How to check if a source supports a claim
Reduce your sentence to the proposition it actually asserts, separate the scope from the claim, then open the source and find the passage that would settle it. Compare the population, the direction of the effect and any numbers you have used. You will end up with one of four answers — supported, partly supported, contradicted, or not establishable — and the difference between the last two matters more than anything else on this page.
The problem nothing else checks
A source can clear every check normally available to you. It exists. It is a peer-reviewed article in a real journal. The reference is formatted correctly and has a matching in-text citation. It is on the right topic and was published in the right decade.
And it can still fail to support the specific sentence citing it — because it studied a different population, measured a different outcome, found a weaker effect than you claimed, found the opposite, or simply never addressed the point you attached to it. Every tool that stops at the reference is checking whether the citation is well-formed. This is a different question, and the only way to answer it is to read the source.
It is also the failure most likely to be noticed by the person marking your work, because they often know the literature. A well-formatted citation to a paper that does not say what you claim reads worse than no citation at all.
A worked example
Your sentence
“Spaced repetition substantially improves long-term retention in undergraduates (Author, 2021).”
What the source actually reports
A study of 62 second-year medical students, measuring recall after two weeks, finding a modest improvement that did not reach significance in one of the two conditions tested.
Where it comes apart
- Population: medical students, not undergraduates generally.
- Timescale:two weeks is not obviously “long-term”.
- Magnitude:“modest” does not carry “substantially”.
- Reliability: a null result in one condition weakens the general claim.
The honest verdict
Partly supported. The direction is right and the source is relevant — but the sentence claims more than the evidence carries, in three separate ways. Narrowing it costs one clause and makes it defensible.
The method, step by step
1. Isolate the proposition you are actually asserting
Strip the sentence to the claim a reader would be entitled to challenge. "Recent work suggests that spaced repetition substantially improves long-term retention in undergraduates" asserts a direction (improves), a magnitude (substantially), a population (undergraduates) and a timescale (long-term). Each of those is separately checkable, and each can separately fail.
2. Separate the scope from the claim
Scope is the conditions under which you are asserting it — the population, the setting, the timeframe. Writers usually check the claim and forget the scope, which is why a finding about medical students ends up supporting a sentence about all undergraduates.
3. Open the source itself
Not the abstract, not the database summary, not what another paper said about it. Abstracts systematically omit the qualifications — sample restrictions, effect sizes, caveats — that decide whether your sentence is accurate.
4. Find the passage that would settle it
Search the source for the specific outcome you named. If the only relevant material is in the introduction or the discussion, you may be citing the source's summary of someone else's work — in which case the real source is the one it cites, and that is what you should be reading.
5. Compare the population and the setting
Who was studied, where, and under what conditions. A result from one country's schooling system, one clinical population or one professional group may not transfer to the group your sentence names, and the source will normally say so if you look.
6. Compare the direction of the relationship
Does the source find the effect you claim, the opposite, or no reliable effect at all? Reversed direction is more common than people expect, usually from remembering a paper's topic rather than its finding.
7. Compare the numbers you have used
Percentages, sample sizes, effect sizes, dates. If your sentence quantifies anything, that quantity has to appear in the source with the same meaning. Watch for a figure that describes a subgroup being used as if it described everyone.
8. Check the strength of your wording against the strength of the finding
"Shows", "suggests", "is associated with" and "proves" are different claims. A correlational study does not establish that one thing causes another, however strong the association, and overstating is one of the easiest ways to lose credibility with a marker who knows the literature.
9. Decide which of four things you have
Not two. The useful outcomes are supported, partly supported, contradicted, and cannot be established — and collapsing them is where the damage happens.
The four outcomes, and why four
Two outcomes — passed or failed — destroys the information you need. Anything that is not a contradiction gets read as a tick, so a partly supported claim goes in unchanged. And an unverifiable claim gets read as a problem with your writing when it is nothing of the kind.
Supported
The passage asserts what your sentence asserts, for a comparable population and in the same direction, at a strength your wording matches.
What to do: Nothing to do. This is the only one of the four that is a clean result.
Partly supported
The source backs part of the sentence and is silent on the rest — often the qualifier. It supports that retention improves, but not that the improvement is substantial, or not for the group you named.
What to do: Narrow the sentence to what the source carries, or add a second source for the remainder. Treating this as a pass is how the unsupported half ends up in your submission.
Contradicted
The source points the other way, or reports no reliable effect where your sentence claims one.
What to do: Fix the sentence. If you are citing it deliberately as a counterpoint, make that explicit in the text so it does not read as a misreading.
Cannot be established
Nothing available settled it — the source could not be retrieved, or its text does not address the point closely enough to judge. This is an absence of evidence, not a finding against you.
What to do: Read it yourself if you can get it, or cite something you can verify. What you must not do is treat it as a failure and delete a sound citation, or as a pass and move on.
The fourth is the one to be careful with. “Could not be established” and “the source does not support this” are completely different statements: one is an absence of evidence, the other is a finding. Any tool that reports the first as the second is telling you something it does not know.
How StudyRef does this, and where it stops
A Full Assignment Scan, or a Pro plan, runs this check across your document. For each claim it takes the sentence and the source it cites, retrieves the source’s full text where that text can lawfully be obtained — in practice, open-access work it can fetch — selects the passages that bear on the claim, and compares them against what you wrote. The result comes back as supported, partly supported, contradicted or not verifiable, with the passage it was judged against, so you can see the reasoning rather than take a verdict on trust.
The boundary is retrieval, and it is a real one. A great deal of academic work sits behind paywalls. Where the text cannot be obtained, or where what is available does not address the claim closely enough to judge, the result says the claim could not be checked and why. It does not fall back to guessing from a title and an abstract, and it does not produce a tick to avoid an awkward gap — a false clean result on the claim your argument rests on is worse than no result at all.
Two more limits worth stating. This checks whether a source supports your sentence; it does not check whether the source is correct, and a well-conducted study can still be wrong. And it does not replace your marker’s judgement about whether the evidence you chose was the right evidence to bring.
Check the evidence behind your claims
Upload your assignment. Claims are checked against the sources they cite, with the passage each judgement was made against — and a clear answer where a source could not be reached.
Check my assignment’s evidenceBefore you get here
Evidence checking is the slowest thing on this list, so it is worth clearing the mechanical problems first — a citation with no reference is a faster fix than a claim with no support.
- Match your in-text citations to your reference list — the bookkeeping check, and the one to do first.
- Check the draft against the brief — including whether the sources you used are the kind that was asked for.
- The full pre-submission checklist — everything else worth verifying before you upload.