- Your literature synthesis should ideally be organized by idea, not by source.
- In every paragraph make 1 claim, and cite at least 2 studies (preferably 3-5) to support it.
- Run this test: delete every citation in the paragraph. If it makes sense, you’ve actually synthesized the evidence. If it reads like nonsense, you’ve made a summary instead.
- The fastest way to find which sources support, extend, or conflict with each other is to make a synthesis matrix, with sources as rows and themes or findings as columns.
What is literature synthesis?
Literature synthesis is the process of combining findings across multiple studies to produce an integrated understanding of a topic that no single source states on its own, organizing the writing by idea rather than by source. It’s an essential part of writing any literature review, be it narrative, scoping, etc.
This article provides useful tips and steps to synthesize literature without sounding like a summary of sources or bibliography.
Before You Synthesize: The Prep Step
Synthesis fails when the reading was done without taking notes. Before you can group studies, you need comparable notes on each one, captured in the same fields every time. Read every study for the argument and the underlying assumptions, not just the reported results.
In every source you read, highlight or annotate these 7 fields as you go along.
| Field to extract | Why it matters for synthesis |
| Research question | Tells you which studies are actually answering the same question. |
| Sample and setting | Explains contradictions: different populations (e.g., institutionalized vs community-dwelling older adults) and different settings (e.g., Title 1 school vs private charter school) often produce different findings |
| Design and method | Lets you weight evidence and spot method clustering across the field. |
| Key findings | The raw material for your claims; record effect size or direction, not just statistical significance. |
| Stated limitations | Often becomes your gap statement when the same limitation repeats. |
| Theoretical framework (if available) | Can help you spot theories that are under-considered or those that are heavily relied on |
| Rough observations on study quality | Your own notes on any flaws or issues you see in the study, such as not reporting power in a quantitative study or no discussion of reflexivity and positionality in a qualitative study. This is not a formal quality appraisal and it’s okay if you don’t have such notes. |
2. Build a Synthesis Matrix
A synthesis matrix is a grid: sources run down the left as rows, and themes or study characteristics run across the top as columns. Each cell records what that source says about that theme, or stays blank if it says nothing.
How to set up the columns
- Start with 3 to 5 provisional themes or study characteristics drawn from your first pass through the sources.
- Fill the grid, then split columns that are doing 2 jobs and merge columns that keep producing the same content.
- Add columns for method, sample, and year: these explain contradictions later.
- Expect to rename most columns at least once. Themes emerge from the matrix; they are not decided in advance.
How to read it
Reading across a row tells you about 1 study, which is summary. Reading down a column tells you what the field collectively says or does, which is synthesis. Empty cells in a column are just as informative as full ones.
| What you see in a column | What it means |
| Several full cells that agree | A settled finding you can state in 1 sentence with grouped citations. |
| Cells that point in opposite directions | A contradiction to explain, usually through the method or sample column. |
| Full cells clustered in recent years only | An emerging line of inquiry worth flagging as current. |
| Mostly empty cells | A gap, and often the justification for your own study. |
3. The 5 Patterns to Look For
Ask the same 5 questions of every column in your matrix. Most of what a good literature review says about a field falls into these 5 categories.
| Pattern | Question to ask the column | What it gives you |
| Convergence | Where do independent studies agree? | What doesn’t need further research. You know what research already agrees upon. (Agreement across different methods is stronger than agreement across identical ones.) |
| Contradiction | Where do findings conflict, and why? | A potential gap that you can attempt to resolve through your own study. Check if this conflict appears to stem from sampling, instruments, or population/settings. |
| Population clustering | Who has been studied? What settings have been examined? | A limitation of the field as a whole if you can find notable minority groups excluded or if research concentrates in a few, similar countries or in similar settings (e.g., in tech firms but not manufacturing) |
| Method clustering | Do supporting studies share a design or instrument? | A limitation of the field as a whole, not of any single paper. Again, this could be a useful gap for your own study to fill. |
| Multiple empty cells | What has nobody asked? | One or more gaps that you could try filling. |
If you’re writing a standalone literature review, you may find it most useful to focus on reasons for contradictions. If you’re planning to conduct original research, focus more on the empty cells and also on the contradictions. But remember to also
4. Literature Synthesis Example
This worked example follows 6 illustrative studies on a common nursing management question: does nurse manager leadership style affect whether staff nurses stay? The sources below are fabricated for the purpose of this article, but the pattern they show is typical of the real evidence base.
4.1 The mini-corpus
| Study | Design and sample | Key finding |
| Alvarez & Kim (2018) | Cross-sectional survey; 412 staff nurses; 6 acute-care units | Higher transformational leadership scores were inversely related to intent to leave. |
| Osei et al. (2019) | Cross-sectional survey; 1,208 nurses; 14 hospitals | Authentic leadership predicted job satisfaction; intent to leave measured with 1 item. |
| Brennan (2020) | Longitudinal, 24 months; 3 units; resignation records | Leadership style explained little variance once staffing ratios entered the model. |
| Duarte & Ferreira (2021) | Qualitative; 22 interviews with nurses who had resigned | Nurses described manager visibility and scheduling fairness, not leadership style. |
| Nakamura et al. (2022) | Systematic review; 31 studies | Consistent association with intent to leave; weak association with actual turnover. |
| Whitfield & Adeyemi (2023) | Quasi-experimental; 9 units; 18-month follow-up | Turnover fell 6% in intervention units; a concurrent pay increase was not controlled. |
4.2 The matrix for these 6 sources
| Study | How leadership was measured | Outcome measured | Evidence strength |
| Alvarez & Kim (2018) | Validated self-report scale | Intent to leave | Low: single time point |
| Osei et al. (2019) | Validated self-report scale | Intent to leave, 1 item | Low: single time point |
| Brennan (2020) | Validated self-report scale | Actual resignations | Moderate: staffing controlled |
| Duarte & Ferreira (2021) | Not measured; described by nurses | Reasons given after leaving | Moderate: explanatory depth |
| Nakamura et al. (2022) | Mixed across 31 studies | Both outcomes | High: pooled evidence |
| Whitfield & Adeyemi (2023) | Training exposure | Actual turnover | Moderate: confounded |
4.3 What the columns reveal
- Convergence: every study that measures intent to leave finds an association in the same direction.
- Contradiction: studies measuring actual resignations find much weaker effects than studies measuring intent.
- The explanation lies in the outcome column. The field is largely measuring an attitude, not a behavior.
- Method clustering: 4 of the 6 studies rely on self-report surveys of currently employed nurses.
- What is absent: no study covers long-term care, community nursing, or the cost of a leadership development program.
4.4 The unsynthesized version: what not to write
Alvarez and Kim (2018) surveyed 412 staff nurses and found that transformational leadership was inversely related to intent to leave. Osei et al. (2019) surveyed 1,208 nurses and found that authentic leadership predicted job satisfaction. Brennan (2020) followed 3 units for 24 months and found a small effect. Duarte and Ferreira (2021) interviewed 22 nurses who had resigned. Nakamura et al. (2022) reviewed 31 studies and found a consistent association with intent to leave but a weak association with actual turnover. Whitfield and Adeyemi (2023) evaluated a leadership development program and reported a 6% drop in turnover.
What is wrong:
- Every sentence begins with an author.
- The reader learns 6 facts without any logical linking.
- Your opinion and interpretation is absent.
4.5 The synthesized version
Cross-sectional surveys consistently report that nurse manager leadership affects intent to leave (Alvarez & Kim, 2018; Osei et al., 2019), and a review of 31 studies confirms the pattern (Nakamura et al., 2022). However, comparatively few studies have actually measured turnover. Studies that measured turnover find that leadership has weaker effects: leadership explained little variance once staffing ratios were modeled (Brennan, 2020), and nurses who had resigned cited scheduling fairness instead (Duarte & Ferreira, 2021). Thus, leadership development should not be the sole focus of measures to improve retention in nurses.
What makes this synthesis useful:
- You’ve found a valid contradiction in the field and an explanation: some studies measure “intent to leave” and some measure actual leaving.
- You’ve grouped similar studies together.
- You’ve also given a direct practical implication: if administrators are trying to improve nurse retention, they should invest in more than just leadership development for nurse managers.
Gaps I can find in the literature from this synthesis
- No longitudinal study
- All studies conducted at large, urban hospitals. No studies in rural areas or small clinics.
- All units/wards lumped together. No study has looked at units where turnover has the most critical impact on patient care (e.g., ER, ICUs, NICUs).
- No study has examined turnover in RNs vs LPNs. Some studies have also included CNAs into their sample.
Now, I can choose to address one of the above gaps in my own study, of course, after evaluating my chosen research question against the FINER criteria.
5. Choosing an Organizing Structure
Once the matrix is full, decide how the synthesized sections will be sequenced. Pick the structure that matches where the interesting variation sits.
| Structure | Use it when | Main risk |
| Thematic | Findings cluster around concepts. This is the default choice for most reviews. | Themes that are really just topic labels, such as “Studies on leadership.” |
| Methodological | The field is split by how the question is studied or by conflicting operational definitions. | Turning the review into a methods lecture rather than an evidence argument. |
| Theoretical | Competing frameworks explain the same phenomenon. | Comparing theories without ever comparing findings. |
| Chronological | The change over time is itself the argument. | The most frequently chosen and least frequently justified structure. |
6. Tips to Write a Good Synthesis
- Open each paragraph with a claim, not with an author name.
- Place citations at the end of the sentence as much as possible. Use phrasing like “XYZ et al. (2020) found …” only for 1-2 seminal works that have defined your topic.
- Use a variety of reporting verbs. Saying that PQRS et al. “challenged” or “refuted” or “corroborated” is more interesting than saying that they “found” or “showed”.
- Aim for 1 claim per paragraph supported by 2 or more sources. A paragraph with 1 citation is usually a summary in disguise. The exception is a seminal or landmark study, but you will have usually 1 of these (maximum 2) in a literature review.
- Keep paragraphs to roughly 80 to 120 words so that each paragraph makes a single, checkable point.
Choose signal transitional words or phrases as follows:
| Purpose | Phrases that do real work |
| Show agreement | consistent with; in line with; both studies report; this pattern holds across |
| Show contrast | in contrast to; the reverse appears when; this holds only when |
| Explain a conflict | the difference is attributable to; once X is controlled; measured differently |
| Flag a shared weakness | both rely on; all 4 studies recruit from; each measures X by self-report |
| Name a gap | no study to date has examined; the evidence is confined to; remains untested in |
Frequently Asked Questions
How many sources do you need for a literature synthesis?
There is no fixed number. You have enough when new sources stop changing your themes. In practice, an undergraduate course assignment may synthesize 10 to 15 sources, a dissertation chapter 40 to 80, and a systematic review whatever the protocol returns.
What is the difference between a summary and a synthesis?
A summary just describes each study and doesn’t critically evaluate them or look for themes and patterns in the literature as a whole. The following table explains key differences between a summary and a synthesis.
| Summary | Synthesis | |
| Organized by | Source | Theme or concept |
| Paragraph starts with | An author’s name | A claim |
| Citations per paragraph | Usually one | Usually several |
| Reader learns | What each study said | What the field collectively knows |
| Your voice | Absent | Drives the paragraph |
What is the difference between synthesis and meta-analysis?
Synthesis is narrative and conceptual. You integrate findings in prose to build an argument. Meta-analysis is statistical: you pool effect sizes from studies that are similar enough to combine numerically. Every meta-analysis includes synthesis, but many syntheses cannot support a meta-analysis because the studies differ way too much in their design, outcomes, and measures (this is called heterogeneity).
Can you synthesize qualitative and quantitative studies together?
Yes, and it is often the strongest thing you can do. Treat the qualitative work as the explanation for the quantitative pattern. For example, survey findings can show an association and a focus group or phenomenology study can shed light on people’s perceptions or real-life experiences. When you’re synthesizing such studies, mention the design so that the reader can see which evidence type supports each claim. This kind of synthesis is especially valuable in an integrative review.
How do you write a synthesis paragraph in a literature review?
Start with a claim about the field. Support it with 2 or more citations grouped in a single set of parentheses. Introduce contrasting evidence and explain the contrast through method, sample, or measure. Close with the implication. If you can delete every author name and the paragraph still makes sense, the structure is right.
What is a synthesis matrix and how do you make one in Excel?
Put each source in a row and each theme in a column, then add columns for design, sample, and year. Use 1 sheet per review question, freeze the top row and the first column, and wrap text so cells stay readable. You can also color-code cells by direction of finding so that agreement and conflict are visible at a glance.
What do you do when studies contradict each other?
Do not average them and do not pick a side. Compare the conflicting studies on sample, setting, measurement, design, and year, and test whether any of those factors explains the split. If none does, say so plainly and name the disagreement as an open question. An explained contradiction is often the most original part of a review.
How many paragraphs should there be in a literature review synthesis?
Allocate space by the weight of the evidence, not by the number of sources. A theme with 12 converging studies may need 1 paragraph; a single well-designed study that contradicts them may need 1 paragraph of its own. As a rough guide, expect 1 paragraph per theme in an undergraduate course assignment and 3 to 5 per theme in the literature review chapter of a dissertation.
Comment