Advanced vocabulary evidence is not a thesaurus ladder: suggests, indicates, supports, shows, demonstrates, confirms, establishes and proves do not simply mean “weak evidence” through “strong evidence” in fixed steps. Each verb changes what the writer claims the evidence is doing, and the right choice depends on the evidence, method, alternatives, scope and discipline.

Maya reads a sentence in a report:

The results prove that the revision method improves examination performance.

Hana looks at the study description.

“One class used the method for two weeks. There was no comparison group.”

Jia Jun reaches for a softer verb.

“Then say the results suggest it.”

Ethan stops him.

“Not automatically. First ask what the results actually support. Maybe the scores improved after the method. That is evidence of change. The causal claim needs another bridge.”

The problem is not vocabulary decoration.

It is evidence calibration.

Cambridge Dictionary defines suggest, in one evidence-related sense, as communicating or showing an idea without stating it directly or supplying proof. Indicate means to show or make something clear. Support, in an evidential sense, means to help show that something is true. Demonstrate can mean to show that something is true. Confirm can mean to state or show that something previously uncertain is true. Prove means to make it clear that something is true.

Those dictionary meanings overlap.

Academic and scientific writing therefore cannot be controlled by dictionary definitions alone.

The writer must also ask:

• What is the exact claim?
• What evidence supports it?
• How directly does the evidence bear on the claim?
• What alternative explanations remain?
• Does the method justify the inference?
• Is the statement descriptive, predictive, causal, mathematical, legal, historical or interpretive?
• How far does the evidence travel beyond the observed case?

Purdue OWL makes the argument structure explicit: evidence supports a claim through a warrant or bridge explaining why the evidence bears on that claim. That bridge matters here because evidence verbs can hide it.

The graph shows that attendance rose.

The graph may directly display the rise.

The graph proves that the new policy caused attendance to rise.

Now the verb is carrying a causal interpretation that the graph alone may not earn.

This article owns one narrow applied job in the eduKatePunggol advanced vocabulary chain: how readers and writers audit the evidential force of verbs such as suggests, indicates, supports, shows, demonstrates, confirms, establishes, corroborates, verifies and proves without turning them into a fake certainty scale.

It is distinct from Advanced Vocabulary Argument | The Qualification Audit, which owns words such as often, usually, mainly, partly and other claim qualifiers; Advanced Vocabulary Causality | The Cause–Mechanism Audit, which owns causal relation language; Advanced Vocabulary Reading | The Assumption Audit, which owns hidden commitments; and the eduKateSG reporting-verb owner, which handles source stance in words such as says, claims, argues, suggests when attributing speech.

The Evidence-Strength Audit asks one question first:

What exactly has the evidence earned?

Then it asks:

1. What is directly observed?
2. What claim is inferred?
3. What bridge connects them?
4. What alternatives remain?
5. Which evidence verb represents that relationship without inflation or needless weakening?

The examples below are constructed instructional material. The external references support the general language and argument principles; the Evidence-Strength Audit is an original classroom application.

The fifty-second route: evidence → claim → bridge → alternative → verb

1. Name the evidence. Observation, measurement, source, example, experiment, comparison, testimony, proof step?
2. Name the claim. What sentence is the evidence meant to justify?
3. Write the warrant. Why should this evidence increase confidence in that claim?
4. Generate one alternative. What else could fit the same evidence?
5. Check scope. One case, this sample, this task, this population, all cases?
6. Choose the verb last. Suggests, indicates, supports, shows, demonstrates, confirms, establishes, corroborates, verifies, proves?

Reader routes: Not a fixed ladder · Suggests and indicates · Supports · Shows and demonstrates · Confirms, establishes and verifies · Proves · Corroboration · AI evidence drift · Case files · Practice · Sources.

First rule: these verbs are not a universal evidence-strength ladder

It is tempting to teach:

suggests < indicates < supports < shows < demonstrates < confirms < proves.

That can be a useful first approximation.

It becomes dangerous when treated as a rule.

Why?

• Different disciplines use the verbs differently.
• Writers can choose cautious wording even when evidence is strong.
• A strong verb can be used for a narrow descriptive fact but not for a broader causal claim.
• Confirm can refer to official verification rather than statistical strength.
• Demonstrate can mean show how to perform a procedure rather than establish a proposition.
• Proof in mathematics has a different status from “proof” in ordinary empirical discussion.
• Support can be strong if many independent lines of evidence converge, even though the verb sounds moderate.

Therefore the real skill is not memorising a ranking.

It is matching the verb to the inferential job.

The five evidence jobs

1. Pointing: the evidence is consistent with or points toward a possibility.
2. Supporting: the evidence gives the claim positive weight without settling every alternative.
3. Displaying: the evidence directly presents the reported pattern or fact.
4. Discriminating: the evidence rules out important competing explanations.
5. Establishing: within a defined system of proof, verification or decision, the claim has met the required standard.

The same verb can sometimes perform more than one job.

The audit therefore begins with function, not prestige.

Laboratory A: suggests and indicates — pointing without pretending the alternatives are gone

Suggests is one of the most useful verbs in academic writing because it allows evidence to point toward an interpretation without pretending the interpretation is the only one possible.

Cambridge Dictionary explicitly gives an evidence-related sense in which suggest communicates or shows an idea without stating it directly or giving proof.

That makes suggests useful when:

• the pattern is compatible with the claim;
• the evidence is preliminary;
• other explanations remain;
• the sample is narrow;
• the method is exploratory;
• the writer wants to separate observation from interpretation.

Case 1. A pattern suggests a bottleneck

Constructed observation:

• Maya answers familiar comprehension questions accurately;
• she fails mainly when a question contains unfamiliar abstract vocabulary;
• after the crucial word is explained, her later reasoning is usually correct.

Reasonable:

The pattern suggests that vocabulary access is one important bottleneck in these items.

Why not proves?

Because the observed repair is informative but may not isolate every cause. The explanation supplied by the tutor can also change attention, confidence or task representation.

Case 2. Suggests can still support an important claim

Some learners mistakenly hear suggests as “weak and unimportant”.

That is wrong.

A large, well-designed body of evidence may still be described as suggesting a conclusion because the authors are preserving epistemic caution or because the conclusion is broader than any single design can establish completely.

Verb strength is not a numerical confidence score.

It is part of the writer’s representation of the evidence-claim relationship.

Indicates: evidence points more directly, but still does not explain itself

Indicates often sounds firmer than suggests because it presents the evidence as pointing toward or making something clear.

It works especially well when the evidence functions as a sign or measurement:

• a thermometer indicates temperature;
• a graph indicates an upward pattern;
• a survey result indicates the distribution of reported answers in the sample;
• a performance pattern indicates a possible learning state.

Case 3. The graph indicates a rise; it does not indicate the cause of the rise

Suppose a graph displays average score by week:

• Week 1: 61;
• Week 2: 64;
• Week 3: 68;
• Week 4: 71.

Defensible:

The graph indicates a sustained rise in average score across the four weeks.

Not earned by the graph alone:

The graph indicates that the new revision routine caused the rise.

The first claim is descriptive.

The second adds a causal theory.

One verb can therefore be appropriate for the pattern and inappropriate for the explanation.

The suggests–indicates audit

1. Is the evidence pointing toward an interpretation or directly displaying the reported feature?
2. Does the sentence preserve alternative explanations?
3. Is the verb being asked to carry a causal or generalising claim beyond the observation?
4. Would “suggests” better preserve uncertainty, or would that falsely weaken a direct descriptive observation?

Laboratory B: supports — positive evidence without automatic closure

In argument, evidence supports a claim when it gives the claim weight.

This does not mean the claim becomes certain.

Purdue OWL’s argument model makes the bridge visible: data or grounds support a claim through a warrant explaining why the evidence matters.

That means supports is especially useful because it can remain true even when:

• the evidence is not conclusive;
• other evidence points the other way;
• the claim is probabilistic;
• the claim is one explanation among several;
• more evidence is still needed.

Case 4. One result can support a claim without settling it

A new vocabulary routine is followed by improved delayed recall in one class.

Safe:

The result supports the hypothesis that the routine can improve delayed recall under these conditions.

The sentence does not pretend one class settles every population, mechanism or implementation question.

Case 5. Evidence can support two competing claims

Hana takes longer on difficult questions but answers more accurately.

The same pattern can support:

• Hana uses additional time to reason carefully;
• Hana lacks automaticity on difficult items.

More evidence is needed to discriminate.

Support is therefore not exclusive ownership of the data.

Case 6. “Supports” needs an object with the right scope

Weak scope:

The three correct answers support the claim that Ethan has mastered algebra.

Better:

The three correct answers support the narrower claim that Ethan can solve this equation form independently.

The evidence did not get stronger.

The claim got smaller enough to fit it.

Support is relational

Evidence is not “strong” in the abstract.

It is strong for a particular claim.

A video recording can be excellent evidence that a learner used a particular strategy on one occasion.

It is weak evidence that the learner will use the strategy independently in every future context.

Always read:

evidence strength = evidence × claim × context.

Laboratory C: shows and demonstrates — direct display versus interpretive overreach

Shows is common because it feels plain and decisive.

It is safest when the evidence directly displays the proposition being stated.

Case 7. The table shows a difference

Table:

• Group A average: 72;
• Group B average: 64.

Defensible:

The table shows an 8-point difference in the reported averages.

Not automatically:

The table shows that Method A is better.

Better requires a criterion and may require causal evidence about the method, not merely group difference.

Case 8. The response shows knowledge on one item

A learner correctly explains why despite introduces a concessive relation in one sentence.

Defensible:

The response shows accurate understanding of the relation in this example.

Too broad:

The response shows mastery of concessive language.

One item can display one success.

Mastery is a pattern across variation.

Demonstrates: stronger when performance itself instantiates the capability

Demonstrates is often useful when a performance directly exhibits the quality being claimed.

Example:

By identifying the hidden assumption and revising the sentence without prompting, Hana demonstrates the target editing skill on this item.

The performance itself instantiates the skill.

But the scope still matters.

One successful demonstration does not automatically establish durable, generalised mastery.

Case 9. Demonstrates can be too strong when the method is indirect

Suppose students who report more reading also have larger vocabularies.

The survey demonstrates that reading builds vocabulary.

Overstated if the design is observational and reverse causation or common causes remain.

Better:

The survey shows an association between reported reading frequency and vocabulary score and supports further investigation of a possible reciprocal relationship.

The show–demonstrate audit

1. Does the evidence directly display the stated fact?
2. Is the sentence adding an explanation not visible in the evidence?
3. Is a one-item performance being expanded into mastery?
4. Does “demonstrates” describe exhibited capability, or is it being used to give prestige to an indirect association?

Laboratory D: confirms, establishes and verifies — closure depends on the question and the standard

Confirms often sounds like the argument is over.

Sometimes that is appropriate.

Sometimes it is not.

Cambridge gives confirm a sense of making something previously uncertain certain or stating/showing that it is true. In real writing, that force depends on what exactly is being confirmed.

Case 10. Confirming identity is not the same as confirming an explanation

A school office checks an identification number against its records.

The record confirms the student’s registration.

Appropriate if the record is authoritative for that administrative fact.

Now compare:

The student’s low score confirms that anxiety caused the performance drop.

The score cannot confirm its own explanation.

The first sentence verifies a bounded status claim.

The second jumps from outcome to cause.

Case 11. “Confirms our hypothesis” can become confirmation bias in sentence form

A learner expects a new strategy to help.

One result improves.

They write:

This confirms our hypothesis.

But they ignore:

• two outcomes that did not improve;
• a task-difficulty change;
• a competing explanation;
• the absence of a comparison group.

Better:

The result is consistent with the hypothesis and provides some support for it, but the current design does not rule out several alternatives.

Confirm should not be used merely because the result points in the expected direction.

Establishes: strong, but often narrow and procedural

Establishes can be excellent when a method settles a defined question under accepted rules.

Examples:

• a document establishes that a date was recorded;
• a controlled experiment can establish a treatment effect under specified conditions;
• a proof establishes a theorem within a formal system;
• a sequence of observations can establish that a pattern occurred in the measured sample.

The verb is strongest when the closure rule is clear.

Case 12. Establishing a narrow fact without establishing the whole theory

Constructed experiment:

• students are randomly assigned to two study conditions;
• the retrieval-practice group has higher delayed recall;
• implementation and measurement are sound.

Defensible:

The experiment establishes a causal effect of assignment to retrieval practice on delayed recall under the tested conditions.

Not automatically:

The experiment establishes that retrieval practice is the best study method for every learner and subject.

The first claim matches the design.

The second adds universal scope and comparative superiority.

Verify: check against a criterion, record or independently reproducible condition

Verify is useful when the task is checking whether a statement, value, result or identity matches an external criterion or independent procedure.

Examples:

• verify a calculation by reversing it;
• verify a quotation against the source;
• verify a measurement with an independent instrument;
• verify that a file hash matches;
• verify that a learner can reproduce the result without the original scaffold.

Verification is therefore different from explanation.

You can verify that an answer is correct without knowing why the learner got it right.

Case 13. Verified answer, unverified understanding

Ethan’s final answer matches the answer key.

Verified:

the final answer is correct under the key.

Not verified:

• the reasoning was valid;
• the method was independently chosen;
• the knowledge will transfer;
• the result was not copied.

Verification always needs an object:

verify what?

The confirm–establish–verify audit

1. What exact uncertainty is being closed?
2. What standard or authority permits closure?
3. Is the verb confirming a fact, a status, a pattern, a cause or a theory?
4. Does the method actually discriminate the claim from its alternatives?
5. Would a narrower object make the verb defensible?

Laboratory E: corroborates, converges with and is consistent with — agreement is not always independent confirmation

Corroborates usually means that another piece or source of evidence strengthens support for a claim.

But corroboration is valuable only to the extent that the new evidence adds genuinely new information.

Case 14. Three witnesses repeating one rumour are not three independent confirmations

Source A hears a claim from Source X.

Source B also hears it from Source X.

Source C quotes Source B.

The story now appears in three places.

It may still rest on one original source.

Numerical repetition is not evidential independence.

Case 15. Three school indicators can be one underlying measure in disguise

A report says:

• quiz score improved;
• percentage correct improved;
• number correct improved.

These may all be transformations of the same underlying responses.

They do not provide three independent confirmations of learning.

Stronger corroboration would come from a different surface or method:

• delayed recall;
• new-context transfer;
• independent explanation;
• school performance outside the trained worksheet.

“Consistent with” is a compatibility claim, not a victory

One of the safest and most misunderstood phrases is:

The evidence is consistent with the hypothesis.

This means the hypothesis can accommodate the evidence.

It does not mean the hypothesis uniquely explains the evidence.

Many rival explanations may also be consistent with the same result.

Case 16. Wet pavement

Wet pavement is consistent with:

• rain;
• a sprinkler;
• street cleaning;
• a burst pipe.

Therefore:

The wet pavement is consistent with recent rain.

is weaker than:

The wet pavement proves that it rained.

The difference is not mere politeness.

It is the number of surviving explanations.

The corroboration audit

1. Is the second source genuinely independent?
2. Does it use a different method or merely re-express the same data?
3. Does it test the same claim under a new condition?
4. Does the evidence merely fit the hypothesis, or does it discriminate against alternatives?

Laboratory F: proves — one word, radically different standards across domains

Proves is the verb most likely to overheat a school essay.

It is not banned.

It belongs where the relevant proof standard is actually met.

Mathematics: proof is deductive

In mathematics, a proof derives a conclusion from accepted definitions, axioms and previously established results through valid reasoning.

If the proof is valid, the theorem follows within the formal system.

That is a much stronger relation than ordinary empirical support.

Empirical science: evidence usually supports models under uncertainty

Scientific experiments can establish highly reliable empirical relationships and causal effects under well-defined conditions.

But scientific theories remain open to refinement as measurements, methods and scope change.

This is why careful scientific prose often prefers supports, demonstrates, provides evidence for, is consistent with or establishes under the tested conditions rather than treating every empirical result as mathematical proof.

Legal settings: proof depends on a defined evidential standard

Legal systems use standards of proof rather than mathematical certainty.

The precise standard depends on jurisdiction and proceeding.

The general lesson for vocabulary is that proved can mean “met the required legal standard,” not “eliminated every logically imaginable possibility.”

History and humanities: evidence can establish facts without proving one total interpretation

A dated official document can establish that a meeting occurred.

It may not prove the motive of every participant.

A diary can establish what the writer recorded.

It may not establish that every recorded belief was factually correct.

Strong humanities work separates:

• fact established by source;
• interpretation supported by source;
• motive inferred from source;
• broader explanation constructed from several sources.

Classroom argument: “prove” often means “support strongly” and should be tested

A student writes:

This quote proves that the character is selfish.

Ask:

• Could the quote also fit fear?
• Could the act be strategic rather than selfish?
• Is one episode representative of the character?
• Does context alter the interpretation?

If alternatives survive, supports the interpretation that may be more accurate.

The prove audit

1. What domain are we in?
2. What counts as proof here?
3. Is the inference deductive, empirical, legal, historical or interpretive?
4. Has the relevant standard actually been met?
5. Could the sentence keep its force with “supports”, “shows” or “establishes” instead?

The evidence-verb field guide: the relationship matters more than the prestige of the verb

The verbs below are not ranked from weakest to strongest. They describe different evidence jobs.

Suggests

The evidence points toward an interpretation while alternatives remain. Useful for exploratory patterns, preliminary evidence and cautious inference.

Indicates

The evidence functions as a sign or points more directly to a state or pattern. Often useful for measured trends, signals and descriptive findings.

Supports

The evidence gives positive weight to a claim. It does not necessarily eliminate rival claims.

Provides evidence for

A transparent phrase that foregrounds the evidential relationship without pretending the issue is closed.

Is consistent with

The evidence fits the claim. It does not show that the claim is uniquely supported because other explanations may fit too.

Points to

Directional language similar to indicates, often slightly less technical. It says where the evidence points, not whether the destination is uniquely determined.

Shows

Best when the evidence directly displays the stated fact or pattern. It becomes too strong when the sentence adds explanation, causation or universal scope not visible in the evidence.

Reveals

Suggests that something previously hidden became visible. Strong and often rhetorical. Ask whether the evidence truly uncovers a hidden state or merely supports one interpretation.

Demonstrates

Can mean directly exhibits a capability or shows a proposition to be true. Strongest when the observed performance itself instantiates the claimed skill or the method directly tests the proposition.

Illustrates

Provides an example that makes an idea visible. Illustration is not proof. One vivid example can explain a mechanism without establishing frequency or generality.

Corroborates

Adds supporting evidence from another source or route. Stronger when the new source is genuinely independent.

Confirms

Closes a previously uncertain fact or status when the evidence or authority is sufficient for that question. Dangerous when used merely to mean that the expected result occurred.

Establishes

Strong closure language for a defined proposition under an accepted method or standard. Scope should remain explicit.

Verifies

Checks a statement, value, identity, result or performance against an external criterion or independent procedure.

Rules out

Eliminates an alternative under specified assumptions or test sensitivity. Strong discrimination language; a weak or insensitive test may fail to rule out much.

Refutes

Provides evidence against a claim strongly enough that the claim, as stated, no longer survives. A counterexample can refute a universal statement while leaving a weaker statement intact.

Proves

Shows something to be true according to the governing proof standard. Its strongest ordinary home is formal proof; empirical uses require careful attention to method, scope and domain.

The verb-object rule: audit the whole phrase, not the verb alone

Compare:

The data show a 12-point increase.

Potentially direct.

The data show that the programme works.

Much broader.

The experiment demonstrates lower error rates under the tested condition.

Bounded.

The experiment demonstrates superior teaching quality.

Now the sentence imports a broad evaluative construct.

A strong verb can be accurate when its object is narrow. A cautious verb can still mislead when its object is enormous.

Laboratory G: AI evidence drift — when fluent rewriting changes what the evidence has earned

AI rewriting often improves fluency by replacing repeated verbs.

That is exactly where evidential force can change unnoticed.

Case 17. “Supports” becomes “proves”

Source:

The findings support the hypothesis that repeated retrieval improves delayed recall under the tested conditions.

AI rewrite:

The findings prove that repeated retrieval improves memory.

The rewrite expands three things at once:

• support becomes proof;
• delayed recall becomes memory generally;
• tested conditions disappear.

Case 18. “Suggests” becomes “shows”

Source:

The response pattern suggests that unfamiliar vocabulary may be contributing to the errors.

AI concise rewrite:

The response pattern shows that vocabulary causes the errors.

The model removes uncertainty, partial contribution and alternative causes.

Case 19. “Shows” becomes “suggests” and throws away direct evidence

Source:

The attendance register shows that 28 of 30 students were present.

AI cautious rewrite:

The register suggests that most students may have been present.

If the register is the authoritative record and the sentence merely reports it, the rewrite is needlessly weak.

Caution is not accuracy when direct evidence has been earned.

Case 20. “Verified” becomes “understood”

Source:

The learner’s final answer was verified against the answer key.

AI report:

The learner demonstrated full understanding.

Correctness of the final answer was verified. Understanding was not.

Case 21. “Consistent with” becomes “confirms”

Source:

The pattern is consistent with the possibility that fatigue contributed to the late-paper errors.

AI rewrite:

The pattern confirms that fatigue caused the late-paper errors.

Compatibility became causal closure.

Case 22. “Establishes under these conditions” becomes “always works”

Source:

The randomised comparison establishes a reduction in error rate under the tested conditions.

AI promotional rewrite:

The method is proven to work for students.

The rewrite drops the specific outcome, comparison, tested conditions, population and scope.

A practical AI prompt for evidence-verb preservation

Rewrite for clarity without changing evidential force. Preserve the distinction among suggests, indicates, supports, is consistent with, shows, demonstrates, corroborates, confirms, establishes, verifies, rules out, refutes and proves. Do not replace a moderate verb with a stronger one unless the underlying evidence and inference justify the change. Do not weaken a directly verified or established fact into vague suggestion merely to sound cautious. Preserve the exact object, scope, population, outcome, conditions, alternatives and causal status attached to the verb. Flag every evidence-verb substitution and explain what inferential commitment changed.

The AI evidence-delta check

1. Verb: did the evidence verb change?
2. Object: did the thing supposedly shown or proved become broader?
3. Scope: did this sample become all students?
4. Outcome: did delayed recall become memory?
5. Causality: did association become cause?
6. Alternatives: did “consistent with” become “confirms”?
7. Conditions: did “under tested conditions” disappear?
8. Domain: did empirical evidence acquire mathematical-proof language?

Extended evidence case files: what the evidence earns, and what it does not

The most dangerous evidence verbs appear in sentences that are mostly true.

The problem is often not the data.

It is the extra distance between the data and the conclusion.

Case file 1. The comprehension repair

Constructed record:

• Maya answers a comprehension item incorrectly.
• The item depends on the word reluctant.
• Maya defines reluctant incorrectly.
• The tutor supplies the correct meaning only.
• Maya then identifies the relevant sentence and answers correctly without further help.

Possible sentence A:

The episode proves that vocabulary caused the comprehension error.

Too strong.

The repair strongly implicates vocabulary access in this item, but giving the word meaning may also have redirected attention or clarified the problem representation.

Possible sentence B:

The episode supports the interpretation that vocabulary access was a critical bottleneck in this item.

Better.

Possible sentence C:

The episode shows that Maya initially misinterpreted the target word and answered correctly after its meaning was supplied.

Direct descriptive claim.

Notice the pattern:

• shows fits the observed sequence;
• supports fits the bottleneck interpretation;
• proves overstates the causal diagnosis.

Case file 2. The class average rises after a new routine

Constructed data:

• average before routine: 64;
• average after four weeks: 71;
• no comparison group;
• students also completed more timed practice;
• the later assessment may have been easier.

Claim:

The results demonstrate that the routine improves scores.

Problem:

The data demonstrate—or simply show—the score increase in this group. They do not isolate the cause of the increase.

Better:

The class average rose from 64 to 71 after the routine was introduced. The change is consistent with the possibility that the routine contributed, but additional changes and the lack of a comparison group limit causal interpretation.

Sometimes the best evidence writing uses two sentences because one verb cannot carry the whole epistemic structure honestly.

Case file 3. The quotation in a literature essay

Passage:

A character refuses to share food during a crisis.

Student sentence:

This proves that he is selfish.

Alternative readings:

• selfishness;
• fear of shortage;
• responsibility for a dependent person;
• mistrust of the group;
• previous betrayal.

One action can support a character interpretation without proving a stable trait.

Better:

The refusal supports the interpretation that the character acts self-protectively in this moment. Whether this reflects selfishness as a stable trait depends on the wider pattern of his actions and motives.

The second sentence protects the distinction between episode and identity.

Case file 4. The science investigation

Constructed design:

• identical plants begin at comparable height;
• light exposure differs between groups;
• water, soil and temperature are controlled;
• growth is measured by height after two weeks;
• high-light plants grow more on average.

What can the experiment demonstrate?

Under the tested conditions, the light treatment affected height growth.

What can it not prove automatically?

• all plant species respond the same way;
• more light always improves every dimension of plant health;
• the same effect continues indefinitely;
• height is identical to total biomass.

Good sentence:

The experiment demonstrates greater height growth under the higher-light condition over the two-week period.

Excellent because the verb is strong and the scope is narrow.

Case file 5. The mathematical proof and the empirical check

Ethan proves algebraically that the sum of two odd integers is even.

He also checks five numerical examples.

Which evidence proves the general statement?

The valid algebraic proof.

What do the examples do?

They illustrate the pattern and may help detect obvious mistakes.

Five examples do not prove the universal theorem.

This is a crucial cross-subject lesson:

example evidence and deductive proof are different kinds of support.

Case file 6. The parent report

Parent message:

She no longer asks me to check every answer, so tuition is definitely making her independent.

Useful observation:

requests for parent checking declined.

Possible interpretations:

• greater independence;
• greater confidence;
• easier homework;
• support moved to school or tutor;
• the learner stopped checking altogether.

Good tutor response:

The reduction in parent checking supports the possibility of greater independence. We should corroborate it with whether she now selects methods, checks errors and completes unfamiliar tasks successfully without replacement help.

The parent’s observation matters.

It does not have to carry the entire construct alone.

Case file 7. The tuition marketing sentence

Constructed claim:

Our students’ score gains prove that small-group tuition works.

Evidence questions:

• Which students are included?
• Were starting scores known?
• How were gains measured?
• Were papers comparable?
• Was there attrition?
• Were results independently verified?
• What else changed?
• What does “works” mean—score, confidence, independence, retention, examination outcome?

A responsible sentence could be:

Among the students included in this progress sample, scores increased over the measured period. These results provide evidence of improvement in the reported outcomes, but they do not by themselves isolate the causal effect of tuition.

This is less glamorous.

It is also more trustworthy.

Case file 8. Three sources that are really one source

A student finds three websites repeating the same statistic.

They write:

Three independent sources confirm the figure.

On inspection:

• Site A cites Report X;
• Site B cites Site A;
• Site C paraphrases Report X.

There is one underlying evidence source.

Better:

The figure is repeated across three pages, but all trace back to the same original report. The repetition does not provide three independent confirmations.

Corroboration is about independence and added information, not search-result count.

Case file 9. The negative result

A small study does not find a difference between two teaching methods.

Student sentence:

The study disproves any difference between the methods.

Too strong.

Failure to detect a difference can arise because:

• the true difference is near zero;
• the sample is too small;
• measurement is noisy;
• implementation is inconsistent;
• the tested outcome misses the real difference.

Better:

The study did not detect a difference on the measured outcome. This result does not establish that the methods are equivalent unless the design had adequate ability to detect differences of practical interest.

Not detected is not the same as disproved.

Case file 10. The one dramatic example

A learner who previously scored 42 later scores 78 after changing study method.

The story is compelling.

It illustrates that large improvement can occur after the change.

It may suggest a useful hypothesis.

It does not prove that the method will produce the same gain in other learners.

Nor does it prove the method caused the entire gain in this learner.

Anecdotes can be excellent for:

• showing what is possible;
• illustrating a mechanism;
• generating a hypothesis;
• making an abstract process concrete.

They are weaker for estimating average effects or generalising broadly.

The evidence-fit matrix

Use this mental matrix rather than a fixed synonym ladder:

• Directly displays a measured fact: shows, documents, verifies, establishes—depending on the fact and authority.
• Points toward an interpretation: suggests, indicates, is consistent with, points to.
• Gives positive argumentative weight: supports, provides evidence for, corroborates.
• Exhibits a capability: demonstrates, shows, provides evidence of.
• Closes a bounded uncertainty: confirms, verifies, establishes.
• Eliminates a competing claim: rules out, refutes, disproves—only when the method truly does so.
• Deductively establishes truth within a formal system: proves.

The full Evidence-Strength Audit

1. Evidence type: observation, measurement, testimony, document, experiment, example, calculation, proof?
2. Direct fact: what can be stated without interpretation?
3. Claim: what proposition is the writer trying to establish?
4. Warrant: why should the evidence support that proposition?
5. Alternative: what other explanation still fits?
6. Independence: are multiple sources truly independent?
7. Design: what alternatives did the method rule out?
8. Scope: one item, one learner, one sample, one population, all cases?
9. Domain: mathematical, empirical, historical, legal, administrative, interpretive?
10. Object: what exactly follows the verb?
11. Verb fit: does the chosen verb match the relationship?
12. Verb inflation: was a moderate relation upgraded for drama?
13. Verb deflation: was direct evidence weakened unnecessarily?
14. Conditions: did “under these conditions” disappear?
15. Closure standard: what would justify confirm, establish, rule out or prove?

Independent practice: thirty evidence-strength decisions

For each item, identify:

1. the direct evidence;
2. the claim;
3. the missing warrant or alternative;
4. the most defensible evidence verb or phrase.

Practice set A: suggests, indicates and supports

1. Maya answers correctly after one crucial word is explained. Is suggests a vocabulary bottleneck more defensible than proves vocabulary caused the error? Why?

2. A graph shows weekly averages of 61, 64, 68 and 71. Write one claim the graph shows directly and one claim it only supports indirectly.

3. Students who report more reading also have larger vocabulary scores. Choose between indicates an association, demonstrates causation and proves reading works.

4. A student’s errors occur mainly under time pressure. What additional evidence would strengthen the sentence The pattern suggests time pressure contributes to the errors?

5. One source says a teaching method helped. Two independent replications find similar results. Which verb becomes more defensible: illustrates, corroborates or refutes?

6. Why can supports be appropriate even when the evidence is strong?

Practice set B: shows and demonstrates

7. A table reports Group A = 72 and Group B = 64. Complete: The table shows ______. Keep the object narrow.

8. A learner identifies a hidden assumption and repairs the sentence independently. Is demonstrates the target skill on this item defensible?

9. One learner uses an advanced word correctly once. Why is demonstrates mastery of advanced vocabulary too broad?

10. A photograph directly displays that a notice was posted on a door. Which is better: shows the notice was present or suggests a notice may have been present?

11. A survey finds that 68% of respondents prefer Method A. What does the survey show directly, and what does it not show about the whole population?

12. A teacher says, This example demonstrates that all students benefit from the method. What scope problem appears?

Practice set C: confirms, establishes and verifies

13. An official register lists a student’s enrolment status. When is confirms enrolment appropriate?

14. A correct answer matches the key. What exactly has been verified?

15. A randomised experiment finds lower error rate in the treatment group. Write a bounded sentence using establishes.

16. A hypothesis predicts improvement; one measured outcome improves while two do not. Why is confirms the hypothesis unsafe?

17. Three websites repeat one statistic but all trace to the same report. Does the repetition confirm the number independently?

18. A calculation is checked by reversing the operation and reproducing the starting value. Which verb fits well?

Practice set D: proves, refutes and rules out

19. Five examples show that odd + odd is even. Do the examples prove the universal statement?

20. A valid algebraic derivation covers arbitrary odd integers. Which verb is appropriate?

21. A small study fails to detect a difference. Does that disprove any difference?

22. One counterexample is found to the claim All learners improve under Method A. What can the counterexample refute?

23. A sensitive test finds no trace of a specific contaminant above the detection threshold. What must be stated before saying the result rules out contamination?

24. A diary proves its author wrote a particular sentence. Does it prove the event described actually happened exactly as written?

Practice set E: AI evidence drift

25. Source: The findings support a possible benefit under the tested conditions. AI: The findings prove the method works. List three changes.

26. Source: The register shows 28 of 30 students were present. AI: The register suggests most students may have attended. What useful certainty was lost?

27. Source: The pattern is consistent with fatigue contributing. AI: The pattern confirms fatigue caused the errors. What inferential upgrades occurred?

28. Source: The final answer was verified. AI: The learner demonstrated full understanding. Which object changed?

29. Source: The experiment establishes lower error rate in this sample under these conditions. AI: The method is proven superior for students. What scope and evaluation changes occurred?

30. Write a one-sentence instruction to an AI editor that protects evidential force during rewriting.

Answer commentary

1. Yes. The repair provides evidence that the misread word mattered in this item, but the intervention may also have redirected attention or clarified the task. Supports the interpretation that vocabulary was a bottleneck is also defensible.

2. Direct: The graph shows the weekly average increased from 61 to 71 across the displayed period. Indirect: it may support a hypothesis about improvement over time, but it does not by itself identify the cause.

3. Indicates an association is the safest of the three unless the design provides causal identification. The observed relationship alone does not establish intervention effect.

4. Compare performance under similar tasks with and without time pressure, track where errors appear in the paper, check fatigue and difficulty, and look for repeated within-learner patterns. Stronger discrimination makes the causal interpretation more defensible.

5. Corroborates becomes more defensible if the replications genuinely provide independent evidence. They add support rather than merely illustrating the original finding.

6. Because supports describes a positive evidential relationship, not a low confidence score. A large body of convergent evidence can strongly support a claim without the writer adopting absolute closure language.

7. Example: The table shows an 8-point difference between the reported group averages. It does not automatically show why the difference exists or which method is better.

8. Yes, if the target skill is precisely the behaviour performed and the sentence keeps the scope to this item. One item still does not demonstrate durable mastery across contexts.

9. One use demonstrates one successful use under one context. Mastery requires stable appropriate use across multiple words, contexts and time.

10. Shows the notice was present is better if the photograph directly and reliably records the relevant door and time. Unnecessary hedging can erase direct evidence.

11. Directly, 68% of the sampled respondents selected Method A under the survey procedure. Generalising to the whole population requires sampling and representativeness assumptions.

12. One example is being expanded into a universal claim about all students. The evidence may illustrate a possibility; it cannot establish population-wide benefit alone.

13. When the register is authoritative for that status, correctly identifies the student and is current enough for the claim. The object should remain the administrative fact: enrolment.

14. The final answer’s agreement with the answer key. The reasoning, independence, transfer and understanding remain separate questions.

15. Example: The randomised comparison establishes that assignment to the treatment condition reduced the measured error rate relative to control in this sample under the tested conditions.

16. The evidence is mixed. One predicted outcome occurred while two did not. A selective confirms statement hides disconfirming or non-supportive evidence.

17. No. The pages are not three independent evidence sources if they all inherit the same original report. Repetition can create visibility without additional corroboration.

18. Verifies fits well: the reverse operation independently checks consistency with the starting value.

19. No. The examples illustrate or support the pattern, but finitely many examples cannot prove the universal statement.

20. Proves is appropriate if the derivation is valid and covers arbitrary odd integers under the formal definitions.

21. No. A non-detection can reflect low power, noisy measurement or an effect below the detection capability. Equivalence requires an appropriate design and criterion.

22. It refutes the universal form all learners improve. It does not refute weaker statements such as many learners improve or average performance improves.

23. State the detection threshold, sampling conditions and what kinds or amounts of contamination the method could detect. A result can rule out contamination only within the sensitivity and scope of the test.

24. No. It establishes or documents what the author recorded. The truth of the reported event requires other evidential considerations.

25. Support became proof; tested conditions disappeared; a possible benefit became the global statement that the method works. Population and outcome may also have broadened.

26. The direct administrative evidence of 28 recorded attendees. If the register is authoritative, the rewrite introduces unnecessary uncertainty.

27. Compatibility became confirmation, contribution became sole/direct causation, and uncertainty disappeared.

28. The source object is answer correctness. The AI substitutes the much broader construct of understanding.

29. The sample becomes students generally; lower error rate becomes overall superiority; conditions disappear; establishes becomes proven; one outcome becomes a global method judgement.

30. Example: Preserve every evidence verb’s inferential force, object, scope, conditions and alternatives; flag rather than silently strengthen or weaken any evidence-to-claim relationship.

The evidence-language independence test

• separates direct observation from interpretation;
• recognises that evidence strength is claim-relative;
• uses suggests without treating it as automatically weak;
• uses indicates for signals and measured patterns without smuggling in causes;
• uses supports for positive weight without false closure;
• uses shows for directly displayed facts or patterns;
• uses demonstrates when performance genuinely exhibits the target capability;
• uses confirms only when the relevant uncertainty is actually closed;
• uses establishes with explicit scope and standard;
• uses verifies for checks against a criterion or independent procedure;
• tests whether corroborating sources are independent;
• understands that consistent with does not mean uniquely explained by;
• distinguishes example from proof;
• distinguishes non-detection from refutation;
• recognises domain-specific proof standards;
• distinguishes mathematical proof from empirical support;
• recognises that a historical source can establish a record without establishing motive;
• audits the object after the evidence verb;
• spots verb inflation in AI rewriting;
• spots unnecessary verb deflation;
• preserves conditions, population and outcome during summarisation;
• reduces claim scope when the evidence is narrow rather than merely hedging the verb;
• asks what evidence would rule out the strongest surviving alternative.

A parent route: ask what the evidence earns before turning one result into a story about the child

Family conclusions often move too quickly from one observation to one explanation.

She got 82, so the new routine clearly works.

He finished without asking me, so he has finally become independent.

She spoke up today, which proves her confidence is back.

The better family sequence is:

1. State what happened. Keep the first sentence descriptive.
2. Name the narrow claim it supports. Do not jump straight to personality, mastery or cause.
3. Ask for one alternative explanation.
4. Ask what next observation would discriminate between them.
5. Upgrade the verb only after the evidence earns it.

Family example: one strong paper

Observation:

Maya scored 82 on this paper, up from 70 on the previous paper.

Narrow interpretation:

The result shows substantially stronger performance on this paper and supports the possibility that recent changes are helping.

Alternative:

the paper may have been easier, or the stronger score may reflect topic mix, sleep, familiarity or ordinary variation.

Next evidence:

repeat comparable performance across several papers and inspect the specific error types that changed.

The family does not need to pretend uncertainty means ignorance.

One result can be encouraging evidence without becoming final proof.

A tutor route: separate episode evidence, pattern evidence, capability evidence and causal evidence

Tutors often have excellent local evidence but use one verb for all of it.

A cleaner system has four levels.

Level 1. Episode evidence

The learner solved this unfamiliar item without prompting.

This can show or demonstrate successful performance on the item.

It does not establish stable mastery.

Level 2. Pattern evidence

The learner solved eight of ten comparable unfamiliar items across three lessons without prompting.

This supports a stronger claim about stable capability in that task family.

Level 3. Transfer evidence

The learner used the skill accurately in school homework, a new passage and a delayed test without the original scaffold.

This can demonstrate broader transfer and corroborate the capability claim through different contexts.

Level 4. Causal evidence

Performance improved because of the tutoring intervention.

This is a different claim.

It requires evidence about what would likely have happened otherwise and what competing changes occurred.

A tutor may have strong evidence that a learner improved and weaker evidence about how much of the improvement was caused by tuition.

That distinction increases credibility rather than reducing it.

The smallest sufficient evidence verb

When several verbs could fit, choose the smallest one that still preserves the useful information.

Do not write suggests when an authoritative record directly shows a fact.

Do not write proves when an observation merely supports one interpretation among several.

Do not write confirms mastery when the evidence verifies one correct answer.

Do not write may indicate when a valid mathematical proof has already established the theorem.

The goal is neither maximum confidence nor maximum caution.

The goal is minimum distortion.

One final evidence conversation in Punggol

The four students are looking at a progress report.

The latest results prove that the new checking routine has transformed Maya’s examination performance.

Jia Jun grins.

“Strong sentence.”

Hana asks:

“What do the results directly show?”

Ethan reads the record.

• three comparable timed papers;
• checking errors fell from 7, 6 and 8 before the routine to 3, 2 and 2 after it;
• overall scores rose;
• the learner used the routine on every later paper;
• sleep and revision volume also improved during the same period.

Maya says:

“So the records show fewer checking errors.”

Hana adds:

“And the repeated pattern supports the idea that the routine helped.”

Ethan says:

“But because other things changed too, we should not assign the whole score gain to the routine.”

Jia Jun deletes proves and transformed.

They write:

Across three comparable timed papers, Maya’s checking errors fell after the new routine was introduced, and she used the routine consistently on each later paper. The repeated pattern supports the conclusion that the routine contributed to better checking performance, although other study and sleep changes during the same period mean the routine’s individual contribution to the overall score gain cannot be isolated from these records alone.

Longer.

Less theatrical.

More informative.

The real goal: calibrated closure

Evidence language should not be permanently timid.

When a record directly establishes a fact, say so.

When a performance genuinely demonstrates the target skill, say so.

When a valid proof proves the theorem, say so.

When an experiment establishes a causal effect under its tested conditions, do not retreat into vague language merely to sound scholarly.

But when alternatives remain, keep them alive in the language.

When one example only illustrates possibility, do not turn it into population proof.

When several webpages trace back to one source, do not count repetition as independent confirmation.

When a negative result fails to detect a difference, do not automatically call the difference disproved.

The strongest writer is not the writer who always uses the strongest verb.

The strongest writer knows when closure has been earned.

Strong evidence deserves strong wording. Incomplete evidence deserves bounded wording. Neither inflation nor unnecessary hedging is precision.

Research notes and further reading

The sources below support the evidence-verb meanings and argument structure. The Evidence-Strength Audit, resident-character case files, family/tutor routes and practice sequence are original instructional material.

• Cambridge Dictionary: suggest — evidence can point toward an idea without supplying proof.
• Cambridge Dictionary: indicate — showing or making something clear.
• Cambridge Dictionary: support — helping to show that something is true.
• Cambridge Dictionary: show — making something clear or visible.
• Cambridge Dictionary: demonstrate — showing clearly that something is true or that a capability can be performed.
• Cambridge Dictionary: confirm — showing or stating that something previously uncertain is true.
• Cambridge Dictionary: prove — showing something to be true.
• Cambridge Dictionary: proof — evidence or argument used to establish truth.
• Purdue OWL: Research and Evidence — choosing, presenting and connecting evidence to an argument.
• Purdue OWL: Organizing Your Argument — claim, grounds, warrant, qualifier and rebuttal structure.

Continue through the advanced eduKatePunggol vocabulary chain

• Advanced Vocabulary Measurement | The Operationalisation Audit
• Advanced Vocabulary Comparison | The Baseline Audit
• Advanced Vocabulary Causality | The Cause–Mechanism Audit
• Advanced Vocabulary Argument | The Qualification Audit
• Advanced Vocabulary Reading | The Assumption Audit
• Advanced Vocabulary Editing | The Counterexample Test
• Education: Regarding Vocabulary