprecise.

62. Narrow interval can support precision within assumptions | 区间窄提高 estimate precision

It does not automatically remove design bias.

63. Precision is not validity | precise ≠ valid

A precisely estimated biased measure can still be wrong.

64. Statistical model assumptions belong to certainty | Model assumption 也影响 claim

If result depends strongly on one specification, language should reflect that fragility.

65. Sensitivity analysis can strengthen calibration | Sensitivity analysis 让 robust claim 更有依据

The conclusion remains similar across alternative specifications.

66. Robustness language should say what was tested | robustness 要具体

Robust to excluding two outliers.

Robust across three model specifications.

Consistent across two independent samples.

67. Risk language needs absolute context | Risk certainty 也需要基线

Risk doubled.

From what to what?

Relative language can sound more certain and dramatic than the practical change warrants.

68. Probability words can be misunderstood | likely / unlikely 不同读者理解不同

If a precise numerical probability is available and relevant, consider giving the number alongside the verbal label.

69. Do not invent numerical probability for verbal hedges | 不要把 may 硬对应成 30%

Unless a field defines a formal scale, natural-language probability terms do not carry universal fixed percentages.

70. Formal probability scales are domain-specific | 某些领域会定义概率词

When a professional standard defines likely or very likely, follow that standard and identify it.

71. Limitation language should identify consequence | Limitation 不只是列问题

Weak:

The sample was small.

Better:

The small sample makes the subgroup estimate imprecise, so the apparent difference should be treated as preliminary.

72. Limitation changes claim strength | Limitation 必须影响 interpretation

If you list a limitation but your conclusion does not change at all, ask whether you actually understood its consequence.

73. Not every limitation matters equally | Limitation 也有优先级

Focus on limitations that change:

• validity;
• precision;
• causal interpretation;
• generalisation;
• decision usefulness.

74. “Despite limitations” can become empty | despite these limitations 不能自动恢复强结论

Explain why the conclusion survives.

75. “Nevertheless” is not evidence | nevertheless 只是逻辑连接词

The sentence after it still needs support.

76. Discussion sections carry more hedging because interpretation increases | Discussion 通常比 Results 更需要 calibration

Manchester Academic Phrasebank notes that research discussion sections are especially rich in hedging when writers propose explanations.

77. Results section should not smuggle interpretation | Results 先写观察

Results:

Mean response time was 18% lower.

Discussion:

This reduction may reflect the simplified workflow.

78. Observation and interpretation need different verbs | Observation 与 interpretation 用词不同

records, shows, increased, decreased for direct measured states.

suggests, may reflect, is consistent with for interpretations.

79. Explanation is usually weaker than finding | “为什么”往往比“发生了什么”更不确定

Keep that asymmetry visible.

80. Abstracts are high-risk compression zones | Abstract 容易 overclaim

When compressing, do not delete:

• population;
• design;
• major uncertainty;
• causal boundary.

81. Titles are also claims | 标题也是 claim

Body:

Use was associated with higher performance.

Bad title:

Tool Improves Performance.

标题不能比正文更因果、更普遍或更肯定。

82. Executive summaries are also high-risk compression zones | Executive summary 也容易 certainty drift

Busy readers may never see the caveat buried later.

83. Conclusion certainty should match body certainty | Conclusion 不能突然 stronger

If body says:

may contribute

conclusion should not become:

causes.

84. Recommendations have a different modality system | Recommendation 不是 evidence statement

The evidence suggests X.

The team should do Y.

Second sentence adds a normative/decision layer.

85. Recommendation strength depends on evidence plus consequence | 建议强度 = evidence + risk + reversibility

Weak evidence may justify a low-risk pilot.

The same evidence may not justify irreversible full adoption.

86. “Should” needs a criterion | should 要说明为什么

The team should run a pilot because the expected learning value is high and the action is low-cost and reversible.

87. Distinguish epistemic and deontic modality | 区分“可能是真的”与“应该去做”

May be effective = knowledge status.

Should be implemented = action judgement.

88. Mandarin “可能” can map to several English choices | 中文“可能”不是只有 may

Depending on function:

• may;
• might;
• could;
• possibly;
• is likely to;
• is consistent with;
• cannot rule out.

先判断证据关系，不要先找翻译词。

89. Mandarin “应该” is even more dangerous | “应该”可能是三种英语

Expectation:

The results should arrive tomorrow.

Recommendation:

The trial should continue.

Inference:

This should be sufficient to explain the difference.

90. “一定” can signal confidence, logic or rhetoric | “一定”不能自动译 definitely

Ask whether the source means:

• logically necessary;
• very likely;
• strong personal insistence;
• policy requirement.

91. “证明了” is often overtranslated as proves | “证明了”要小心 proves

Possible English:

• shows;
• demonstrates;
• supports;
• provides evidence for;
• is consistent with.

根据研究设计选择，不根据中文表面词选择。

92. “由此可见” often overstates inference | “由此可见”不要自动 translate as it can be clearly seen

Better depending evidence:

These findings suggest…

Taken together, the evidence supports…

The result indicates…

93. “不排除” maps well to cannot rule out—but check function | 不排除

Use when the evidence cannot exclude a possibility.

Do not use it merely as vague politeness.

94. “有一定影响” is often too vague in English too | “有一定影响”需要解包

Ask:

• positive or negative?
• how large?
• which outcome?
• under what conditions?

95. “在一定程度上” should identify the degree if possible | to some extent 常常太空

Replace with the exact boundary when available.

96. Mandarin topic-comment structure can hide source ownership | 中文话题结构可能让 source ownership 变模糊

English academic writing often benefits from explicit attribution:

In this sample, the data suggest…

The authors argue…

Our analysis indicates…

97. Do not overuse “we believe” | we believe 不是万能 hedge

If evidence supports the sentence, state the evidence relationship.

98. “We hypothesise” is different | hypothesise 表示可测试 explanation

Use when proposing a hypothesis, not when reporting an established finding.

99. “We speculate” is deliberately weak | speculate 更弱

Appropriate for low-evidence explanatory possibility.

100. “We argue” is not uncertainty | argue 表示提出论点

An argument can be strongly supported or weakly supported. The verb itself does not settle certainty.

101. “We propose” can introduce model or mechanism | propose 适合模型/建议

Again, separate proposal status from evidence strength.

102. Over-hedging is a real error | Over-hedging 会让文章失去判断

Weak:

It might perhaps be possible that the findings could possibly indicate…

Reader cannot tell whether you have a conclusion.

103. One precise hedge is stronger than three vague hedges | 一个精准 hedge 胜过三个模糊 hedge

The findings suggest…

Then state the exact limitation.

104. Hedge the uncertain dimension, not the entire sentence | 只 hedge 真正不确定的部分

Weak:

There may possibly be some evidence that response time decreased.

If the decrease was directly measured:

Response time decreased; the cause of the reduction remains uncertain.

105. Do not hedge measured facts unnecessarily | 直接 observation 不需要全部变软

Thirty-two participants completed the survey.

Not:

Approximately 32 participants appear to have completed…

unless the record itself is uncertain.

106. Do not boost interpretations merely because the result is important | Importance ≠ certainty

A high-stakes result may require more caution, not less.

107. High consequence should increase transparency | Consequence 高时，unknowns 更要明确

State:

• what is known;
• what is likely;
• what is possible;
• what remains unknown;
• what action is justified despite uncertainty.

108. Uncertainty does not prevent decisions | 有 uncertainty 也可以行动

Decision language can be:

Evidence is incomplete, but a limited reversible pilot is justified because the potential benefit is material and the downside is controlled.

109. Scientific uncertainty and indecision are different | 不确定 ≠ 无法决定

Policy can operate under uncertainty using risk, reversibility and thresholds.

110. Certainty can be decision-specific | 对不同决策，证据门槛不同

The evidence needed to run a small pilot is lower than the evidence needed to mandate permanent adoption.

111. Build an evidence-to-language table | 建立 evidence → language 表

Evidence state	Language
direct measured fact	was / increased / decreased / records
replicated pattern	consistently shows / is well supported
moderate converging evidence	suggests / indicates / supports
plausible explanation	may reflect / is consistent with
open possibility	may / might / could
not excluded	cannot rule out
unresolved conflict	remains uncertain / evidence is mixed at…

112. But the table is not a phrase vending machine | 表格不是自动选词机器

Always check claim type, scope and discipline.

113. Build a claim-strength audit | Claim strength 审计

For every important sentence, mark:

1. claim type;
2. evidence source;
3. design strength;
4. scope;
5. uncertain dimension;
6. chosen verb/modality.

114. The “verb swap” test | Verb swap 测试

Try replacing:

suggests → shows → demonstrates.

At which step does evidence no longer justify the sentence?

115. The “scope removal” test | Scope removal 测试

Remove:

in this sample

Does the claim become too broad?

116. The “alternative explanation” test | 替代解释测试

If a plausible alternative remains, causal wording should reflect that.

117. The “replication” test | Replication 测试

Would you use the same strength if this were the only study?

118. The “directness” test | Directness 测试

Does evidence measure your exact outcome or a proxy?

119. The “precision” test | Precision 测试

Is the magnitude tightly estimated?

120. The “source distance” test | 来源距离测试

Are you reporting:

• your data;
• another study;
• a review;
• a news summary of a study?

Each step can increase distance from primary evidence.

121. The “headline” test | 标题测试

Is the title stronger than the strongest sentence the evidence supports?

122. The “abstract” test | 摘要测试

Did compression delete the qualifier?

123. The “conclusion” test | 结论测试

Did repeated drafting accidentally turn may into does?

124. The “Chinese back-translation” test | 中文回译测试

Translate your English claim back into Chinese.

Did suggests become “证明” in your mental interpretation?

If yes, your internal certainty is drifting.

125. The “reader reconstruction” test | 读者重建测试

Ask a reader:

How certain does this sentence sound to you?

Compare their answer with the intended evidence state.

126. Worked case: fictional feedback intervention | 完整案例：虚构 feedback intervention

All data below are fictional teaching material.

Study A: randomised trial, 240 learners. Guided feedback group scores 6 points higher on an immediate revision task.

Study B: independent replication, 310 learners. Difference is 4 points.

Study C: six-month follow-up, 180 learners. Difference narrows to 1 point and is imprecise.

Interviews: learners report that guided feedback helps them identify the next revision action.

127. Direct observation | 直接观察

In Studies A and B, the guided-feedback groups achieved higher immediate revision scores.

This can be stated directly.

128. Replicated immediate effect | 重复出现的即时效果

The two independent trials provide consistent evidence of an immediate performance advantage.

129. Magnitude uncertainty | 幅度仍有 uncertainty

The estimated size of the advantage varies across studies.

130. Long-term uncertainty | 长期效果不确定

Evidence for persistence is weak: at six months, the estimated difference is small and imprecise.

131. Mechanism language | 机制语言

Interviews:

Learner reports are consistent with the possibility that guided feedback helps by making the next action more explicit.

Do not write:

The interviews prove that action clarity is the mechanism.

132. Calibrated overall conclusion | 校准后的总判断

Guided feedback has a replicated short-term advantage for revision performance, but the durability of the effect remains uncertain. Learner reports suggest that action clarity may contribute to the benefit, although the mechanism is not established by the interview evidence alone.

133. Stronger but invalid conclusion | 看起来有力但不合法

Guided feedback clearly causes lasting improvement because learners understand what to do next.

Problems:

• lasting effect unsupported;
• mechanism overclaimed;
• clearly adds rhetoric, not evidence.

134. Over-hedged version | 过度 hedge

Guided feedback may perhaps possibly have some short-term benefit and could maybe be related to action clarity.

Problem: hides established immediate evidence.

135. Calibration means different strength inside one paragraph | 同一段里可以有不同 certainty

Fact strong.

Long-term claim weak.

Mechanism plausible.

Good academic prose makes those differences visible.

136. Practice A: choose the verb | 练习 A

A survey finds that 82% of respondents prefer Option B.

Choose:

proves / shows / suggests / demonstrates

for the proposition:

82% of respondents preferred B.

137. Practice B: do not overgeneralise | 练习 B

Source: 82% of 50 volunteer respondents preferred B.

Repair:

Users prefer B.

138. Practice C: causal calibration | 练习 C

Performance improved after a training programme with no control group.

Write one observation and one cautious causal interpretation.

139. Practice D: null result | 练习 D

A small study finds no statistically significant difference.

Do not write:

The methods are equal.

Write a calibrated version.

140. Practice E: mechanism | 练习 E

Interviews suggest one mechanism, but no design isolates it.

Write:

is consistent with…

141. Practice F: generalisation | 练习 F

One high-resource school shows a strong effect.

Write internal conclusion + external uncertainty.

142. Practice G: over-hedging | 练习 G

Repair:

It may perhaps be possible that…

143. Practice H: over-boosting | 练习 H

Repair:

It is obvious that the results clearly prove…

144. Practice I: Chinese “证明了” | 练习 I

Chinese intent:

数据证明了两组平均时间不同。

If the numbers directly record a difference but do not establish cause, write a precise English version.

145. Practice J: Chinese “应该” | 练习 J

Translate three functions:

• expected result;
• recommendation;
• logical inference.

146. Practice K: limitation consequence | 练习 K

Source limitation: only 18 participants.

Write why that matters to certainty.

147. Practice L: mixed certainty sentence | 练习 L

Direction replicated; size uncertain; long-term effect unknown.

Write one sentence preserving all three states.

148. Model answers | 示范解析

A: The survey shows that 82% of respondents preferred Option B. “Shows” is appropriate for the directly recorded response pattern, not for a broader claim about all users.

B: Among the 50 volunteer respondents, 82% preferred Option B; the result should not be assumed to represent all users.

C: Average performance increased after the training programme. The timing is consistent with a training effect, but the absence of a control group prevents a firm causal conclusion.

D: The study did not detect a clear difference between the methods, but the estimate is too imprecise to establish equivalence.

E: The interview pattern is consistent with the possibility that learners benefit from clearer next-action guidance.

F: The intervention produced a strong improvement in this school; whether the same effect would occur in lower-support settings remains uncertain.

G: The result may reflect…

H: The results provide strong evidence that… only if evidence is genuinely strong; otherwise use a weaker form.

I: The data show a difference in mean completion time between the two groups.

J expectation: The result should be available tomorrow. Recommendation: The team should repeat the test. Inference: This should be sufficient to explain the discrepancy.

K: The small sample makes the estimate imprecise, so the apparent subgroup difference should be treated as preliminary.

L: The direction of the effect is consistent across replications, but its magnitude remains imprecise and long-term persistence has not yet been established.

149. The 20-minute certainty drill | 20 分钟 calibration 训练

Time	Task
3 min	read one result/discussion extract
4 min	mark direct facts vs interpretations
4 min	map uncertain dimensions
5 min	rewrite with calibrated language
4 min	audit scope/verb/causal wording

150. The 45-minute growth session | 45 分钟增长模式

Time	Task
8 min	evidence-state profile
8 min	verb/modality ladder
10 min	results vs interpretation rewrite
10 min	limitations/generalisation
9 min	Chinese-English calibration audit

151. The 90-minute deep session | 90 分钟深度模式

Time	Task
15 min	read 3–4 sources on one claim
15 min	map direction/magnitude/causation/scope/persistence
15 min	write strong/medium/weak versions
15 min	counterevidence and replication audit
15 min	abstract + conclusion calibration
15 min	full language audit

152. Seven-day certainty cycle | 七天训练循环

1. Day 1: modal verbs and lexical verbs.
2. Day 2: source attribution and reporting verbs.
3. Day 3: causal language.
4. Day 4: statistics, precision and null results.
5. Day 5: limitations, scope and generalisation.
6. Day 6: Chinese-English certainty transfer.
7. Day 7: unseen full calibration benchmark.

153. Twelve-week C1–C2 certainty progression | 12 周路线

Weeks	Focus	Output
1–2	certainty dimensions + modality	calibration tables
3–4	reporting verbs + attribution	source-based paragraphs
5–6	causal/alternative explanation language	discussion sections
7–8	statistics, precision, robustness	results-to-discussion rewrites
9–10	limitations, scope, generalisation	calibrated conclusions
11–12	bilingual independent transfer	full academic portfolio

154. Monthly benchmark | 每月基准任务

Choose one unfamiliar academic article and produce:

1. five direct findings;
2. five interpretations;
3. one evidence-state profile;
4. one causal-language audit;
5. one scope/generalisation audit;
6. one limitations-consequence paragraph;
7. one 150-word abstract rewrite;
8. one 200-word conclusion rewrite;
9. one Chinese explanation preserving the English certainty states;
10. one English reconstruction from the Chinese without looking at the original.

155. Evaluate more than grammar | 不要只看 grammar

Track:

• certainty accuracy;
• claim-evidence match;
• causal calibration;
• scope control;
• source attribution;
• limitation consequence;
• precision language;
• over-hedging;
• over-boosting;
• bilingual meaning preservation.

156. First weak-link diagnosis | 第一个薄弱环节

Symptom	Likely weak link	Repair
everything uses may	no evidence-state distinctions	map dimensions first
everything sounds certain	boosting/translation drift	verb + scope audit
cause language from association	causal gate missing	alternative-explanation test
limitations listed but ignored	consequence link missing	state how limitation changes claim
abstract stronger than body	compression drift	restore qualifiers
“research shows” everywhere	source resolution weak	name evidence type/source cluster
Chinese “证明” becomes proves	surface translation	rebuild from evidence state
hedges stacked repeatedly	uncertainty not diagnosed	hedge exact dimension only

157. What not to do | 不要这样控制 certainty

• Do not map every hedge to a fixed probability unless your field formally defines one.
• Do not use proves because the Chinese source says 证明.
• Do not use clearly/obviously as substitutes for evidence.
• Do not hedge direct measured facts unnecessarily.
• Do not use shows for causal claims your design cannot establish.
• Do not interpret non-significance as equivalence.
• Do not let a narrow sample become a universal conclusion.
• Do not bury critical uncertainty outside the abstract/title.
• Do not stack may/perhaps/possibly to sound academic.
• Do not let recommendation language masquerade as evidence language.

158. Research and reference floor | 研究与参考基础

• Council of Europe | CEFR Companion Volume — propositional precision, including precise qualification of certainty/uncertainty and finer shades of meaning at C1–C2.
• Council of Europe | CEFR Global Scale — C2 precision and differentiation of finer shades of meaning.
• University of Manchester Academic Phrasebank | Being Cautious — hedging and epistemological strength in academic claims.
• Academic Phrasebank | Referring to Sources — source attribution and reporting language.
• Academic Phrasebank | Being Critical — disciplined evaluation of claims and limitations.
• University of Toronto | Critical Reading Towards Critical Writing — evaluating claims, evidence, methods, assumptions and alternative interpretations.
• George Mason University | Counterarguments — useful for matching certainty to surviving objections.

159. Canonical eduKate routes | eduKate 主页面路由

• How Modality and Certainty Calibrate English Claims
• Lesson 009 | Evidence-Based Reports and Recommendations
• Lesson 011 | Multi-Source Synthesis
• Lesson 012 | Literature Review as Scholarly Conversation
• Lesson 013 | Academic Argument That Survives Counterevidence

160. SEO language map | 本课自然覆盖的搜索意图

This lesson naturally serves learners searching for academic hedging, hedging in academic writing, epistemic stance, certainty in academic writing, cautious language, modality academic English, reporting verbs academic writing, causal language, research limitations writing, academic certainty words, C1 academic writing, C2 academic English, academic English for Chinese speakers, English for Mandarin speakers, 学术英语 hedge, 学术英语确定程度, 学术写作语气, 英语不确定性表达, 研究结论怎么写, suggests shows demonstrates 区别, may might academic writing, 中文母语学术英语 and C1 C2 academic writing.

161. The one-page certainty-calibration operating system | 一页 certainty calibration 系统

1. Identify claim type. Observation / interpretation / cause / generalisation / recommendation。
2. Map the evidence state. Direction / magnitude / precision / causation / scope / persistence / mechanism。
3. Separate fact from interpretation. 直接 observation 可以直接写。
4. Check alternative explanations. causal wording 前必须检查。
5. Check replication and independence. 不把同一数据当多次验证。
6. Check scope. 人群、情境、时间。
7. Check precision. estimate 是否稳定。
8. Choose reporting verb. records / shows / suggests / supports / is consistent with。
9. Add exact qualifier. hedge 真正不确定的维度。
10. State limitation consequence. limitation 必须改变 interpretation。
11. Protect source ownership. 谁说的？谁的数据？
12. Audit title/abstract/conclusion. 压缩时不丢 certainty。
13. Audit Chinese-English transfer. 不把 证明 自动变 proves，不把 应该 全部变 should。
14. Remove stacked hedges. 精准，不模糊。
15. State what remains unknown. uncertainty 有具体对象。
16. State what evidence would strengthen the claim. 让 claim 可更新。

162. Final assignment | 最终作业

Choose one research article or evidence-based professional report. Use a source you are allowed to access and cite.

Complete this chain:

1. Extract five direct findings.
2. Extract five interpretations.
3. Classify each claim by type.
4. Build an evidence-state profile for the main claim.
5. Mark every causal verb.
6. Mark every reporting verb.
7. Mark every modal/hedging device.
8. Identify one place where the source may overclaim.
9. Identify one place where the source may be unnecessarily cautious.
10. Rewrite one paragraph with better calibration.
11. Rewrite the abstract in 120–180 words without certainty drift.
12. Write a Chinese explanation preserving all major uncertainty states.
13. Rebuild the explanation in English without looking at the original.
14. Compare the two English versions for strength drift.
15. Write the strongest defensible one-sentence conclusion.
16. Write one sentence explaining what remains uncertain.
17. Write one sentence naming the next evidence that would most increase confidence.

Then ask:

If a reader changed only my modal verbs and reporting verbs, could they make my evidence sound much stronger or weaker than it really is?

如果别人只改我的 modal verbs 和 reporting verbs，就能让同一份证据听起来强很多或弱很多吗？

If yes, those words are carrying epistemic weight. Choose them deliberately.

---

Continue | 继续

The next lesson will move from sentence-level certainty into full research-section control: how to keep Methods, Results and Discussion doing different jobs so observation, procedure and interpretation do not leak into one another.

下一课把 certainty control 放进完整 research architecture：怎样让 Methods、Results 与 Discussion 各自完成自己的任务，避免 procedure、observation 与 interpretation 混在一起。

Next: EDKS-ADV-ZH-0015 · Lesson No.015 · Keep Methods, Results and Discussion Doing Different Jobs · 让方法、结果与讨论各司其职

Back to Advanced English Chinese Edition Hub · 返回高级英语中文版主页