How to synthesize sources instead of listing summaries

Choose the question the sources must answer

A broad subject cannot organise a synthesis. Write a question specific enough that two sources could give meaningfully different answers: ‘Which conditions make asynchronous review useful?’ works; ‘collaboration research’ does not. Define important terms before comparing sources that may use them differently.

Write your provisional answer too. It may be wrong, but it gives the reading a hypothesis to test. The aim is not to force sources into agreement; it is to notice which evidence strengthens, narrows, or overturns that starting position.

Narrow the question until sources can genuinely disagree about the answer. 'What does the literature say about feedback?' is a filing category. 'When does delayed feedback improve revision for novice writers?' identifies a condition, outcome, and group. The narrower question tells you which studies are comparable and which are contextual. It also keeps an impressive but irrelevant source from taking over simply because you read it recently.

Build a synthesis matrix

Put sources in rows and the dimensions of the question in columns. Useful columns include claim, population or setting, method, evidence, definition, limitation, and implication. Fill cells in your own words and keep a locator beside anything you may cite.

The blank cells are information. They reveal that a confident source never measured the outcome you care about, or that two apparently contradictory papers studied different populations. A matrix makes those differences visible before prose hides them inside smooth transitions.

Give the matrix columns that force comparison: claim, population or setting, evidence type, mechanism, boundary, and relevance to your question. Put sources in rows, then read down each column. Empty cells are informative. They may show that authors assume a mechanism, use incompatible definitions, or never test the group your project concerns. A matrix earns its place when it reveals a pattern you could not see in separate source summaries.

Summary reports what each source contains. Synthesis explains what becomes visible when the sources are read together.

Code relationships, not merely topics

Mark where sources agree, extend one another, disagree, use different definitions, share data, or leave the same gap. Also mark differences in evidence quality. Three opinion essays do not outweigh one strong study simply because they occupy three rows.

Look for conditional patterns: an approach works in one setting and fails in another; a finding appears only with a particular measure; a disagreement dissolves after terms are aligned. These conditions are often the actual contribution of your synthesis.

Use precise relationship codes in the margin: agrees, extends, qualifies, contradicts, uses a different measure, or addresses another level. 'Same topic' is not a relationship. Two sources can recommend the same practice for different reasons, while two apparently conflicting results may come from different populations. Coding that distinction prevents a false debate and helps you decide whether the paragraph needs contrast, sequence, or combination.

Write a pattern sentence before a paragraph

Start each paragraph with what the sources collectively show, not with a source name. A useful pattern sentence is precise enough to be disputed: ‘Across the four studies, review speed improves only when one editor owns the final decision.’ Then use sources as evidence, exceptions, or qualifications for that sentence.

If you cannot write the pattern sentence, return to the matrix. You may have gathered related material without deciding what it means. More quotations will not solve that decision.

A strong pattern sentence includes the shared result and its boundary: 'Across the three classroom studies, rapid feedback improved first revisions, but only the study with a second review measured whether the change lasted.' That sentence gives the paragraph a claim to prove. The following evidence can then be selected for its role instead of presented one source at a time in the order you happened to read it.

Make disagreement legible and fair

Do not manufacture consensus by averaging incompatible claims. State the disagreement, then investigate whether it comes from method, population, definition, time period, or genuine interpretation. Give the strongest version of each position and indicate what evidence would resolve the dispute.

Avoid turning your preferred source into the neutral baseline. Describe comparable features in comparable language. If one source has a limitation, check the others for the same limitation rather than using asymmetrical scrutiny to produce the conclusion you wanted.

Represent disagreement at the level where it occurs. Authors may dispute a definition, method, causal explanation, value judgment, or practical recommendation. State the shared ground before the split, then give comparable evidence to each side. If one side has weaker evidence, say so without making it foolish. Fair synthesis helps the reader see what would need to be measured or decided to resolve the conflict.

End with the implication and remaining gap

A synthesis earns its place by changing what the reader should believe, do, or investigate. Close the section with that implication. Then name what the available sources cannot establish, especially when the gap limits your recommendation.

Keep the matrix with the draft. During revision, it becomes a fact-checking map: every citation has a cell and locator, every pattern can be tested against the sources, and a newly found source has an obvious place to enter the comparison.

End the synthesis with a coverage audit. Ask which populations, periods, settings, and outcomes the combined evidence actually covers, then name what remains outside it. The remaining gap should follow from the matrix, not from a ritual sentence that 'more research is needed'. In practical writing, the implication may be a bounded recommendation, a decision rule, or a reason not to generalise beyond the studied setting.