From source notes to literature review synthesis

Here is a perfectly respectable set of source notes:

Van Ockenburg and colleagues reviewed interventions designed to improve synthesis writing. Luna and colleagues compared process-focused instruction with instruction about the formal features of the finished text. Robledo and Álvarez studied a longer program based on strategic and self-regulated instruction. Luo and Kiewra ran two experiments on a strategy called SOAR.

Nothing in this block is necessarily wrong. It simply never tells us what the four sources add up to.

The sentences follow the safest available structure: paper, description, next paper. A reader who wants to know whether the studies reached the same conclusion, measured the same ability, or qualify one another must reconstruct that relationship alone.

This is where synthesis in a literature review often stalls. The writer has understood the papers separately and may have grouped them under a sensible topic. The more difficult work begins when those papers have to support a sentence that none of them wrote individually.

A synthesis matrix can make that work visible. Its cells do not decide what the sources collectively support.

The missing sentence is not another summary

Each sentence in the opening block answers a source-level question: What did this paper examine?

The absent sentence answers a different one: What do these papers allow us to say together?

That distinction separates several operations that often occupy the same literature review:

Summary remains necessary. A writer cannot compare two studies’ methods or findings after misunderstanding either study. If identifying one paper’s question, method, or conclusion is still the problem, start with how to reconstruct a research paper’s argument before comparing it with other sources. The problem addressed here is the draft that keeps every source inside its own accurate miniature.

The UNC Writing Center describes literature reviews as usually combining summary and synthesis, while Purdue OWL defines synthesis through relationships and broader conclusions across sources. These are academic-writing conventions rather than experiments showing that one paragraph structure causes better reasoning.

Research on multiple-text integration supplies a related but more technical distinction. In their review of 50 empirical studies, Liron Primor and Tami Katzir found that assessments variously asked readers to select relevant information, form relationships between texts, and make further inferences. The tasks and scoring systems differed enough that integration did not refer to one uniform performance.

A literature review can still devote substantial space to one source. A foundational definition, unusual method, or historical turning point may need to be understood before comparison becomes possible. The difficulty arises when the order of the reading list quietly becomes the architecture of the whole review.

Nor must every review defend a startling new theory. Here, an argument can be a reasoned cross-source answer: what the literature supports, the conditions under which it supports it, and what it leaves unresolved.

Let the review question choose the columns

Suppose the four papers from the opening are entered into a spreadsheet with a column called Main result. That column will fill quickly. The comparison may remain as muddy as before.

The problem is not a shortage of notes. It is that “main” depends on the question the review is trying to answer.

For this example, the question is:

Which kinds of instruction improve university students’ synthesis writing, and what remains difficult after instruction?

Now several distinctions become relevant:

• who completed the writing task;

• what kinds and how many source texts they used;

• what instruction or materials they received;

• what the comparison condition received;

• whether the outcome concerned selection, organization, source connection, argumentative integration, or overall text quality;

• whether performance was tested immediately or later.

Formal review-method guidance makes the underlying principle explicit: the research question guides which information should be extracted and analyzed. For a student narrative review, that principle is useful without importing the full procedure of a systematic review.

For this task, a useful literature review matrix can separate two layers.

The separation matters. If a writer’s interpretation sits in the same undifferentiated “Notes” cell as the study result, it can later reappear in the draft as though the paper itself had said it.

The fields should also change with the review. Population and measured outcome may be indispensable when comparing experiments. A review of theories might need intellectual context, assumptions, conceptual claims, and interpretive differences instead. No universal spreadsheet can determine in advance which distinctions will matter.

Four papers on one topic can still be four different questions

Return to the opening notes. All four sources concern synthesis instruction. That shared topic creates a reason to read them together, but it does not yet create a fair comparison.

Before marking two sources as agreeing or contradicting one another, inspect the basis of the comparison:

• Question: Are they trying to establish the same thing?

• Construct: Do they use terms such as synthesis or integration in the same way?

• Population and context: Are the learners and settings sufficiently similar?

• Evidence type: Is each source a review, experiment, theoretical argument, or descriptive account?

• Outcome: Did they assess the same aspect of writing?

• Timing: Are immediate and delayed results being kept separate?

• Independence: Do the papers use distinct participants, data, or research lines?

The sources in this example begin to separate as soon as those questions are applied.

The 2019 intervention review included 16 papers and 19 experimental conditions drawn from work located through a search conducted in September 2016. It combines interventions involving different ages, tasks, durations, comparisons, and measures. Its job is to map an earlier and heterogeneous instructional literature, not to provide one clean estimate of a matrix’s effect.

The Luna et al. experiment involved 66 first- and second-year Education and Psychology students at a Spanish distance university. Students wrote from two texts presenting conflicting positions and were assessed immediately after one of two active online interventions. The outcome was the degree to which the final text integrated arguments and counterarguments.

The Robledo and Álvarez study involved 84 first-year Early Childhood Education students in four intact class groups. It examined a longer strategic and self-regulated program, assessed several writing products and processes, and returned for a one-year follow-up.

The SOAR paper contains two controlled college-writing experiments. In the first, the researchers supplied carefully prepared SOAR materials. In the second, students received brief training and generated their own notes, matrices, associations, and outlines before another immediate writing task.

The dates also matter. The three primary articles postdate the 2019 review’s search, so they are not included in it as publications. The review does, however, include an earlier 2016 intervention from Robledo’s research line. The later Robledo study extends related work rather than providing a wholly independent confirmation.

These sources belong in the same discussion because they illuminate a broader question about teaching synthesis. They are not four replications of one intervention, and counting them as four votes would erase precisely the distinctions the review needs to explain.

A full matrix can still leave the argument blank

Several university writing resources teach synthesis matrices. The University of Sheffield, for example, defines one as a spreadsheet organized by theme with space for cross-source interpretation. An older NC State handout hosted by FIU shows writers moving vertically through individual sources and horizontally across a shared idea.

These resources show what the matrix is meant to organize. Whether a blank spreadsheet changes the quality of an authentic literature review is an empirical question, and the available intervention studies do not isolate it.

In Luna and colleagues’ experiment, both instructional groups produced more integrated writing after training, and the process-focused group improved more. Yet that group reached only the study’s minimum integration level on average. Two students formulated a compromise solution in the instructional guide and then wrote final texts that supported only one position. They had located a possible relationship but did not carry it into prose.

Robledo and Álvarez found a different split. Compared with the control classes, the experimental classes improved more in overall text quality, organization, note-taking, and planning. The program did not produce significant improvements in information selection, source connection, or writing review. Both groups also improved by the one-year follow-up, although the experimental advantage in text quality remained.

SOAR examines matrix-centered instruction most directly, and its two experiments do not tell one simple story. When the researchers supplied an ideal matrix and association prompts in Experiment 1, the SOAR condition did better on information selection, categorical organization, and intertextual connections. In Experiment 2, after brief training, both groups included more source ideas in their later essays; the SOAR group did not show a differential selection gain. Its clear advantage concerned categorical organization, and there was no corresponding improvement in intertextual connections.

The clearest result is organizational: students more often arranged source material into categories. Because the interventions also included prompts, explanation, practice, planning, or self-regulation, the spreadsheet’s independent contribution remains unknown. The tasks used short, prepared source sets rather than thesis-length literature reviews.

A spreadsheet can therefore be complete while the argument remains blank. The next step is not another column of source facts. It is a judgment about the relationship among them.

The relationship cell is where the writing starts

Before drafting the paragraph, give the relationship a working name.

These labels are working vocabulary, not a formal taxonomy. Their purpose is to produce a more exact sentence than “the literature is mixed.”

Here is a worked example of literature review synthesis, compressed to fit the page:

“Several studies examined synthesis instruction” would accurately describe the table while doing almost no useful work with it.

“Matrix use improves synthesis writing” goes too far in the other direction.

A bounded cross-source claim is more useful:

In this small and methodologically varied set of studies, structured instruction produced more consistent gains in organization and planning than in measured cross-source connection.

That sentence still needs a boundary. The studies use different interventions, outcomes, populations, and scoring systems. It is not a pooled estimate showing that organization is statistically easier to improve than integration. It identifies a recurring distinction that the review should preserve.

Let the claim, not the reading order, run the paragraph

Once the cross-source claim exists, it determines which source appears first and how much detail each source receives.

The original notes followed author order. A synthesis paragraph can follow the evidence:

Across these studies, organizational gains are clearer than gains in cross-source integration. In Luo and Kiewra’s first experiment, students given fully prepared SOAR materials improved in information selection, categorical organization, and intertextual connections. In the second experiment, brief SOAR training clearly shifted organization but produced no differential improvement in information selection or intertextual connections. Robledo and Álvarez found a similar separation in a longer program: the experimental classes improved more in text quality, organization, note-taking, and planning, while selection, source connection, and writing review did not improve significantly. Luna and colleagues provide a partial counterpoint. Process-focused instruction improved integration more than product-focused instruction, yet the process group reached only the minimum integration category on average, and two students who articulated an integrative solution in the guide did not reproduce it in their final text. Together, the studies support structured comparison as a scaffold, especially for organization, while providing no basis for treating a matrix as a substitute for forming and expressing the cross-source claim.

The paragraph begins with the pattern rather than the first paper in the notes. Luo and Kiewra establish the clearest matrix-related result. Robledo and Álvarez show a similar outcome split in a different and longer instructional package. Luna and colleagues prevent the paragraph from hardening into the claim that integration cannot improve; their study found an improvement while also showing how limited the final integration remained.

The sources receive unequal space because they perform unequal jobs. That is the writer’s contribution: not an unsupported opinion, but an accountable decision about which comparison matters, which source changes it, and where the conclusion must stop.

Words such as similarly or however can help signal that structure after the relationship has been established. In the original notes, adding them between the four summaries would merely decorate an inference the paragraph had not yet made.

Download the literature review matrix template (XLSX)

Download the two blank matrix tables as UTF-8 CSV files (ZIP)

Open the online matrix in Google Sheets

Use the blank workbook alongside the completed example from this article. It includes separate extraction and synthesis sheets, a comparability check, a relationship key, and empirical and conceptual-review variants. The ZIP contains the source-extraction and cross-source-synthesis tables for use outside Excel.

Change the matrix when the review changes

The workflow in this article is designed for narrative, explanatory, critical, and proposal-oriented student literature reviews. It does not replace the methodology of a systematic, scoping, realist, or meta-analytic review.

Formal evidence syntheses commonly require explicit procedures for searching, eligibility decisions, screening, extraction, appraisal, synthesis, and reporting. PRISMA 2020 supplies reporting materials for systematic reviews. The SWiM guideline provides nine reporting items for systematic reviews that synthesize quantitative intervention effects without meta-analysis. Neither is a generic method for converting a handful of papers into one literature-review paragraph.

The fields also need to change when the sources do. In a theoretical review, population or experimental outcome may have no sensible equivalent. One possible adaptation is to record intellectual context, central proposition, analytical approach, assumptions, scope, and relationships among interpretations. That is an editorial starting point, not a validated universal template for the humanities.

At the other end of the scale, a short review of three closely related theoretical papers may not need a spreadsheet at all. The tool becomes useful when it helps the writer see a relationship that is otherwise difficult to hold and test. Completing the tool is not the task.

Before you call it synthesis

For any paragraph that draws on several sources, check four things:

1. What question does this paragraph answer?

2. Which sources are comparable for that question?

3. What relationship do they support?

4. Where does the resulting claim stop?

The four summaries from the opening can stay in the notes. The paragraph needs the sentence that explains why those sources belong together—and where that sentence stops.

References

• UNC Writing Center. “Literature reviews.” University of North Carolina at Chapel Hill.

• Purdue Online Writing Lab. “Synthesizing sources.” Purdue University.

• Primor, L., and Katzir, T. (2018). “Measuring multiple text integration: A review.” Frontiers in Psychology, 9, 2294.

• Paré, G., and Kitsiou, S. (2017). “Methods for literature reviews.” In F. Lau and C. Kuziemsky (Eds.), Handbook of eHealth Evaluation: An Evidence-based Approach. University of Victoria.

• van Ockenburg, L., van Weijen, D., and Rijlaarsdam, G. (2019). “Learning to write synthesis texts: A review of intervention studies.” Journal of Writing Research, 10(3), 401–428.

• Luna, M., Villalón, R., Martínez-Álvarez, I., and Mateos, M. (2023). “Online interventions to help college students to improve the degree of integration of their argumentative synthesis.” Reading and Writing, 36, 937–963. Published online in 2022.

• Robledo, P., and Álvarez, M. L. (2024). “Strategic and self-regulated instruction in synthesis tasks in the university context.” Frontiers in Psychology, 15, 1386907.

• Luo, L., and Kiewra, K. A. (2019). “Soaring to successful synthesis writing.” Journal of Writing Research, 11(1), 163–209.

• University of Sheffield. “How to synthesise information.”

• North Carolina State University Writing and Speaking Tutorial Service. (2006). “Writing a literature review and using a synthesis matrix.” Hosted by Florida International University.

• PRISMA. “PRISMA 2020 statement.”

• Campbell, M., McKenzie, J. E., Sowden, A., et al. (2020). “Synthesis without meta-analysis (SWiM) in systematic reviews: Reporting guideline.” BMJ, 368, l6890.