1. Choose the kind of review before you search
Different review types answer different questions and require different levels of transparency. Choosing the review type early prevents you from collecting literature without knowing what the final product is supposed to do.
| Review type | Use it when... | Minimum transparency |
|---|---|---|
| Narrative review | You are building conceptual background across adjacent literatures. | Explain anchor papers, how the search expanded, inclusion judgment, and how sources were prioritized. |
| Structured review | You need stronger transparency without a full systematic review. | List databases, search dates, search strings, screening criteria, and synthesis categories. |
| Scoping review | You are mapping the breadth of evidence and methods. | Define concepts, eligibility, charting fields, and the structure of the evidence map. |
| Systematic review | You are answering a focused, reproducible question. | Use a protocol, search strategy, screening flow, extraction plan, and risk/quality appraisal. |
| Meta-analysis | You are statistically pooling comparable effects. | Define the effect metric, dependency handling, heterogeneity, outlier, and sensitivity rules. |
2. Build the search from concept blocks
Instead of typing a full research question into one database, break the question into concept blocks. For each concept, list several synonyms or closely related terms. Combine synonyms within a block with OR, then combine major concept blocks with AND. This protects the search from missing studies simply because different authors use different language.
| Concept | Example terms | Why it helps |
|---|---|---|
| Population | undergraduate OR college student OR university student OR adolescent OR youth | Captures different population labels. |
| Exposure / intervention | research training OR inquiry-based learning OR undergraduate research OR scholarly writing | Captures both formal programs and related mechanisms. |
| Outcome | publication OR poster presentation OR research self-efficacy OR academic identity | Allows developmental outcomes as well as tangible outputs. |
| Context | higher education OR STEM OR humanities OR social sciences | Can narrow the search, but should be used carefully to avoid excluding relevant work. |
3. Decide what belongs before you read deeply
Create clear reasons for exclusion before you become attached to individual papers. Common reasons include the wrong population, outcome, design, insufficient detail, or duplicate/companion publication. If a famous or foundational source falls outside the review criteria, you can still recognize it as background while keeping it separate from the evidence set. Borderline decisions should be documented at the moment they are made.
4. Build an evidence matrix instead of relying on memory
For every retained source, record the same core information. A consistent evidence matrix makes synthesis much easier because you can compare studies across population, design, measurement, and findings rather than reopening every PDF repeatedly.
| What to record | What it means |
|---|---|
| Citation key | A short AuthorYear identifier that prevents citation chaos. |
| Purpose | The study purpose written in your own words. |
| Population or corpus | Age, setting, discipline, sample, text, or other boundary. |
| Design or method | RCT, survey, interview, thematic analysis, close reading, or another method. |
| Measure or evidence | The instrument, textual unit, database, coding frame, or other source of evidence. |
| Finding direction | A pattern, contradiction, or null result. |
| Use in your paper | Background, direct evidence, method precedent, or counterexample. |
5. Write by pattern, not by paper
Once the matrix is built, organize the review around patterns in the evidence. A paragraph should make a controlling claim and use multiple sources to support, complicate, or limit that claim. Useful synthesis moves include convergence, contradiction, measurement critique, design critique, and gap identification.
| Synthesis move | Example frame |
|---|---|
| Convergence | Across studies using [method/population], [pattern] appears consistently, although [boundary]. |
| Contradiction | Evidence is divided: [study type] suggests [A], whereas [study type] suggests [B]. |
| Measurement critique | [Construct] is often operationalized as [measure], which captures [part] but misses [part]. |
| Design critique | Because most evidence is cross-sectional, the literature supports association rather than temporal ordering. |
| Gap opening | What remains underdeveloped is not whether [topic] matters, but how [mechanism] operates in [context]. |
Helpful standards & references
These are starting points, not substitutes for the requirements of a specific journal, conference, institution, or research team.
PRISMA 2020: https://www.equator-network.org/reporting-guidelines/prisma/
EQUATOR Network reporting guidelines: https://www.equator-network.org/reporting-guidelines/