Choose evidence that actually supports a claim

A credible source can still be the wrong evidence for a particular sentence. An authoritative report about student wellbeing does not automatically support a claim about one college’s timetable. A survey about preferences cannot by itself establish that a policy caused better results. A quotation that mentions the same topic may not answer the point you are making.

The practical question is not only “Is this source reliable?” It is also:

What, exactly, can this evidence establish?

This guide shows how to match the kind and scope of evidence to the claim. It focuses on evidence fit across an argument rather than on building the sentences of one paragraph. For that separate task, use the guide to building a paragraph around a claim and evidence.

Write the claim before judging the evidence

You cannot assess fit while the claim remains a topic label. “Equipment access”, “student satisfaction” and “online learning” do not state what must be supported.

Turn the topic into a testable sentence:

Evening equipment collection was used by students on the certificate course.
Most eligible students preferred evening collection.
Evening collection caused more students to submit practical work on time.
The collection window can be sustained within the department’s current staffing.
The pilot should continue for another term.

These claims concern different matters: use, prevalence, causation, capacity and recommendation. They may appear in one essay, but they do not need the same evidence.

Yale’s elements of the academic paper distinguishes claims, evidence and analysis. It also notes that evidence often supports smaller claims that develop the paper’s main claim, and that the connection usually needs to be explained. Begin by isolating the smaller assertion you want a piece of evidence to support.

Identify the job the evidence must do

Ask what kind of judgment the sentence makes.

Description: What happened, existed or was said?
Prevalence: How common was it, and among which group?
Comparison: How did one condition or group differ from another?
Causation: Did one factor produce a change?
Experience or interpretation: How did people understand an event, text or process?
Feasibility: What time, access, skills or resources would implementation require?
Evaluation: How well did something meet stated criteria?
Recommendation: What action follows after benefits, limits, alternatives and costs are considered?

UNC Writing Center’s evidence guide emphasises that disciplines recognise different kinds of evidence. A passage from a novel may be central evidence in literary analysis, while numerical data may be needed for a quantitative claim. The assignment, disciplinary conventions and instructor’s requirements therefore come before any universal evidence hierarchy.

Do not ask whether interviews are “better” than statistics in the abstract. Ask whether interviews can answer this claim. Interviews may illuminate experience or process. They usually cannot, without a suitable sampling design, tell you how common that experience is across a population.

Run five evidence-fit tests

A useful source must pass more than a credibility check.

On a narrow screen, scroll the table sideways to read every column.

Questions to ask before treating information as supporting evidence
Fit test	Question	Typical mismatch
Construct	Did the evidence measure or examine the thing named in the claim?	File openings are treated as learning, or attendance is treated as engagement
Population	Do the people, texts, places or cases match the group named in the claim?	Volunteers from one class are used to describe all students
Comparison	Does the evidence provide the baseline or alternative required by the claim?	A post-pilot count is called an improvement without a comparable earlier count
Inference	Can the design support description, association or causation?	A difference between cohorts is presented as an effect of the intervention
Amount and variation	Is there enough relevant evidence, including exceptions or uncertainty, for the strength of the claim?	One example becomes “always”, or a small voluntary sample becomes “most”

A sixth question sits across all five: What qualification belongs in the sentence? George Mason University’s guide to softening academic claims shows how words such as “may”, “tend” and “some” can mark uncertainty or limited scope. Qualification is not a cosmetic way to make weak writing sound cautious. It should state the boundary the evidence actually creates.

Work through a fictional case pack

Eastmere Technical College, its equipment-loan pilot and every figure below are fictional teaching material. They are not evidence about a real institution.

Imagine this assignment:

Evaluate whether Eastmere should keep an evening collection window for equipment used on certificate-level practical courses.

The fictional case pack states:

One department opened an equipment counter from 17:00 to 19:00 on eight Tuesdays.
The department had 180 students eligible to borrow equipment.
The booking system recorded 84 evening collections by 53 distinct student accounts.
Thirty-one of those 53 accounts had previously collected equipment during daytime hours.
Logs record completed collections, not why students chose the time or whether access would otherwise have been impossible.
A voluntary online survey received 41 responses; 30 respondents said the evening window was easier to use than daytime collection.
The survey link was sent to students who used the evening window, not to all 180 eligible students.
Seventy-eight per cent of the previous year’s cohort submitted the practical assignment by the deadline; 82 per cent of the current cohort did so.
The cohorts were different, and the case pack gives no adjustment for assignment changes, student characteristics or other support.
Staff recorded 12 hours of additional paid time across the eight evenings.
Nine late returns were recorded in the eight pilot weeks, compared with eight in the preceding eight weeks.
The case pack does not report the number of loans in the earlier period, so late-return rates cannot be compared.
No interviews or open comments were collected about why the window helped or created difficulties.

These facts can support several bounded statements. They cannot support every attractive conclusion about access, preference, achievement or cost.

Audit each claim against the case

On a narrow screen, scroll the table sideways to compare each claim with its evidence.

What the fictional Eastmere evidence does and does not establish
Draft claim	Available evidence	Fit judgment	Defensible revision
Students used the evening window	84 collections by 53 distinct accounts	Directly supported as a descriptive claim	Fifty-three student accounts used the evening window at least once during the eight-week pilot
Most students preferred evening collection	30 favourable responses from 41 users who answered a voluntary survey	Population mismatch; non-users and most eligible students were not surveyed	Thirty of 41 voluntary respondents who had used the window said it was easier than daytime collection
The pilot increased on-time submission	82% in the current cohort versus 78% in the previous cohort	Comparison exists, but the unmatched cohorts do not isolate the pilot's effect	On-time submission was four percentage points higher in the pilot cohort, but the case pack cannot attribute the difference to evening collection
The pilot did not increase workload	12 additional paid staff hours	Contradicted by the recorded staffing measure	The pilot required 12 additional paid staff hours; its wider administrative workload was not measured
Evening opening increased late returns	Nine late returns during the pilot and eight before it	No comparable denominator; counts cannot establish a higher rate	The records show one additional late return, but different loan volumes may explain the count
The college should continue the service permanently	Use, voluntary feedback, staffing time and incomplete outcome data	The recommendation outruns the short, single-department pilot	Continue a time-limited trial while collecting comparable demand, cost and access evidence before making the window permanent

The audit does not reject useful evidence because it is imperfect. It assigns each item the job it can perform.

Do not substitute topic match for logical support

Evidence may repeat key words from a claim without supporting its reasoning.

Suppose the essay states:

Evening collection improves academic performance because flexible access helps students complete practical work.

A student comment that “the evening counter was convenient” is about access, but it does not measure academic performance. The submission comparison mentions performance, but the design cannot show that the counter caused the difference. Combining the two does not repair the missing causal bridge.

George Mason University’s discussion of missing the point describes the problem as support that does not actually establish the claim it accompanies. Test the logical relationship in a complete sentence:

If this evidence is accurate, it gives the reader reason to accept the claim because …

If the sentence requires a new assumption—such as “convenience always improves grades”—you have found a missing warrant, not completed the argument.

Purdue OWL’s argument-organisation guidance separates grounds from the warrant that explains how those grounds support a claim. You do not have to label each part in the essay, but the connection must survive scrutiny.

Separate counts, proportions and explanations

Numbers can look more decisive than they are. Check the denominator and the unit.

“Thirty students preferred the service” means something different when 41 users responded than when all 180 eligible students answered. “Nine late returns” cannot show that performance worsened unless you know how many loans could have been returned late in each period. “Eighty-four collections” is not the same as 84 students because one account may collect more than once.

Qualitative evidence has different limits. An interview could explain why an evening window mattered to a student managing work or caring commitments. It could also reveal a problem the booking logs did not capture. It would not automatically show how frequently that explanation applies across the college.

Use each form of evidence for its appropriate contribution:

counts and logs for recorded activity;
proportions with clear denominators for prevalence in the measured group;
interviews or open responses for experience, meaning and mechanisms;
suitable comparisons for differences;
stronger designs and alternative-explanation checks for causal claims;
resource records for feasibility; and
several relevant criteria for a recommendation.

Revise the claim when better evidence is unavailable

Finding more evidence is not always possible. The ethical alternative is to narrow the claim, not to disguise the gap.

A useful revision sequence is:

Underline the precise assertion.
Name the evidence needed to establish it.
Write what the available evidence actually measures.
Identify the missing population, comparison, mechanism or denominator.
Either find suitable evidence or rewrite the assertion at the supported level.
Explain why the bounded result still matters to the assignment.

For Eastmere, “the pilot improved completion” becomes “on-time submission was slightly higher in the pilot cohort, but the available comparison cannot establish why.” The revised sentence is less dramatic and more useful: it reports the pattern without turning it into a causal finding.

Check evidence fit before polishing the paragraph

For each important claim, ask:

What kind of claim is this: description, prevalence, comparison, cause, experience, feasibility, evaluation or recommendation?
Does the evidence measure the construct I named?
Does its population match the population in my sentence?
Is the denominator visible where a count or proportion needs one?
Does the comparison justify words such as “higher”, “improved” or “more effective”?
Am I describing an association as though it proved a cause?
Have I explained the warrant between evidence and claim?
Would “some”, “in this case”, “was associated with” or another qualification state the limit honestly?
Does contrary or missing evidence require me to revise the claim?
Would my discipline or assignment expect another kind of evidence?

Once the match is sound, use the claim-and-evidence paragraph guide to present the point, evidence and analysis clearly. If the mismatch appears across several paragraphs, move from description to critical analysis before editing sentence style.

If you want guided feedback on whether evidence in your own draft supports the claims you are making, academic writing development & revision can help you test the reasoning and revise your work while you retain authorship of the submission.