How to Turn Paper Summaries into a Real Synthesis

A stack of paper summaries is not a synthesis. A synthesis is an argument about what a body of work collectively shows, where it disagrees, and what remains unresolved — and it is built from decisions you make about scope, comparison, and evidence, not from stringing abstracts together. If you are finishing a course paper, a thesis chapter, or a first manuscript and staring at a folder of notes that will not cohere, this guide gives you a complete sequence: diagnose where you actually are, produce the artifacts a synthesis needs, check your work at defined checkpoints, and know when to stop.

A summary answers "what does this paper say?" A synthesis answers "what does this collection of papers establish, contest, or leave open?" The gap between them is structural. Summaries are organized paper by paper; synthesis is organized claim by claim. When you read your notes in the order you wrote them, you are reading a list, not an argument.

Rigorous research practice treats careful assessment of prior knowledge, strengths, weaknesses, and knowledge gaps as core work rather than an afterthought. That framing is useful here: a synthesis is precisely a documented assessment of prior knowledge, and it needs the same transparency you would apply to methods. If your notes do not record how you found and selected papers, you cannot yet make claims about the state of the field — only about the papers you happened to read.

Most stalled synthesis projects are stuck at one of four recognizable states. Identify yours honestly, because each needs a different next move.

• State 1 — Collection without criteria. You have papers but no written rule for what belongs in scope. Next move: write eligibility criteria before adding anything else.
• State 2 — Summaries without comparison. You can describe each paper but cannot say how any two relate. Next move: build a comparison matrix (below).
• State 3 — Comparison without claims. You have a matrix but no articulated positions. Next move: draft claim statements and test each against your matrix.
• State 4 — Claims without provenance. You have an argument but cannot trace each claim back to specific papers and specific passages. Next move: attach evidence to every claim before drafting prose.

A quick diagnostic: try to write three sentences of the form "The literature agrees that X, disagrees about Y, and has not established Z." If you cannot fill X, Y, and Z from your notes, you are in State 1 or 2. If you can fill them but cannot name the papers behind each, you are in State 4.

Cochrane guidance recommends planning searches from eligibility criteria and documenting the process. That principle scales down to a thesis chapter or a course paper: decide what counts as in-scope before you search, and write it down.

Your criteria should specify the population or problem, the kinds of studies or sources you will accept, the time window, and the languages or formats you can actually read. Keep the criteria visible in the same document as your notes so that every paper you add can be checked against them.

This is also where a scholarly search interface earns its place. If you are still assembling the candidate set, run your criteria-based queries through WisPaper search so that results arrive in a workspace you can keep, rather than in a browser tab you will lose. For readers who want a public entry point to the same activity, the WisPaper Scholar Search page explains the paper-finding starting point.

Checkpoint 1: You can state your eligibility criteria in under 100 words, and every paper in your folder passes them.

The single highest-leverage artifact is a matrix with one row per paper and one column per comparison dimension. Choose dimensions that match your research question — for example: the problem framing, the method or approach, the evidence base, the main finding, the stated limitation, and the relationship to your own question.

Fill the matrix from the papers themselves, not from memory. Where a cell is empty, mark it as unknown rather than guessing; unknown cells are legitimate findings about the state of reporting in your area.

Once the matrix is populated, read down the columns rather than across the rows. Column reading is what turns per-paper summaries into cross-paper patterns: which methods recur, which findings conflict, which limitations are shared.

Checkpoint 2: Every row is complete or explicitly marked unknown, and you can point to at least two columns where papers differ meaningfully.

A synthesis is carried by claims, and each claim needs a scope you can defend. Write each candidate claim as a single sentence with an explicit quantifier: "Most studies in this set use approach A," "Two studies report conflicting results on B," "No study in this set examines C under condition D."

Then test each claim against the matrix. If the matrix does not support the quantifier, narrow the claim until it does. A claim that says "several" when your matrix shows two is honest; a claim that says "the literature shows" when your matrix shows two is not.

This is where a library-based question-and-answer step can save time. If your papers are saved in one place, you can ask targeted questions across them rather than re-reading each one. Library QA supports asking questions based on papers uploaded to My Library or added from search results — useful for checking whether a claim you are drafting is actually reflected in your saved set before you commit it to prose.

Checkpoint 3: Each claim in your draft outline is traceable to specific rows and columns of the matrix.

Provenance is what separates a synthesis from an opinion. For each claim, record the papers that support it, the specific finding or passage relied on, and any papers that qualify or contradict it.

Cochrane guidance also recommends linking multiple reports of one study. This matters for synthesis integrity: if two of your rows turn out to be reports of the same underlying study, your apparent agreement is inflated. Check for this before you count papers as independent support.

If your reference list was assembled partly from memory or from secondary sources, screen it before submission. TrueCite checks BibTeX against real academic databases to flag hallucinated references — a concrete way to catch entries that would otherwise undermine an otherwise sound synthesis.

Checkpoint 4: Every claim has at least one named source, and no source is counted twice as independent evidence.

Suppose you are a master's student starting a thesis chapter in January, after a winter break spent collecting PDFs. Your folder has 40 papers and 40 one-paragraph summaries. Here is the sequence in practice.

Week 1 — scope. You write eligibility criteria: empirical studies of your phenomenon, published in the last decade, in English, with a method section you can evaluate. Twelve papers fail and move to a separate "background only" folder. You run supplementary queries through WisPaper search to check for obvious gaps against your criteria.

Week 2 — matrix. You build a six-column matrix and populate it from the papers. Three cells stay unknown because the papers do not report the needed detail. You note this as a reporting gap.

Week 3 — claims. Reading down the method column, you see two clusters. You draft: "Studies in this set divide into two methodological clusters," and check that every row fits one cluster or is explicitly mixed. One paper does not fit; you revise the claim to name the exception rather than smoothing it over.

Week 4 — provenance and drafting. You attach sources to each claim, screen your BibTeX, and draft the chapter around the claims rather than around the papers.

The result is a chapter organized by what the field shows and disputes, with the paper-by-paper detail relegated to supporting paragraphs.

Keep this checklist in the same document as your matrix and run it before every drafting session.

Documenting the search and selection process is also what makes transparent reporting possible; PRISMA 2020 provides reporting checklists and flow-diagram templates for systematic reviews, and its emphasis on transparent reporting of search, selection, and synthesis methods is a useful standard even for a smaller project. You do not need to follow a systematic-review template, but you do need to be able to say how you found and chose your papers.

A synthesis that lives across five tools will fragment. Keep your criteria, matrix, claims, and provenance map in one project space. If you are starting fresh, Research Projects is the WisPaper area for creating and organizing a research project, which gives the matrix and claim map a home.

Keep the papers themselves in a library you can query. My Library lets you save or upload papers and organize them, which is the precondition for asking questions across your set later. For longer literature-review tasks, Scholar Agent is the WisPaper area for extended research work with literature review use cases.

Checkpoint 5: A collaborator could open your project space and reconstruct your argument from the criteria, matrix, and claim map alone.

A synthesis is ready when four conditions hold. First, every claim is traceable to named sources. Second, disagreements in the literature are stated rather than averaged away. Third, gaps are described as gaps in the evidence you reviewed, not as absences in the field. Fourth, your search and selection process is documented well enough to be repeated.

Stop and escalate rather than push through in these situations:

• Stop collecting when new searches return only papers that fail your criteria. Adding more will not change your claims.
• Stop drafting and return to the matrix when a paragraph cannot be traced to a claim.
• Escalate to your supervisor or instructor when your claims depend on a body of work you cannot evaluate with your current methods training, or when two of your core sources conflict and you cannot adjudicate between them from the evidence alone.
• Escalate when your eligibility criteria exclude most of the relevant literature — that is a scope decision, not a writing problem.