Series ID: EDKS-ADV-ZH-0016 · How to Learn Advanced English (Chinese Edition) · Lesson No.016 · C1 → C2

Write an Abstract That Compresses a Study Without Distorting It | 把完整研究压缩成不失真的摘要

An abstract is not a trailer that makes the study sound exciting. It is a compressed model of the study that must remain faithful when most of the paper has disappeared.

Abstract 不是电影预告片。它不是负责“把研究说得更精彩”。它是一份高度压缩的研究模型：当大部分正文都被删掉以后，剩下的部分仍然必须保持真实。

This makes abstract writing unusually difficult. You are not simply shortening sentences. You are making decisions about what information is structurally necessary.

你必须判断：

What problem must remain?
Which design feature changes interpretation?
Which sample boundary must survive?
Which result deserves the limited space?
Which number is essential?
Which limitation changes the conclusion?
How strong can the final sentence honestly be?

The core task is:

full study → invariant research state → compressed abstract.

完整研究 → 不可丢失的研究状态 → 摘要。

Before we begin: the abstract is where Lessons 009–015 meet | Abstract 是前面多课能力的压缩测试

A strong abstract depends on skills already built across this series:

Lesson 009 — evidence-based reporting.
Lesson 010 — audience adaptation without meaning loss.
Lesson 011 — source synthesis.
Lesson 013 — defensible academic argument.
Lesson 014 — certainty calibration.
Lesson 015 — Methods/Results/Discussion job separation.

Lesson 016 asks whether you can preserve all those relationships under severe space pressure.

1. Purdue: an effective abstract must stand alone | Abstract 必须能够独立工作

Purdue OWL’s current scientific-abstract guidance describes effective abstracts as concise, coherent, stand-alone summaries that let readers understand the paper’s essential content without reading the full article. The guidance identifies purpose/research question, methods, findings and conclusions/recommendations as common abstract components.

Purdue OWL | Writing Scientific Abstracts

2. Elsevier: concise, factual, purpose → principal results → major conclusions | 摘要必须 factual

Elsevier’s current author guidance states that an abstract should be concise and factual, briefly state the purpose of the research, principal results and major conclusions, and be able to stand alone because it is often presented separately from the article.

Elsevier | Guide for Authors

3. Nature treats title and abstract as core manuscript skills | Abstract 也是 publication interface

Nature Masterclasses includes title and abstract writing in its research-paper course because these short forms determine how editors and readers first encounter the work.

Nature Masterclasses | Writing a Research Paper

4. Reporting guidelines can require design-specific abstract content | 不同研究设计可能有专门 checklist

PRISMA 2020 for Abstracts, for example, asks systematic reviews to include objectives, eligibility criteria, information sources, risk-of-bias methods, synthesis methods, included studies, main synthesis results, limitations of evidence, interpretation, funding and registration.

PRISMA 2020 for Abstracts

This is an important lesson:

There is no single abstract template for every research design.

5. First rule: read the target venue’s abstract instructions | 第一条规则：先看 venue

Before drafting, check:

word limit;
structured or unstructured format;
required headings;
whether references are allowed;
whether trial registration/funding must appear;
keyword requirements;
whether numerical results are expected;
whether graphical abstract/highlights are separate.

Do not learn one “universal 250-word formula” and apply it everywhere.

6. Structured and unstructured abstracts | Structured / unstructured 两种常见形式

Structured abstract:

Background / Objective
Methods
Results
Conclusion

Unstructured abstract:

One or more paragraphs that perform the same intellectual jobs without visible headings.

7. Structured does not automatically mean better | 有 heading 不自动更高级

Structured abstracts improve visibility of required elements in many research contexts.

But disciplines and journals differ. Follow the venue.

8. The abstract is a compression problem, not a shortening problem | Abstract 不是“每段缩短一点”

Bad method:

Take one sentence from Introduction, one from Methods, one from Results, one from Discussion.

This often produces a disjointed collage.

Better:

reconstruct the study’s invariant state first.

9. Build an abstract invariant map | 建立 abstract invariant map

State	Question
Problem	Why was the study needed?
Objective	What exactly did it test/ask?
Design	What type of study was it?
Population/data	Who/what was studied?
Main method	What critical comparison/analysis was used?
Primary result	What is the central finding?
Magnitude/precision	How large/precise?
Boundary	What scope/limitation changes interpretation?
Conclusion	What is the strongest defensible answer?

10. Not every state always appears | 不是每个 abstract 都要九项平均分配

A methods paper may need more methodological detail.

A systematic review may need search/synthesis details.

A qualitative study may need context and analytic approach.

A conference abstract may include required programme-specific headings.

11. But every omitted state is a decision | 删除某项必须有理由

Ask:

If I remove this, could the reader misunderstand what kind of evidence the study provides?

12. The background must earn its space | Background 每一个词都很贵

Weak:

Education is very important in modern society, and students face many challenges.

This consumes space without defining the research problem.

13. Use background to locate the exact problem | Background 的任务是定位研究问题

Better:

Automated writing feedback can accelerate correction, but whether these gains transfer to independent revision remains unclear.

One sentence gives:

known state;
unresolved state;
research need.

14. Abstract background should rarely become a mini literature review | 不要把 abstract 变 literature review

The reader needs enough context to understand the question, not the history of the field.

15. Research question/objective must be explicit | Objective 不能隐身

This study tested whether structured feedback improves independent revision among advanced bilingual learners.

16. Avoid empty purpose language | 不要写空 purpose

Weak:

This study aims to explore issues related to feedback.

Better:

This study compared structured and standard feedback on one-week independent revision performance.

17. “Explore” should mean exploration | explore 不要当万能 academic verb

If the study tested a specific hypothesis, say tested.

If it estimated an association, say examined the association.

If it elicited lived experience, say investigated how participants experienced…

18. Abstract method detail should identify evidential status | Methods 要告诉读者“这是什么 evidence”

You usually do not need every procedural step.

You often do need:

design;
sample/population;
main comparison/intervention;
primary outcome;
analysis type where it changes interpretation.

19. Design words carry huge information density | Design label 是高密度词

randomised trial

prospective cohort

cross-sectional survey

qualitative interview study

systematic review and meta-analysis

These compress many methodological assumptions.

20. Use the correct design label | 不要给研究“升级”

A before-after study is not a randomised trial.

A narrative review is not a systematic review.

A convenience survey is not a population survey.

21. Sample size often belongs | Sample size 常常值得保留

We randomised 240 learners…

This immediately informs scale.

22. Sample boundary may matter more than sample size | Population boundary 可能比 n 更重要

advanced bilingual university learners

may be essential because it constrains generalisation.

23. Setting may be essential | Setting 有时不能删

A laboratory result, one-school trial and nationwide registry study carry different external meaning.

24. Intervention/comparator must be recognisable | 比较对象不能模糊

Weak:

Two groups were compared.

Better:

Learners were randomised to structured action-oriented feedback or standard evaluative feedback.

25. Outcome must be clear | 研究到底 measured 什么

writing performance may be too broad.

one-week independent revision score is more informative.

26. Methods detail should stop before it becomes a procedure manual | Abstract Methods 不复制整段 Methods

Cut:

routine software version unless essential;
minor instrument details;
administrative sequence;
secondary analytic choices that do not affect central interpretation.

27. Results are the centre of gravity | Abstract 最重要的通常是 actual results

One of the most common weak abstracts spends most of its space on background and ends with:

Results will be discussed.

For a completed study, this fails the central informational job.

28. Report the principal result, not merely “a significant difference” | 要给 actual result

Weak:

There was a significant difference between groups.

Better:

The structured-feedback group scored 6.1 points higher on the independent revision task.

29. Numbers protect against rhetorical inflation | 数字可以保护意义

Improved dramatically is interpretation.

Mean score increased from 68.1 to 74.2 is more inspectable.

30. Use effect size/estimate where relevant | 用真正有意义的 estimate

Depending on field:

mean difference;
risk ratio;
odds ratio;
correlation;
regression coefficient;
confidence interval;
theme prevalence only where appropriate.

31. Statistical significance is not the abstract result | p-value 不是 result 的全部

Readers need direction and magnitude.

32. If uncertainty matters, preserve it numerically | 重要 uncertainty 不要全部翻成 vague prose

difference 4.2 points, 95% CI 0.8–7.6

can be more informative than:

the result was statistically significant.

33. Report denominator when proportions matter | 百分比需要 denominator context

42 of 50 participants (84%)

may be more transparent than 84% alone.

34. Main result should answer the objective | Result 必须对应 objective

If objective asks long-term retention, do not headline satisfaction.

35. Secondary outcomes should enter only if they change interpretation | 次要结果要有理由

Include a secondary result when it:

qualifies the main finding;
shows harm/trade-off;
answers a stated secondary objective;
prevents misleading interpretation.

36. Null results can be abstract-critical | Null result 可能必须保留

If immediate outcome improved but delayed outcome did not, omitting the delayed null result creates distortion.

37. Adverse/unexpected results may be essential | 不漂亮的 result 不能因为空间紧就删

Compression cannot become selective reporting.

38. Negative evidence can define the real conclusion | 反证可能决定 abstract 结论

Near-term gains were observed, but no clear difference remained at six months.

This sentence may be more important than three positive secondary outcomes.

39. The abstract conclusion must be downstream of the result | Conclusion 必须从 result 长出来

Bad:

This innovative method should transform education worldwide.

Nothing in a small six-week study can usually support that scope.

40. Conclusion should answer the research question | Conclusion 回到 objective

Question:

Does structured feedback improve independent revision?

Calibrated conclusion:

Structured feedback improved near-term independent revision in this sample, but long-term transfer remains untested.

41. Do not introduce a new claim in the final sentence | 最后一句不能突然开新战线

No new mechanism.

No new policy.

No new population.

42. Distinguish conclusion from recommendation | Conclusion ≠ Recommendation

The intervention improved short-term performance.

is not the same as:

All schools should adopt the intervention.

43. Recommendation requires decision criteria | Recommendation 需要 cost/risk/feasibility

Abstracts often do not have room to justify large policy recommendations.

44. If recommendation is required, keep it proportional | 如果 venue 要 recommendation，控制 scope

The findings support a larger multi-site trial.

may be justified where full implementation is not.

45. Limitations: not every abstract needs a separate sentence, but some absolutely do | Limitation 是否写，看它是否改变结论

Ask:

If the reader never sees the full paper, what limitation could make the abstract conclusion misleading?

46. PRISMA explicitly requires evidence limitations in systematic-review abstracts | Systematic review 的 limitation 可能是 mandatory move

PRISMA 2020 for Abstracts asks authors to provide a brief summary of limitations in the included evidence, such as risk of bias, inconsistency and imprecision.

47. Limitation should be consequence-bearing | Abstract limitation 不写 generic apology

Weak:

The study had several limitations.

Better:

The single-site sample limits generalisation beyond high-support settings.

48. One good limitation sentence can prevent major distortion | 一句可以救整个摘要

The design was observational, so causal interpretation remains limited.

49. Abstract certainty must equal paper certainty | Certainty conservation

Full paper:

may contribute to

Abstract:

causes

= distortion.

50. Abstract compression creates certainty drift because qualifiers look expendable | Hedge 常在压缩时最先被删

But may, in this sample, short-term, associated with can carry the scientific boundary.

51. Do not delete a word because it is small | 小词可能承担大 epistemic job

Delete decorative words before deleting scope/uncertainty.

52. The “claim conservation” rule | Claim conservation

Every major claim in the abstract should be no stronger than the corresponding claim in the paper.

53. The “scope conservation” rule | Scope conservation

If paper says:

among 96 advanced bilingual learners

abstract should not silently become:

students.

54. The “time conservation” rule | Time conservation

one-week improvement cannot become:

lasting improvement.

55. The “design conservation” rule | Design conservation

observational association cannot become:

intervention effect.

56. The “authority conservation” rule | Authority conservation

authors recommend cannot become:

evidence proves policy should change.

57. The “outcome conservation” rule | Outcome conservation

revision score cannot automatically become:

English proficiency.

58. The “population conservation” rule | Population conservation

volunteer university learners cannot become:

all learners.

59. The “uncertainty conservation” rule | Uncertainty conservation

Wide interval, conflicting replication or unresolved confounding must not disappear if they materially change interpretation.

60. Abstracts often fail through semantic expansion | 摘要常见错误不是删太多，而是删完以后 claim 变大

Example:

Full paper:

Participants reported greater confidence.

Abstract:

The intervention improved competence.

Confidence ≠ competence.

61. Avoid proxy inflation | Proxy 不要升格成真正 outcome

Self-report intention ≠ actual behaviour.

Task performance ≠ durable learning.

Click rate ≠ comprehension.

62. Avoid category inflation | Category 不要扩大

grammar accuracy ≠ writing quality.

63. Avoid causal inflation | Causation 不要升级

associated with ≠ improves.

64. Avoid policy inflation | Study result ≠ policy conclusion

Small pilot ≠ national recommendation.

65. Avoid novelty inflation | “First ever” 是高风险 claim

Unless you have a defensible search basis, avoid:

the first study ever;
the only research;
completely novel.

66. Avoid significance inflation | significant 有多个意义

Statistically significant?

Practically important?

Conceptually important?

Write the intended meaning.

67. The title–abstract contract | 标题与 abstract 也要一致

If title claims Improves, abstract must support causation.

If abstract only shows association, title should not upgrade it.

68. Abstract should usually be written late—but designed early | 摘要最好晚写，但核心状态可以早定义

Write the final abstract after results and discussion are stable.

But define the intended abstract components early to keep the study question clear.

69. Do not copy the introduction into the abstract | Abstract 不是 Introduction 的浓缩版

Completed research abstracts must include what was found.

70. Do not copy the conclusion into the abstract | 也不是 Conclusion 的扩写

The reader needs enough method and result to evaluate that conclusion.

71. Write from a study map, not by cutting sentences | 从 map 写，不从正文剪

Recommended process:

Close the manuscript.
Write the invariant map from memory.
Reopen and verify every item.
Draft the abstract from the map.
Compare abstract claims against full-paper claims.

72. Why write from memory first? | 为什么先不看正文？

It forces you to identify the central study rather than the most elegant sentences.

Then verification prevents memory distortion.

73. Draft long, then compress | 先写完整再压缩

Write a 400–500 word accurate version first if necessary.

Then reduce to the venue limit.

This is often safer than trying to write 150–250 words immediately.

74. Compression order | 建议删减顺序

repeated background;
decorative adjectives;
minor method details;
secondary results that do not change interpretation;
duplicated conclusion language;
nonessential transitions.

Delete scope/uncertainty last, not first.

75. Replace phrases with precise compact forms | 压缩靠结构，不靠删逻辑

Long:

Participants who were enrolled in the study were randomly assigned into one of two different groups.

Compact:

Participants were randomised to two groups.

76. Replace abstract nouns with direct verbs when appropriate | 动词可以节省空间

an examination of the relationship between

→ examined the relationship between

77. Remove empty framing | 删掉空框架

It is important to note that…

Usually delete.

78. Remove self-commentary | 不要描述“本文将做什么” if study already completed

This paper will discuss…

Replace with what the study did/found.

79. Avoid long abbreviation systems | Abstract 不适合建立复杂缩写系统

Elsevier recommends avoiding non-standard/uncommon abbreviations unless essential, and defining them at first mention.

80. One-use abbreviations waste space | 只用一次就别缩写

If a term appears once, the abbreviation often costs more than it saves.

81. References are often avoided | Abstract 通常不放 references

Elsevier notes that abstracts often stand separately from the article, so references should generally be avoided unless essential and the venue allows them.

82. Never invent citations to make an abstract sound scholarly | 不要用 citation 装饰摘要

The abstract’s job is your study, not a mini bibliography.

83. Keywords are not a second abstract | Keywords 是检索接口

Choose terms that accurately represent:

core phenomenon;
population/context where relevant;
method/design where useful;
discipline-specific indexing language.

84. Avoid keywords that are too broad | education / English / study 往往太宽

Use phrases readers actually search for.

85. Avoid duplicating every title word | Keywords 可以扩展检索面

But follow venue indexing rules.

86. Abstract audience is wider than the full paper audience | 摘要读者可能只读这一段

Editors.

Reviewers.

Search-engine users.

Researchers scanning databases.

Policy/professional readers.

87. Stand-alone means define enough context | Stand-alone 不等于解释所有背景

A reader should understand the study’s central logic without needing Section 2.3.

88. Technical terms can remain when they are necessary | 不要为了“可读性”删掉核心术语

Define or choose accessible phrasing where needed.

Lesson 010 applies here: adapt without changing the intellectual object.

89. Avoid insider shorthand | 内部项目名/缩写不应成为 abstract 门槛

Pipeline B using V3 protocol

is meaningless to external readers unless defined.

90. Abstracts for qualitative studies need design-appropriate compression | Qualitative abstract 不要硬装 quantitative

Often include:

problem/objective;
context/participants;
data-generation method;
analytic approach;
major themes/findings;
interpretive conclusion.

91. Do not fake numeric precision in qualitative work | Qualitative 不需要为了“科学感”乱加数字

Frequency may be useful in some designs, but thematic importance is not always proportional to count.

92. Abstracts for methods papers need performance criteria | Methods paper 的 result 是方法表现

State:

what method does;
what it was compared with;
which benchmark/outcome;
how performance differed;
where it may be useful.

93. Abstracts for systematic reviews need synthesis state | Systematic review abstract 需要 review-specific 信息

PRISMA 2020 for Abstracts provides a design-specific checklist including search sources, eligibility, synthesis, included studies, main outcomes, limitations, interpretation, funding and registration.

94. Review abstract cannot simply say “the literature was reviewed” | Review Methods 要可识别

What sources?

What eligibility?

How many included studies?

How synthesised?

95. Abstracts for randomised trials may have reporting-guideline requirements | Trial abstract 也有 design-specific reporting

The EQUATOR Network maintains reporting guidelines including CONSORT guidance for randomised-trial abstracts and the updated CONSORT 2025 framework for trial reporting.

EQUATOR Network | CONSORT for Abstracts

EQUATOR Network | CONSORT 2025

96. Reporting guideline is a floor, not a prose template | Checklist ≠ writing style

A checklist tells you what information should be present.

You still need coherent, efficient language.

97. Conference abstracts can have different jobs | Conference abstract 可能用于 acceptance decision

They may have:

strict headings;
character limits;
required preliminary results;
different publication status.

Read the call-for-papers instructions.

98. Proposal abstracts differ from completed-study abstracts | Proposal abstract 不应该假装已经有 Results

If the work is planned:

will examine may be correct.

If completed:

examined / found is appropriate.

99. Do not fabricate results to fit a completed-study template | 没有结果就不要造结果

100. Thesis/dissertation abstracts may have broader summary needs | Dissertation abstract 可能需要更大范围

Institutional requirements vary.

Follow graduate-school guidance.

101. Humanities abstracts can use different architecture | Humanities abstract 可能没有 Methods/Results heading

But still needs:

problem/question;
object/corpus;
approach;
central argument;
contribution.

102. Do not force empirical vocabulary onto interpretive work | Humanities 不需要假装 experiment

This essay argues… through a comparative reading of…

may be exactly right.

103. The abstract should represent the paper’s actual genre | Genre conservation

Review → review.

Protocol → protocol.

Case study → case study.

Replication → replication.

104. Avoid generic “This paper discusses…” | Genre 具体化

This systematic review synthesises…

This qualitative study examines…

This conceptual article proposes…

105. Abstract cohesion matters | 摘要不是九条 checklist 机械拼接

Transitions should preserve the study’s logic:

problem → question → design → result → conclusion.

106. Pronoun reference must remain clear under compression | 压缩以后 pronoun 更容易失去 antecedent

Avoid:

This showed that…

when “this” could refer to the method, result or previous sentence.

107. Sentence density must remain readable | 不要把 300 words 压成 4 个超长句

Compression is not syntactic packing.

108. Use one sentence for one main move when possible | 每句承担一个主要 move

Especially under C1–C2 clarity goals.

109. Avoid noun-stack overload | Academic abstract 容易 noun pile-up

learner feedback action implementation transfer effectiveness evaluation

Break the relation into clauses.

110. Avoid nominalisation when it hides actor/action | 名词化有用，但不能让逻辑消失

the implementation of the intervention resulted in an improvement…

may become:

The intervention improved…

only if causation is actually established.

111. Compact does not mean simplistic | 简短也可以精确

Scores increased after training, but the uncontrolled design prevents causal attribution.

Fourteen words can preserve result + limitation.

112. Abstract should preserve negative space | 摘要也要告诉读者“没有建立什么”

Useful:

The study establishes short-term performance differences but not long-term transfer.

113. “No evidence of” must be used carefully | no evidence of ≠ evidence of no

If a study was underpowered, saying no effect can distort.

114. “Did not detect” can be more precise | did not detect

The study did not detect a clear between-group difference at six months.

115. If equivalence was tested, say so | Equivalence 需要对应 design

Do not infer equivalence from ordinary non-significance.

116. Confidence intervals are abstract-efficient | CI 是高信息密度表达

One interval can communicate direction, magnitude and precision.

117. But do not overload the abstract with every statistic | 统计数字要服务 main result

Select the estimates needed to understand the principal conclusion.

118. One study can have more than one main result—but hierarchy matters | 多 outcome 要有 primary/secondary 层级

Readers should know what was central.

119. If primary outcome is null, do not headline a positive secondary outcome | 不要 secondary cherry-pick

Unless clearly labelled and justified, this distorts the study.

120. Abstracts are vulnerable to spin | Abstract 最容易产生 spin

Spin includes:

causal wording from non-causal design;
highlighting secondary positives over primary nulls;
omitting important harms;
generalising beyond sample;
stronger conclusion than results;
headline claims unsupported by the study.

121. The anti-spin audit | Anti-spin 审计

Ask:

What was the primary outcome?
Did I report it?
Did I preserve the design?
Did I preserve the population?
Did I preserve time horizon?
Did I preserve null/adverse findings?
Did I preserve causal boundaries?

122. The “abstract vs full paper” claim table | 建立 claim table

Full paper claim	Abstract claim	Match?
associated with	improves	No — causal inflation
short-term	lasting	No — time inflation
volunteer sample	users	No — population inflation
may reflect	is caused by	No — mechanism inflation

123. The “sentence parent” audit | 每个 abstract 句子要有正文 parent

For every abstract sentence ask:

Where in the paper is this supported?

If nowhere, remove or rewrite.

124. The “paper child” audit | 正文的重要结论要在 abstract 留下 child

For every central paper conclusion ask:

Does the abstract represent this?

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

Give only the abstract to someone.

Ask them:

What did the researchers do?
Who/what did they study?
What did they find?
How strong is the conclusion?
What can the study not establish?

If their reconstruction is wrong, the abstract is misleading or incomplete.

126. The “word budget” audit | Word budget 审计

Mark every sentence:

B = Background
O = Objective
M = Methods
R = Results
C = Conclusion
L = Limitation

If half the abstract is B and only one sentence is R, rebalance.

127. The “number density” audit | 数字密度审计

Too few numbers can make findings vague.

Too many numbers can make the abstract unreadable.

Keep the estimates that determine the conclusion.

128. The “abbreviation” audit | 缩写审计

Delete abbreviations used only once.

Define essential uncommon abbreviations.

129. The “reference” audit | 引用审计

Remove references unless essential and venue-permitted.

130. The “keyword” audit | Keyword 审计

Can the intended reader discover the study using these terms?

131. Worked case: fictional advanced-English feedback study | 完整案例：虚构研究

All details below are fictional teaching material.

Full-study state:

Question: Does structured feedback improve independent revision among advanced bilingual learners?
Design: randomised two-group trial.
Sample: 96 advanced bilingual learners.
Intervention: structured action-oriented feedback vs standard evaluative feedback.
Primary outcome: unseen independent revision one week later.
Result: 74.2 vs 68.1, difference 6.1 points.
Secondary: largest difference in evidence integration; sentence accuracy similar.
Limitation: single site; no long-term follow-up.
Conclusion: improves near-term independent revision in this sample; durability/generalisation unresolved.

132. Weak abstract version | 弱版本

Feedback is very important for language learners. This study explored structured feedback. Ninety-six learners took part and several analyses were conducted. The results were significant and showed that structured feedback is superior. These findings prove that schools should adopt structured feedback to improve English.

133. Diagnose the weak version | 为什么弱

background generic;
objective vague;
method insufficient;
result has no actual number;
significant ambiguous;
“superior” too broad;
“prove” overclaims;
policy recommendation unsupported;
scope/limitation absent.

134. Calibrated abstract version | 校准版本

Whether structured feedback improves independent revision rather than only supported performance remains unclear. We randomised 96 advanced bilingual learners to structured action-oriented feedback or standard evaluative feedback and assessed independent revision on an unseen task one week later. Mean revision scores were 74.2 in the structured group and 68.1 in the standard group, a 6.1-point difference; the largest difference occurred in evidence integration, while sentence-level accuracy was similar. Structured feedback improved near-term independent revision in this sample, particularly for evidence integration. The single-site design and one-week follow-up leave generalisation and long-term durability unresolved.

135. What survived compression? | 什么被保留下来？

problem;
design;
population;
comparison;
time;
actual primary result;
important secondary pattern;
scope;
limitation.

136. What was deleted? | 什么被删掉？

minor procedural details;
full scoring rubric;
software;
secondary descriptive statistics;
speculative mechanism;
large policy recommendation.

This is disciplined compression.

137. Structured version of the same study | 同一研究的 structured abstract

Objective: To test whether structured action-oriented feedback improves independent revision among advanced bilingual learners.

Methods: Ninety-six learners were randomised to structured or standard feedback and completed an unseen independent revision task one week later.

Results: Mean scores were 74.2 versus 68.1, a 6.1-point difference. The largest difference occurred in evidence integration; sentence-level accuracy was similar.

Conclusion: Structured feedback improved near-term independent revision in this sample. Generalisation and durability remain uncertain because the study was conducted at one site with a one-week follow-up.

138. Same study, different venue = different surface form | Venue 改变 form，不改变 state

The information hierarchy can stay stable even when headings/word limit change.

139. Practice A: cut generic background | 练习 A

Compress:

Education is changing rapidly in the modern world, and technology is becoming increasingly important in many educational contexts.

into one problem-specific sentence.

140. Practice B: identify missing method state | 练习 B

We studied 200 students and found higher scores.

What critical design information may be missing?

141. Practice C: repair result vagueness | 练习 C

Repair:

The treatment significantly improved outcomes.

142. Practice D: repair causal inflation | 练习 D

Observational study:

Tool use improved grades.

143. Practice E: repair scope inflation | 练习 E

Study: 60 volunteer adult learners.

Abstract conclusion:

The method works for language learners.

144. Practice F: preserve null result | 练习 F

Immediate result positive; six-month result null.

Write one compact Results sentence.

145. Practice G: limitation consequence | 练习 G

Single high-resource school.

Write one limitation clause.

146. Practice H: abstract word budget | 练习 H

Your 200-word abstract has:

100 words background;
35 methods;
25 results;
40 conclusion.

What is wrong?

147. Practice I: systematic-review abstract | 练习 I

List five PRISMA abstract states you would protect under a strict word limit.

148. Practice J: proposal vs completed study | 练习 J

Rewrite will investigate after the study has been completed.

149. Practice K: keyword selection | 练习 K

Title already includes structured feedback and bilingual writing.

Choose additional discoverability terms without repeating generic words.

150. Model answers | 示范答案

A: Whether AI-assisted feedback improves independent revision rather than only supported performance remains unclear.

B: Design/comparison, sampling, intervention/exposure, outcome timing and analysis may all be necessary to interpret “higher scores”.

C: Mean delayed-test scores were 5.4 points higher in the treatment group than in the comparison group.

D: More frequent tool use was associated with higher grades.

E: The method improved performance among the volunteer adult learners studied; generalisation to other learner groups remains uncertain.

F: The intervention improved immediate performance, but no clear between-group difference remained at six months.

G: The single high-resource setting limits generalisation to lower-support schools.

H: The abstract spends half its space before the study begins and gives too little evidence. Reduce generic background and expand actual Methods/Results.

I: Objective/question; eligibility/information sources/synthesis method; included studies; main synthesis result; evidence limitation/interpretation. Exact content depends on the review and checklist requirements.

151. The 20-minute abstract drill | 20 分钟 abstract 训练

Time	Task
3 min	write invariant map
5 min	draft accurate long abstract
5 min	compress to venue limit
4 min	claim/scope/certainty audit
3 min	stand-alone reader test

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

Time	Task
8 min	analyse three abstracts from target journal
8 min	identify move/word budgets
10 min	draft structured version
10 min	convert to unstructured version
9 min	anti-spin + certainty audit

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

Time	Task
15 min	target-venue requirements + examples
15 min	full study invariant map
20 min	400-word accurate draft
15 min	compress to required limit
10 min	numbers/limitations/negative-result audit
15 min	title + abstract + keywords consistency audit

154. Seven-day abstract cycle | 七天 abstract cycle

Day 1: problem/objective compression.
Day 2: method-state compression.
Day 3: numeric result reporting.
Day 4: conclusion/limitation calibration.
Day 5: structured ↔ unstructured transformation.
Day 6: anti-spin/title/keyword audit.
Day 7: unseen paper → abstract benchmark.

155. Twelve-week C1–C2 abstract progression | 12 周路线

Weeks	Focus	Output
1–2	study-state reconstruction	invariant maps
3–4	method + result compression	accurate mini-abstracts
5–6	numbers, uncertainty, limitations	evidence-rich abstracts
7–8	genre/design-specific abstracts	structured/unstructured sets
9–10	anti-spin, titles, keywords	submission-ready packages
11–12	disciplinary independent transfer	full abstract portfolio

156. Monthly benchmark | 每月基准任务

Choose one full research paper you are allowed to use.

Produce:

study invariant map;
400-word accurate summary;
250-word abstract;
150-word abstract;
structured version;
unstructured version;
six keywords;
one title;
claim-conservation table;
anti-spin audit;
reader reconstruction test;
one paragraph explaining what information had to survive every compression level.

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

Symptom	Likely weak link	Repair
abstract mostly background	study hierarchy unclear	result-first word budget audit
methods too vague	design state missing	design + sample + comparison + outcome
results say “significant” only	numeric result missing	report direction/magnitude
conclusion stronger than paper	certainty drift	claim conservation table
sample disappears	scope inflation	restore population boundary
positive secondary result dominates	spin/selective reporting	restore primary/null outcome
abstract reads like sentence collage	cut-and-paste drafting	write from invariant map
word limit impossible	details not prioritised	delete procedure/decorative text before boundaries

158. What not to do | 不要这样写 abstract

Do not spend most of the abstract proving the topic is important.
Do not hide the actual study design.
Do not report “significant” without the main result.
Do not remove the population/time boundary to save words.
Do not upgrade association into causation.
Do not omit an important null/adverse result because space is tight.
Do not let the title become stronger than the abstract.
Do not let the abstract become stronger than the paper.
Do not force one abstract structure onto every discipline/design.
Do not write from copied sentences when the study’s real hierarchy is unclear.

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

Purdue OWL | Writing Scientific Abstracts — abstract purpose, stand-alone quality and common purpose/method/findings/conclusion structure.
Elsevier | Guide for Authors — concise factual abstract, purpose, principal results, major conclusions, stand-alone requirement, references/abbreviations caution.
Nature Masterclasses | Writing a Research Paper — effective titles/abstracts as part of publication-ready research writing.
PRISMA 2020 for Abstracts — design-specific systematic-review abstract reporting requirements.
EQUATOR Network | CONSORT for Abstracts — trial-abstract reporting guidance.
EQUATOR Network | CONSORT 2025 — updated reporting framework for randomised trials.
University of Manchester Academic Phrasebank | Reporting Results — result-reporting language.
Academic Phrasebank | Being Cautious — certainty control.

160. Canonical eduKate routes | eduKate 主页面路由

161. SEO language map | 本课自然覆盖的搜索意图

This lesson naturally serves learners searching for how to write an abstract, research abstract, scientific abstract, structured abstract, unstructured abstract, abstract writing examples, abstract results, abstract methods, abstract conclusion, research paper abstract, systematic review abstract, PRISMA abstract, C1 academic writing, C2 academic English, academic English for Chinese speakers, English for Mandarin speakers, 摘要怎么写, 学术英语摘要, 研究摘要, structured abstract, abstract Results 怎么写, 摘要结论怎么写, 如何压缩论文, 中文母语学术英语 and C1 C2 research writing.

162. The one-page abstract operating system | 一页 abstract 操作系统

Read venue requirements first. Word limit / headings / checklist。
Build the invariant map. Problem / objective / design / population / method / result / boundary / conclusion。
Write from the map, not by cutting sentences.
Protect design identity. 不给研究升级。
Protect population and time. 不扩大 scope。
Report actual principal results. Direction + magnitude + precision where needed。
Protect important null/adverse findings.
Keep conclusion downstream of results.
Preserve certainty. Abstract 不比 paper stronger。
Preserve causal boundary. Association 不变 causation。
Include limitation if omission would mislead.
Compress procedure before logic.
Use only essential abbreviations.
Avoid references unless required/permitted.
Choose discoverable, accurate keywords.
Run anti-spin audit.
Run claim-conservation table.
Run reader reconstruction test.

163. Final assignment | 最终作业

Choose one complete study or research paper you are allowed to work with.

Complete this chain:

Read the target venue’s abstract requirements.
Build a nine-state invariant map.
Write a 400-word accurate study summary.
Compress it to 250 words.
Compress it again to 150 words.
Create one structured abstract.
Create one unstructured abstract.
Mark every numerical result.
Mark every scope word.
Mark every certainty word.
Mark every causal verb.
Identify the primary outcome.
Check that no positive secondary outcome has displaced a primary null result.
Add one limitation sentence if its omission would mislead.
Create a claim-conservation table comparing paper and abstract.
Run the anti-spin audit.
Give the abstract to a reader who has not seen the paper.
Ask them to reconstruct design, population, main result and conclusion.
Repair every reconstruction error.
Choose final keywords and a title whose certainty matches the abstract.

Then ask:

If someone read only my abstract and never opened the paper, would they hold a materially different belief about what the study established?

如果一个人只读我的 abstract、永远不打开全文，他对“这项研究到底建立了什么”的理解，会不会和读完整篇 paper 后明显不同？

If yes, compression has become distortion.

If no, the abstract is doing its job.

Continue | 继续

The next lesson will move from the compressed research interface back to the front of the full paper: how to write an Introduction that makes one research question intellectually necessary without manufacturing a gap.

下一课回到完整 research paper 的前门：怎样写 Introduction，让一个研究问题自然变得“有必要回答”，同时不制造假的 research gap。

Next: EDKS-ADV-ZH-0017 · Lesson No.017 · Write a Research Introduction That Makes the Question Necessary · 把研究引言写到问题自然变得必要

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

Small Group Tutorials

How to Learn Advanced English (Chinese Edition) | Lesson No.016 | Write an Abstract That Compresses a Study Without Distorting It | 第016课：把完整研究压缩成不失真的摘要

Write an Abstract That Compresses a Study Without Distorting It | 把完整研究压缩成不失真的摘要

Before we begin: the abstract is where Lessons 009–015 meet | Abstract 是前面多课能力的压缩测试

1. Purdue: an effective abstract must stand alone | Abstract 必须能够独立工作

2. Elsevier: concise, factual, purpose → principal results → major conclusions | 摘要必须 factual

3. Nature treats title and abstract as core manuscript skills | Abstract 也是 publication interface

4. Reporting guidelines can require design-specific abstract content | 不同研究设计可能有专门 checklist

5. First rule: read the target venue’s abstract instructions | 第一条规则：先看 venue

6. Structured and unstructured abstracts | Structured / unstructured 两种常见形式

7. Structured does not automatically mean better | 有 heading 不自动更高级

8. The abstract is a compression problem, not a shortening problem | Abstract 不是“每段缩短一点”

9. Build an abstract invariant map | 建立 abstract invariant map

10. Not every state always appears | 不是每个 abstract 都要九项平均分配

11. But every omitted state is a decision | 删除某项必须有理由

12. The background must earn its space | Background 每一个词都很贵

13. Use background to locate the exact problem | Background 的任务是定位研究问题

14. Abstract background should rarely become a mini literature review | 不要把 abstract 变 literature review

15. Research question/objective must be explicit | Objective 不能隐身

16. Avoid empty purpose language | 不要写空 purpose

17. “Explore” should mean exploration | explore 不要当万能 academic verb

18. Abstract method detail should identify evidential status | Methods 要告诉读者“这是什么 evidence”

19. Design words carry huge information density | Design label 是高密度词

20. Use the correct design label | 不要给研究“升级”

21. Sample size often belongs | Sample size 常常值得保留

22. Sample boundary may matter more than sample size | Population boundary 可能比 n 更重要

23. Setting may be essential | Setting 有时不能删

24. Intervention/comparator must be recognisable | 比较对象不能模糊

25. Outcome must be clear | 研究到底 measured 什么

26. Methods detail should stop before it becomes a procedure manual | Abstract Methods 不复制整段 Methods

27. Results are the centre of gravity | Abstract 最重要的通常是 actual results

28. Report the principal result, not merely “a significant difference” | 要给 actual result

29. Numbers protect against rhetorical inflation | 数字可以保护意义

30. Use effect size/estimate where relevant | 用真正有意义的 estimate

31. Statistical significance is not the abstract result | p-value 不是 result 的全部

32. If uncertainty matters, preserve it numerically | 重要 uncertainty 不要全部翻成 vague prose

33. Report denominator when proportions matter | 百分比需要 denominator context

34. Main result should answer the objective | Result 必须对应 objective

35. Secondary outcomes should enter only if they change interpretation | 次要结果要有理由

36. Null results can be abstract-critical | Null result 可能必须保留

37. Adverse/unexpected results may be essential | 不漂亮的 result 不能因为空间紧就删

38. Negative evidence can define the real conclusion | 反证可能决定 abstract 结论

39. The abstract conclusion must be downstream of the result | Conclusion 必须从 result 长出来

40. Conclusion should answer the research question | Conclusion 回到 objective

41. Do not introduce a new claim in the final sentence | 最后一句不能突然开新战线

42. Distinguish conclusion from recommendation | Conclusion ≠ Recommendation

43. Recommendation requires decision criteria | Recommendation 需要 cost/risk/feasibility

44. If recommendation is required, keep it proportional | 如果 venue 要 recommendation，控制 scope

45. Limitations: not every abstract needs a separate sentence, but some absolutely do | Limitation 是否写，看它是否改变结论

46. PRISMA explicitly requires evidence limitations in systematic-review abstracts | Systematic review 的 limitation 可能是 mandatory move

47. Limitation should be consequence-bearing | Abstract limitation 不写 generic apology

48. One good limitation sentence can prevent major distortion | 一句可以救整个摘要

49. Abstract certainty must equal paper certainty | Certainty conservation

50. Abstract compression creates certainty drift because qualifiers look expendable | Hedge 常在压缩时最先被删

51. Do not delete a word because it is small | 小词可能承担大 epistemic job

52. The “claim conservation” rule | Claim conservation

53. The “scope conservation” rule | Scope conservation

54. The “time conservation” rule | Time conservation

55. The “design conservation” rule | Design conservation

56. The “authority conservation” rule | Authority conservation

57. The “outcome conservation” rule | Outcome conservation

58. The “population conservation” rule | Population conservation

59. The “uncertainty conservation” rule | Uncertainty conservation

60. Abstracts often fail through semantic expansion | 摘要常见错误不是删太多，而是删完以后 claim 变大

61. Avoid proxy inflation | Proxy 不要升格成真正 outcome

62. Avoid category inflation | Category 不要扩大

63. Avoid causal inflation | Causation 不要升级

64. Avoid policy inflation | Study result ≠ policy conclusion

65. Avoid novelty inflation | “First ever” 是高风险 claim

66. Avoid significance inflation | significant 有多个意义

67. The title–abstract contract | 标题与 abstract 也要一致

68. Abstract should usually be written late—but designed early | 摘要最好晚写，但核心状态可以早定义

69. Do not copy the introduction into the abstract | Abstract 不是 Introduction 的浓缩版

70. Do not copy the conclusion into the abstract | 也不是 Conclusion 的扩写

71. Write from a study map, not by cutting sentences | 从 map 写，不从正文剪

72. Why write from memory first? | 为什么先不看正文？

73. Draft long, then compress | 先写完整再压缩

74. Compression order | 建议删减顺序

75. Replace phrases with precise compact forms | 压缩靠结构，不靠删逻辑

76. Replace abstract nouns with direct verbs when appropriate | 动词可以节省空间