Series ID: EDKS-ADV-ZH-0011 · How to Learn Advanced English (Chinese Edition) · Lesson No.011 · C1 → C2

Synthesize Multiple Sources Without Losing Their Differences | 综合多来源而不抹掉分歧

Synthesis is not the art of making several sources sound as if they all said the same thing.

综合，不是把几份来源写成“大家其实都一样”。

Real synthesis is harder. You must discover where sources genuinely converge, where they answer different questions, where methods produce different kinds of evidence, where populations or time periods differ, where one source qualifies another, and where disagreement remains unresolved.

真正的 synthesis 更难：你要判断哪些来源真的一致，哪些其实在回答不同问题；哪些方法产生的是不同类型证据；哪些研究对象、时间范围或定义不同；一个来源怎样限制另一个来源；以及哪些分歧目前仍然无法解决。

Then you have to build a new text whose structure belongs to your question, not to the order in which you found the sources.

最后，你要写出一个新的文本。它的结构由你的问题控制，而不是由“我先读了 Source A，再读 Source B，再读 Source C”控制。

---

Before we begin: this lesson owns advanced multi-source synthesis | 开始之前：本课负责高级多来源综合

eduKate Sengkang already has strong mechanism and school-stage owners for synthesis, source credibility and comparative reading. They remain intact.

eduKate Sengkang 已有 synthesis、source credibility、comparative reading 等主页面。本课不抢这些 owner。

• How Students Synthesize Evidence Into a Larger Meaning — general synthesis mechanism.
• Secondary 3 English Learning Guide | Synthesis Across Sources and Viewpoints — school-stage owner.
• Secondary 3 English Learning Guide | Comparative Reading and Multiple-Text Judgement — comparative-reading owner.
• Secondary 3 English Learning Guide | Source Credibility, Reliability and Evidence Evaluation — stage-specific evidence-quality owner.
• Lesson 004 | Read Deeply, Think Clearly, Write Precisely — deep reading and reasoning.
• Lesson 009 | Write Evidence-Based Reports and Recommendations — evidence-to-action writing.
• Lesson 010 | Adapt Complex English for Different Audiences — mediation and meaning preservation.

Lesson 011 owns the C1–C2 integration layer: how to reconstruct a field of evidence from multiple sources while preserving genuine differences.

第 011 课负责：怎样从多个来源重建一个证据领域，同时不抹掉真正差异。

1. CEFR places multi-source reconstruction near the top of proficiency | CEFR 把多来源重建放在高水平能力里

The CEFR C2 global descriptor includes the ability to summarise information from different spoken and written sources and reconstruct arguments and accounts in a coherent presentation. CEFR mediation descriptors also include comparing, contrasting and synthesising information and viewpoints from academic and professional publications.

CEFR 的 C2 global descriptor 明确提到：从不同口头与书面来源总结信息，并把 arguments 与 accounts 重建成 coherent presentation。Mediation descriptors 也包含 compare、contrast 和 synthesise academic/professional sources 的能力。

Council of Europe | CEFR Global Scale

Council of Europe | Mediation

2. A source list is not synthesis | 来源清单不是 synthesis

Weak:

Smith says X. Jones says Y. Chen says Z. Kumar says W.

This shows that you found sources. It does not yet show what the sources mean together.

真正的 synthesis 会问：

• Which sources address the same question?
• Which use comparable definitions?
• Which methods are compatible?
• Where do findings converge?
• Where do they diverge?
• Can the divergence be explained?
• What remains unresolved?

3. George Mason’s Writing Center gives a useful warning | 文献综述不是“一篇一篇摘要”

George Mason University’s Writing Center emphasises that a literature review is not a sequence of source summaries. It should be organised by topics, themes, approaches or responses to the research question, so that sources are examined in relation to one another.

George Mason University Writing Center | Writing a Literature Review

这和我们的核心原则一致：source 是 evidence node，不是文章结构。

4. Purdue frames scholarship as a conversation | 学术来源之间存在“对话”

Purdue’s graduate-writing materials describe synthesis as noticing how scholars build on, dispute or depart from one another, and suggest keeping notes that help identify themes, patterns, disagreements and changes over time.

Purdue OWL | Active Reading as a Graduate Student

所以你不是把四个作者排队。你是在画出他们之间的关系。

5. Start with one controlling question | 先确定一个控制问题

Before reading deeply, write:

What exactly am I trying to know?

Examples:

• Does retrieval practice improve long-term retention?
• Under what conditions does remote work improve productivity?
• Why do studies disagree about class size?
• What evidence supports the proposed policy?
• Which implementation risks recur across reports?

问题越模糊，source pile 越容易变成信息仓库。

6. One topic can contain several different questions | 同一个 topic 里可能有多个问题

“AI in education” may contain:

• Does it improve learning?
• Does it save teacher time?
• Does it change assessment validity?
• Does it widen or reduce access gaps?
• What risks arise from inaccurate output?

如果 Source A 回答 learning，Source B 回答 workload，就不能简单写“研究结果矛盾”。

7. The first synthesis skill is question alignment | 第一项技能：对齐问题

Before comparing results, ask:

Are these sources actually answering the same question?

很多假冲突来自问题不一致。

8. Build a source matrix | 建立 source matrix

Source	Question	Population/context	Method	Main finding	Limit
A	learning outcome	first-year students	controlled experiment	moderate benefit	short follow-up
B	student preference	volunteer users	survey	high satisfaction	self-selection
C	teacher workload	three schools	interviews	mixed workload effects	small context
D	long-term outcome	multiple cohorts	longitudinal	benefit weakens	non-random exposure

Source matrix 的目标不是把研究变成表格，而是让可比较维度同时可见。

9. A matrix protects you from memory blending | Matrix 防止来源在脑中混在一起

After reading ten sources, learners often remember:

“Research says…”

但已经忘了：

• which source;
• which group;
• which measure;
• which limitation;
• which result.

高级 synthesis 需要 traceability。

10. Separate claim from evidence type | 主张与证据类型分开

A survey can provide evidence about reported preferences.

An experiment can provide stronger evidence about causal effects under controlled conditions.

An interview can reveal mechanism, experience or implementation constraints.

A longitudinal study can reveal persistence and change over time.

不同证据不是简单排成“强 → 弱”；它们往往回答不同问题。

11. Do not ask one method to do another method’s job | 不要让一种方法承担另一种方法的任务

“95% of respondents liked it” does not establish learning.

“Test scores improved” does not explain why users hated the process.

“Teachers reported heavy workload” does not quantify system-wide cost.

Synthesis begins by respecting what each source can and cannot establish.

12. Operational definitions can create apparent disagreement | 定义不同会制造“表面冲突”

Source A defines success as test score.

Source B defines success as task completion.

Source C defines success as user satisfaction.

They can produce different results without truly contradicting one another.

13. Put definitions into the matrix | 把定义写进 matrix

Construct	Source A	Source B	Source C
engagement	attendance	self-report interest	time on task
performance	exam score	task accuracy	manager rating

同一个词可能隐藏不同 measurement。

14. Population matters | 人群不同，结果可能合理不同

A method may help beginners but not experts.

A policy may work in high-resource schools but fail in low-resource settings.

A treatment may affect one age group differently from another.

所以：

different result ≠ bad research.

15. Time matters | 时间尺度不同会改变结论

Immediate effect:

performance improves today.

Delayed effect:

benefit disappears after one month.

Both can be true.

16. Context matters | 情境也会改变结果

Laboratory, classroom, workplace and public-system contexts impose different constraints.

不要把一个 tightly controlled result 直接当作 universal implementation result。

17. Method quality matters—but method labels are not enough | 方法质量重要，但不能只看标签

“Randomised trial” sounds strong, but you still need to inspect:

• sample size;
• attrition;
• measurement validity;
• implementation fidelity;
• follow-up length;
• analysis choices.

同样，“qualitative” 不等于 weak。方法要与问题匹配。

18. Source prestige does not remove the need to read | 权威来源也要读内容

A strong journal, university or institution can improve baseline trust, but it does not tell you whether a particular claim is supported by the particular evidence you need.

19. Publication type changes what you can infer | 出版类型改变证据角色

Different roles:

• primary empirical study;
• systematic review;
• meta-analysis;
• narrative review;
• policy report;
• commentary;
• news report;
• company white paper.

不要把 commentary 和 empirical study 写成同一种 evidence。

20. Secondary sources can be powerful maps | Secondary source 可以是地图

A high-quality review can help you see:

• major themes;
• methods used;
• areas of consensus;
• contradictions;
• research gaps.

但重要结论仍要知道 review 如何选、评、综合 source。

21. Systematic review and meta-analysis are not synonyms | Systematic review 与 meta-analysis 不同

A systematic review uses an explicit process to identify and evaluate relevant studies.

A meta-analysis statistically combines compatible quantitative results.

Some systematic reviews include meta-analysis; some do not.

不要把 “meta-analysis” 当作“高级 review”的同义词。

22. Compatibility matters before pooling | 合并之前先问能不能合并

If studies measure different outcomes, populations or interventions, a single combined number may obscure meaningful differences.

23. Agreement can occur at one level and disagreement at another | 同意与不同意可以同时存在

Sources may agree:

there is an effect.

but disagree:

how large it is.

or:

why it occurs.

or:

whether it lasts.

24. Diagnose the level of disagreement | 诊断分歧层级

Level	Question
Fact	Did the event/result occur?
Magnitude	How large is it?
Mechanism	Why did it happen?
Scope	Where/for whom does it apply?
Persistence	How long does it last?
Value	Is the trade-off acceptable?
Action	What should be done?

高级 synthesis 不说 “the literature is mixed” 就结束。它要说明 mixed where?

25. “Mixed evidence” is often too vague | mixed evidence 常常太模糊

Better:

Studies consistently report a short-term improvement, but evidence for long-term retention is inconsistent.

这才有信息。

26. “Studies disagree” can hide methodological structure | “研究不同意”可能隐藏方法结构

Example:

Experimental studies tend to report a benefit under controlled conditions, whereas field studies show more variable effects when implementation quality differs.

分歧开始变成可解释模式。

27. Build a disagreement map | 建立 disagreement map

Source	Claim	Confidence	Possible reason for difference
A	strong positive effect	moderate	high-support setting
B	small effect	high	larger diverse sample
C	no delayed effect	moderate	longer follow-up
D	benefit only for beginners	moderate	subgroup interaction

28. Preserve minority findings when they are informative | 少数结果不能自动删除

If eight sources report X and two report Y, do not automatically write:

Research proves X.

Ask whether the two different studies involve:

• better methods;
• different populations;
• longer follow-up;
• a boundary condition.

29. Consensus is not a vote count | Consensus 不是简单数票

Ten weak repetitions do not automatically outweigh one strong contradictory study.

Evidence synthesis requires weighting, not counting alone.

30. Weighting should be explicit | Evidence weighting 要透明

Possible dimensions:

• design fit;
• sample relevance;
• measurement quality;
• replication;
• precision;
• follow-up length;
• independence of source;
• risk of bias;
• directness to your question.

31. Do not invent a fake universal score | 不要自己发明万能 evidence score

A numerical score can create false precision if the weighting scheme is arbitrary.

For many student/professional synthesis tasks, a transparent qualitative comparison is better.

32. Directness matters | 与问题的直接程度很重要

A beautifully designed study of a different population may be less directly useful than a moderate study in the population you actually need to understand.

33. Recency matters only when the underlying system changes | 新不一定更好

Recent evidence deserves extra attention when technology, policy, clinical practice, language use or context has changed.

For stable historical or theoretical questions, older foundational work may remain central.

34. Chronology can reveal why studies differ | 时间顺序有时能解释分歧

Later studies may:

• use better measures;
• include broader populations;
• respond to earlier criticism;
• study a changed environment.

不要只写 newest = correct。

35. Citation chains can reveal dependence | Citation chain 可能显示来源依赖

If five articles all cite the same original small study, you may not have five independent pieces of evidence.

36. Repeated claims can originate from one source | 一个说法可以被“重复成共识”

Trace important claims backwards when possible.

Ask:

Where did this claim first become evidence?

37. Source independence matters | 来源独立性重要

Company reports that all use the same internal dataset are not independent replications.

38. Funding and incentives are context, not automatic disqualification | 利益关系是 context，不是自动判死刑

A conflict of interest should increase scrutiny. It does not by itself prove the result false.

Inspect methods, transparency and replication.

39. Negative evidence and absence of evidence differ | negative evidence 与 no evidence 不同

No study found an effect may mean evidence against an effect.

No study examined the effect means an evidence gap.

这两个 synthesis conclusion 完全不同。

40. “No significant difference” does not automatically mean “same” | no significant difference 不等于相同

The result could reflect insufficient precision, small sample or genuinely small differences.

报告来源时不要把 statistical non-significance 自动改写成 equivalence。

41. Confidence belongs in synthesis | 综合结论也要有 confidence

Possible forms:

• strongly supported across contexts;
• supported but context-dependent;
• promising but preliminary;
• mixed with identifiable moderators;
• unresolved;
• insufficiently studied.

42. Build a claim matrix | 从 source matrix 升级到 claim matrix

Claim	A	B	C	D
short-term benefit	supports	supports	supports	supports
long-term benefit	not tested	weak	does not support	supports subgroup only
low implementation cost	not tested	supports	contradicts	not tested

现在你开始按 claim 组织，而不是按 source 组织。

43. The claim matrix often reveals the paragraph structure | Claim matrix 会自然长出段落结构

Paragraph 1: short-term effect.

Paragraph 2: persistence.

Paragraph 3: implementation cost.

每段调用多个来源。

44. Your paragraph should have an analytical spine | 每段需要 analytical spine

Useful pattern:

SYNTHESIS CLAIM → SOURCE RELATION → DIFFERENCE → EXPLANATION → LIMIT → IMPLICATION.

45. Example synthesis paragraph | 示例 synthesis 段落

Across the four studies, the short-term effect is relatively consistent, but evidence for persistence is weaker. Studies A and B report immediate gains, while Study C finds that the difference is no longer clear at the delayed follow-up. Study D partly reconciles these findings by reporting persistence only among beginners. Taken together, the evidence supports a short-term benefit more strongly than a universal long-term effect, and suggests that starting level may be an important moderator.

注意：不是 A summary + B summary + C summary + D summary。

46. The paragraph begins with the relationship | 段落开头先给 source relationship

Across the four studies…

读者先知道“我们现在在比较什么”。

47. Source names should serve the analysis | 作者名字服务分析，不主导结构

Use author-prominent writing when the scholar/theory matters.

Use information-prominent writing when the pattern matters.

Academic Phrasebank notes that disciplines differ in how prominently authors appear.

University of Manchester Academic Phrasebank | Referring to Sources

48. Signal agreement precisely | 精准表达一致

• is consistent with;
• converges with;
• supports;
• replicates;
• extends;
• corroborates;
• reports a similar pattern.

这些词不是完全同义。

49. “Supports” is not “proves” | supports 不等于 proves

Keep evidence strength proportional.

50. Signal disagreement precisely | 精准表达分歧

• contrasts with;
• does not replicate;
• reports a smaller effect;
• finds no delayed effect;
• questions the proposed mechanism;
• applies a narrower interpretation.

Academic Phrasebank offers multiple ways of synthesising contrasting evidence; use them as language resources, not as templates to paste mechanically.

Academic Phrasebank | Compare and Contrast

51. Signal partial agreement | 部分一致很常见

Both studies identify the same direction of change, although they differ substantially in estimated magnitude.

52. Signal boundary conditions | 写出 boundary condition

The benefit appears consistently in novice groups but not in participants with high baseline proficiency.

这比 “results are mixed” 更有解释力。

53. Signal methodological explanations cautiously | 方法解释要谨慎

One possible explanation for the difference is the longer follow-up used in Study C.

不要直接写：

Study C is different because it used a longer follow-up.

除非 evidence 真能建立因果。

54. Use cautious language when reconciling conflict | 调和分歧也需要 hedge

Academic Phrasebank’s cautious-language resources are useful for separating plausible explanation from established explanation.

Academic Phrasebank | Being Cautious

55. Not every disagreement should be reconciled | 不是每个冲突都必须解释掉

Sometimes the correct synthesis is:

The available evidence does not yet explain why the two high-quality studies reach different conclusions.

高级写作允许 unresolved。

56. Do not manufacture harmony | 不要制造假和谐

Weak:

Although the studies differ, they all show the importance of the topic.

This often avoids the actual conflict.

57. Do not manufacture conflict either | 也不要为了“批判性”制造冲突

If two sources measure different outcomes, do not frame them as opposing camps merely to make the literature review dramatic.

58. Synthesis requires a taxonomy of relationships | 建立 source relationship taxonomy

Relationship	Question
Agreement	Do they support the same claim?
Extension	Does one add population/time/mechanism?
Qualification	Does one narrow another?
Contradiction	Do they make incompatible claims?
Method complement	Do they answer different parts with different methods?
Chronological development	Does later work revise earlier work?
Gap	What important question is still not answered?

59. Extension is one of the most valuable relations | Extension 常被忽略

Study B may not disagree with A. It may simply show:

A’s result also appears in another population.

or:

A’s effect disappears after six months.

Extension can strengthen or qualify.

60. Qualifying is not refuting | qualification 不等于 refutation

Source A:

The intervention works.

Source B:

It works only under high-support conditions.

B may qualify A rather than completely reject it.

61. Mechanism evidence can connect outcome studies | Mechanism evidence 可以连接 outcome studies

If A shows an effect and B explains a plausible process, synthesis can connect them—but do not claim B proves A’s mechanism unless the designs support that inference.

62. Qualitative evidence can explain implementation heterogeneity | 定性证据可以解释“为什么不同地点结果不同”

Example:

Quantitative studies show variable impact.

Interviews show that staff training quality differs dramatically.

This may suggest an implementation moderator.

63. Case studies can identify possibilities, not prevalence | Case study 可以揭示机制，但不自动代表普遍频率

Preserve that evidence role in synthesis.

64. Cross-sectional and longitudinal sources answer different time questions | 横断面与纵向研究不要混成一种

Cross-sectional:

What differs now?

Longitudinal:

What changes over time?

65. Observational and experimental evidence differ in causal leverage | observational 与 experimental 因果能力不同

Do not erase this difference in a combined paragraph.

66. The source matrix should include causal status | Matrix 加一列 causal status

Source	Design	Causal leverage	Main use
A	randomised trial	higher within design	effect estimate
B	observational cohort	limited by confounding	real-world association
C	interviews	not causal estimate	mechanism/experience

67. Synthesis can change your original question | 综合可能迫使你改问题

You begin:

Does X work?

After reading:

For whom, under what conditions and for how long does X work?

That is progress.

68. A better question often appears after disagreement | 分歧常常帮助形成更好的问题

Do not treat disagreement only as noise. It can reveal a moderator or hidden assumption.

69. Theme-based organisation | 按主题组织

Useful when sources address recurring dimensions:

• effectiveness;
• equity;
• cost;
• implementation;
• long-term sustainability.

70. Method-based organisation | 按方法组织

Useful when the methodological contrast itself explains the field:

• experimental studies;
• observational evidence;
• qualitative implementation studies.

71. Chronological organisation | 按时间组织

Useful when the field evolves:

• early theory;
• initial evidence;
• critical replication;
• modern refinement.

72. Population-based organisation | 按人群组织

Useful when effects differ:

• children;
• adolescents;
• adults;
• specialist groups.

73. Mechanism-based organisation | 按机制组织

Useful when competing explanations drive the literature.

74. Do not combine all organisational schemes at once | 不要一篇文章同时用五种主结构

Choose one primary spine. Use other dimensions inside sections.

75. The literature-review paragraph is not a source container | 文献综述段落不是 source container

A paragraph should make one analytical move.

Sources enter because they help complete that move.

76. Use topic sentences that synthesise | Topic sentence 直接综合

Weak:

Many studies have investigated remote work.

Stronger:

Evidence for productivity gains is strongest in roles with high task autonomy, while collaborative roles show more variable outcomes.

第二句已经给读者 synthesis result。

77. Use evidence clusters | 把相似证据 cluster 起来

Instead of three separate sentences:

A found X. B found X. C found X.

Try:

Three studies across different settings report the same directional effect, although the estimated magnitude varies.

Then cite the sources appropriately.

78. But do not cluster away important differences | Cluster 不能吞掉重要差别

If C has a different population or weaker method, say so.

79. Source boundaries must remain visible | 来源边界要清楚

George Mason’s Writing Center emphasises marking the boundary between your prose and source material. In synthesis, this becomes especially important because several sources may appear in one paragraph.

George Mason University Writing Center | Marking the Boundaries

80. Avoid citation fog | 避免 citation fog

Weak:

Several studies show that the effect is large, temporary and especially strong for beginners (A; B; C; D).

Which study supports which part?

Clarify the mapping.

81. Multi-citation sentences are useful when sources support the same proposition | 多来源 citation 要支持同一 proposition

If the studies support different parts, split or signal the relation.

82. Quote less when synthesis is the task | Synthesis 任务通常少 quote

Direct quotation preserves wording but can make multi-source integration harder.

Use quotation when:

• wording itself matters;
• a definition is contested;
• rhetorical language is evidence;
• the original formulation is analytically important.

83. Paraphrase does not mean synonym replacement | Paraphrase 不是同义词替换

Reconstruct the proposition, scope and certainty.

Then express it naturally.

84. Preserve reporting distance | 保留 reporting distance

Smith argues, the study reports, the data suggest and it is established carry different commitments.

85. Separate author claim from your synthesis | 作者结论与自己的 synthesis 分开

Example:

Study A interprets the change as a motivation effect. However, when A is read alongside B and C, workload and prior proficiency also emerge as plausible moderators.

第二句是你的 synthesis，不是 A 的原话。

86. Critical writing means inspecting, not attacking | Critical 不等于挑毛病

Academic Phrasebank describes being critical as questioning what you read and examining weaknesses, limitations or alternative interpretations.

Academic Phrasebank | Being Critical

你可以得出：

This study is useful but narrow.

而不是：

This study is bad.

87. Strength and limitation can coexist | 优点与限制可以同时存在

The study offers a strong causal design but a narrow population.

The survey provides broad coverage but relies on self-report.

这种句子更适合 evidence weighting。

88. Do not criticise a source for not doing a different study | 不要因为 source 没做另一项研究就随便批评

A qualitative study may intentionally pursue depth rather than prevalence.

Critique whether it succeeds at its own question.

89. Distinguish limitation from flaw | limitation 与 flaw 不同

A limitation may be inherent to scope.

A flaw may undermine validity.

不要把所有 caveat 都写成 fatal problem。

90. A synthesis conclusion must be narrower than its strongest unsupported leap | 最终结论不能超出证据

If evidence supports:

short-term benefit in supported settings

do not conclude:

the method improves outcomes everywhere.

91. Use layered conclusions | 分层写 synthesis conclusion

Layer 1: what is well supported.

Layer 2: what is conditionally supported.

Layer 3: what remains uncertain.

Layer 4: what evidence would reduce uncertainty.

92. Example layered conclusion | 示例分层结论

Across the reviewed studies, there is relatively consistent evidence of short-term improvement. The size and persistence of the effect appear to depend on baseline proficiency and implementation support. Evidence for long-term benefit remains limited because few studies include delayed follow-up. A useful next step would be a larger study that measures both immediate and delayed performance across different starting levels.

93. Synthesis should generate a better next question | 综合应该产生更好的下一问

If your conclusion merely repeats the original question, the synthesis may not have done enough work.

94. Literature gaps should be specific | “研究不足”要具体

Weak:

More research is needed.

Stronger:

Long-term outcomes remain uncertain because only one study follows participants beyond eight weeks.

95. Gap ≠ importance | gap 不自动等于值得研究

A topic can be unstudied because it is unimportant.

Explain why the gap matters to the question.

96. Contradiction can be a research gap | 未解释的冲突本身也是 gap

Two well-designed studies reach different conclusions under apparently similar conditions; the source of this divergence remains unresolved.

97. Missing population can be a gap | 缺少某人群也是 gap

But only if that population matters to the intended generalisation.

98. Missing outcome can be a gap | 缺少 outcome 也可能是 gap

Example:

Studies measure satisfaction but not learning.

99. Missing duration can be a gap | 时间不足也是 gap

Immediate post-test evidence cannot answer persistence.

100. Worked case: four fictional studies of an English-learning intervention | 完整案例：四个虚构高级英语研究

All studies below are fictional teaching materials.

Source A: 120 university learners were randomly assigned to standard practice or retrieval-based practice for six weeks. Retrieval group scored higher on an immediate transfer test. No delayed test.

Source B: 860 students across 12 schools were observed for one term. More frequent retrieval activities were associated with higher end-of-term scores, but prior achievement and teacher differences may confound the result.

Source C: 38 teachers were interviewed. Many reported that retrieval activities were useful when questions required explanation rather than simple recall; several reported high preparation burden.

Source D: 210 learners were followed for six months. The initial benefit was still visible among lower-baseline learners but was small among high-baseline learners.

101. First synthesis: what can we say strongly? | 第一层：什么结论比较强？

Source A provides causal evidence for an immediate benefit under its experimental conditions.

Source B adds broader real-world association but weaker causal leverage.

Source D adds persistence and subgroup information.

Source C adds implementation mechanism and workload trade-off.

102. What should not be combined into one number? | 什么不应该硬合成一个数字？

The four studies use different designs, outcomes and purposes. A single pooled effect would not be justified from this fictional description.

103. What converges? | 哪些地方 converges？

A, B and D all point toward a positive relationship between retrieval-oriented practice and performance.

C does not measure the same performance outcome, but its teacher reports are compatible with the possibility that task design matters.

104. What qualifies the positive story? | 什么限制了“正面故事”？

• A lacks delayed follow-up.
• B is observational.
• C identifies workload cost.
• D suggests baseline proficiency moderates persistence.

105. Model synthesis paragraph | 示例综合段落

The four fictional studies provide converging but non-equivalent evidence that retrieval-oriented practice can support performance. The randomised experiment offers the clearest evidence of an immediate effect, while the larger observational study shows that the same directional relationship appears across multiple schools but cannot rule out teacher and baseline differences. The six-month study further qualifies the claim by suggesting that persistence is stronger among learners with lower initial proficiency. Teacher interviews add an implementation constraint: perceived benefit depends partly on question design and may require additional preparation time. Taken together, the evidence supports retrieval practice as a promising strategy rather than a uniform effect that can be assumed across learners, time periods and implementation conditions.

106. Why this paragraph is synthesis | 为什么这是真 synthesis

It:

• makes one overall claim;
• gives each source a role;
• preserves design differences;
• preserves subgroup differences;
• adds implementation cost;
• ends with a calibrated conclusion.

107. A weaker paragraph | 弱版本

Source A found retrieval practice effective. Source B also found it effective. Source C said teachers liked it. Source D found it worked for some learners. Therefore retrieval practice is effective.

问题：

• method differences vanish;
• C is misrepresented;
• persistence disappears;
• conditionality disappears;
• conclusion becomes too broad.

108. Build a source-role sentence | 给每个 source 定角色

Before drafting, finish:

• A establishes…
• B extends…
• C explains/qualifies…
• D limits/refines…

这会减少 source-by-source writing。

109. Practice A: identify false conflict | 练习 A：找假冲突

Source 1 reports high satisfaction.

Source 2 reports no significant learning gain.

Are they contradictory?

Answer: not necessarily. They measure different outcomes.

110. Practice B: identify scope conflict | 练习 B：找 scope 差异

Source 1: benefit for beginners.

Source 2: no benefit among experts.

Possible synthesis:

The effect may depend on baseline proficiency.

111. Practice C: identify time conflict | 练习 C：找时间差异

Immediate test positive; six-month follow-up null.

Synthesis:

Evidence supports acquisition more strongly than persistence.

112. Practice D: identify method complement | 练习 D：方法互补

Experiment shows effect.

Interview identifies why implementation sometimes fails.

Do not choose one as “better”; use both for different questions.

113. Practice E: repair “research is mixed” | 练习 E：修 mixed evidence

Rewrite:

Research is mixed.

into a sentence that states where and why results differ.

114. Practice F: repair vote-count synthesis | 练习 F：修“数票式综合”

Seven weak studies support X; two strong studies do not.

Do not write “most studies support X” and stop.

Discuss design quality and directness.

115. Practice G: build a claim matrix | 练习 G：建立 claim matrix

Use four sources and three claims. Mark:

• supports;
• contradicts;
• not tested;
• qualifies.

116. Practice H: write one synthesis topic sentence | 练习 H：写 synthesis topic sentence

Do not name an author in the first sentence.

State the relationship across the field.

117. Practice I: preserve unresolved conflict | 练习 I：保留 unresolved

Two strong sources disagree and no clear moderator is visible.

Write a conclusion that does not force reconciliation.

118. Practice J: separate source claim from your inference | 练习 J：区分来源与自己的推论

Write two sentences:

1. what Source A claims;
2. what you infer after reading A with B and C.

119. Model answers | 示范解析

E: Studies consistently report an immediate improvement, but findings diverge at delayed follow-up, with longer studies showing weaker persistence.

F: Although a numerical majority of studies report a positive association, the two most direct controlled studies do not reproduce the effect, so the current evidence cannot be evaluated by study count alone.

H: Across the literature, evidence is strongest for short-term performance gains and weaker for durable transfer.

I: Two well-designed studies reach different conclusions under similar conditions, and the current evidence does not identify a clear explanation for the divergence.

J: Source A attributes the improvement to increased practice frequency. Read alongside B and C, however, the evidence also suggests that feedback quality and baseline proficiency may influence the observed effect.

120. The 20-minute synthesis drill | 20 分钟综合训练

Time	Task
4 min	read 3 short sources
4 min	fill source matrix
4 min	build one claim matrix
5 min	write one synthesis paragraph
3 min	audit scope/certainty/source boundaries

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

Time	Task
8 min	define controlling question
10 min	source matrix
8 min	relationship map
12 min	two synthesis paragraphs
7 min	critical/cautious language audit

122. The 90-minute deep session | 90 分钟深度训练

Time	Task
15 min	read 5–6 sources
15 min	question + source matrix
15 min	claim/disagreement map
20 min	theme-based synthesis draft
10 min	source-boundary/citation audit
15 min	conclusion + gap + next question

123. Seven-day synthesis cycle | 七天 synthesis 循环

1. Day 1: source matrix.
2. Day 2: definitions and measurement.
3. Day 3: population/time/context differences.
4. Day 4: disagreement diagnosis.
5. Day 5: evidence weighting.
6. Day 6: synthesis paragraphs.
7. Day 7: unseen multi-source benchmark.

124. Twelve-week C1–C2 synthesis progression | 12 周高级 synthesis 路线

Weeks	Focus	Evidence
1–2	source reconstruction + matrix	source maps
3–4	definitions, methods, populations	comparison grids
5–6	agreement, qualification, contradiction	relationship maps
7–8	evidence weighting + cautious judgement	claim matrices
9–10	literature-review structure	multi-paragraph synthesis
11–12	independent multi-source transfer	full synthesis portfolio

125. Monthly benchmark | 每月基准任务

Use four unseen sources that include:

• two apparently conflicting findings;
• one different method;
• one population difference;
• one important limitation.

Produce:

1. source matrix;
2. three-claim matrix;
3. relationship map;
4. two synthesis paragraphs;
5. 150-word overall conclusion;
6. one unresolved question;
7. one explanation you considered but cannot yet establish.

126. Measure more than citation count | 不要用“引用了几篇”评价 synthesis

Track:

• question alignment;
• source-role accuracy;
• method awareness;
• definition/measurement control;
• scope preservation;
• disagreement diagnosis;
• evidence weighting;
• source boundaries;
• paragraph-level integration;
• calibrated conclusion.

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

Symptom	Likely weak link	Repair
one paragraph per source	no claim matrix	organise by question/theme
everything looks contradictory	question/definition mismatch	align constructs first
everything looks consistent	differences erased	map scope/method/time
“research is mixed” repeated	no disagreement diagnosis	name level and moderator
conclusion too strong	certainty drift	layer support/uncertainty
many citations but unclear source roles	citation fog	map claim → source
criticism sounds hostile	evaluation language weak	critique method/scope precisely

128. What not to do | 不要这样做多来源综合

• Do not write one mini-summary per source.
• Do not call different outcomes “contradictory” automatically.
• Do not count studies without considering quality and directness.
• Do not pool incompatible results.
• Do not erase minority findings.
• Do not force reconciliation when conflict remains unresolved.
• Do not confuse lack of evidence with evidence of no effect.
• Do not let citation clusters obscure who supports what.
• Do not criticise sources for answering a different legitimate question.
• Do not conclude beyond the narrowest defensible synthesis.

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

• Council of Europe | CEFR Global Scale — C2 summarising information from different spoken and written sources and reconstructing arguments coherently.
• Council of Europe | Mediation — constructing meaning and conveying ideas/information.
• CEFR Companion Volume | Mediating a Text — comparing, contrasting and synthesising information/viewpoints in academic and professional texts.
• George Mason University Writing Center | Writing a Literature Review — synthesis organised by themes/questions rather than source-by-source summaries.
• George Mason University | Organizing Literature Reviews — select, arrange, group and order material across sources.
• Purdue OWL | Active Reading as a Graduate Student — connecting multiple sources, themes, disagreements and scholarly conversation.
• University of Manchester Academic Phrasebank | Referring to Sources — source integration and synthesising contrasting evidence.
• Academic Phrasebank | Being Critical — disciplined evaluation of source limitations and claims.
• Academic Phrasebank | Being Cautious — calibrated interpretation when explanations remain uncertain.

130. Canonical eduKate routes | eduKate 主页面路由

• How Students Synthesize Evidence Into a Larger Meaning
• Secondary 3 Synthesis Across Sources and Viewpoints
• Secondary 3 Comparative Reading and Multiple-Text Judgement
• Source Credibility, Reliability and Evidence Evaluation
• Lesson 004 | Read Deeply, Think Clearly, Write Precisely
• Lesson 009 | Write Evidence-Based Reports and Recommendations
• Lesson 010 | Adapt Complex English for Different Audiences

131. SEO language map | 本课自然覆盖的搜索意图

This lesson naturally serves learners searching for how to synthesize sources, synthesis writing, multiple source synthesis, literature review synthesis, compare sources, integrate sources in academic writing, C1 academic writing, C2 academic English, source matrix, synthesis matrix, literature review for Chinese students, academic English for Mandarin speakers, evidence synthesis, conflicting research findings, critical literature review, how to write synthesis paragraphs, 高级英语综合, 多来源综合, 文献综述怎么写, 英语文献综述, 来源比较, 研究结果冲突, source matrix, synthesis matrix, 中文母语学术英语 and C1 C2 academic writing.

132. The one-page multi-source synthesis operating system | 一页多来源综合操作系统

1. Write one controlling question. 一个控制问题。
2. Reconstruct each source accurately. 先读懂每个来源。
3. Build a source matrix. 问题、人群、方法、结果、限制。
4. Align definitions and measures. 先确认是否在测同一件事。
5. Build a claim matrix. 按 claim，不按 source。
6. Diagnose disagreement. fact / magnitude / mechanism / scope / persistence / value / action。
7. Weight evidence transparently. 看 directness、design、measurement、replication。
8. Choose one organising spine. theme / method / time / population / mechanism。
9. Write synthesis topic sentences. 先给关系，再给来源。
10. Preserve source boundaries. 谁说什么必须清楚。
11. Use cautious language for explanations. plausible 不等于 established。
12. Layer the conclusion. supported / conditional / uncertain / next evidence。
13. State unresolved conflict honestly. 不制造假和谐。
14. Generate the next better question. 综合以后问题应该更好。

133. Final assignment | 最终作业

Choose four to six credible sources on one serious question. Use sources you are permitted to access and cite.

选择 4–6 个可信来源，围绕一个真实问题。使用允许访问和引用的材料。

Complete this chain:

1. Write the controlling question in one sentence.
2. Build a source matrix containing question, population, method, finding and limitation.
3. Identify at least three important claims.
4. Build a claim matrix.
5. Mark every apparent disagreement.
6. Classify each disagreement by level.
7. Identify one difference that is probably explained by population, method, measurement or time.
8. Identify one difference you cannot responsibly explain.
9. Write two synthesis paragraphs organised by claim/theme.
10. Write a 150–250 word overall synthesis conclusion.
11. State what is well supported.
12. State what is conditionally supported.
13. State what remains unresolved.
14. State one specific next question or evidence need.
15. Audit every important sentence: which source supports which part?

Then ask:

If I removed the author names and citations, would the structure of my writing still show the relationships among the evidence?

如果把作者名字和引用暂时拿掉，我的文章结构还看得出证据之间的关系吗？

If yes, you are synthesising.

If no, you may still be listing sources.

---

Continue | 继续

The next lesson will move from synthesis into a more demanding academic skill: building a literature review that maps a field, identifies schools of thought, tracks change over time and shows exactly where your own research question enters the conversation.

下一课进入更高级的学术写作：怎样把 literature review 写成“领域地图”，识别主要观点群、追踪研究怎样变化，并准确指出自己的问题在哪里进入学术对话。

Next: EDKS-ADV-ZH-0012 · Lesson No.012 · Build a Literature Review as a Scholarly Conversation · 把文献综述写成学术对话

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