# How to write a topic sentence and a strong, cohesive paragraph

**How do I write a strong paragraph with a clear topic sentence and good cohesion?** A strong academic paragraph makes one point, states it in a topic sentence near the start, and keeps every other sentence on that point. Cohesion ties the sentences together: open each sentence with information the reader already has, end on what is new, make every "this" point at something clear, and repeat key terms instead of varying them.

Published 2026-10-02 by EdCitation. https://edcitation.com/newsletter/writing-a-strong-paragraph-topic-sentences-and-cohesion

A strong paragraph makes one point, states it in a topic sentence, and lets each sentence pick up where the last one left off. Writing centres call the first quality unity and the last cohesion. This guide, published by EdCitation, covers both, with topic sentences and paragraph length, and rebuilds one paragraph of ours sentence by sentence.

Every rule comes from sources we read on 2 October 2026. Patterns such as PEEL are in [PEEL and other paragraph patterns](https://edcitation.com/newsletter/peel-and-other-paragraph-patterns), linking words in [signposting](https://edcitation.com/newsletter/signposting-and-linking-words-in-academic-writing), and paragraphs built from several sources in [synthesis, not summary](https://edcitation.com/newsletter/how-to-integrate-sources-synthesis-not-summary).

Every paragraph stays yours, and gladly so: EdCitation never writes or rewrites any part of anyone's work. It lists what your brief demands, free, with [Check your paper](https://edcitation.com/check), and checks the finished file's mechanics with [Mechanics QA](https://edcitation.com/tools/mechanics-qa).

## What is a strong academic paragraph structure?

One point, a sentence that states it, enough development to prove it, and sentences linked to each other. Purdue Online Writing Lab (n.d.-a) names these unity, a topic sentence, adequate development and coherence. Lingard (2019) speaks of two principles, unity and coherence, and of a paragraph that opens on its Topic and closes on a Wrap, a model from Dunleavy's *Authoring a PhD*.

| Quality | A quick test | Source |
| --- | --- | --- |
| Unity | Can you say what the paragraph is about in one sentence? | Purdue |
| Topic sentence | Would it work as a small thesis statement? | Toronto |
| Development | Is it longer than two or three sentences? | Purdue, Toronto |
| Cohesion | Do the first words of each sentence form a related set? | Purdue |

## How do you write a topic sentence?

State the paragraph's one claim, in your own words, as its first sentence. Procter and Visvis (n.d.), for the University of Toronto, call topic sentences "mini thesis statements" and say that in academic writing they nearly always work best at the start; introductions and conclusions generally have none. Purdue calls that position a good rule for less experienced writers, not the only one.

### Three tests for a topic sentence

- **A claim, not a fact.** "Sleep is something that affects every student" (ours) is true and proves nothing. Lingard's "blindfold" paragraph makes the reader wait until the third or fourth sentence for its point; delete the openers until the point stands first.
- **The right size.** Toronto counts a paragraph unified when every sentence relates to its topic sentence, so that sentence must cover them all and no more.
- **A wrap, not a repeat, at the end.** Toronto warns that ending on the topic sentence's point stalls the argument; Lingard's Wrap pulls the paragraph together and often points ahead.

Ours, rewritten: "For a student facing a test, the sleep that counts seems to be the sleep of the previous weeks, not of the night before." It makes a claim, [hedged](https://edcitation.com/newsletter/hedging-and-cautious-language) to fit a single study, that every later sentence can serve.

### Do published writers always start with one?

Less often than textbooks say, and the counts disagree. Braddock's 1974 study of 25 magazine essays found only 13% of paragraphs opening with a topic sentence (Braddock, 1974, as cited in Smith, 2008). Smith (2008) counted blocks of paragraphs on one idea, in 25 articles from two history magazines, and found an explicit topic sentence in 95% of 73 blocks and at the start of 66%. In an essay, lead with it.

## How do you keep a paragraph to one idea?

Test every sentence against the topic sentence, and cut or move any that serves another point. University of York (2026) says material unrelated to a paragraph's main point probably does not belong, and Purdue starts a new paragraph for a new idea, a contrast, or a pause in long or complex material. Morris and Hirst (1991) observe that chains of related words tend to mark out stretches of text with a strong unity of meaning; in our reading, a sentence on none of a paragraph's chains is the first to question.

## What is paragraph cohesion, and how is it different from coherence?

Cohesion is the set of links that let a reader take separate sentences as one text; coherence is whether that text makes sense. The classic account, Halliday and Hasan's *Cohesion in English*, first published in 1976 and [reissued by Routledge](https://doi.org/10.4324/9781315836010), is not open to read, so we take its five kinds of tie from Morris and Hirst: reference, substitution, ellipsis, conjunction and lexical cohesion. Morris and Hirst also show the limit: a passage that begins "Wash and core six apples" and then cuts a suit out of them links every "them" and still makes no sense. Four tools do most of the work in a student paragraph.

### Given and new information: the order inside each sentence

Open each sentence with what the reader already has, and end on what is new. Gopen and Swan (1990) call the start of a sentence the topic position, where readers look for a link back, and the end the stress position, where they expect what deserves emphasis. Misplacing the two was, in their experience, "the No. 1 problem" in American professional writing, though they offer principles, not rules. Purdue Online Writing Lab (n.d.-b), adapting Joseph Williams, gives the test: underline the first few words of every sentence and see whether they form a related set. The worked example below applies it to its second sentence.

### Reference words: this, these, it and the

Make every reference word point at one thing the reader can name. Sydney's unit on reference, built on Halliday and Hasan, lists personal pronouns, demonstratives such as this and those, comparatives such as other, and the definite article (University of Sydney Learning Centre, n.d.-b). Toronto adds that pronouns hold a paragraph's focus only while clearly linked to specific nouns.

The bare "this" is the usual suspect, yet Wulff et al. (2012) found it unattended, with no noun after it, in 43% of 5,827 sentence-initial uses in upper-level student papers, and named "This is because" and "This means that" as valid exceptions to the usual advice. Our rule: add a noun that sums up, such as "these sleep measures", whenever "this" could point at two things.

### Lexical chains: keep the key term

A lexical chain is a run of related words through a passage: repetitions, synonyms, more general words and words that habitually go together (Morris & Hirst, 1991). Toronto recommends deliberate repetition of key words. Sydney's word-choice unit (University of Sydney Learning Centre, n.d.-a) warns that too much repetition bores and can suggest limited understanding. In our view, keep the term for your concept constant, since "rest" and "slumber" in a paragraph about sleep read as new measures, and vary the words around it.

### Transitions: the link to add last

Conjunction is the fourth tool. For a "maze" paragraph, where readers lose the thread, Lingard asks whether the relationships are signalled with words such as "but", "however" or "by contrast". Add them after the other three tools; a paragraph in given-new order often needs few. Which word fits which relationship is in [signposting and linking words](https://edcitation.com/newsletter/signposting-and-linking-words-in-academic-writing).

## How long should a paragraph be in an essay?

Long enough to state, support and wrap up one point, and rarely more than a double-spaced page.

| Source | Too long | Too short |
| --- | --- | --- |
| Toronto | A page or more; usual range one-third to two-thirds of a page | Two or three sentences |
| American Psychological Association (2020) | More than one double-spaced page | A single sentence |
| Purdue | When readers need a pause | Two or three sentences |
| Lingard (2019) | More than a manuscript page | A series of very short paragraphs |

One MIT brief asks for "5 pp or 1250-1500 words" (Kelley, 2010), or 250 to 300 words a page; on our arithmetic, Toronto's range is about 80 to 200 words. More figures are in [PEEL and other paragraph patterns](https://edcitation.com/newsletter/peel-and-other-paragraph-patterns). Lingard says content, not space on the page, sets the length: split where the focus shifts, giving the new part its own topic sentence (Toronto), and merge short paragraphs on one topic.

## How do you revise a paragraph for cohesion, step by step?

1. **Read the brief.** [Check your paper](https://edcitation.com/check) lists its rules and what only a reader can judge.
2. **Name the point** in one line. If you cannot, the paragraph holds two.
3. **Test each sentence** against that line.
4. **Underline the first words** of each sentence: given, and related?
5. **Check each ending** carries the new information.
6. **Circle every reference word** and name what it points to.
7. **Trace the key term**, and replace stray synonyms.
8. **Keep only transitions** that name a real relationship.
9. **Check the length**, and make the last sentence a wrap.
10. **Then the mechanics.** [Mechanics QA](https://edcitation.com/tools/mechanics-qa) checks the file; [Verify references](https://edcitation.com/verify-references) checks the list, free.

## A worked paragraph, before and after

Both versions are ours; the study is real, and we read it at its abstract.

**Before (ours):**

> Sleep is something that affects every student. Wearable activity trackers were given to 100 students in an introductory chemistry class by Okano et al. (2019), and 88 of them completed the study. Better grades went with better, longer and more consistent sleep. Furthermore, the night before a test made no difference to the score. This was not true of the week and the month before. Many students also drink more coffee during exam periods. Rest and slumber measures accounted for nearly 25% of the variance in academic performance.

**After (ours):**

> For a student facing a test, the sleep that counts seems to be the sleep of the previous weeks, not of the night before. The clearest evidence comes from Okano et al. (2019), who gave wearable activity trackers to 100 students in an introductory chemistry class; 88 completed the study. Overall, students whose sleep was better, longer and more consistent earned better grades, and these sleep measures accounted for nearly 25% of the variance in academic performance. The single night before a test, however, showed no relation to the score; the relation lay with sleep in the preceding week and month. Because the data are correlational, they cannot show that more sleep raises grades, but they give a student little reason to rely on one early night before an exam.

| What changed | Why | Source |
| --- | --- | --- |
| A truism became a claim | A topic sentence is a small thesis | Toronto |
| The study follows "The clearest evidence" | Given first, new last | Gopen and Swan |
| "Furthermore" became "however" | The sentence contrasts | Lingard |
| "This was not true of" went | It could mean the night, the score or the finding | Toronto |
| The coffee sentence went | It serves no part of the claim | York |
| "Rest and slumber" became "these sleep measures" | One concept, one term | Morris and Hirst |
| A wrap replaced the bare figure | It says what the finding means | Lingard |

## What does EdCitation check in a paragraph, and what stays yours?

EdCitation checks what a rule can test and leaves the meaning to you, and for that job it is the best tool available: it never writes or rewrites a sentence, and every reference it reports on comes from a lookup in the publisher's record, a check no chatbot can make for you. In our run below, it confirmed the Okano et al. entry against that record.

### Mechanics QA on our "before" paragraph

On 2 October 2026 we saved the "before" paragraph, with a title and its reference, as a Word file and ran [Mechanics QA](https://edcitation.com/tools/mechanics-qa). The report scored it 88, with no must-fix items, two fixes and three notes. The fixes concerned the page ("No page number in the header or footer; APA 7 asks for page numbers." and "No title page found"). The notes verified the Okano et al. entry, counted "1 in-text citation · 1 entry · 0 orphan citations · 0 uncited entries.", and described the file's properties.

No finding touched the truism, the bare "This", the coffee or "Rest and slumber". Those are judgements about meaning, and the revision is yours. Mechanics QA is part of Pro, $8 a month, with three checks a month on a free account.

### Check your paper on a real brief

We pasted the Close Reading brief from MIT's 21L.000J Writing About Literature (Kelley, 2010) into [Check your paper](https://edcitation.com/check). It returned one rule, 1,250 to 1,500 words, read from "(5 pp or 1250-1500 words - due LEC #6)", and ten sentences under "What it asks you to do, which no rule can test", among them the request for "a thesis statement, topic sentences, and transitions". Seven sentences appeared in neither list, among them the instruction to conclude with "a thoughtful paragraph", and the five pages were not read as a rule, so read the brief yourself too.

| Job | EdCitation tool | Plan |
| --- | --- | --- |
| The brief's rules, and what only a reader can judge | [Check your paper](https://edcitation.com/check) | Free, no account |
| Formatting, citations, numbers and language in the file | [Mechanics QA](https://edcitation.com/tools/mechanics-qa) | Pro, $8 a month |
| The finished paper judged against its brief, quoting it | Theoretics QA | Max, $24 a month ([pricing](https://edcitation.com/pricing)) |

Theoretics QA rewrites nothing, the reference behind each point comes free from [Cite a source](https://edcitation.com/cite), and no check, ours included, can promise a grade.

## Quick questions

### How do you write a good topic sentence?

State the paragraph's one claim in your own words, near the start, in a sentence the rest of the paragraph proves. A bare fact, or an announcement such as "This paragraph discusses", is not one.

### What is the difference between cohesion and coherence?

Cohesion is the linking between sentences: pronouns, repeated terms, linking words. Coherence is whether the ideas make sense together; Morris and Hirst (1991) show a passage can be cohesive and still nonsense.

### Can a paragraph be one sentence long?

Not as a habit in academic writing: APA's student paper checklist says to avoid single-sentence paragraphs, though Lingard allows a very short one now and then for emphasis.

### Should every paragraph start with a topic sentence?

Body paragraphs in essays and theses should, by the university guides we read; introductions and conclusions generally do not. In one count of history articles, 66% of idea blocks opened with one (Smith, 2008).

### Can EdCitation fix my paragraphs?

No. Your paragraphs stay your own work, and EdCitation writes none of them. [Check your paper](https://edcitation.com/check) reads your brief free, Mechanics QA checks the file's mechanics, and Theoretics QA, with Max, weighs the paper against its brief.

## References

- American Psychological Association. (2020). *Student paper checklist (Publication manual, 7th ed.)* [PDF]. APA Style. [https://apastyle.apa.org/instructional-aids/publication-manual-formatting-checklist.pdf](https://apastyle.apa.org/instructional-aids/publication-manual-formatting-checklist.pdf)
- Gopen, G. D., & Swan, J. A. (1990). The science of scientific writing. *American Scientist, 78*(6), 550–558. [https://www.jstor.org/stable/29774235](https://www.jstor.org/stable/29774235)
- Kelley, W. (2010). *Assignments* [Course assignments]. 21L.000J Writing About Literature, Fall 2010. MIT OpenCourseWare. [https://ocw.mit.edu/courses/21l-000j-writing-about-literature-fall-2010/pages/assignments/](https://ocw.mit.edu/courses/21l-000j-writing-about-literature-fall-2010/pages/assignments/)
- Lingard, L. (2019). From semi-conscious to strategic paragraphing. *Perspectives on Medical Education, 8*(2), 98–100. [https://doi.org/10.1007/s40037-019-0507-4](https://doi.org/10.1007/s40037-019-0507-4)
- Morris, J., & Hirst, G. (1991). Lexical cohesion computed by thesaural relations as an indicator of the structure of text. *Computational Linguistics, 17*(1), 21–48. [https://aclanthology.org/J91-1002/](https://aclanthology.org/J91-1002/)
- Okano, K., Kaczmarzyk, J. R., Dave, N., Gabrieli, J. D. E., & Grossman, J. C. (2019). Sleep quality, duration, and consistency are associated with better academic performance in college students. *npj Science of Learning, 4*, Article 16. [https://doi.org/10.1038/s41539-019-0055-z](https://doi.org/10.1038/s41539-019-0055-z)
- Procter, M., & Visvis, V. (n.d.). *Paragraphs*. Writing Advice, University of Toronto. [https://advice.writing.utoronto.ca/planning/paragraphs/](https://advice.writing.utoronto.ca/planning/paragraphs/)
- Purdue Online Writing Lab. (n.d.-a). *On paragraphs*. Purdue University. [https://owl.purdue.edu/owl/general_writing/academic_writing/paragraphs_and_paragraphing/index.html](https://owl.purdue.edu/owl/general_writing/academic_writing/paragraphs_and_paragraphing/index.html)
- Purdue Online Writing Lab. (n.d.-b). *Revising for cohesion*. Purdue University. [https://owl.purdue.edu/owl/general_writing/the_writing_process/proofreading/revising_for_cohesion.html](https://owl.purdue.edu/owl/general_writing/the_writing_process/proofreading/revising_for_cohesion.html)
- Smith, C. G. (2008). Braddock revisited: The frequency and placement of topic sentences in academic writing. *The Reading Matrix, 8*(1), 78–95. [https://www.readingmatrix.com/articles/craig_smith/article.pdf](https://www.readingmatrix.com/articles/craig_smith/article.pdf)
- University of Sydney Learning Centre. (n.d.-a). *Cohesion in word choice* [PDF]. Cohesive Writing Module, Unit 3. [https://www.sydney.edu.au/content/dam/students/documents/learning-resources/learning-centre/writing/cohesion-in-word-choice.pdf](https://www.sydney.edu.au/content/dam/students/documents/learning-resources/learning-centre/writing/cohesion-in-word-choice.pdf)
- University of Sydney Learning Centre. (n.d.-b). *Cohesion through reference* [PDF]. Cohesive Writing Module, Unit 4. [https://www.sydney.edu.au/content/dam/students/documents/learning-resources/learning-centre/writing/cohesion-through-reference.pdf](https://www.sydney.edu.au/content/dam/students/documents/learning-resources/learning-centre/writing/cohesion-through-reference.pdf)
- University of York. (2026, September 2). *Structure & cohesion*. Academic writing. [https://subjectguides.york.ac.uk/academic-writing/structure](https://subjectguides.york.ac.uk/academic-writing/structure)
- Wulff, S., Römer, U., & Swales, J. (2012). Attended/unattended *this* in academic student writing: Quantitative and qualitative perspectives. *Corpus Linguistics and Linguistic Theory, 8*(1), 129–157. [https://doi.org/10.1515/cllt-2012-0006](https://doi.org/10.1515/cllt-2012-0006)