SciSpace 发布文献综述写作指南：附注释范例与七步流程

You have read the papers, but your literature review still sounds like one study after another.

A good example shows what changes: the writing is organised around an idea, the studies are compared, and the research gap follows from the evidence.

This guide breaks down a short literature review example and shows how to build a similar synthesis from your own sources. The approach combines guidance from 3 SciSpace webinars on planning a review, organising evidence and keeping the researcher in control of the interpretation.

The process at a glance

Literature review example

(The studies below are fictional teaching examples created for this guide.)

Review question: How do postgraduate students organise and retrieve notes from research papers during academic writing?

Worked synthesis paragraph

Digital note-taking tools can make research notes easier to retrieve, but the evidence suggests that the organisation system matters more than the presence of a tool alone. In interviews, postgraduate students reported that topic tags helped them relocate ideas during writing (Study A). An observed retrieval study, however, found that inconsistent labels slowed retrieval even when tagging was available (Study B). These studies address related but different parts of the problem: one reports perceived usefulness, while the other measures retrieval behaviour. A survey showing high satisfaction with a note-taking application cannot yet resolve this difference because it did not directly measure retrieval performance (Study C). Together, the evidence points to a need for research that compares students' perceptions with observed retrieval under clearly defined tagging systems.

What each sentence does

The example does more than report what three studies found. It explains how the findings relate, where the evidence differs and what still needs to be studied.

1. Define what your literature review needs to explain

Write the question your review needs to answer before collecting more papers. The question helps you decide which evidence belongs and which details can remain outside the review.

Dr. Lyndon Walker explains that the depth of a literature review should match its purpose. A PhD review, a journal article and a rapid briefing all require accurate evidence, but they do not require the same breadth or level of synthesis.

For a focused review section, complete these three lines:

1. Topic: This review examines [subject].
2. Question: It asks [specific question].
3. Purpose: The answer is needed to [explain the study, justify a decision or establish a gap].

A question such as “How do students use digital notes?” is still broad. “How do postgraduate students organise and retrieve notes from research papers during academic writing?” specifies the group, behaviour and context.

Keep the question visible while reading. Include a source when it helps answer that question, challenges an emerging explanation or provides context the reader genuinely needs.

Try this

2. Organise the evidence before you start writing

A literature review becomes difficult to write when every paper is recorded differently. Extract the same core details from each source so that you can compare them later.

Dr. Ethan Gantana demonstrates this by restricting SciSpace to a selected Zotero collection and specifying the fields he wants from every paper. His table includes the study design, findings, methodology and critical appraisal. He then uses the table as material for his own synthesis.

Start with fields that answer your review question:

Keep the authors' findings separate from your interpretation. This prevents a note such as “participants liked the tool” from gradually becoming the stronger claim “the tool improved performance.”

Also record whether you reviewed the full text or only an abstract. Ethan's demonstration distinguishes papers with accessible PDFs from papers summarised from abstracts. An abstract may identify the headline finding, but it may not contain the methodological detail required for comparison.

The synthesis remains my responsibility, but SciSpace can help extract the information for me.

Dr. Ethan Gantana

3. Check whether the studies are actually comparable

Shared keywords do not guarantee that studies examine the same thing. Before combining findings, compare the definitions, populations, methods and outcomes behind them.

Prof. Benita Olivier gives the example of “adherence.” In one paper, adherence may describe whether participants completed a programme over several weeks. In another, it may describe whether they performed an exercise correctly during one session. A fluent summary can flatten those different constructs into one result unless the researcher notices the distinction.

Use these checks:

In the worked example, Study A reports what students say about tags. Study B observes how quickly they retrieve notes. Both are relevant, but they measure different forms of evidence. Study C reports satisfaction, which cannot be treated as a direct measure of retrieval.

This is where critical reading adds value. The goal is not to make every paper fit the same conclusion. It is to explain what can and cannot be inferred from the comparison.

4. Group the studies around ideas, not author names

Once the evidence is checked, group studies by the point they help you explain. A useful theme may be a shared finding, a methodological difference, a debate or an unresolved question.

Avoid organising the review as: Study A found this. Study B found that. Study C found something else.

That structure reports the reading but leaves the reader to work out why the studies belong together.

Instead, create a paragraph plan:

• Point to explain: What does this paragraph need to establish?
• Supporting evidence: Which studies address the point?
• Comparison: Where do they agree, differ or examine different parts of the issue?
• Interpretation: What does the relationship mean for the review question?
• Unresolved point: What remains unknown?

Prof. Olivier recommends creating your own outline before asking AI for one. Starting with your reasoning makes it easier to notice what an AI-generated structure has omitted. She also recommends examining strong completed reviews for their flow, headings, frameworks and use of summary tables, without copying their content.

For the example above, the theme is not “Study A, Study B and Study C.” The theme is how the organisation of notes affects retrieval. The studies are brought in because each contributes a different piece of evidence to that idea.

5. Write a synthesis, not a list of summaries

Synthesis explains the relationship between sources. It uses the evidence to make a point the reader could not get from any one paper alone.

Summary-only version:

Study A found that students considered tags useful. Study B found that inconsistent labels slowed retrieval. Study C found that students were satisfied with a note-taking application.

Every sentence may be accurate, but the paragraph does not explain the relationship between the findings.

Synthesised version:

Students' perceptions of a note-taking system and their observed ability to retrieve notes address different parts of the problem. Study A suggests that topic tags are perceived as useful, whereas Study B shows that inconsistent labelling can still make retrieval difficult. Study C reports satisfaction with a note-taking application, but because retrieval was not directly measured, it provides limited evidence for the question examined here.

The second version does four things:

1. Organises the paragraph around an idea.
2. Compares the sources rather than listing them.
3. Preserves differences in methods and outcomes.
4. Qualifies the conclusion at the strength the evidence supports.

Prof. Olivier emphasises that the researcher should read the core papers and write the synthesis in their own voice. AI can help surface themes and debates, but the final explanation must reflect the researcher's understanding.

6. Make the research gap follow from the evidence

A research gap is not simply a topic with few papers. It is a meaningful question that the existing evidence does not yet answer well.

The worked example does not claim that nobody has studied digital note-taking. It identifies a narrower gap: existing studies may describe students' perceptions or satisfaction, but the selected evidence does not yet show how those perceptions relate to observed retrieval under consistent and inconsistent tagging systems.

Use this structure:

1. What is known: Existing studies show [supported pattern].
2. What is uncertain: The studies differ or provide limited evidence about [specific issue].
3. Why it matters: Resolving this would help [research or practice consequence].
4. What should be studied next: Research should examine [focused unanswered question].

Dr. Ethan uses AI-generated gap suggestions as questions to investigate, not conclusions to accept. Lyndon makes the same broader point: AI can refine the researcher's thinking, but it cannot supply the intellectual ownership of the review.

Check the proposed gap against the original papers. Look for recent studies, alternative terminology and evidence outside the first search. If the answer already exists, refine the question rather than preserving a convenient gap.

7. Check every claim and improve the flow

Read the draft as a connected argument. Each paragraph should make one clear point, support it with appropriate sources and explain why the comparison matters.

Use this final check:

• Question: Does each section help answer the review question?
• Evidence: Does every factual claim have appropriate support?
• Comparison: Have you explained why studies agree, differ or cannot be combined?
• Qualification: Is the wording as strong or as cautious as the evidence requires?
• Gap: Does the gap follow from the reviewed evidence?
• Flow: Does each paragraph lead logically to the next?

Return to the original paper whenever a sentence depends on a specific method, population, result or limitation. A real citation can still be used inaccurately if the finding is misquoted, the wrong population is described or an abstract is treated as a full-text analysis.

Benita's final reminder is useful here: AI can assist, but the researcher remains in charge. Read the papers you plan to cite, verify the claims and keep the analytical voice yours.

Put the example into practice

A strong literature review makes three things clear: what the evidence shows, where the studies differ and what remains unresolved. Start with one paragraph. Build the evidence table behind it, check the comparison and write the relationship in your own words.

Use the SciSpace tasks linked throughout this guide to organise sources, compare studies or review a draft then keep only the output you can verify and defend.

The expert guidance behind this guide

Prof. Benita Olivier is a professor at Oxford Brookes University and the Allied Health Professionals Research Lead at Oxford Health NHS Foundation Trust. Her webinar focuses on planning a narrative literature review, studying strong examples, reading core papers and keeping the final synthesis in the researcher's voice.

Dr. Ethan Gantana is a senior lecturer at Stellenbosch University and a pathologist with the National Health Laboratory Service. His session demonstrates a researcher-led workflow for curating papers, building a cross-paper evidence table and checking the details before synthesis.

Dr. Lyndon Walker explains how the depth and rigour of a literature review should match its purpose, and why AI should support exploration, verification and understanding without replacing the researcher's thinking.

Frequently asked questions

注意转载授权 · 阅读原文