03 · What You Need to Know
How to move from summarizing studies to synthesizing evidence
First, understand what synthesis actually means
Summary and synthesis are related but different intellectual tasks.
Summary
Explains the question, methods, findings, argument, or contribution of an individual source.
Synthesis
Combines information from multiple sources to explain a larger pattern, relationship, disagreement, development, or conclusion.
Suppose three studies examine the relationship between remote work and employee well-being.
A summary-oriented review might tell the reader what Study A found, followed by Study B, followed by Study C.
A synthesis-oriented review asks what happens when those studies are compared. Perhaps all three report better job satisfaction but only one finds lower psychological distress. Perhaps the studies define remote work differently. Perhaps two are cross-sectional while one follows employees over time. Those relationships are the substance of the synthesis.
Start with a question the literature needs to answer
Synthesis is much easier when you know what you are trying to explain.
Instead of approaching a stack of papers with the vague instruction “write the literature review,” formulate questions such as:
- What does the literature suggest about the relationship between X and Y?
- Under what conditions does the relationship change?
- Why have studies reached different conclusions?
- How has the dominant explanation developed over time?
- Which methods have been used to investigate the question?
- What are the strongest and weakest parts of the evidence base?
- Which theoretical explanations are supported?
- What remains genuinely uncertain?
Your literature review can contain several such questions. Each one gives you a reason to bring particular studies together.
Do the synthesis before you start polishing paragraphs
Trying to discover relationships among papers while simultaneously writing elegant prose is difficult.
First externalize the comparison.
A synthesis matrix can place studies in rows and important dimensions in columns. Depending on your question, those dimensions might include:
- research design;
- sample or population;
- theoretical framework;
- definitions;
- measurement;
- intervention or exposure;
- outcome;
- main finding;
- effect direction;
- important limitation;
- risk of bias;
- context; and
- contribution to a theme.
A workflow for organizing papers and building a synthesis matrix can make these comparisons easier to maintain.
Once the studies are visible side by side, patterns that were difficult to notice while reading PDFs individually often become obvious.
Look for convergence
Start by asking where studies point in the same direction.
Agreement can involve more than identical results. Studies may converge on:
- the existence of a relationship;
- the direction of an association;
- a theoretical mechanism;
- a methodological limitation;
- a useful measurement approach;
- a subgroup difference;
- an implementation challenge; or
- an unresolved question.
Do not merely write that “several studies agree.” Explain what they agree about and how strong that convergence is.
Three small cross-sectional studies using nearly identical samples provide a different kind of convergence from evidence reproduced across longitudinal, experimental, qualitative, and geographically diverse studies.
Look for disagreement
Conflicting findings are not something to hide so that the review appears cleaner.
They are often where synthesis becomes most valuable.
When studies disagree, ask:
- Did they study different populations?
- Did they define the key concept differently?
- Were different measures used?
- Did the studies use different designs?
- Were follow-up periods different?
- Were the interventions actually comparable?
- Did analytical choices differ?
- Could bias or confounding explain part of the difference?
- Were the estimates genuinely incompatible, or is the apparent disagreement based only on different p-values?
Your job is not necessarily to force the disagreement into one answer. Sometimes the correct synthesis is that the evidence remains inconsistent and the reason is not yet clear.
Do not use statistical significance as the definition of agreement
Imagine two studies estimating nearly identical effects. One reports p < 0.05 and the other reports p > 0.05 because its estimate is less precise.
Writing that one study “found an effect” while the other “found no effect” may exaggerate their disagreement.
Compare effect estimates, uncertainty, design, and context rather than classifying studies solely by whether each crossed a significance threshold.
Similarly, a null or statistically non-significant result does not automatically demonstrate that no meaningful effect exists.
Look for methodological patterns
Sometimes the most important synthesis is not that studies disagree, but that particular methods tend to produce particular conclusions.
For example:
- cross-sectional studies may report stronger associations than longitudinal studies;
- self-reported exposure may produce different results from objectively measured exposure;
- small convenience samples may show more variable estimates;
- different instruments may operationalize the same construct differently; or
- adjusted analyses may produce weaker associations than unadjusted analyses.
These patterns can help explain an apparently chaotic evidence base.
But be careful not to invent an explanation merely because it sounds plausible. If you believe methodology explains disagreement, show how the studies support that interpretation and preserve uncertainty when they do not.
Look for differences in definitions
Two papers can appear to study the same concept while operationalizing it very differently.
Consider “social media use.” One study may measure minutes per day. Another may distinguish active communication from passive browsing. Another may examine one platform. Another may measure problematic use with a psychometric scale.
Combining these under one label without discussing the difference can produce a misleading review.
Definitions themselves may become a synthesis theme:
The apparent inconsistency in findings partly reflects the fact that studies are not measuring the same form of the exposure.
That statement does more analytical work than listing the definitions one paper at a time.
Look for theoretical relationships
Synthesis is not limited to empirical results.
If several theories attempt to explain the same phenomenon, compare:
- their assumptions;
- the mechanisms they propose;
- their predictions;
- the evidence supporting each;
- where their predictions overlap;
- where they conflict; and
- whether later theories extend or revise earlier ones.
A literature review can therefore show how an explanation evolved rather than simply stating that Author A proposed Theory A and Author B proposed Theory B.
Use chronology only when time explains something
Chronological organization can be useful when the development of the field is part of the argument.
For example:
early descriptive studies → initial theoretical model → methodological criticism → longitudinal tests → revised theory
That chronology tells an intellectual story.
Simply ordering papers from 2015, 2016, 2017, and 2018 because that is when they were published does not necessarily create synthesis.
Use time as an organizing principle when it helps explain change.
Identify the foundational work without letting it dominate everything
Seminal papers can establish where an idea originated, but the literature review should also show what happened afterward.
A useful pattern is:
original proposal → early evidence → replication or extension → criticism → current understanding
This connects foundational work to the present evidence rather than treating an influential old paper as the final authority.
If you are still identifying those foundational sources, trace the papers that introduced or substantially shaped the important ideas, methods, and findings in the field.
Evaluate evidence while you synthesize it
Synthesis is not vote counting.
Suppose seven weak observational studies support an association while two well-designed longitudinal studies do not. Writing “seven studies found a relationship and two did not” treats all nine as interchangeable units.
A better synthesis asks what kind of evidence each study provides.
Consider:
- study design;
- risk of bias;
- measurement quality;
- sample appropriateness;
- confounding;
- precision;
- directness to your question;
- consistency; and
- important analytical choices.
This does not mean casually declaring one paper “high quality” and another “low quality.” Explain which methodological features affect how strongly each finding should influence your interpretation.
A more detailed critical appraisal of each paper can help determine how much weight its findings deserve in the synthesis.
Distinguish synthesis from meta-analysis
Narrative synthesis
Integrates findings through structured comparison and explanation in words, often supported by tables or other displays.
Meta-analysis
Uses statistical methods to combine quantitative estimates from sufficiently comparable studies under specified assumptions.
You can synthesize literature without performing a meta-analysis. Conversely, calculating a pooled estimate does not remove the need to explain heterogeneity, study characteristics, limitations, and applicability.
For systematic reviews without meta-analysis, the SWiM reporting guideline provides guidance for reporting synthesis methods when quantitative results are synthesized without statistical meta-analysis.
Organize sections around themes or claims
Once you understand the patterns, create an outline that reflects them.
A weak outline might be:
- Smith et al.
- Jones et al.
- Garcia et al.
- Lee et al.
A stronger outline might be:
- Evidence for the proposed relationship
- Evidence that the relationship depends on context
- Methodological reasons findings may differ
- Populations for which evidence remains limited
The second structure gives each section an intellectual job.
Make the paragraph about a claim
The same principle applies at paragraph level.
A synthesis paragraph often begins with an interpretive statement rather than an author's name.
Instead of:
Smith et al. found that remote work was associated with higher job satisfaction.
consider a topic sentence such as:
Evidence generally suggests that remote work is associated with higher job satisfaction, although the strength of the relationship varies with how remote work is defined and who is studied.
The rest of the paragraph can then provide evidence for that claim, introduce exceptions, explain differences, and arrive at a qualified interpretation.
Bring multiple studies into the same paragraph
A useful synthesis paragraph often contains several sources because the paragraph is about a relationship among evidence.
A practical structure is:
claim → supporting evidence → contrasting or qualifying evidence → explanation of the pattern → implication
Not every paragraph needs all five elements, and the sequence should not become mechanical. But the pattern helps prevent the review from becoming a catalogue of studies.
Use author names when the author actually matters
Author-focused sentences are not inherently bad.
Use them when:
- identifying who introduced a theory;
- discussing a seminal study;
- contrasting explicit scholarly positions;
- describing a distinctive method;
- tracing an intellectual debate; or
- attributing an interpretation to particular researchers.
The problem arises when every paragraph is organized around authors simply because citations contain author names.
Compare studies explicitly
Do not make the reader infer the relationship.
Useful synthesis language includes:
- similarly;
- in contrast;
- consistent with;
- unlike;
- extends;
- qualifies;
- challenges;
- replicates;
- converges with;
- differs primarily in;
- may reflect;
- appears strongest when; and
- remains uncertain because.
These words are useful only when the relationship is real. Do not add transition language to create an artificial appearance of synthesis.
Explain why studies may disagree
A literature review becomes analytically useful when it moves from identifying inconsistency to investigating it.
For example:
Although cross-sectional studies generally report a negative association between X and Y, longitudinal studies provide weaker and less consistent evidence. This difference may partly reflect temporal ambiguity in cross-sectional designs and differences in baseline adjustment.
That sentence does three things: identifies a pattern, distinguishes designs, and proposes a reason that can be evaluated against the studies.
Use cautious language such as “may,” “could,” or “appears” when the explanation is an inference rather than something directly demonstrated.
Do not manufacture a research gap
A synthesis can reveal a gap, but the gap should emerge from the evidence.
Examples include:
- a population rarely studied;
- an important theory that has not been tested under certain conditions;
- heavy dependence on one research design;
- inconsistent definitions preventing comparison;
- limited longitudinal evidence;
- poor measurement of an important construct;
- contradictory findings that existing studies cannot explain; or
- a new context in which established findings remain uncertain.
A gap is not automatically “no one has studied my exact combination of variables.” It should represent an uncertainty whose resolution would improve understanding.
Distinguish absence of evidence from evidence of absence
If few studies have investigated a question, you may be able to say that evidence is limited.
You generally cannot conclude from that alone that the relationship does not exist.
Similarly, if studies produce imprecise or statistically non-significant estimates, do not automatically synthesize them as proof of “no effect.” Examine what effect sizes remain compatible with the evidence and whether the studies were capable of detecting effects that would matter.
Use tables to support synthesis, not replace it
A study-characteristics table can efficiently show:
- authors and year;
- design;
- sample;
- setting;
- measures;
- intervention or exposure;
- outcomes; and
- main findings.
That can reduce the need to repeat descriptive information in prose.
But the table does not explain why findings differ or what the evidence collectively means. That interpretive work belongs in the review itself.
Keep source-level notes separate from synthesis-level notes
This distinction can make writing much easier.
| Source-level note |
Synthesis-level note |
| Study A used a cross-sectional survey of 1,200 students. |
Most evidence in this theme comes from cross-sectional student surveys. |
| Study B found a small negative association after adjustment. |
Adjusted estimates are generally smaller than unadjusted associations. |
| Study C used a different scale for the exposure. |
Measurement inconsistency may contribute to heterogeneous findings. |
| Study D included only one university. |
Evidence is concentrated in institution-specific samples, limiting broader generalization. |
The right-hand column is much closer to what belongs in a literature review.
Write provisional synthesis statements before drafting
Before writing polished paragraphs, try completing sentences such as:
- Most studies suggest...
- Evidence is strongest for...
- Findings are inconsistent when...
- The apparent disagreement may reflect...
- Earlier research assumed..., whereas later studies...
- Most studies rely on..., which limits...
- Evidence remains sparse for...
- The literature therefore supports..., but not...
These are not templates to repeat mechanically in the final manuscript. They force you to articulate what the evidence collectively means.
Keep citation density proportional to the claim
A synthesis claim about a body of literature may require several citations. A specific methodological statement about one study may require one.
Do not attach ten references to a vague sentence simply to make it appear authoritative. Make sure the cited sources actually support the claim and represent the evidence fairly.
Likewise, avoid citing only one convenient paper for a claim that you present as a consensus across an entire field.
Do not cite a review when you are making a claim about an original study you have not checked
Reviews are invaluable for discovering and synthesizing literature, but secondary citation can distort what original papers actually reported.
If a particular original study is important to your argument, retrieve and read it before describing its methods or findings whenever reasonably possible.
This is particularly important for foundational claims, surprising findings, disputed interpretations, and details central to your own research rationale.
Revise for synthesis after the first draft
Even a well-planned literature review can drift toward summary during drafting.
On revision, inspect each paragraph and ask:
- What is this paragraph actually about?
- Does it contain more than one source where comparison is appropriate?
- Have I explained how the sources relate?
- Am I evaluating differences or merely listing them?
- Does the topic sentence express an idea rather than simply name a study?
- Does the final sentence tell the reader what the evidence means?
- Could descriptive details be moved to a table?
If the paragraph can be divided cleanly into “sentence about Study A, sentence about Study B, sentence about Study C,” it may still need synthesis.