Series ID: EDKS-ADV-ZH-0015 · How to Learn Advanced English (Chinese Edition) · Lesson No.015 · C1 → C2

Keep Methods, Results and Discussion Doing Different Jobs | 让方法、结果与讨论各司其职

A research paper becomes difficult to trust when procedure, observation and interpretation are mixed so thoroughly that the reader can no longer tell what was done, what was found and what the writer thinks it means.

一篇研究文章最危险的混乱之一，就是 procedure、observation 和 interpretation 混在一起，让读者分不清：你做了什么、你看到了什么，以及你认为这些结果意味着什么。

At advanced level, research writing is not only about correct grammar or formal vocabulary. It is about architectural control.

You need to keep several kinds of statement distinct:

Procedure: what was done.
Observation: what was found.
Interpretation: what the finding may mean.
Evaluation: how strong or weak the evidence is.
Implication: what may follow from the finding.
Limitation: what the design cannot establish.

这些 statement 都可以出现在同一篇文章里，但它们不应该失去身份。

Before we begin: IMRaD is a useful architecture, not a universal law | IMRaD 是重要结构，但不是万能定律

Purdue Writing Lab notes that many academic papers use IMRaD—Introduction, Methods, Results and Discussion—but also emphasises that document structures vary by discipline, genre and publication venue. Some theoretical papers do not use IMRaD. Humanities papers may use entirely different structures. Some research articles combine Results and Discussion.

Purdue Writing Lab | Organization at the Document Level

所以，本课不是教你“所有研究都必须长成同一个样子”。本课教的是 section job control。

Even when the headings change, the intellectual jobs still exist.

1. The three core jobs | 三个核心任务

Section	Primary question	Main job
Methods	What did you do?	Make the research process inspectable.
Results	What did you find?	Report the relevant evidence clearly.
Discussion	What does it mean?	Interpret, evaluate, connect and limit the findings.

如果你只记住这三个问题，research architecture 已经清楚很多。

2. Manchester separates these jobs explicitly | Manchester 的 Academic Phrasebank 也按 section job 组织

The University of Manchester Academic Phrasebank maintains separate resources for Describing Methods, Reporting Results and Discussing Findings. It treats them as distinct communicative functions rather than one generic academic-writing task.

3. Elsevier states the same broad division | Elsevier 的 author guidance 也强调 section role

Elsevier’s general author guidance describes Experimental/Methods as the place for enough detail to understand or reproduce the work where appropriate, Results as clear and concise findings, and Discussion as exploration of the significance of those findings rather than simple repetition.

Elsevier | Guide for Authors: Article Structure

4. Section identity reduces epistemic confusion | Section identity 降低知识混乱

Imagine reading:

Participants completed the intervention for six weeks, which clearly proved the strategy was effective because average scores rose by 8%.

This one sentence mixes:

method;
result;
causal interpretation;
rhetorical overclaim.

Better separation:

Methods: Participants completed the intervention for six weeks.

Results: Mean scores were 8% higher at post-test.

Discussion: The increase is consistent with a beneficial intervention effect, although the absence of a control group prevents a firm causal conclusion.

5. Section separation makes argument auditing possible | 分开以后，读者才能 audit

The reader can now ask:

Was six weeks enough?
How was score measured?
Was the 8% difference precise?
What alternative explanations remain?

Research writing should make those questions easier to ask, not harder.

6. Methods is not a diary | Methods 不是实验日记

Weak Methods writing records everything in chronological order simply because it happened that way.

Strong Methods writing selects the information another knowledgeable reader needs to understand, evaluate and—where relevant—replicate or reproduce the procedure.

7. Manchester’s core Methods principle | Methods 的核心：可检查、可理解、足够详细

Academic Phrasebank states that the Methods section explains how research was carried out and should be clear and detailed enough for an experienced reader to follow the process. The amount of detail increases when methods are unfamiliar, novel or controversial, or when readers come from multiple disciplines.

8. Replicability is field-dependent | “可重复”也要看 discipline

A laboratory experiment may support close procedural replication.

An ethnography cannot be reproduced as if human context were a chemical recipe.

An archival study may depend on access conditions.

A computational study may require code, model version and data provenance.

所以，不要把 “replicable” 机械理解成“别人做一遍就必须得到完全一样的结果”。

9. Methods should make the inferential chain inspectable | Methods 要让 inference chain 可检查

The reader needs to understand:

who or what was studied;
how cases were selected;
what variables or constructs were measured;
what intervention/exposure occurred;
what comparison existed;
how data were collected;
how data were processed;
how analysis was performed;
which exclusions occurred;
which ethical/procedural constraints mattered.

10. The Methods section is an argument about trustworthiness | Methods 其实也在回答“为什么应该信这些结果”

But this argument is usually made through transparent procedure rather than rhetorical praise.

Weak:

We used a very reliable and excellent method.

Better:

Two independent coders classified each response using the preregistered rubric; disagreements were resolved by a third coder.

11. Name the design | 先说 design 是什么

Examples:

randomised controlled trial;
quasi-experiment;
cross-sectional survey;
longitudinal cohort;
case study;
ethnography;
corpus study;
mixed-methods design;
systematic review;
simulation study.

The design label helps readers activate the right expectations.

12. But design labels do not replace procedural detail | Design 名字不能代替真正方法

We conducted a survey.

Still leaves:

who;
how recruited;
what questionnaire;
when;
response rate;
analysis.

13. Participants/population belong in Methods | 谁被研究，属于 Methods

State relevant inclusion/exclusion characteristics without adding interpretation prematurely.

14. Recruitment can affect the meaning of Results | Recruitment 会改变结果可推广性

If participants volunteered, say so.

If sampling was convenience-based, say so.

If selection was random, explain from what population.

15. Sample-size rationale belongs near design decisions | Sample size rationale 属于 Methods

Where relevant, explain power analysis, feasibility constraint or sampling rationale.

16. Do not put the observed outcome into the sample rationale | 不要把 result 倒灌进 Methods

Weak:

We selected 50 participants, which was enough to show a significant effect.

That sentence uses the result to justify a prior method decision.

17. Variables and constructs must be operationalised | Construct 必须变成可测量对象

Engagement is not a method until you define how it was measured.

Attendance?

Time on task?

Self-report?

Observation rubric?

18. Operational definition prevents result ambiguity | Operational definition 防止 Results 变模糊

If two papers measure “engagement” differently, their Results may not be directly comparable.

19. Instruments belong in Methods | Instrument 属于 Methods

Questionnaire, sensor, rubric, test, interview protocol, software pipeline, coding scheme.

20. Instrument validity/reliability may need justification | 工具是否合适，需要说明

How much depends on discipline and whether the instrument is established.

21. Procedure should follow reader logic | Procedure 不一定照现实时间流水账

Chronological order is useful when process order matters.

Conceptual grouping is useful when several procedures run in parallel.

22. Subheadings can reduce cognitive load | Methods 可用 subheadings

Common examples:

Participants;
Materials;
Procedure;
Measures;
Analysis.

But follow field/journal conventions.

23. Tense in Methods is functional, not mechanical | Methods 时态不是死规则

Manchester notes that simple past is common for many methods functions because writers describe completed procedures.

But established facts or instrument properties may take present tense.

24. Passive voice is not automatically “academic” | Passive voice 不是高级写作标签

Samples were analysed… can be useful when procedure matters more than actor.

We analysed the samples… can be clearer where author agency matters and field conventions permit.

25. Methods should not hide agency when agency matters | 需要责任时，不要藏 actor

If researcher judgement affected coding, say who coded and how disagreement was handled.

26. Analysis plan belongs in Methods | Analysis plan 属于 Methods

Readers should know how raw material became the reported result.

27. Statistical procedure should answer the design | 分析方法必须匹配问题

Do not list software commands without explaining what question the analysis addressed.

28. Software is not the method | SPSS / R / Python 不是分析逻辑

Data were analysed in R.

is incomplete without the actual analytical approach.

29. Preprocessing can materially affect results | Preprocessing 也属于 inferential chain

Missing-data handling, exclusions, transformations, normalisation, tokenisation, coding rules, outlier treatment.

30. If preprocessing changes the evidence, document it | 会改变 evidence 的处理必须可追踪

31. Qualitative Methods need analytical transparency too | Qualitative Methods 也要透明

State:

how data were generated;
how transcripts/fieldnotes were handled;
how coding/themes were developed;
who participated in interpretation;
how reflexivity/credibility was addressed where relevant.

32. “Thematic analysis was performed” is often too thin | 只写 thematic analysis 不够

Which form?

How were codes produced?

Inductive/deductive?

How were themes reviewed?

33. Mixed Methods requires integration logic | Mixed Methods 不只是做两种数据

Explain:

sequence;
priority;
where strands connect;
how integration occurs.

34. Methods limitations can appear in Methods or Discussion depending genre | 方法限制放哪里看功能

Procedural constraints may be mentioned in Methods.

Their consequences for interpretation usually belong in Discussion.

35. The result of a manipulation check may appear in Results | Manipulation check 是 result

Do not put it in Methods merely because it concerns the intervention procedure.

36. Pilot-study outcomes can blur the boundary | Pilot 信息要判断角色

If pilot results determined the final method, some pilot information may legitimately appear in Methods.

But distinguish development evidence from the main study Results.

37. Results answers “what emerged?” | Results 的主问题：观察到了什么？

Manchester describes Results as systematic, detailed presentation and description of findings, with tables/figures and commentary that directs readers to important patterns.

38. Results is not raw data dump | Results 不是 data warehouse

Your job is to select and organise evidence relevant to the research questions.

39. Data, result and interpretation are different | Data / result / interpretation 不同

Data:

12, 15, 17, 21…

Result:

Mean response time fell by 18%.

Interpretation:

The streamlined interface may have reduced decision friction.

40. Results usually transform data into informative patterns | Results 是 data 的结构化描述

So “no interpretation whatsoever” is too rigid.

Choosing what pattern matters and describing it already involves analytical judgement.

41. But Results should not outrun the evidence | Results 不应该提前写完整解释

Weak Results:

The treatment group scored higher, proving that the intervention improved deep learning.

Better Results:

The treatment group scored 7.2 points higher on the transfer task.

42. Organise Results around research questions or hypotheses | Results 结构要跟 question 对齐

If the study has three questions, readers should be able to see which findings answer each one.

43. Purdue recommends alignment between Methods and Results | Methods–Results 要有 alignment

Purdue’s graduate-writing guide notes that each result should have a clear relation to the method that produced it.

Purdue Writing Lab | Graduate Writers Guide

44. Mirror structures can help | Methods 与 Results 可形成 mirror

Methods:

Primary outcome
Secondary outcome
Interview analysis

Results:

Primary outcome
Secondary outcome
Interview themes

This reduces reader reconstruction.

45. Report important negative/null findings | 不支持 hypothesis 的结果也属于 Results

Elsevier’s writing guidance explicitly recommends reporting negative results where relevant rather than hiding findings that do not support the hypothesis.

Elsevier | Writing the Results Section

46. Selective reporting destroys trust | 只报“漂亮结果”会破坏可信度

Do not make the Results section a success story.

47. Primary outcomes deserve priority | 主 outcome 要先清楚

Do not bury the main answer beneath dozens of secondary analyses.

48. Exploratory analyses should be marked as exploratory | Exploratory ≠ preregistered confirmatory

The language and placement should preserve that distinction.

49. Result order can follow importance, question order or analytic sequence | Results 顺序有多种合法方案

Choose one logic and keep it visible.

50. Tables and figures should carry evidence, not decoration | Table/Figure 是 evidence surface

Use them when they reveal:

patterns;
comparisons;
distribution;
time trends;
relationships;
model estimates.

51. Text should not repeat every table cell | 正文不要逐格抄表

Manchester describes a common Results move: locate the table/figure, then highlight the relevant pattern.

52. Results text answers “what should I notice?” | 正文告诉读者看什么

Example:

As shown in Table 2, the largest difference occurred at delayed follow-up rather than immediately after training.

53. Do not use figures to hide uncertainty | 图表也要显示 uncertainty

Error bars, confidence intervals, sample sizes or uncertainty information may be essential depending on field.

54. Visual scale can exaggerate effects | 坐标轴也会制造 rhetoric

A truncated axis can make a small difference appear dramatic.

55. Captions should be self-sufficient enough to orient the reader | Caption 也有 job

State what is shown, units/groups and necessary interpretive keys.

56. Avoid evaluation adjectives in Results unless defined | Results 少用“excellent / impressive / disappointing”

Those often belong to interpretation.

57. “Significant” must mean what you intend | significant 是高风险词

Statistical significance?

Practical importance?

Conceptual importance?

Do not let one word hide three meanings.

58. Descriptive statistics should describe | Descriptive statistics 回答“数据长什么样”

Mean, median, spread, range, frequency, distribution.

59. Inferential statistics should answer an inferential question | Inferential result 要对应 inferential question

Difference?

Association?

Model estimate?

Prediction?

60. Exact reporting conventions are discipline-specific | 数字格式看 discipline/style guide

Follow your target journal, style manual and field convention.

61. Results certainty should usually be higher for direct observations | 直接 observation 可以更直接

Mean score increased from 61 to 68.

No hedge needed if the recorded values are correct.

62. Interpretation certainty should be calibrated separately | 原因解释另外 calibration

The increase may reflect improved retrieval.

63. Keep measured fact and explanation in separate grammatical units | Fact 与 explanation 分开写

This makes certainty easier to control.

64. Qualitative Results are not “just quotes” | Qualitative Results 不是 quote collection

You still need:

theme/pattern;
analytic category;
representative evidence;
variation/negative cases where relevant.

65. Quote should illustrate an analytic claim | Quote 服务于 theme

Do not insert long participant quotations and expect the reader to infer the theme.

66. Negative cases matter in qualitative Results | 反例/偏离案例不能全部删掉

They may refine the theme.

67. Frequency language needs caution in qualitative work | “most / many / some” 要有依据

Follow field conventions; qualitative importance is not always frequency.

68. Mixed-methods Results require strand clarity | Mixed methods 要让两种 evidence 不混成 fake score

Quantitative result and qualitative explanation are not interchangeable units.

69. Integration can occur in Results, Discussion or both | Mixed Methods integration 看设计

State where integration happens.

70. Discussion begins where interpretation becomes central | Discussion 从“它意味着什么”开始

Manchester calls Discussion one of the most complex sections because it typically cycles through finding, interpretation, comparison, implication, limitation and future work.

71. Discussion should not merely repeat Results | Discussion 不是 Results 第二遍

Elsevier’s general guidance explicitly warns against simply repeating the Results and instead asks authors to explore their significance.

72. Start with the finding, then move beyond it | Discussion 可以重述关键 result，但要立刻解释

Weak:

The treatment group scored higher. The treatment group scored 7 points higher.

Stronger:

The higher transfer score suggests that the intervention affected more than task familiarity, although the absence of delayed testing leaves durability unresolved.

73. Discussion has several recurring jobs | Discussion 的常见 moves

Restate important finding.
Interpret its meaning.
Compare with previous literature.
Explain unexpected results.
Evaluate alternative explanations.
State implications.
State limitations and their consequences.
Identify future work.

74. Not every Discussion needs every move | 不是每一项都必须出现

Use moves that serve the research question and field convention.

75. Discussion cycles can repeat per major finding | 每个重要 finding 可以有一个 discussion cycle

Finding A → interpretation → literature → limitation.

Finding B → interpretation → literature → implication.

76. Interpretation should answer “why might this pattern exist?” | Interpretation 不是简单重述

But separate plausible explanation from established mechanism.

77. Lesson 014 controls the language here | Discussion certainty 要回到 Lesson 014

suggests, may reflect, is consistent with, cannot rule out.

Lesson 014 | Calibrate Certainty Without Becoming Vague

78. Compare findings with previous literature | Discussion 负责放回 field

Ask:

Does this replicate?
Extend?
Qualify?
Contradict?
Explain a previous inconsistency?

79. Do not force agreement with literature | 不要为了“看起来合理”制造一致

If your result differs, say so and diagnose the difference.

80. Unexpected results belong in Discussion | Unexpected result 不是要隐藏的 failure

They may reveal:

measurement issue;
boundary condition;
new mechanism;
chance;
implementation difference.

81. Do not invent post-hoc certainty | 事后解释最容易过强

Unexpected result → plausible explanation.

Not:

unexpected result → “obviously caused by…”

82. Alternative explanations belong in Discussion | Rival explanations 在 Discussion 比较

The difference may reflect X; however, Y remains plausible because…

83. Strong Discussion compares explanatory models | Discussion 不只是讲一个故事

Ask which explanation best fits all observed results, including anomalies.

84. Limitation is not apology | Limitation 不是“自我批评段”

It is a statement about what inference the design cannot support.

85. State limitation consequence | Limitation 后面写 consequence

Weak:

The sample was small.

Strong:

The small subgroup sample produces a wide estimate, so the apparent interaction should be treated as preliminary.

86. Method limitation and result limitation are different | Method weakness 与 result uncertainty 分开

Method limitation:

No random comparison group.

Inference consequence:

Causal attribution remains uncertain.

87. Generalisation limitation | 外部推广限制

The trial was conducted in one high-support setting, so generalisation to lower-support contexts remains uncertain.

88. Measurement limitation | 测量限制

Performance was measured immediately after instruction, so the study cannot establish durable retention.

89. Attrition limitation | Attrition 限制

Higher dropout in one group may bias the final comparison.

90. Self-report limitation | Self-report 适合某些问题，但有边界

It can capture perceived experience.

It may not directly establish behaviour or learning.

91. Future work should follow from uncertainty | Future work 不是“more research is needed”

Strong:

A delayed follow-up with equal teacher-contact time would help distinguish persistence from attention effects.

92. Future study should test the unresolved mechanism | Next study 应该降低当前 uncertainty

Do not propose unrelated future work merely to fill a paragraph.

93. Implication is not recommendation automatically | implication ≠ should

Finding:

Structured feedback improved near-transfer performance.

Implication:

Feedback usability may matter to transfer.

Recommendation:

Programmes should include structured feedback.

第三句增加了 decision criteria。

94. Practical implication needs feasibility | Practical implication 需要 context

Effectiveness alone does not settle cost, staffing, equity or risk.

95. The Discussion should not create new primary results | Discussion 不应该突然出现主要新数据

If an analysis materially supports a claim, it should usually be reported transparently as a result, even if exploratory.

96. Post-hoc analyses need labels | 事后分析要标 exploratory / post-hoc

This protects inferential status.

97. Introduction and Discussion form a loop | Introduction 与 Discussion 是一个 loop

Introduction asks the research problem.

Discussion returns to that problem with evidence.

98. Purdue describes Introduction and Discussion as connected | 前后要对齐

The introduction previews the problem; the discussion interprets the answer and implications.

99. Research question alignment is an architectural invariant | RQ alignment 是整篇文章的 invariant

Research question	Method	Result	Discussion
Does X improve Y?	design comparison	effect estimate	meaning/causality/scope
How do users experience X?	interviews	themes	interpretation/context
What predicts Y?	model	associations/estimates	explanation limits

100. Every major result should have a method parent | 每个 Result 都应该能追溯到 Method

Ask:

Where was this result generated?

101. Every major Discussion claim should have a result parent | 每个 Discussion claim 都应该能追溯到 Result

Ask:

Which finding is this interpretation explaining?

102. The lineage test | 建立 research lineage

Question → Method → Result → Interpretation → Implication.

If a sentence has no parent, inspect it.

103. Orphan result | Orphan Result

A finding appears with no described method.

104. Orphan method | Orphan Method

A measure/procedure is described but never produces a reported result.

105. Orphan interpretation | Orphan Interpretation

A Discussion claim has no matching evidence.

106. Orphan limitation | Orphan Limitation

A limitation is listed but never changes any conclusion.

107. Orphan recommendation | Orphan Recommendation

A strong action proposal appears without decision logic.

108. The leakage test | Section leakage 测试

Highlight:

green = procedure;
blue = observation;
orange = interpretation;
red = implication/recommendation.

Then inspect where colours cluster.

109. Methods leakage into Discussion | Discussion 里塞 procedure

Some method recap may be necessary to explain a limitation.

But long procedural repetition usually signals poor section economy.

110. Discussion leakage into Results | Results 提前解释原因

One sentence of orientation may be field-acceptable.

Full causal story belongs later unless Results and Discussion are intentionally combined.

111. Results leakage into Methods | Methods 偷偷告诉读者结果

Example:

We selected the scale because it successfully detected the expected improvement.

This uses outcome knowledge to justify procedure.

112. Introduction leakage into Results | Results 重讲背景

Results should not reopen the literature review before every table.

113. Combined Results & Discussion is legitimate | 合并 Results & Discussion 完全可能是对的

Manchester notes that research articles sometimes combine Results and Discussion, and Elsevier says a combined section is often appropriate.

The solution is not artificial separation. The solution is move-level clarity:

result → interpretation → literature relation → next result.

114. Combined sections need stronger signalling | 合并以后更需要 signposting

Readers must still know which sentence is evidence and which is interpretation.

115. Use linguistic signals | 用语言显示 move

Result:

The analysis found…

Interpretation:

This pattern suggests…

Comparison:

This is consistent with…

Limitation:

However, the estimate should be interpreted cautiously because…

116. Qualitative papers may integrate analysis more tightly | Qualitative Results/Discussion 边界更可能融合

In many qualitative traditions, presenting a theme already involves interpretive analysis.

Do not force a false quantitative-style separation.

117. But evidence and interpretation should still be traceable | 融合不等于失去边界

The reader should still see:

theme → supporting material → analytic interpretation.

118. Humanities research may not use Methods/Results at all | Humanities 可能完全不用这三个 heading

Yet analogous jobs still exist:

source/method/framework;
textual/historical evidence;
interpretation/argument.

119. Engineering papers may use Experimental / Validation / Performance | 工程领域可能换 heading

Again, job identity matters more than label.

120. Computational papers may need data/model/reproducibility detail | Computational Methods 有额外责任

Dataset version.

Preprocessing.

Model architecture.

Hyperparameters.

Random seeds.

Evaluation metrics.

Hardware/software versions where materially relevant.

121. AI-assisted research requires model/version traceability | AI workflow 也要版本可追踪

If a model materially affected data generation, coding, analysis or transformation, document the relevant system/version/prompt/procedure according to disciplinary and venue rules.

122. Do not treat AI output as a result without validation | AI output 不是自动 evidence

Methods must explain how outputs were checked and what human/technical validation occurred.

123. Secondary-data studies need provenance | Secondary data 需要 provenance

Where data came from.

When collected.

Who was represented.

What transformations occurred.

124. Systematic reviews have their own Methods/Results distinction | Systematic review 也有 section job

Methods:

search strategy, eligibility, screening, extraction, appraisal, synthesis method.

Results:

how many studies, study characteristics, findings/synthesis.

Discussion:

meaning, strength, limitations, implications.

125. Literature review as narrative essay differs from systematic review | Narrative literature review 不能假装自己有 systematic method

If your search is selective, say so honestly.

126. Case study Results may look like narrative | Case study 可以是 rich description

But distinguish case evidence from interpretive claim.

127. Ethnographic writing may distribute interpretation across sections | Ethnography 结构会不同

Follow disciplinary exemplars.

128. Genre analysis is part of advanced academic English | C1–C2 不是套模板，而是识别 genre

Purdue explicitly advises writers to examine conventions in their own field and publication venue.

129. Read three strong papers from your target venue | 写之前读 target journal

Look for:

section names;
section order;
method detail;
results commentary;
discussion cycles;
table/figure conventions;
limitation placement.

130. Build a section-function matrix | 建立 section-function matrix

Move	Methods	Results	Discussion
Describe sample	primary	descriptive outcome if relevant	only when interpreting scope
Explain procedure	primary	rare	only if relevant to interpretation
Report estimate	rare	primary	may restate key result
Explain mechanism	hypothesis/procedure rationale only	usually limited	primary
Compare literature	method precedent sometimes	usually limited	primary
State limitation	procedural constraint possible	rare	inference consequence primary
Recommend action	no	no	possible if warranted

131. “Primary” is not “exclusive” | Primary job ≠ absolute firewall

This is the central sophistication of this lesson.

132. The Methods–Results handoff | Methods → Results 交接

Methods promises:

Here is how I will answer the question.

Results delivers:

Here is what that process produced.

133. The Results–Discussion handoff | Results → Discussion 交接

Results says:

Here is the pattern.

Discussion says:

Here is what the pattern may mean and how far we can take it.

134. The Discussion–Conclusion handoff | Discussion → Conclusion 交接

Discussion explores.

Conclusion compresses the final defensible answer.

135. Do not let Conclusion become a second Discussion | Conclusion 不要重新展开全部解释

State:

main answer;
scope;
important uncertainty;
implication/next step.

136. Abstract compresses all section jobs | Abstract 是整个 research system 的压缩

A strong abstract typically contains compressed versions of:

problem;
method;
result;
conclusion/implication.

That is why abstract writing is difficult.

137. Abstract certainty must match the full paper | Abstract 不能 certainty drift

The next lesson will focus on this compression problem.

138. Worked case: fictional advanced-English feedback study | 完整案例：虚构 advanced-English feedback study

All study details below are fictional teaching material.

Question:

Does structured feedback improve independent revision quality among advanced bilingual learners?

139. Bad mixed paragraph | 错误混合版本

We randomly assigned 96 learners to structured or standard feedback, and the structured group performed much better because the explicit revision steps helped them think more clearly, proving that structured feedback is superior for advanced bilingual learners.

This sentence contains:

design;
result;
mechanism;
causal conclusion;
universal recommendation.

140. Methods version | Methods 版本

Ninety-six advanced bilingual learners were randomly assigned to structured-feedback or standard-feedback conditions. Both groups received the same source text and revision time. The structured condition received feedback organised into claim, evidence and revision-action prompts; the standard condition received evaluative comments without required action prompts. Independent revision quality was assessed one week later using an unseen task scored by two blinded raters.

141. Why this belongs in Methods | 为什么这属于 Methods

It explains:

allocation;
comparison;
intervention difference;
outcome timing;
measurement.

142. Results version | Results 版本

The structured-feedback group achieved a mean independent-revision score of 74.2 compared with 68.1 in the standard-feedback group, a difference of 6.1 points. Inter-rater agreement was high. The largest group difference occurred in evidence integration; differences in sentence-level accuracy were small.

143. Why this belongs in Results | 为什么这属于 Results

It tells the reader what the procedure produced.

It does not yet tell a full causal/mechanistic story.

144. Discussion version | Discussion 版本

The higher unseen-task score suggests that structured feedback supported some transfer beyond the practised text. The concentration of the effect in evidence integration rather than sentence accuracy is consistent with the possibility that explicit revision prompts helped learners organise higher-order decisions. However, the study does not isolate whether the benefit came from the feedback structure itself or from the additional cognitive effort required by the prompts. The one-week follow-up also leaves long-term persistence unresolved.

145. Why this belongs in Discussion | 为什么这属于 Discussion

It interprets:

transfer;
possible mechanism;
alternative explanation;
time limitation.

146. Conclusion version | Conclusion 版本

Structured feedback improved near-term independent revision in this sample, particularly for evidence integration. The result supports further testing of action-oriented feedback designs, but does not yet establish the mechanism or long-term durability of the effect.

147. Notice how the claim narrows as interpretation increases | 解释越多，越要控制 certainty

Methods = what happened procedurally.

Results = what happened empirically.

Discussion = what may explain it.

Conclusion = strongest defensible compressed answer.

148. Practice A: classify the sentence | 练习 A

Participants completed two 20-minute retrieval sessions.

Method, Result or Discussion?

149. Practice B: classify the sentence | 练习 B

Mean delayed-test performance was 11% higher in the retrieval group.

150. Practice C: classify the sentence | 练习 C

The delayed advantage may reflect stronger retrieval routes rather than increased exposure alone.

151. Practice D: repair Results leakage | 练习 D

Repair:

The intervention clearly improved deep learning, as scores were 8% higher.

Write a Results sentence only.

152. Practice E: repair Methods leakage | 练习 E

Repair:

We used the questionnaire because it successfully showed the treatment was effective.

153. Practice F: write limitation consequence | 练习 F

Method: no delayed follow-up.

Discussion consequence: ______.

154. Practice G: link result to method | 练习 G

Method includes accuracy, speed and interview data.

Results report speed only.

What architectural problem exists?

155. Practice H: identify orphan interpretation | 练习 H

Discussion says learners developed better metacognition, but no measure or interview addressed metacognition.

Diagnosis?

156. Practice I: combined Results & Discussion | 练习 I

Write four sentences:

result;
interpretation;
literature comparison;
limitation.

Use signals so the reader can tell each move apart.

157. Practice J: qualitative result | 练习 J

Create:

theme → one supporting quote/paraphrase → analytic interpretation.

158. Practice K: figure commentary | 练习 K

Write one sentence that highlights the main pattern without repeating every value.

159. Practice L: Discussion comparison | 练习 L

Your finding differs from a major previous study.

Write one cautious sentence proposing a possible reason.

160. Model answers | 示范解析

A: Methods.

B: Results.

C: Discussion.

D: Mean post-test scores were 8% higher in the intervention group than in the comparison group.

E: The questionnaire was selected because it measures the target construct using the same scale employed in prior validation studies.

F: The study cannot establish whether the observed advantage persists beyond the immediate assessment period.

G: Two measures were described but not reported; either report them, explain omission transparently, or revise the stated Methods if they were not part of the final analysis.

H: Orphan interpretation. The Discussion claim has no evidential parent.

I: Mean accuracy increased by 6 percentage points. This pattern suggests that the intervention affected task performance rather than only speed. The result is consistent with earlier work reporting improved accuracy under structured prompting. However, the short follow-up prevents conclusions about persistence.

L: One possible explanation for the difference is the longer training period used in the present study, although the designs do not allow this explanation to be isolated directly.

161. The section-colour audit | Section colour 审计

Take a full draft and colour:

Methods = green;
Results = blue;
Discussion = orange;
Recommendation = red.

Then inspect unexpected colour clusters.

162. The verb audit | Verb 审计

Methods verbs:

selected, assigned, measured, coded, analysed.

Results verbs:

increased, decreased, differed, reported, revealed.

Discussion verbs:

suggests, may reflect, supports, challenges, implies.

This is not a rulebook; it is a diagnostic clue.

163. The tense audit | 时态审计

Completed methods often use past tense.

Results often report past observations.

Discussion may shift between past findings and present-tense interpretation/literature claims.

Follow discipline conventions.

164. The parent-child audit | Parent–child 审计

For every major Results paragraph:

Which Method produced this?

For every major Discussion paragraph:

Which Result does this interpret?

165. The question audit | Research question 审计

Can every major section be traced to a research question/objective?

166. The table audit | Table 审计

Does every table answer a question?

Does the text highlight rather than duplicate?

167. The limitation audit | Limitation 审计

Does each important limitation change a claim, certainty level, scope or recommendation?

168. The certainty audit | Certainty 审计

Results state direct observations directly.

Discussion calibrates interpretation.

Conclusion does not grow stronger than Discussion.

169. The negative-result audit | Null/negative result 审计

Did you omit a result because it made the story less neat?

170. The exploratory-analysis audit | Exploratory 审计

Are post-hoc/exploratory analyses labelled honestly?

171. The reporting completeness audit | Reporting completeness 审计

Can another reader understand what data were collected, which analyses were performed, and which outcomes were reported?

172. The Discussion inflation audit | Discussion inflation 审计

Highlight:

proves;
causes;
clearly;
universally;
effective;
should.

Check whether each is warranted.

173. The section-deletion test | 删除 section label 测试

If headings disappeared, could a knowledgeable reader still infer which job each paragraph performs?

If not, the prose itself may lack functional signalling.

174. The combined-section test | 合并 section 测试

If Results and Discussion are combined, can the reader distinguish:

observation;
interpretation;
literature relation;
limitation?

175. The discipline-fit test | Discipline fit 测试

Does your structure resemble strong recent papers in the target field/venue?

176. The 20-minute section drill | 20 分钟训练

Time	Task
4 min	classify 12 sentences by job
4 min	repair section leakage
4 min	map Method → Result lineage
4 min	map Result → Discussion lineage
4 min	certainty/limitation audit

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

Time	Task
8 min	read one target-journal paper
8 min	section-function map
10 min	rewrite weak Methods paragraph
10 min	rewrite weak Results paragraph
9 min	rewrite weak Discussion paragraph

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

Time	Task
15 min	analyse two papers from same field
15 min	build research lineage map
15 min	draft Methods
15 min	draft Results with table/figure commentary
20 min	draft Discussion cycles
10 min	cross-section audit

179. Seven-day research-section cycle | 七天 section cycle

Day 1: genre/venue analysis.
Day 2: Methods transparency.
Day 3: Results selection and ordering.
Day 4: tables/figures and highlighting.
Day 5: Discussion interpretation cycles.
Day 6: limitations/alternatives/implications.
Day 7: full cross-section benchmark.

180. Twelve-week C1–C2 research-writing progression | 12 周路线

Weeks	Focus	Output
1–2	IMRaD/genre architecture	section maps
3–4	Methods transparency	reproducible/evaluable method descriptions
5–6	Results selection + visual reporting	results sections
7–8	Discussion interpretation + literature	discussion cycles
9–10	limitations, alternatives, implications	calibrated discussions
11–12	discipline-specific independent transfer	full research paper section portfolio

181. Monthly benchmark | 每月基准任务

Take one small research question and a fictional or permitted dataset.

Produce:

one clear research question;
Methods section;
one table or figure;
Results section;
Discussion section;
limitations paragraph;
one future-work sentence;
one conclusion;
Method → Result lineage map;
Result → Discussion lineage map;
section-colour audit;
certainty audit.

182. Evaluate architecture, not only grammar | 不要只评 grammar

Track:

method transparency;
research-question alignment;
measure/analysis clarity;
results completeness;
null-result honesty;
table/figure efficiency;
observation–interpretation separation;
discussion explanatory quality;
limitation consequences;
certainty calibration;
cross-section lineage;
discipline/venue fit.

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

Symptom	Likely weak link	Repair
Methods feels like diary	reader job unclear	organise by design/measure/analysis
Results repeats tables	highlighting weak	state main pattern only
Results sounds argumentative	interpretation leakage	move mechanism/implication to Discussion
Discussion repeats numbers	interpretation move missing	finding → meaning → literature → limit
limitation list feels generic	consequence missing	state inference affected
new data appear in Discussion	reporting lineage broken	report analysis transparently in Results
result has no method parent	orphan result	add/clarify method or remove claim
interpretation has no result parent	orphan interpretation	anchor to evidence

184. What not to do | 不要这样写 Methods / Results / Discussion

Do not turn Methods into a chronological diary.
Do not use software names as substitutes for analysis descriptions.
Do not hide recruitment, exclusions or preprocessing that affect inference.
Do not report only results that support the hypothesis.
Do not repeat every table value in prose.
Do not turn Results into a causal story before causal evidence is evaluated.
Do not make Discussion a second Results section.
Do not list limitations without explaining their consequences.
Do not introduce important new primary evidence only in Discussion.
Do not force IMRaD onto disciplines that use different research genres.

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

Purdue Writing Lab | Organization at the Document Level — IMRaD as a common but discipline-dependent structure; sections serve different purposes.
Purdue Writing Lab | Graduate Writers Guide — Methods as analytical tools/procedure, Results as findings aligned to methods, Discussion as interpretation and implications.
University of Manchester Academic Phrasebank | Describing Methods — method description, procedural detail and rationale.
Academic Phrasebank | Reporting Results — systematic presentation, tables/figures and highlighting statements.
Academic Phrasebank | Discussing Findings — result restatement, interpretation, implications, literature relations and future research.
Academic Phrasebank | Being Cautious — calibrated interpretation in Discussion.
Elsevier | Guide for Authors: Article Structure — Methods/Experimental, Results and Discussion roles, including legitimate combined Results and Discussion sections.
Elsevier | Writing the Results Section — negative results, tables/figures and distinction between results and broader discussion.

186. Canonical eduKate routes | eduKate 主页面路由

187. SEO language map | 本课自然覆盖的搜索意图

This lesson naturally serves learners searching for how to write methods section, how to write results section, how to write discussion section, methods results discussion, IMRaD structure, research paper structure, results vs discussion, methods vs results, discussion section academic writing, reporting research results, interpreting findings, research limitations writing, academic research writing for Chinese students, C1 academic writing, C2 academic English, Methods 怎么写, Results 怎么写, Discussion 怎么写, 研究论文结构, IMRaD, 学术英语研究写作, 中文母语学术英语 and C1 C2 research writing.

188. The one-page Methods–Results–Discussion operating system | 一页 research-section 操作系统

Start with the research question. 所有 section 回到 question。
Name the genre/discipline convention. 不假设 universal IMRaD。
Methods: make the process inspectable. 谁、什么、怎样、如何分析。
Results: report what emerged. 不筛掉不漂亮的结果。
Discussion: interpret what the pattern means. 机制、literature、alternative、limit、implication。
Map Method → Result lineage. 每个主要 Result 有 Method parent。
Map Result → Discussion lineage. 每个主要 interpretation 有 evidence parent。
Separate observation from explanation. Fact 与 interpretation 分开。
Use tables/figures to carry patterns. 正文 highlight，不逐格复读。
Report null/negative results honestly. Results 不是成功故事。
State limitation consequence. 不只列 limitation。
Label exploratory/post-hoc analysis. 保留 inferential status。
Calibrate Discussion certainty. 回到 Lesson 014。
Allow legitimate section blending. Combined Results & Discussion 要 move 清楚。
Audit title/abstract/conclusion. 不让压缩制造 certainty drift。
Read target-journal exemplars. C1–C2 写作要有 genre awareness。

189. Final assignment | 最终作业

Choose a small research question and either a permitted dataset, a classroom-safe fictional dataset, or a study of your own work where you are authorised to analyse the material.

Complete this chain:

Write one research question.
Write a 250–400 word Methods section.
Identify every variable/construct and how it is operationalised.
Create one table or figure.
Write a 200–350 word Results section.
Include one null, negative or unexpected finding if it exists.
Write a 350–600 word Discussion section.
Include one literature comparison.
Include one alternative explanation.
Include one limitation and its inference consequence.
Include one specific future-study need.
Write a 100–150 word conclusion.
Create a Method → Result lineage map.
Create a Result → Discussion lineage map.
Run the section-colour audit.
Run the orphan-result audit.
Run the orphan-interpretation audit.
Run the certainty audit.
Compare your structure with three strong papers from the target field or venue.

Then ask:

If I removed the headings, could a careful reader still distinguish what I did, what I found and what I think the findings mean?

如果我暂时删掉 Methods、Results、Discussion 的标题，一个仔细的读者还分得清我做了什么、发现了什么、以及我认为这些发现意味着什么吗？

If yes, your research architecture is becoming clear.

If no, the problem is not only formatting. The intellectual jobs are probably still mixed.

Continue | 继续

The next lesson will compress the whole research architecture into one of the hardest short forms in academic English: the abstract.

下一课进入 academic writing 最困难的短文本之一：怎样把完整研究压缩成一个 abstract，同时不丢掉 design、main result、scope、certainty 与 conclusion。

Next: EDKS-ADV-ZH-0016 · Lesson No.016 · Write an Abstract That Compresses a Study Without Distorting It · 把完整研究压缩成不失真的摘要

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.015 | Keep Methods, Results and Discussion Doing Different Jobs | 第015课：让方法、结果与讨论各司其职

Keep Methods, Results and Discussion Doing Different Jobs | 让方法、结果与讨论各司其职

Before we begin: IMRaD is a useful architecture, not a universal law | IMRaD 是重要结构，但不是万能定律

1. The three core jobs | 三个核心任务

2. Manchester separates these jobs explicitly | Manchester 的 Academic Phrasebank 也按 section job 组织

3. Elsevier states the same broad division | Elsevier 的 author guidance 也强调 section role

4. Section identity reduces epistemic confusion | Section identity 降低知识混乱

5. Section separation makes argument auditing possible | 分开以后，读者才能 audit

6. Methods is not a diary | Methods 不是实验日记

7. Manchester’s core Methods principle | Methods 的核心：可检查、可理解、足够详细

8. Replicability is field-dependent | “可重复”也要看 discipline

9. Methods should make the inferential chain inspectable | Methods 要让 inference chain 可检查

10. The Methods section is an argument about trustworthiness | Methods 其实也在回答“为什么应该信这些结果”

11. Name the design | 先说 design 是什么

12. But design labels do not replace procedural detail | Design 名字不能代替真正方法

13. Participants/population belong in Methods | 谁被研究，属于 Methods

14. Recruitment can affect the meaning of Results | Recruitment 会改变结果可推广性

15. Sample-size rationale belongs near design decisions | Sample size rationale 属于 Methods

16. Do not put the observed outcome into the sample rationale | 不要把 result 倒灌进 Methods

17. Variables and constructs must be operationalised | Construct 必须变成可测量对象

18. Operational definition prevents result ambiguity | Operational definition 防止 Results 变模糊

19. Instruments belong in Methods | Instrument 属于 Methods

20. Instrument validity/reliability may need justification | 工具是否合适，需要说明

21. Procedure should follow reader logic | Procedure 不一定照现实时间流水账

22. Subheadings can reduce cognitive load | Methods 可用 subheadings

23. Tense in Methods is functional, not mechanical | Methods 时态不是死规则

24. Passive voice is not automatically “academic” | Passive voice 不是高级写作标签

25. Methods should not hide agency when agency matters | 需要责任时，不要藏 actor

26. Analysis plan belongs in Methods | Analysis plan 属于 Methods

27. Statistical procedure should answer the design | 分析方法必须匹配问题

28. Software is not the method | SPSS / R / Python 不是分析逻辑

29. Preprocessing can materially affect results | Preprocessing 也属于 inferential chain

30. If preprocessing changes the evidence, document it | 会改变 evidence 的处理必须可追踪

31. Qualitative Methods need analytical transparency too | Qualitative Methods 也要透明

32. “Thematic analysis was performed” is often too thin | 只写 thematic analysis 不够

33. Mixed Methods requires integration logic | Mixed Methods 不只是做两种数据

34. Methods limitations can appear in Methods or Discussion depending genre | 方法限制放哪里看功能

35. The result of a manipulation check may appear in Results | Manipulation check 是 result

36. Pilot-study outcomes can blur the boundary | Pilot 信息要判断角色

37. Results answers “what emerged?” | Results 的主问题：观察到了什么？

38. Results is not raw data dump | Results 不是 data warehouse

39. Data, result and interpretation are different | Data / result / interpretation 不同

40. Results usually transform data into informative patterns | Results 是 data 的结构化描述

41. But Results should not outrun the evidence | Results 不应该提前写完整解释

42. Organise Results around research questions or hypotheses | Results 结构要跟 question 对齐

43. Purdue recommends alignment between Methods and Results | Methods–Results 要有 alignment

44. Mirror structures can help | Methods 与 Results 可形成 mirror

45. Report important negative/null findings | 不支持 hypothesis 的结果也属于 Results

46. Selective reporting destroys trust | 只报“漂亮结果”会破坏可信度

47. Primary outcomes deserve priority | 主 outcome 要先清楚

48. Exploratory analyses should be marked as exploratory | Exploratory ≠ preregistered confirmatory

49. Result order can follow importance, question order or analytic sequence | Results 顺序有多种合法方案

50. Tables and figures should carry evidence, not decoration | Table/Figure 是 evidence surface

51. Text should not repeat every table cell | 正文不要逐格抄表

52. Results text answers “what should I notice?” | 正文告诉读者看什么

53. Do not use figures to hide uncertainty | 图表也要显示 uncertainty

54. Visual scale can exaggerate effects | 坐标轴也会制造 rhetoric

55. Captions should be self-sufficient enough to orient the reader | Caption 也有 job

56. Avoid evaluation adjectives in Results unless defined | Results 少用“excellent / impressive / disappointing”

57. “Significant” must mean what you intend | significant 是高风险词

58. Descriptive statistics should describe | Descriptive statistics 回答“数据长什么样”

59. Inferential statistics should answer an inferential question | Inferential result 要对应 inferential question

60. Exact reporting conventions are discipline-specific | 数字格式看 discipline/style guide

61. Results certainty should usually be higher for direct observations | 直接 observation 可以更直接

62. Interpretation certainty should be calibrated separately | 原因解释另外 calibration

63. Keep measured fact and explanation in separate grammatical units | Fact 与 explanation 分开写

64. Qualitative Results are not “just quotes” | Qualitative Results 不是 quote collection

65. Quote should illustrate an analytic claim | Quote 服务于 theme

66. Negative cases matter in qualitative Results | 反例/偏离案例不能全部删掉

67. Frequency language needs caution in qualitative work | “most / many / some” 要有依据

68. Mixed-methods Results require strand clarity | Mixed methods 要让两种 evidence 不混成 fake score

69. Integration can occur in Results, Discussion or both | Mixed Methods integration 看设计

70. Discussion begins where interpretation becomes central | Discussion 从“它意味着什么”开始

71. Discussion should not merely repeat Results | Discussion 不是 Results 第二遍

72. Start with the finding, then move beyond it | Discussion 可以重述关键 result，但要立刻解释

73. Discussion has several recurring jobs | Discussion 的常见 moves

74. Not every Discussion needs every move | 不是每一项都必须出现

75. Discussion cycles can repeat per major finding | 每个重要 finding 可以有一个 discussion cycle

76. Interpretation should answer “why might this pattern exist?” | Interpretation 不是简单重述