How do I turn a synthesis matrix into literature review paragraphs?

The gap between a finished matrix and a written chapter feels larger than it is, and the reason is usually that people try to write from the whole matrix at once. Do not. Filter it down to a single theme, hide everything else, and work only from what is in front of you. Twelve rows is a paragraph. The whole grid is paralysis.

With one theme visible, read the rows and ask what they add up to. Sometimes they agree and the sentence is straightforward. Sometimes they split, and the sentence names the split, which is a better paragraph. Sometimes they turn out not to belong together, which means the theme was a topic rather than a question and needs splitting before you write anything.

Thomas and Harden describe that move in three stages, free line-by-line coding, grouping codes into descriptive themes, then developing analytical themes that go past what any single study says (Thomas and Harden, 2008, p. 4). A completed matrix is already at the second stage. In their own review, the themes came out as school-based learning, hands-on experience, treating fruit and vegetables differently, and avoiding health-warning messages, and the interventions that matched those preferences were the more effective ones (Thomas and Harden, 2008, p. 1). That is what a theme looks like when it is ready to write from, a named pattern with a consequence attached, not a set of studies filed under the same topic.

Write the topic sentence before you look at the citations again. It should be a claim you would defend in conversation, not a description of the literature. "Mentoring effects are strongest in the first three years" is a claim. "Several studies have examined mentoring effects" is a table of contents entry.

Then place the evidence. The rows give you the sources, the pages, and your own strength judgments, so the paragraph writes itself in the order strong, supporting, qualifying. When the paragraph is done, the matrix rows you used can be marked, which is how you avoid citing the same three studies in four different sections.

Expect the first version of the section to be wrong in structure, and budget for that. Pautasso is direct about it, a rough draft needs rewriting, restructuring, and rethinking before it holds a coherent argument, and recommends reference-management software plus careful marking of quotations so attributions survive the reshuffling (Pautasso, 2013, p. 2). The revision test is mechanical. Underline the controlling claim in each paragraph, circle the evidence, mark the sentence that says why the comparison matters. A paragraph with only circles has reproduced the matrix instead of converting it.

What is the shape of a good synthesis paragraph?

Claim, evidence, qualification, consequence. State what the literature supports, cite the studies that support it, name the limit or the dissenting finding, then say what follows for your argument. Four moves, usually four to six sentences, and the last one is the one most drafts leave out.

The consequence sentence is what makes a review argue rather than report. Without it the reader gets a well-cited paragraph and no reason to have read it. With it, each paragraph pushes the chapter a step closer to your research question.

The verb you attach to each source carries the comparison. Lingard sorts reporting verbs by the relation they signal, acknowledges, confirms, and supports for agreement, refutes, questions, and challenges for conflict, and extends, refines, and builds on for a study that develops an earlier one (Lingard, 2018, Part II, p. 134). Lingard also warns against inflating what an author did, since a tentative reflection should be reported with suggests or explored, not with the stronger argues (Lingard, 2018, Part II, p. 135). Three sentences beginning Smith found, Jones found, Lee found are not a synthesis paragraph, they are the matrix read aloud.

The qualification is the second most-skipped move, and skipping it is what makes a chapter read as uncritical. A paragraph that admits its own limits is more persuasive than one that does not, because the reader stops looking for what you left out.

Weight that qualification by study quality, not by how many papers sit on each side. When Thomas and Harden checked which papers drove their synthesis, the poorer-quality studies contributed comparatively little and produced few themes the stronger studies did not already carry (Thomas and Harden, 2008, p. 8). Pautasso comes at the same task from the other direction, asking a review to name the major achievements, the areas of debate, the methodological problems, and the questions still open (Pautasso, 2013, p. 3). One caution on the consequence sentence. If your reading comes from putting the descriptive themes through a framework the original studies never used, Thomas and Harden treat that as a separate analytical step (Thomas and Harden, 2008, p. 9), so attribute it to yourself rather than to any study in the row.

How many sources belong in one paragraph?

Two to five for a synthesis paragraph. One source means you are summarizing a paper, not synthesizing a field. More than about six and the citations stop being read, so group them: cite three representative works and say the pattern holds across the rest.

There is one honest exception. When a single study is genuinely foundational, or when it is the only work that has done the thing, a paragraph built on it is correct. Say so explicitly, since "the only study to test this directly" is useful information and often a gap statement in disguise.

Watch for the opposite failure too. A paragraph carrying nine citations for an uncontroversial claim is padding, and it reads as though you are proving you did the reading rather than making an argument.