Every time you cite someone you choose a verb, and that verb commits you to a position on their work whether you meant it or not. Write “Mkhize (2021) claims that…” and you have told your examiner you doubt Mkhize. Write “demonstrates” and you have accepted the finding as proved, and made it yours to defend. In a literature review carrying 75 sources — the median for a South African master’s dissertation — that is 75 small decisions nobody ever taught you to make.
The comparison table comes first, because it is the part worth keeping open in another tab while you write. The ranked families, the discipline conventions and the twelve-verb recommendation follow it.
The verb bank, compared
| Verb | What it commits you to | Strength | Where it belongs | Where it costs you |
|---|---|---|---|---|
| states, reports, describes | The source said this. You take no position. | Neutral | Background, literature review | As a default it flattens a review into a list |
| notes, observes | The source said this and it is worth noticing | Neutral, faintly positive | Literature review | Mild endorsement you may not intend |
| presents, outlines, sets out | The source offers a framework or model | Neutral | Theoretical framework | Wrong for empirical findings |
| identifies, measures, estimates | The source did a specific technical thing | Neutral | Methodology, literature review | None. Badly underused |
| found, reported | The source produced evidence | Medium | Literature review, discussion | Wrong for a purely conceptual paper |
| suggests, indicates | The evidence points this way without closing the question | Medium-low | Discussion, literature review | None. The safest general-purpose verb |
| implies, points to | The inference is yours, not the author’s | Low | Discussion | Do not attribute your inference to the author |
| shows, showed | The evidence is good and you accept it | Medium-high | Discussion | You now own the finding as well |
| demonstrates | The evidence is strong and the point is made | High | Discussion, sparingly | An examiner may ask whether the study earns the word |
| establishes | The question is settled | High | Only for a large replicated literature | Almost never true of a single study |
| confirms, corroborates | The source agrees with something already known | Medium | Discussion | Requires a prior finding you have actually cited |
| proposes, posits | A theoretical move, no evidence claimed | Medium | Theoretical framework | Wrong for empirical results |
| argues, contends, maintains | The source takes a position and defends it | Medium-high | Literature review, discussion | Only for a source that actually argues |
| holds, takes the view that | A position, reported without heat | Medium | Humanities, policy, theory | None |
| challenges, questions, disputes | The source disagrees with other work | Medium | Literature review | Name who is being disagreed with |
| refutes, contradicts | The source proved something wrong | High | Discussion, rarely | Refute means disproved, not disagreed with |
| assumes | The source did not justify this | Negative, mild | Critique | Legitimate only where it is true |
| asserts | Stated without support | Negative, mild | Critique | Reads as criticism even when unintended |
| claims | You doubt it | Negative, strong | Critique only | The commonest accidental insult in a dissertation |
| alleges, purports | You distrust it | Negative, very strong | Contested or retracted work | Reserve for genuine disputes of fact |
The five families, ranked by how often you should reach for them
- Tentative-evidential — suggests, indicates, found. Suits almost every sentence in a literature review and most of the discussion. Falls short when a finding really is decisive and you make it sound provisional; a genuinely strong result deserves a stronger verb once.
- Neutral report — states, reports, notes, describes, outlines. Suits background, definitions and methodological precedents, where your job is to record rather than judge. Falls short as the default for a whole chapter, because a review built entirely from neutral verbs takes no position and reads as an annotated bibliography.
- Argumentative — argues, contends, maintains, holds. Suits policy, theory and any source making a case rather than reporting data. Falls short when applied to an empirical paper: a randomised trial does not argue, it finds.
- Strong evidential — demonstrates, shows, establishes, confirms. Suits the three or four findings your own argument leans on hardest. Falls short through repetition. Used twenty times it stops meaning anything, and each use is a claim an examiner may test.
- Distancing — claims, asserts, assumes, alleges. Suits deliberate critique, which every good literature review contains some of. Falls short catastrophically when used by accident, because the reader assumes you meant it.
Why “claims” causes more trouble than any other verb
In everyday South African English — and in journalism everywhere — claims is close to neutral. “The department claims the backlog has been cleared” is ordinary reporting. In academic English it is not neutral at all. It signals that the writer regards the statement as unsupported.
So a literature review that says “Naidoo (2019) claims that participative management improves retention” has, to an examiner, taken a position against Naidoo without a single sentence of justification for that position. If you meant it, the sentence is fine and you owe the reader your grounds in the next clause. If you did not, you have picked a fight you did not know you were in — possibly with somebody who supervises down the corridor, or who is about to be appointed as your external examiner.
The repair is one search. Find every claims, asserts and alleges in your document and ask, one at a time: do I actually doubt this? Keep the ones where the answer is yes and put the grounds beside them. Replace the rest with argues, states, reports or suggests, according to what the source was doing.
Two rules that are not preferences
The first is about proves. A single study does not prove a hypothesis; it produces evidence consistent with it. That is not a stylistic nicety but a statement about what inferential statistics can support, and it is why demonstrates and establishes should be rare rather than routine in a master’s dissertation. Correlational findings in particular describe associations, not effects.
The second is about who is doing the acting. A study cannot argue, believe or conclude — researchers can. “The study argues” is weaker than “Naidoo argues”, and the habit of attributing agency to documents produces sentences in which nobody is responsible for anything. Name the person.
Neither rule comes from your referencing style, which is a separate question and a confusing one in South Africa, because Harvard is not a single style while APA is — the practical consequences are set out in our comparison of Harvard and APA referencing at South African universities. Your faculty guide governs the format of the citation. It almost certainly says nothing at all about the verb in front of it.
Does the citation format change the verb?
It changes how visible the author is, which changes how often you need a verb at all.
Author-date formats — Harvard and APA, which between them cover most South African faculties — make the author a grammatical subject easily: Naidoo (2019) found that…. This is the integral or author-prominent citation, and it is the form that requires a reporting verb.
Numeric formats — Vancouver in the health sciences, IEEE in engineering — reduce the citation to a bracketed number, so the natural sentence becomes information-prominent: Adaptive routing reduces latency under load [12]. There is no reporting verb at all, and the stance migrates into the main verb of your own sentence. That is why engineering and health dissertations can quietly state another team’s finding as though it were settled fact. When a source deserves attribution, name the author in the sentence: Rao et al. propose an adaptive scheme [12].
The working rule: use an integral citation with a reporting verb when the who matters — a landmark study, a position you will argue with, a method you are borrowing. Use a non-integral citation when the what matters and three sources say the same thing.
Which verbs does your discipline expect?
- Engineering and computer science: propose, present, introduce, implement, evaluate, report, outperform. Argument verbs are rare; the literature is a sequence of contributions rather than a debate.
- Management, commerce, public administration: find, examine, investigate, argue, posit, model, test. A hybrid literature, so let the individual source decide the family.
- Education, sociology, psychology: argue, contend, suggest, report, find, theorise, problematise. Argument verbs are normal because much of the literature is arguing.
- Nursing and health sciences: report, observe, find, show, associate — with a strong preference for association language over causal language. Note that demonstrate is used more freely here than elsewhere.
- Humanities and law: argue, hold, read, interpret, contend, trace. The present tense is normal for a text’s argument, and your own position is expected to be visible.
The most reliable calibration is not this list. Take three recent articles from the journal you would most like to publish in, highlight every reporting verb on the first two pages, and count the families. That is your field’s distribution, measured rather than assumed — and it takes twenty minutes.
Six words to strike from a South African dissertation
- opined. Legal and journalistic register; it almost never appears in international academic writing. Use argued or held.
- said, told. These report speech, and a journal article is not speech. Use wrote, reported or argued. The exception is genuine interview data, where said is exactly right.
- talks about, deals with, looks at. Vague, informal, and carrying no stance whatsoever. Name what the source does: tests, models, reviews, compares.
- sheds light on, throws light on. A metaphor doing the work a verb should do. Say what the source showed.
- highlighted. Not wrong, but so overused in literature reviews that it has become invisible, and it tells the reader nothing about whether evidence was involved.
- alludes to. Almost always used to mean states. To allude is to refer indirectly without naming, which is rarely what a cited author was doing.
The recommendation: twelve verbs are enough
You do not need the whole table. A dissertation that uses these twelve accurately reads better than one reaching for variety:
reports · notes · describes · presents · found · suggests · indicates · argues · contends · shows · demonstrates · challenges
Add claims, asserts and assumes only where you are deliberately critical, and establishes and confirms only where a replicated literature earns them. Repetition of an accurate verb is not a stylistic fault. Substituting an inaccurate synonym to avoid repetition is — and it is the single most common way a wrong verb enters a chapter.
How to audit one chapter in twenty minutes
- Highlight every reporting verb in one chapter. One chapter, not the whole dissertation, or you will not finish.
- Count the families. If more than about two-thirds are neutral-report verbs, your review is describing rather than synthesising, and the structural repair is in our guide to writing a literature review for a South African dissertation.
- Check every distancing verb. Every claims, asserts, alleges, assumes: intended, or accidental?
- Check every strong evidential verb. For each demonstrates, proves, establishes, ask whether you could defend that word if an examiner stopped on it. If not, drop a rung to shows or suggests.
- Check the verbs attached to your own findings. The discussion chapter is where an overclaiming verb costs most, and the paragraph pattern that keeps interpretation bounded is set out in our guide to writing the discussion chapter.
Expect roughly forty to eighty changes across a full dissertation, almost all of them one word. That is normal, and it says nothing whatsoever about the quality of your research.
Getting the verb right the first time
Auditing verbs after the fact works, but it is a third pass over a chapter you have already written twice. The cheaper version is to have the stance decided at the moment you write the sentence — which means having the source, its evidence and its strength in front of you, rather than in your memory of a PDF you read in March.
Tesify keeps every claim attached to the source it came from while the chapter grows, so the sentence you write about a study sits beside what that study actually found. The judgement about how strongly to put it stays yours, which is the part an examiner is testing. There is a free tier, so you can try it on one section before deciding about anything paid.
Build your literature review with Tesify
Frequently asked questions
What is a reporting verb?
The verb you attach to a cited source to say what that source did — found, argued, suggested, demonstrated. It is the smallest unit of stance in academic writing, because it tells the reader both what the source did and what you think of it, in a single word.
Is it wrong to use “claims” in a literature review?
Not wrong, but rarely what the writer intends. In academic English it signals that you regard the statement as unsupported. Use it deliberately, with your grounds in the next clause, or replace it with argues, states or suggests.
Which reporting verb is safest if I am unsure?
Suggests, for a finding that points in a direction, and reports, for a fact the source recorded. Both are accurate about a wide range of sources, and neither commits you to a position you have not defended.
Should reporting verbs be in the past or the present tense?
Conventions differ by discipline and by faculty guide, and your guide decides. What matters far more is consistency within a section: a paragraph that moves between “Naidoo finds” and “Naidoo found” for no reason reads as unedited.
Can I say a study “proves” something?
Avoid it outside mathematics, where a proof is a technical object and the verb is literal. A single study produces evidence consistent with a hypothesis rather than proof of it, and an examiner who sees “proves” in a master’s dissertation will stop on the word.
How many different reporting verbs should a dissertation use?
Around a dozen, used accurately. Reaching for synonyms to avoid repetition is how an inaccurate verb enters a chapter, and an examiner notices the inaccuracy long before they notice the repetition.
Does my Harvard faculty guide tell me which verbs to use?
Almost certainly not. Faculty guides govern the format of citations and reference lists, not the verbs in front of them. That gap is exactly why reporting verbs go unexamined in so many dissertations.
Do reporting verbs matter in a Vancouver or IEEE dissertation?
They matter differently. Numeric citation makes it easy to state a source’s finding as bare fact, which quietly converts other people’s results into settled knowledge. Where attribution matters, name the author in the sentence rather than relying on the bracket alone.
Is “highlighted” acceptable?
It is grammatical and it is not an error. It is simply worn out, and it carries no information about whether evidence was involved. A more specific verb does more work in the same space.
Will my examiner really notice the verb?
Examiners reliably notice distancing verbs and overclaiming verbs, because each creates a question they can ask. A misplaced neutral verb usually passes unremarked. That is why the audit above checks only two families rather than every sentence.
Does this apply to an honours research report as well?
Yes, though the useful set is narrower. An honours report leans on reports, found, suggests, argues and shows, and rarely needs the distancing family at all.
How many sources am I choosing verbs for?
More than you think. We counted the reference lists of 25 South African master’s dissertations and found a median of 75 sources, with half between 42 and 170 — the full count is in our analysis of how many sources a master’s dissertation cites. Every one of those is a verb decision.