本文由 UniWriterPro 學術規例研究組 verify · Last updated: 2026-08-28 · 本文係按 HKUST CSE FYP guide、PolyU ELC RPG thesis guidelines、HKU CAES General Discussion 同 Manchester Academic Phrasebank 逐條解讀，唔會亂寫。
Supervisor 睇完你 draft discussion chapter，喺 margin 度寫：「You’re just repeating your results」／「Where’s the interpretation?」／「This is a Results section, not Discussion」。你自己讀返都覺得每一段開頭都係「The results show that…」、「This finding indicates that…」——係咪已經算 interpret？
唔係。fyp discussion 點寫呢個問題，關鍵唔係「加多啲字」，係將每個 finding 由 restate（重講一次結果）轉成 interpret（解釋點解會咁樣、對 literature 有咩意義、對 research question 答咗咩）。呢篇文由 HKUST CSE FYP、PolyU ELC、HKU CAES 同 Manchester Academic Phrasebank 官方指引，拆解 5 個具體 rewrite pattern，每個都有 before／after 對照。
如果覺得太難嘅話， 可以直接搵我哋幫手代做—— FYP / Final Year Project 代做全程支援
目錄
- Supervisor 講「只係 repeat results」實際上批緊乜嘢
- Pattern 1：由「finding is X」改成「finding suggests why X」
- Pattern 2：加返 literature 對照——同 whose finding 一致／相反
- Pattern 3：用 hourglass 結構——由 specific finding 拉去 general implication
- Pattern 4：處理 unexpected result——唔好淨係「the finding is unexpected」就完
- Pattern 5：加 tentative signposting——用 phrasebank 語感講 cautious interpretation
- Discussion 同 Conclusion 分別
- Marker 睇到嘅 signal
- FAQ
Supervisor 講「只係 repeat results」實際上批緊乜嘢
官方指引對 discussion chapter 有一個 explicit 定義，你先要知道 supervisor 用咩 baseline 判斷。HKUST CSE FYP guide 開章第一句：
In this section you interpret your results based on your own opinions and experience.
HKUST CSE FYP – Discussion
PolyU ELC 對 discussion chapter 目的講得更 specific：
The fundamental objective of the Discussion chapter in a thesis is to deliberate on the key findings of the research so as to establish their significance and demonstrate how they address the research gap identified by the current study.
PolyU ELC – Guidelines for Thesis Writing
兩個關鍵動詞：interpret 同 establish significance。Restate 係咁：「The results show that X causes Y」；interpret 係咁：「The results show that X causes Y, which suggests [某個 mechanism／理論解釋], and this addresses the research gap identified by [某個 prior study]」。
UCI Libraries 有一句直白到唔可以再直白嘅 rule：
Also do not repeat your results, this is a discussion.
UCI Libraries – Writing a Scientific Paper: Discussion
Supervisor 寫「just repeating results」嗰陣，佢想見到嘅係接住 4 件事其中一件（HKU CAES 講到）：
- 對比返 literature：呢個 finding 支持定推翻邊個 prior study？
- 解釋 mechanism：點解會出現呢個 finding？可能背後嘅 causal 理由係乜？
- 回應 research question：呢個 finding 對你 introduction 提出嘅 question，答咗乜？
- 提 implication：呢個 finding 對 practice／policy／後續研究有咩意義？
HKU CAES 講到 discussion 應該點樣同 introduction 呼應：
It should match the Introduction in the way it moves from a close focus on your own study — to relating that study more broadly, to other work in the field.
HKU CAES – General Discussion
下面 5 個 rewrite pattern，每個對應返其中一種「由 restate 走去 interpret」嘅具體技術。全部用同一個 running example：一個 hypothetical FYP 研究 mobile trading app usability 對 retail investor 交易頻率嘅影響。
好擔心你份 FYP? 即時睇吓每一 part 點做：FYP 題目點揀、FYP Proposal 點寫、FYP 研究方法點揀、Literature Review 點先展示到你嘅 critical engagement? 、SPSS 結果論文點寫、論文剩返兩個禮拜？即睇點應對
Pattern 1：由「finding is X」改成「finding suggests why X」
最常見嘅 restate 陷阱：每段開頭都係「The finding shows…」，跟住只係將 results chapter 嘅數字用另一句話再講一次。Rewrite pattern 係：喺 restate 之後接一個 why-explanation，即係俾個 possible mechanism。
Manchester Academic Phrasebank 對呢類 explanation 有 explicit sentence starter：
A possible explanation for this might be … / An alternative explanation for this result is that it is due to … / It is difficult to explain this result, but it might be related to …
Manchester Academic Phrasebank – Discussing findings
Before：「The results show that respondents who rated the app UI as easier to use had a 35% higher weekly trading frequency than those who rated it as difficult. This finding indicates that UI usability is associated with trading frequency.」
問題：第二句「indicates that UI usability is associated with trading frequency」只係將第一句嘅 correlation 翻譯成句子，冇提供任何 mechanism。呢個係典型 restate。
After：「The results show that respondents who rated the app UI as easier to use had a 35% higher weekly trading frequency than those who rated it as difficult. A possible explanation for this might be that lower cognitive friction during order entry reduces the decision-abandonment rate — a mechanism consistent with Kahneman’s (2011) System 1 processing framework, where lower cognitive load correlates with higher action likelihood.」
後半段做咗兩件事：(1) 提供一個 causal mechanism（cognitive friction → decision abandonment），(2) 將呢個 mechanism 錨定去 established framework（Kahneman System 1）。呢個先叫 interpret。
Pattern 2：加返 literature 對照——同 whose finding 一致／相反
PolyU ELC 對 discussion chapter 開列六大元素，其中一項係：
Comparison of results obtained with those from existing research
PolyU ELC – Guidelines for Thesis Writing
好多 discussion draft 完全冇引 literature，或者只喺文末總結一句「This finding is consistent with previous research」就過，冇講具體係邊份研究、點樣一致。
Manchester Phrasebank 提供咗兩組具體 sentence starter：
Supporting previous findings: This finding was also reported by Smith et al. (1989). / This finding is consistent with that of Smith (2000) who … / Contradicting previous findings: This outcome is contrary to that of Smith et al. (2001) who found … / However, the findings of the current study do not support the previous research.
Manchester Academic Phrasebank – Discussing findings
Before：「The results indicate a significant positive correlation between mobile trading app usage frequency and portfolio churn rate. This is consistent with previous studies.」
問題：「consistent with previous studies」係空字。邊份 study？點樣 consistent？如果 consistent，你篇 FYP 又有咩貢獻？
After：「The results indicate a significant positive correlation between mobile trading app usage frequency and portfolio churn rate (r = 0.42, p < 0.01). This finding is consistent with that of Barber and Odean (2002) who reported that online trading access increased individual investor turnover by 74%, and it further extends their findings to a mobile-first context. However, this outcome is contrary to that of Choi et al. (2019) who found no significant increase in churn among mobile investors under 30, suggesting that the age-cohort effect observed in their sample may not generalise to the demographic profile of the current study.」
Rewrite 之後做咗三件事：(1) 引具體 study 名 + year，(2) 講出對比嘅具體維度（mobile-first context extends；age cohort contradicts），(3) 用「extends」／「contrary to」呢類 verb 表明你自己個 finding 對 literature 嘅 position。
Pattern 3：用 hourglass 結構——由 specific finding 拉去 general implication
HKU CAES 用咗 hourglass model 描述 discussion 應該點樣同 introduction 呼應：
Here is the bottom half of the hourglass model. It illustrates how the Discussion mirrors the Introduction. Note the more inductive shape which corresponds to the unfolding of the Discussion.
HKU CAES – General Discussion
Hourglass 意思：Introduction 由 broad topic 收窄去你 specific research question；Discussion 反轉，由 specific finding 慢慢拉闊返去 general implication。
Manchester Phrasebank 提供咗 implication 呢層嘅 sentence starter：
An implication of this finding is the possibility that … / One of the issues that emerges from these findings is … / These results have implications for … / These results may help us to understand …
Manchester Academic Phrasebank – Discussing findings
Before（只有 specific finding）：「The findings show that push notifications with real-time price alerts increased trading frequency by 22% compared to daily-digest notifications. Users who received real-time alerts also reported higher perceived urgency scores.」
問題：finding 只停留喺樣本層面，冇 zoom out 去 general implication，讀者讀完唔知「so what」。
After：「The findings show that push notifications with real-time price alerts increased trading frequency by 22% compared to daily-digest notifications. At the individual level, this suggests that notification design directly influences retail investor decision cadence. At the platform-design level, an implication of this finding is the possibility that trading apps’ default notification settings function as a soft nudge toward higher-frequency trading — a design choice with potential financial-wellbeing consequences for users who are not aware of the behavioural pattern. This raises broader questions about the regulatory treatment of default notification settings in consumer financial products, an area that has received limited attention in the Hong Kong context.」
三層 hourglass 由窄至闊：(1) 樣本層 finding，(2) 個人／平台層 implication，(3) regulatory／field-level 廣義問題。呢個唔係扮 profound——係按 CAES 講嘅 「moves from a close focus on your own study to relating that study more broadly」。
Pattern 4：處理 unexpected result——唔好淨係「the finding is unexpected」就完
好多 draft 遇到 unexpected result 只寫一句「This finding was unexpected」就跳去下一個 finding，好似 unexpected 本身就係 conclusion。Manchester Phrasebank 拎出 unexpected outcome 做獨立一個 discussion cycle 元素：
Contrary to expectations, this study did not find a significant difference between … / One unanticipated result was that … / This finding was unexpected and suggests that …
Manchester Academic Phrasebank – Discussing findings
關鍵係 「suggests that …」——unexpected 之後要接一個 hypothesised explanation，唔可以停喺「unexpected」。UCI 亦講：
Discuss possible alternate explanations for the results.
UCI Libraries – Writing a Scientific Paper: Discussion
Before：「Contrary to hypothesis, users aged 45+ demonstrated higher app engagement than users aged 25–34. This finding is unexpected.」
問題：宣佈完 unexpected 就冇後續。Marker 會覺得你自己都唔識點解釋自己個 finding。
After：「Contrary to hypothesis, users aged 45+ demonstrated higher app engagement than users aged 25–34. One unanticipated result was that the older cohort exhibited longer session duration (mean = 8.2 mins vs 4.7 mins). Several factors could explain this observation. Firstly, the older cohort in the sample were predominantly semi-retired individuals with more discretionary time, whereas the 25–34 cohort included working professionals with fragmented app-use windows. Secondly, the older cohort reported higher self-directed investment activity, which may have driven more deliberate app engagement rather than casual checking. This finding suggests that time-availability and investment-role variables may confound the age-engagement relationship, and future work with time-diary controls would be needed to isolate the age effect.」
Rewrite 有三層：(1) 承認 unexpected，(2) 提出兩個 possible explanation，(3) 講明呢個 finding 對後續研究嘅 implication。呢個先叫 handle 一個 unexpected result。
Pattern 5：加 tentative signposting——用 phrasebank 語感講 cautious interpretation
Discussion chapter 學術寫作有個 register 特徵：唔可以太 confident。你篇 FYP 樣本可能得 200 人、時間可能得 3 個月、方法可能只係 cross-sectional survey——寫嘅時候要用 tentative language 承認 limitation，唔係扮 finding 係 established fact。
Manchester Phrasebank 對 tentative interpretation 有 explicit reminder：
Note that when offering explanations and suggesting implications the language used is very tentative or cautious.
Manchester Academic Phrasebank – Discussing findings
常用 tentative 動詞 / phrase：
- Suggest（唔用 prove／show／demonstrate）：「These findings suggest that …」
- May／might：「A possible explanation may be …」
- Appears to：「The results appear to indicate …」
- Could：「This could be attributed to …」
- Tentative + limitation：「These results must be interpreted with caution given the sample size of …」
Before：「This study proves that mobile trading app usage causes higher portfolio turnover among retail investors. The findings demonstrate the causal effect of notification frequency on trading behaviour.」
問題：用「proves」／「demonstrate the causal effect」呢類語言，樣本＋方法都撐唔起。Marker 會第一時間圈起呢兩個 verb。
After：「These findings suggest that mobile trading app usage may be associated with higher portfolio turnover among retail investors, and the observed relationship between notification frequency and trading behaviour points to a possible causal pathway. However, these results must be interpreted with caution: the cross-sectional design cannot establish temporal ordering, and the self-reported nature of trading frequency introduces potential recall bias. Longitudinal work with app-usage log data would be needed to confirm whether the notification-trading relationship holds over time.」
Rewrite 動詞由「proves／demonstrate」改成「suggest／may be associated with／points to a possible」。同時加返 limitation（cross-sectional、self-report），最後提出 future work 方向。呢個 register 先係 marker 期待嘅 discussion voice。
Discussion 同 Conclusion 分別
另一個常見混淆：Discussion 同 Conclusion 撈埋。PolyU ELC 講得明兩者分工：
This chapter [Conclusion] be entirely based on the findings and observations of the current study. This means you should not introduce any new ideas for discussion or observations.
PolyU ELC – Guidelines for Thesis Writing
Rule of thumb（呢個係常用 practice，冇官方統一 threshold）：
- Discussion：拎每個 finding 逐個 interpret（引 literature、講 mechanism、提 implication、承認 limitation）。字數通常佔 Final Report 20–30%。
- Conclusion：總結成個 study 答咗咩 research question、有咩 key contribution、有咩 recommendation。冇新 finding、冇新 literature 引入。通常 800–1,500 字。
如果你發現 draft 嘅 conclusion 章突然引咗新 study 或者展開新 argument，通常代表嗰啲內容應該搬去 discussion 章。
Marker 睇到嘅 signal
如果你 rewrite 完想自己 sanity check，marker 讀 discussion 通常會 quick scan 呢幾樣：
- 每段開頭 verb：如果 90% 段落係「The results show / indicate / demonstrate」開頭，就係 restate 為主。健康嘅 discussion 應該混合「suggests / may explain / is consistent with / contrary to」呢類 interpret verb。
- Literature citation 密度：Discussion 章 citation 密度應該接近 literature review 章，唔係得零星幾個。
- Tentative vs assertive language 比例：Discussion 應該以 tentative 為主（may、suggest、appears）——見到大量 proves／demonstrates／confirms 就係 register 問題。
- Hourglass 結構：每個 finding 段落有冇由 specific 拉去 general？定係全部停喺樣本層？
- Unexpected result 處理：有冇解釋定係只係報咗 unexpected？
呢啲 signal 唔係 rubric，係 supervisor 打回頭嘅高頻 pattern。你自己過一次 draft 都可以 catch 到 90% 問題。
FAQ
Discussion chapter 應該幾多字？
搵唔到官方文件講明統一字數。PolyU ELC、HKU CAES、HKUST CSE FYP 均冇指定 discussion chapter 字數。實際字數以你院校 FYP handbook 為準。常見 practice 係 Final Report 總字數嘅 20–30%，例如 10,000 字報告嘅 discussion 大約 2,000–3,000 字，但呢個唔係官方 threshold。
Discussion 要引幾多篇 literature？
搵唔到官方文件講明具體篇數。PolyU ELC 明確要求「Comparison of results obtained with those from existing research」——即係一定要引，但冇規定篇數。實務常見 rule of thumb 係每個 major finding 至少對應 1–2 份 supporting 或 contradicting study。以你 supervisor 要求或者你院校 FYP handbook 為準。
Findings 章同 Discussion 章可唔可以合併？
視乎院校要求。PolyU ELC 明確提到有兩種 model：「unless you have decided to present results and discuss their significance within the same chapter, i.e. Results and Discussion」。即係「Findings and Discussion 合併章」係認可 model，但要一致落成一個 model，唔可以有時分有時合。以你院校 FYP handbook 或者 supervisor 指引為準。
唔明個 finding 嘅 mechanism，可唔可以照寫「possible explanation is X」？
可以，而且係 discussion chapter 應有做法。Manchester Phrasebank 明確提供「It is difficult to explain this result, but it might be related to …」呢個 sentence starter。承認 uncertainty 加上一個 possible mechanism，比扮 confident 講死一個原因更 academic。但 explanation 要合理——例如你可以引 established theory 支撐，唔可以純靠估。
我 discussion 章寫完成篇冇引 literature，係咪一定要重寫？
唔一定係全篇重寫，但 supervisor 好大機會會打回頭。PolyU ELC 明確列 「Comparison of results obtained with those from existing research」做 discussion chapter 六大元素之一——完全冇 literature 對照嘅 discussion 章通常過唔到 marker rubric。修訂方向：喺每個 major finding 段落插入至少一個 supporting 或者 contradicting study 對照。
Discussion 應該包唔包 limitation？定係擺去 Conclusion？
兩種 practice 都有院校用。PolyU ELC 直言：「If the author presents the limitations in the Conclusion, it is done to highlight shortcomings in the research and present them as aspects that future research could address… Students are advised to check with their supervisor to determine the approach to discussing limitations that their specific discipline adopts」。即係擺 discussion 定 conclusion 都有 disciplinary convention 差別，要問 supervisor。
Supervisor 話我 discussion 「lacks depth」，具體要點做？
「Lacks depth」通常係以下三個問題其中之一：(1) restate 為主冇 interpret，(2) 冇同 literature 對照，(3) 停喺樣本層冇拉去 general implication。用本文嘅 Pattern 1（加 mechanism）、Pattern 2（引 literature）、Pattern 3（hourglass）逐段檢查你 draft。搵唔到官方文件講明「depth」統一 rubric，但呢三個 pattern 係最常見 fix。
我用 tentative 語言（「suggest」／「may」）會唔會太 uncertain？
唔會。Manchester Academic Phrasebank 明確指出：「Note that when offering explanations and suggesting implications the language used is very tentative or cautious.」 Tentative 語言係學術 register 特徵，唔係示弱。反而過度 confident 嘅動詞（proves／demonstrates／confirms）先會令 marker 質疑你 methodology。
Rewrite 咗 3 次 supervisor 都話 discussion 唔夠 interpretation？
如果你已經按上面 5 個 pattern 逐段檢查、insert 咗 literature 對照、加咗 mechanism、由 restate 改成 tentative interpret，但 supervisor 仲係第 3 次打回頭話「still descriptive」／「where’s the interpretation」，通常唔係你唔識用 pattern，係 個別 finding 同 literature 嘅具體 anchor point 未搵對——即係你未搵到最適合對照嗰篇 study，或者 mechanism 嘅 theoretical framework 選錯。
呢啲情況 pattern-level rewrite 幫唔到，需要有人 chapter-by-chapter 睇你 finding 逐個匹配返 literature、逐個檢查 mechanism 揀邊個 framework 最 fit。UniWriterPro 提供由 human writer 起草嘅 FYP 代寫，可以由你已經寫嘅 draft 接住做 discussion chapter 逐段 revise，交稿附 Turnitin 原創度報告同 AI 偵測報告。如果你 draft 只係其中一兩段有問題，都可以只交嗰幾段做 學術修改服務，唔一定要成篇重做。
提交時話我哋知：(1) supervisor feedback 原話、(2) 你 draft 而家嘅 discussion chapter、(3) 你 finding 對應嘅 results chapter、(4) deadline。24 小時內覆你一個報價，確認咗先付款。