How to Make a Synthesis Matrix for Research
- How do you make a synthesis matrix for research?
- What does synthesis add to a summary?
- How should you choose the columns?
- What information belongs in a cell?
- What does a small completed example look like?
- How do you write from the example without overclaiming?
- Does an empty cell mean the source disagrees?
- Should a theme appear in every source?
- How do you avoid turning the matrix into a second bibliography?
- When is the matrix ready to support a draft?
- Sources
How do you make a synthesis matrix for research?
Make a synthesis matrix by placing sources on one axis and shared questions or themes on the other. In each relevant cell, record the source's contribution, its exact location and important limits. Then compare across sources before drafting a conclusion. Keep quotations, paraphrases and your own interpretation distinct. The matrix organizes evidence; it does not verify it, establish a complete literature search, or authorize sharing confidential material or drawing high-stakes professional conclusions.
A source log helps you find an item again. A synthesis matrix helps you see what several items say about the same question. You can use both without making either one enormous.
What does synthesis add to a summary?
A summary describes what a source says. Synthesis examines relationships between sources and develops an account of those relationships. Purdue OWL's explanation distinguishes these tasks and cautions against forcing connections or ignoring outlying perspectives.
For example, three descriptions of a catalogue do not automatically make one explanation of how that catalogue changed. You must identify which dates, versions and features can actually be compared.
The question is not “How can I fit every note into a paragraph?” It is “Which relationship can these particular notes support?” Sometimes the answer is agreement. Sometimes it is a difference in scope, an unresolved contradiction or a question that remains unanswered.
How should you choose the columns?
Start with a focused question and themes that help answer it. The University of Illinois-hosted Writing for Inquiry and Research chapter presents a matrix as a way to track themes and relationships, rather than every detail of every article.
Our working layout puts one source in each row. Columns contain comparison questions, followed by a limits note. Keep full bibliographic details in your source log and use its stable source IDs in the matrix.
A column headed “Interesting points” provides little help with comparison. A question such as “What does this item say about revision dates?” tells you what belongs there and what does not. Do not force unrelated material into the nearest empty cell.
You may revise a theme as your reading develops. When you do, review every existing row against the new definition. Otherwise the heading changes while older notes still answer the old question.
What information belongs in a cell?
Use a short note you can trace back to a passage:
- the source's relevant statement, accurately paraphrased;
- its page, section, table or another usable locator;
- any scope condition necessary to understand it;
- a clearly marked uncertainty if the statement is not yet checked.
NC State's synthesis-matrix teaching page emphasizes topic-based notes and retrievable locations. That retrieval detail matters when a compressed phrase becomes the basis of a polished sentence later.
Keep the source's statement separate from your interpretation. “This edition supplies an index” and “an index would help my project” do different work. The first describes the item; the second records your judgment.
For direct quotations, retain exact wording and attribution. A paraphrase still requires attribution, as Purdue's quotation and paraphrase guidance explains. Changing a few words is not a substitute for writing an accurate note in your own language. Respect copying and access restrictions throughout.
What does a small completed example look like?
Everything in the following exercise is invented: the collection, documents, dates, IDs and locators. They are teaching material, not real research or citations.
The fictional question is: “What do these three documents establish about revision information in a small collection catalogue?”
Assume these are the complete relevant details supplied:
| Fictional source | Revision date | Description of changes | Item-level detail | Limit |
|---|---|---|---|---|
| Example A: catalogue introduction, section 1 | States March 2024 revision | Says entries were revised | No individual changes listed | Describes that edition only |
| Example B: change notice, paragraph 2 | States March 2024 revision | Lists wording changes | Identifies entries 4 and 9 | Gives no reasons for the changes |
| Example C: later reader note, paragraph 1 | Does not state a revision date | Mentions seeing different wording | Does not identify entries | Observation cannot be matched to an edition |
The matrix contains three sources and three comparison questions: 3 × 3 = 9 comparison cells. The separate limit column adds context, not a fourth research question. Counting cells checks the layout; it does not quantify evidence quality.
Notice what the exercise does not supply: an independent history of the collection, a test of reader comprehension, or proof that the revision corrected errors. You cannot obtain those conclusions by combining the available cells more confidently.
How do you write from the example without overclaiming?
Begin with one column and compare the relevant cells. Then draft a relationship sentence, keeping the fictional source IDs attached.
A supported synthesis would be:
In this fictional set, A and B identify a March 2024 revision, while B alone identifies the affected entries as 4 and 9. C reports different wording but supplies neither the edition nor the entries, so its observation cannot be matched to that revision from the details given.
This is our original example paragraph. It combines the records while preserving the difference between a dated document and an unmatched observation.
An unsupported version would be: “All three sources prove that the March update improved the catalogue.” The exercise supplies neither agreement from all three nor evidence of improvement. The problem is not merely missing cautious language; the proposed conclusion asks the documents to establish something they do not contain.
Do not silently repair that sentence by changing “prove” to “suggest.” First remove the unsupported relationship and outcome claim. Then state the narrower relationship the records actually support.
Does an empty cell mean the source disagrees?
No. Use explicit working labels so an unfinished note does not become a negative finding:
| Label | What you mean |
|---|---|
| Not yet checked | You have not completed the relevant reading |
| Not discussed in the reviewed material | You checked the stated scope and found no discussion there |
| Not applicable | The question does not fit this source's role |
| Unclear | The wording or evidence needs another check |
These are our suggested labels, not a standardized coding system. Define any additional labels before using them.
“Not discussed in chapter two” is narrower than “the author never discusses this.” Record the scope of your reading. When a source is inaccessible, mark access incomplete rather than filling its row from another writer's description.
A disagreement requires statements that can actually conflict. Different questions, definitions, populations or editions may explain apparent differences; they should remain visible rather than being compressed into yes and no.
Should a theme appear in every source?
No. The Illinois chapter's example makes room for uneven coverage. Purdue also advises against ignoring outliers or forcing relationships.
Our editorial recommendation is to assess a theme's relevance and evidentiary support, not discard it solely because only one source addresses it. One item may supply an important qualification. Equally, a detail appearing everywhere may be peripheral to your actual question.
Keep a note of why a theme stays, changes or leaves. “Outside this project's question” is a more useful reason than “only one filled cell.” You are making a reasoned selection, not a popularity count.
How do you avoid turning the matrix into a second bibliography?
Give each tool a defined job:
- Source log: identity, retrieval, version and access details.
- Matrix: relationships across the selected comparison questions.
- Draft: a readable explanation with appropriate citations.
Do not paste the full log into every row. Keep the ID stable so the three tools remain connected. When you split a source into separate relevant sections, label those sections clearly without implying they are independent documents.
Before drafting, follow one matrix note back to its precise source location. Before sharing, follow one draft claim back through the matrix to that location. Fix any broken connection. This is our suggested traceability check, not a validated guarantee of accuracy.
The source-credibility guide handles the separate question of whether the material is suitable evidence in the first place.
When is the matrix ready to support a draft?
It is useful when you can explain the main relationships, locate their supporting passages and name the remaining limits. A completely filled grid is neither necessary nor sufficient.
Keep the project question and version date beside it. Preserve meaningful changes to your interpretation instead of silently overwriting the reasoning that led to an earlier draft. Store only information you are authorized to retain, in an appropriate system.
For academic work, follow the instructor's or institution's method and citation requirements. Medical, legal, financial, safety and accusatory conclusions require their own professional standards. An ordinary research worksheet does not establish those conclusions or justify publishing personal information.
Continue through Research Notes when you need help keeping evidence attached to the sentences it supports.