How to Do a Literature Review for a Research Paper
A dependency-ordered workflow for moving from a research question and search record to a defensible synthesis rather than a paper-by-paper summary.
Next step
Choose the next useful decision step first.
Use the guide or checklist that matches this page's intent before you ask for a manuscript-level diagnostic.
Quick answer: Do a literature review by defining the exact question and review type, documenting a reproducible search, screening sources with explicit criteria, extracting comparable evidence, synthesizing agreements and conflicts, and writing a bounded gap claim. Do not draft a chain of one-paragraph paper summaries and call it synthesis.
Evidence basis: We reviewed the PRISMA 2020 statement, the PRISMA-S search-reporting extension, and Purdue OWL guidance on synthesizing sources on September 6, 2026. Official sources establish applicable reporting duties; Manusights guidance supplies the working matrix. This workflow cannot predict acceptance or an editorial decision.
Review dependency | Record before writing | Output |
|---|---|---|
Question | Population, system, intervention or exposure, comparator, outcome, context | A bounded review question |
Review type | Narrative, scoping, systematic, or another defined approach | The correct rigor and reporting route |
Search | Databases, dates, strings, filters, and supplementary methods | A reproducible search record |
Selection | Inclusion, exclusion, duplicate handling, and screening decisions | A defensible evidence set |
Extraction | Claim, design, sample, measure, result, uncertainty, and limitation | A cross-study evidence matrix |
Synthesis | Agreement, conflict, heterogeneity, bias, and missing evidence | A supported gap and contribution |
Inputs and expected output
Gather the research question, target journal or course requirements, review type, databases available, date boundary, key concepts and synonyms, inclusion and exclusion criteria, and a reference manager or structured record.
The expected output is not “everything published on the topic.” It is a traceable argument explaining the current evidence, its limits, and the specific unresolved question that the paper can honestly address.
Define the review before searching
Write one sentence that fixes the unit of analysis and the decision the review must support. Then choose the review type. A narrative review can explain a field and its tensions, but it should not imply exhaustive retrieval. A systematic review needs a prespecified, reproducible method and the reporting framework appropriate to its question.
PRISMA 2020 primarily addresses reporting of systematic reviews, with extensions for other evidence-synthesis jobs. Do not add a flow diagram to an informal review merely to make it look systematic.
Build a search record that another person could inspect
- Split the question into concepts and collect genuine synonyms for each.
- Choose databases for the discipline rather than relying on one general search engine.
- Save exact search strings, database names, run dates, filters, and result counts.
- Add citation chasing or expert-recommended sources as separate supplementary methods.
- Record duplicate removal and every later exclusion rule.
Use the peer-reviewed-journal finder to verify unfamiliar source venues. Use the citation verification guide before the final reference list is locked.
Extract evidence by claim, not by paper
Create one row per study and columns that allow real comparison. At minimum capture the study question, design, population or system, sample, exposure or intervention, comparator, outcome definition, main estimate or finding, uncertainty, and limitation.
Then add a claim column. Which sentence in the future review could this study support? A source that is relevant to the topic but supports no planned claim may be background, not central evidence.
Worked example: from summaries to synthesis
Weak draft: “Lee et al. found an association. Gomez et al. found no association. Patel et al. used a larger sample.”
Evidence matrix: Lee studied short follow-up in one hospital; Gomez measured a different outcome in a community cohort; Patel used multiple hospitals but excluded older participants.
Synthesis: “The apparent disagreement tracks the outcome definition and recruitment setting rather than sample size alone. Evidence is consistent for the short-term hospital outcome, but transfer to community and older populations remains uncertain.”
The synthesis makes the boundary visible. It neither votes by paper count nor hides counterevidence.
Turn the matrix into a section structure
Organize sections around claims, mechanisms, methods, populations, or contradictions that answer the question. Within each section:
- State the provisional conclusion.
- Compare the strongest supporting and conflicting evidence.
- Explain whether design, measurement, population, or bias accounts for the difference.
- State the limit that survives the comparison.
- Connect that limit to the paper's question without promising more than the study can deliver.
The research-paper outline guide can convert these evidence dependencies into the manuscript structure.
Failure states and recovery
Symptom | Diagnosis | Recovery |
|---|---|---|
One paragraph per source | Notes were organized by citation | Re-sort the matrix by claim and boundary |
Search never ends | The question and date boundary are unstable | Freeze scope and record an update rule |
Every paper “supports” the thesis | Counterevidence was screened out or softened | Run a contradiction search and add negative findings |
The gap is “few studies exist” | The consequence of missing evidence is unclear | Name the population, measure, mechanism, or decision left unresolved |
A systematic label lacks a method | Reporting language outran the actual process | Apply the appropriate standard or relabel honestly |
Completion checklist
- The question and review type are explicit.
- Databases, strings, dates, filters, and supplementary searches are recorded.
- Inclusion, exclusion, duplicates, and screening decisions are inspectable.
- The matrix captures comparable evidence and limitations.
- Counterevidence and alternative explanations receive a fair test.
- Sections synthesize across studies rather than summarize them serially.
- The gap claim is bounded to what the evidence truly leaves unresolved.
- Citations support the exact adjacent claims.
Before submission, a manuscript evidence-chain review can expose claims whose cited sources do not carry the weight assigned to them.
Sources accessed September 6, 2026.
- PRISMA 2020, PRISMA Executive.
- PRISMA-S reporting guidance, PRISMA Executive.
- Synthesizing Sources, Purdue OWL.
Readiness check
Run the scan while the topic is in front of you.
See score, top issues, and journal-fit signals before you submit.
Submit if / Think twice if
Start drafting when the question, review type, search record, selection rules, evidence matrix, and counterevidence are stable enough to inspect. Think twice when the corpus still changes with every search, one influential paper sets the entire structure, or the proposed gap is merely “not enough studies.” Stop when a systematic label is being used without reproducible retrieval and screening. Either complete the method the label requires or describe the review's actual scope honestly.
Frequently asked questions
It establishes what is known, where findings agree or conflict, what evidence limits remain, and why the paper’s question or contribution is warranted.
No. PRISMA is a reporting guideline for systematic reviews and related evidence syntheses; use the reporting standard appropriate to the review type and target journal.
Organize an evidence matrix by claim, design, population, result, and limitation, then write across studies around agreements, conflicts, and boundary conditions.
Before you upload
Choose the next useful decision step first.
Move from this article into the next decision-support step. The scan works best once the journal and submission plan are clearer.
Use the scan once the manuscript and target journal are concrete enough to evaluate.
Private API processing. Your manuscript is not used to train models.
Put the guidance to work
Choose the practical next step for your research.
Start with the question in front of you, or use the guided learning path when you want a clearer view of the full publishing process.