Contents
The sentence I most want to watch in a draft is often the one that feels obvious. It may begin as a source's narrow observation, pass through my interpretation and emerge as a confident general claim. I would keep those stages separate while taking notes, before polished prose makes the transition difficult to notice.
I will use an invented example about opening a shared reading space. There are no real survey results here. The purpose is to show how a writer can preserve a source's scope, add a reasoned interpretation and make a proposal without pretending the three statements are the same kind of knowledge.
Name the kind of statement
I would begin with three labels: source description, interpretation and proposed action. A description reports what the source contains. An interpretation explains what I think it may mean for a question. An action recommends something to do. All three can belong in a useful brief, but their boundaries should remain visible.
Purdue OWL explains that quotation, paraphrase and summary all require attribution. I would add this working habit: attribution should also make it possible to see where the source ends. A citation attached to an entire paragraph can otherwise appear to support a recommendation the source never made.
Keep the original scope intact
Suppose a fictional report describes feedback collected at one afternoon session. I would not turn that into “Residents prefer afternoon meetings.” The broader statement introduces a population, a preference and a comparison that the description may not establish. A natural-sounding sentence can still contain several unsupported steps.
I would ask who was included, what was observed, how it was collected and under what conditions. Not every note needs a formal methods review. But if the claim depends on those limits, the note needs enough context to prevent me from forgetting them when I draft.
Write the three lines separately
Here is the demonstration in its simplest form. The “source” is a fictional document label; it is not evidence for an actual opening-hours decision. Separating the lines makes the missing work visible without requiring a complicated evidence-management system.
I might later combine them into readable prose. Before doing so, I would check that the wording still signals the inference: “This suggests a question worth testing,” rather than “This proves the new hours will work.” The quality of the reasoning matters more than the confidence of the verb.
Source description: The example report describes feedback from one afternoon session. My interpretation: The feedback could help identify questions for another session. Not established: What non-attendees need, or which time is best. Proposed action: Ask the group what information is still needed before choosing hours.
Let a note hold an unresolved claim
Zotero's notes provide places to keep source-specific comments and broader standalone thinking. I would use that distinction to preserve the source description separately from my evolving proposal. A note can contain an unresolved question without becoming defective or needing an immediate answer.
The key is naming what remains unresolved. “Needs research” is less actionable than “Check whether the feedback includes people who left early.” A precise open question tells the next reader why the current conclusion is limited and what kind of evidence could change it.
Compare sources without averaging away disagreement
If another source points in a different direction, I would first compare the questions and conditions. Perhaps it concerns a different group or a different kind of event. Two apparently conflicting statements may not be answering the same question. Combining them into a vague middle position can hide the useful difference.
If the conflict is genuine and relevant, I would retain it in the brief. A recommendation can be conditional while a question remains open. I would not present a numerical confidence score unless the project has a meaningful basis for it. An explained limitation is often more honest than invented precision.
Check the boundary again in the finished paragraph
My final pass would underline source-supported claims and circle my own inferences. If I cannot tell which is which, the reader probably cannot either. I would inspect the citation placement, the verbs and any sentence that quietly changes “this group” into “people.”
I would then ask whether the proposed action exceeds what I know. It may still be reasonable to try a small, reversible next step. But I should explain that as a decision under uncertainty, not as a result established by the source. That is a useful first-person voice: accountable about why I would choose something.
Examine a paragraph that changes scope while sounding natural
I would draft a deliberately imperfect sentence for the fictional reading-space proposal: “The report's positive feedback shows that residents want more afternoon sessions.” It sounds plausible, but it contains several steps. What does positive mean? Who supplied feedback? Were they residents, attendees or another group? Did the report ask about additional sessions? I would separate those questions before making the prose smoother. A fluent sentence can conceal the exact point where description becomes inference.
My first revision would state only what I can verify from the hypothetical document's actual contents once inspected. The note might identify the group invited to comment and the topic the feedback addressed. I would avoid supplying invented responses in the demonstration. The example is useful without them: it shows that a report of feedback does not automatically establish a preference about time or frequency. Those are additional claims with their own requirements.
The next line would identify my interpretation. Perhaps I think the feedback suggests a question worth discussing with the group. That can be a reasonable editorial judgment, but I should explain the link. I would say why the question matters to the proposed decision and what remains unknown. A citation can help the reader inspect the underlying material; it should not make the source appear responsible for a recommendation it did not make.
Finally, I would write the proposed action separately: ask the group what information is needed before choosing the next schedule. This action may fit the uncertainty, but it is still a decision rather than a finding. I would check whether its wording creates an unintended commitment, such as implying that more sessions are already approved. The paragraph should distinguish investigation from agreement. That distinction makes the subjective voice useful: the writer owns the judgment while leaving the evidence available for inspection.
Build a claim record with a visible boundary
For a central sentence, I would keep a small claim record. It would contain the proposed wording, the source location, the scope supported and the part that is my interpretation. I would not create this record for every ordinary sentence. It is most useful where a draft relies on a source to justify an important choice or where a disagreement has exposed uncertainty. The amount of structure should reflect the consequence of misunderstanding the claim.
In the example, the record might say that the source describes feedback from one session, while the draft asks about scheduling future sessions. That difference identifies the bridge that needs explanation. If another source supplies relevant information, I would add its role without pretending two sources necessarily provide independent confirmation. They may share the same underlying data or concern different questions. A longer reference list does not automatically repair a mismatch between the claim and its basis.
I would keep the most important limitation near the sentence it affects. A general caveat at the end of the document may be too far away to guide interpretation. If the evidence concerns one group under specific conditions, the sentence should retain that boundary. This need not make the writing cumbersome. A precise subject and a carefully chosen verb can do more work than a paragraph of broad warnings added after an overstated claim.
The claim record should also identify what would change the current view. Perhaps the writer needs the report's definition of attendance or clarification of how comments were collected. That is a concrete next check. “More research” is less useful because it does not say what kind of information would matter. Naming the missing connection gives the investigation an end and prevents the writer from collecting more general material while avoiding the one question that affects the draft.
Compare two sources without manufacturing a consensus
Imagine that a second fictional source describes a different event with a different attendance pattern. I would first compare the circumstances. The sources may discuss different populations, purposes or collection methods. If so, they do not necessarily disagree. I would preserve those conditions rather than compress the sources into a statement about what people generally prefer. A broad synthesis is a new claim, and it needs a reasoned basis beyond the existence of two related documents.
If the sources do conflict on a relevant point, I would state the conflict in terms the reader can assess. Which question receives different answers? Are the materials current? Does one concern an official policy while the other describes an informal practice? The proper next step depends on that distinction. An event organizer can clarify an arrangement; another general article cannot necessarily resolve what is authorized in a particular setting. I would avoid treating every disagreement as a contest in which the more convenient source wins.
My interpretation can still be clear. I might prefer a small reversible investigation while the conflict remains open. I would explain why that action fits the uncertainty and what its limits are. I would not call it evidence that the preferred outcome will occur. A useful recommendation can acknowledge incomplete information without becoming indecisive. The reader needs to understand the tradeoff being accepted, not be reassured by stronger words than the material permits.
When the conflict is resolved, I would update all the places where the old interpretation matters. A source note, project brief and final paragraph can drift apart if only one is corrected. I would keep a short explanation of a substantial change when the project's history needs it. That gives the next reader a way to understand why an earlier draft reached another view. The goal is a current, coherent argument with a traceable basis, rather than a collection of individually polished sentences pointing in different directions.
Review the finished draft as a skeptical collaborator
I would read each source-supported paragraph and ask what a skeptical but fair collaborator would need to inspect. Can they locate the passage? Can they distinguish the source's wording from my explanation? Does the sentence retain important conditions? Is the recommendation clearly mine or the group's, rather than attributed vaguely to research? These questions help evaluate the actual argument without assuming that the presence of citations has already answered them.
I would pay particular attention to verbs. “Describes,” “suggests,” “demonstrates” and “requires” do different work. A source may describe what happened in one setting without demonstrating why it happened. A report may suggest an avenue for inquiry without requiring a particular action. I would choose the verb for the relationship I can explain, not for the rhetorical strength I want the sentence to have. That makes the prose more precise without automatically making it longer.
I would also inspect nouns that have become broader during editing. “Respondents” can quietly become “attendees,” and attendees can become “residents.” A wording change may seem stylistic while altering the population covered by the claim. I would compare the final sentence with the claim record and source, especially after a substantial rewrite. The note-taking boundary matters most when the polished draft is far enough from the original reading that its limits are easy to forget.
The last pass would examine the proposed action. Does it follow from the explanation, and does it respect the open questions? A reasonable next step might be asking for clarification, narrowing a proposal or dropping a claim. It need not always be collecting another source. I would finish when the reader can inspect the evidence and understand the judgment separately. That leaves room for disagreement about the recommendation without creating confusion about what the source actually established.
Give a collaborator two things to disagree with separately
I would hand a collaborator the source description and my recommendation as separate statements. They might accept that the report describes one session's feedback while disagreeing that the group should ask another scheduling question. That is a productive disagreement. It lets the conversation focus on the judgment without first arguing about which words belong to the source.
A different response may challenge the source description itself. Perhaps I missed a definition or summarized the passage too broadly. I would inspect that issue before defending the recommendation. The recommendation may still be reasonable, but it should not depend on preserving an inaccurate account of the evidence. Keeping the two lines distinct makes correction possible without requiring every part of the writer's position to stand or fall together.
I would ask the collaborator to identify the step they would change: the observation, its scope, the inference or the proposed action. That turns a broad comment such as “not convincing” into a more useful revision question. It also avoids treating confidence as the deciding factor. The argument improves when the relevant relationship can be explained and inspected, not merely repeated with stronger verbs or more citations.
After revision, I would check the finished paragraph again for a clean boundary. The source should receive credit for what it actually contributes; the writer should own the interpretation and choice. A first-person voice can help by making that responsibility explicit. The result is an article that can express a clear view while allowing the reader to assess its basis, recognize its assumptions and disagree with its conclusion without being misled about the underlying material.
Frequently Asked Questions
Q: Can I include an opinion in a sourced article?
A: Yes. Make clear that it is your interpretation or recommendation and explain its basis. A linked source should not be made to appear responsible for an opinion it does not express.
Q: Does paraphrasing remove the need for a citation?
A: No. Paraphrasing changes the wording, not the origin of the idea or information. Preserve a source pointer and avoid presenting the source’s work as your own.
Q: How much uncertainty should a note include?
A: Enough to prevent the claim from being used beyond its support. Focus on limitations that could change the decision rather than adding a generic warning to every sentence.
Q: What if a source has no clear author?
A: Record the available organization, title, publication details and URL, then assess whether it is suitable for the claim. Missing attribution is a reason to inspect carefully, not permission to guess an author.
Q: Can I make a recommendation with incomplete evidence?
A: Sometimes a conditional next step is reasonable. State what is known, what remains open and why the action fits that uncertainty. Avoid presenting the recommendation as an established finding.
One thing to take forward
I would keep two different lines for “the source says” and “I think,” even if the final article brings them into the same paragraph. The boundary lets a reader assess both the evidence and the judgment, which is more useful than asking a citation to do all the explaining.
Sources & further reading
Something needs correcting? Write to the editors with the article link.



