No. An FYP or a thesis supports a hypothesis; it does not prove one. A statistical test tells you that your data would be unlikely if the null hypothesis were true, on one sample, measured one way, at one point in time. The verbs that fit are supports, indicates and suggests. The verb proves promises something your design cannot deliver.
Why Does the Word Prove Cause Trouble?
Because it claims that three limits do not apply to you, and all three always do.
- The sampling limit. You measured part of a population and drew a conclusion about the whole. The probability that you are wrong is exactly what your alpha level quantifies.
- The design limit. A survey taken at one moment cannot say which variable moved first, so it cannot establish cause.
- The measurement limit. Your instrument measures indicators of a construct, not the construct itself. Job satisfaction is not measured directly; agreement with statements about job satisfaction is.
Every empirical thesis carries all three. Writing this study proves asserts that yours does not, and a panel that reads it has an obvious first question ready before you sit down.
What Ladder Should You Choose From?
Pick the highest rung your evidence supports, then stop. Moving one rung above your evidence is the single most common language fault in a Malaysian FYP.
| Rung | Wording | Use when |
|---|---|---|
| 1. Observation | was recorded, was observed, appeared in the data | Reporting numbers with no interpretation |
| 2. Possibility | may, might, could, it is possible that | A pattern is visible but untested |
| 3. Tendency | tends to, appears to, is consistent with | Direction is stable but the effect is small |
| 4. Support | suggests, indicates, supports, is associated with | A test was significant at the level you declared |
| 5. Strong support | demonstrates, consistently shows | Significant, adequate effect size, consistent with prior work |
| 6. Proof | proves, confirms, establishes | Effectively never in an undergraduate or master’s study |
Rung 4 is where most correctly written FYP findings belong. Rung 5 is available, but you should earn it in the same sentence by naming the effect size or the replication that justifies it.
Which Constructions Overclaim Without Sounding Like It?
Some overclaiming is obvious. The dangerous kind is grammatically modest and still says too much.
- The results show that X influences Y. Influences is a causal verb wearing neutral clothes. From a survey, write was associated with, or predicted Y in the regression model.
- There is no relationship between X and Y. Your test failed to find one, which is a different statement. Write the analysis did not find a significant relationship, and give the p value.
- Employees are more motivated when… Present tense with a bare plural quietly promotes your sample to all employees everywhere. Name the sample.
- This confirms the theory. One study on one sample is consistent with a theory; it does not confirm one.
- The findings are very significant. Significance has no degrees. Report the exact p value, and use an effect size if you want to talk about magnitude.
For the mechanics of laying out the numbers themselves — which tables come in what order, and how to word a non-significant test in Chapter 4 — see the guide to writing the results chapter. This article is about the strength of the claim you build on top of those numbers, which is a Chapter 5 and abstract problem.
How Far Can You Generalise a Malaysian Sample?
As far as your sampling frame reaches, and not one step further. The test is mechanical: copy the sentence from Chapter 3 that defines your population, and attach it to every conclusion in Chapter 5. If a conclusion becomes odd once the limit is attached, the claim was always too wide.
Too wide: Malaysian undergraduates prefer blended learning to fully online delivery.
Correct: Among 214 undergraduates at one public university faculty, blended delivery was rated higher than fully online delivery on all four scale items.
The corrected version is longer by a clause and stronger by a wide margin, because it can be checked. How many respondents you needed in the first place, and what that number licenses, is covered in how many respondents do I need.
Can Hedging Go Too Far in the Other Direction?
Yes, and the result is worse than mild overclaiming because it says nothing at all. Two patterns to watch:
Stacked hedges. It may possibly appear that there could be a potential tendency for X to be somewhat associated with Y. Six hedges, zero content. The rule is one hedge per claim; if one is not enough, move down a rung rather than adding words.
The empty closing sentence. This is an interesting area for further research. It commits to nothing and could close any thesis ever written. Replace it with a claim that could be wrong: A longitudinal design would establish whether the association survives beyond one semester, which this study cannot determine.
How Do Examiners Read an Unhedged Finding?
As an opening. An overclaimed sentence lets a panel challenge you without reading your data — they need only point at your design. Dropping one rung before submission closes that route without costing you anything, because your actual finding is unchanged.
There is a second effect worth knowing. A candidate whose limitations chapter states the boundaries plainly usually finds the viva shifts from testing basic understanding to discussing the findings. That is a much better conversation to be in, and the preparation for it is set out in how to prepare for a viva voce.
What Should the Abstract Say?
The abstract has no room for qualifying clauses, so the verb has to carry all of the caution by itself. One finding sentence, one rung, numbers included:
Results indicate that perceived workload was negatively associated with intention to stay (β = −0.34, p = .002), while supervisor support showed no significant association.
Three things are working there. The verb sits on rung 4. The numbers let the reader judge magnitude without trusting an adjective. And the null result is worded as no significant association rather than no association. Length limits and the rest of the abstract formula are in how to write an abstract.
Which Two Sentences Belong in Your Limitations?
Limitations are not a confession. They are where you set the boundary of your own claim before someone else sets it for you. Two sentences cover most FYP designs:
- The population boundary. Because respondents were drawn from a single faculty at a single institution, the findings should not be generalised to other disciplines without replication.
- The design boundary. Because data were collected at a single point in time, the direction of the observed association cannot be determined.
Write both, then check that no sentence anywhere else in the thesis contradicts them. A conclusion that claims causation while the limitations deny it is the sharpest contradiction a panel can find.
How Do You Clear Overclaiming from a Draft?
Four searches, roughly fifteen minutes for a full thesis:
- Search prove, proves, proved, proof. Nearly all need to come down.
- Search cause, causes, effect of, impact of, influence. Check the design behind each one.
- Search very significant, highly significant, slightly significant. None of these exists.
- Search all, always, every, none. Check each against your sampling frame.
Tesify can flag these constructions across a full draft and offer wording at the rung you specify, which turns a chapter-by-chapter reread into a review pass. What no tool can decide is how strong a claim your evidence justifies — that judgement is yours, and you remain the author answering for it in the viva. Whether a similar wording pass counts as acceptable assistance is a separate question, addressed in can Turnitin detect ChatGPT.
Does the Same Rule Apply to Your Recommendations?
No, and this is the one place students under-claim without needing to. A recommendation proposes an action; it does not describe reality, so it does not carry the same evidential burden. Writing the department may possibly wish to consider reviewing the workload allocation is not caution, it is vagueness — nobody can act on it.
Recommendations should be concrete, addressed to someone specific, and traceable to a finding: Faculties reporting the highest workload scores should review teaching allocation before the next intake, since workload was the only variable associated with intention to leave in this sample. The hedge sits in the final clause, attached to the evidence, while the instruction itself stays direct.
Run one check across your recommendations chapter: every recommendation should name who acts, what they do, and which finding it rests on. Any recommendation missing the third element is either unsupported or belongs in the further-research section instead.
Frequently Asked Questions
Is the word prove ever acceptable in a thesis?
It is acceptable in mathematics and formal logic, where proof has a technical meaning. In empirical work it is safe only when the object is your own hypothesis rather than reality, as in the second hypothesis was not supported. Even there, supported reads better than proved.
What is the safest verb to use in an abstract?
Suggests or indicates for a modest result, found for a clear one. All three signal support without promising certainty, and all three are accepted across engineering, business and the social sciences.
Does hedging make my thesis sound weak?
Not when it is paired with specificity. A hedged claim that names the sample, the measure and the effect size sounds more confident than an unhedged one that names none of them, because the reader can check it.
How many hedges should one sentence carry?
One. Stacking hedges, as in it may possibly appear to suggest a potential tendency, empties the sentence. If one hedge feels insufficient, move down a rung on the ladder instead of adding another word.
Can I say a variable influences another in a survey study?
Only if you attach the model. Influence and effect imply direction in time, which a single measurement cannot establish. Write that the variable was associated with the outcome, or that it predicted the outcome in the regression model.
How far can I generalise from a Malaysian sample?
As far as your sampling frame reaches and no further. If your respondents were students at one faculty at one university, that is your population. Widening it to Malaysian students in the conclusion changes the population after the data were collected.
What does a significant result actually mean?
It means the data would be unlikely if the null hypothesis were true, at the alpha level you declared. It is not the probability that your hypothesis is correct, and it says nothing about how large the effect is.
Is very significant or highly significant acceptable?
No. Significance is a threshold decision and has no degrees. Report the exact p value and an effect size if you want to convey magnitude.
My supervisor asked me to write more assertively. What should I change?
Shorten the sentence and add numbers rather than raising the verb. A claim that carries a coefficient, a p value and a sample size reads as far more assertive than an adjective ever will.
Where in the thesis is strong language legitimate?
In the recommendations, where you are proposing actions rather than describing reality. A recommendation is not an empirical claim, so it does not require the same evidential backing.
Do examiners really question single words?
They question the claim, and the verb is the most visible form the claim takes. An overclaimed abstract is the cheapest possible opening for a panel, because they can challenge it without examining your data.
Does this apply to a qualitative study?
Yes, with different vocabulary. Qualitative work describes, explores and identifies patterns within its cases. It does not prove, and it does not generalise to a population, so claims of prevalence are the main risk to watch.