Cohesion & Coherence in Academic Writing

Cohesion and coherence are the two properties that make a piece of academic writing read as a single, connected argument rather than a sequence of correct-but-disjointed sentences. They are related but not interchangeable. Cohesion is a property of the text itself — the grammatical and lexical ties (pronouns, repeated key terms, conjunctions, parallel structure) that link one sentence to the next. Coherence is a property of the reader’s experience — whether the ideas, taken together, make logical sense as an argument, even if some of the surface-level ties are sparse. A paragraph can be highly cohesive (every sentence linked by explicit connectives) and still incoherent (the underlying logic doesn’t hold together), and a paragraph can be coherent with relatively light cohesive marking, if the idea sequence itself is well ordered.

This distinction goes back to the applied-linguistics literature on text structure — most influentially M.A.K. Halliday and Ruqaiya Hasan’s Cohesion in English (1976), which catalogued the specific grammatical and lexical devices writers use to tie sentences together, and the Prague School work on thematic progression (associated with František Daneš), which describes how the “given” information at the start of one sentence connects to the “new” information introduced in the sentence before it. Both frameworks are still the standard reference points in academic-writing pedagogy and in journals’ own author-facing style guidance.

This guide is deliberately narrower than general writing advice: it covers the specific linguistic mechanisms — reference, substitution, lexical repetition, conjunction, and thematic ordering — that produce cohesion and coherence in a research manuscript, thesis chapter, or grant narrative, and the revision techniques for finding and fixing breaks in them. It is a companion to two related but distinct CASRAI guides: hedging in academic writing covers how word choice signals the certainty of a claim, and signposting language covers explicit, self-referential phrases (“this section examines…”) that orient the reader within the document’s structure. Cohesion and coherence operate underneath both of those — they are what makes prose flow even in the sentences that contain no hedge and no signpost at all.

Cohesion vs. coherence: the operational difference

Use these two tests when deciding which problem you’re actually looking at in a draft:

Cohesion test: Can you point to a specific word or grammatical device — a pronoun, a repeated noun, a connective like “however” or “consequently,” an ellipsis — that links this sentence back to the previous one? If not, the sentence may be cohesively unmarked (not necessarily wrong, but worth checking).
Coherence test: If you strip out all the cohesive devices and read the bare sequence of claims, does the argument still make logical sense — does each idea follow from what came before, and does the paragraph as a whole answer a single implicit question? If not, no amount of added transition words will fix it; the underlying idea sequence needs to be reordered or rethought.

The practical implication: a manuscript that reads as “choppy” sentence-to-sentence usually has a cohesion problem, fixable by editing connective devices. A manuscript where individual sentences and paragraphs read fine in isolation but the section as a whole feels hard to follow usually has a coherence problem, which requires reworking the order or grouping of ideas, not just adding transition words.

The five cohesion devices (Halliday & Hasan)

Halliday and Hasan’s taxonomy remains the standard inventory of sentence-linking devices. All five appear routinely in published research writing:

Reference — a pronoun or determiner that points back (or occasionally forward) to something already named. “The intervention group received weekly counselling. This group also completed a follow-up survey at six months.” (“This group” refers back to “the intervention group.”)
Substitution — replacing a word or phrase with a placeholder to avoid repetition. “The first cohort was recruited by referral; the second was recruited the same way.” (“The same way” substitutes for “by referral.”)
Ellipsis — omitting a word or phrase that the reader can recover from context. “Group A improved on the primary outcome; Group B did not [improve on the primary outcome].”
Conjunction — explicit connectives that signal a logical relationship: addition (“furthermore,” “in addition”), contrast (“however,” “by contrast”), cause/result (“therefore,” “as a result”), or sequence (“subsequently,” “finally”).
Lexical cohesion — repetition of a key term, or use of a synonym, near-synonym, or superordinate category term, to keep the reader anchored on the same referent across a passage (e.g., repeating “the cohort” rather than alternating unpredictably between “the cohort,” “the sample,” and “the participants” without signalling they’re the same group).

A common failure mode in early drafts is over-relying on one device — usually conjunction (stacking “however” and “furthermore” at the start of nearly every sentence) — while under-using lexical cohesion and reference, which do more of the real work of tracking a referent across a paragraph without becoming mechanical.

Coherence techniques: ordering ideas so they hold together

Where cohesion is about links between adjacent sentences, coherence is about the shape of the whole passage. Three techniques do most of the work in research prose:

Given-new (theme-rheme) progression. Each sentence typically opens with information the reader already has (the “given,” often the subject of the previous sentence) and closes with new information (the “new”). When a paragraph keeps introducing new information in a sentence-opening position without first establishing it as given, readers experience it as disjointed even if every sentence is grammatically fine on its own.
One controlling idea per paragraph. A coherent paragraph answers a single implicit question or advances a single claim; supporting sentences elaborate, qualify, or provide evidence for that one idea rather than introducing a second, unrelated claim partway through.
Consistent logical ordering. Chronological, most-important-first, general-to-specific, and problem-to-solution are all legitimate ordering principles — the requirement is picking one and holding to it within a section, rather than switching structuring logic paragraph to paragraph without signalling the switch.

Where cohesion and coherence problems typically surface in a manuscript

Introduction: gaps in the given-new chain are common when a writer moves from broad field context to a specific research gap without an explicit bridging sentence connecting the two; the paragraph can read as two separate mini-essays stapled together.
Methods: cohesion tends to be less of a problem here because procedural writing is naturally sequential, but lexical cohesion failures are common — referring to the same instrument, dataset, or subgroup by different names in different subsections confuses a reader trying to track what was done to what.
Results: coherence problems appear when findings are reported in the order the analyses were run rather than the order that best supports the paper’s argument; reordering results around the research questions (rather than the analysis timeline) usually improves coherence without changing a single fact. See CASRAI’s guide to writing the results section of a research paper for more on structuring this section.
Discussion: this is where both properties matter most, since the discussion has to link findings back to the introduction’s framing, to the literature, and to the study’s limitations, often across the longest unbroken stretch of prose in the paper. See CASRAI’s guide to writing the discussion section of a research paper.

Illustrative example: a coherence revision

The paragraph below is an illustrative composite constructed for this guide, not a quotation from any real manuscript, author, or institution — it demonstrates a coherence problem and its fix, not a documented case.

Before (weak coherence, individually correct sentences): “Sleep quality affects cognitive performance. Our survey used a validated instrument. Twelve studies have examined this relationship in adults. The response rate was 64%. Most prior work has focused on shift workers rather than students.”

After (reordered around one controlling idea per sentence-group, given-new maintained): “Sleep quality is known to affect cognitive performance, and twelve prior studies have examined this relationship in adults — but most of that work has focused on shift workers rather than students, the gap this study addresses. To do so, we administered a validated sleep-quality instrument to a student sample, achieving a 64% response rate.”

No fact changed between the two versions. What changed is the grouping (background claims together, methods claims together) and the ordering (general finding → gap → how this study addresses it), which is a coherence fix, plus added reference and conjunction devices (“but,” “the gap this study addresses,” “to do so”), which are cohesion fixes. Most real revision work involves exactly this combination.

A self-editing checklist

Read only the first sentence of every paragraph in sequence. If that sequence doesn’t sketch a coherent mini-outline of the section’s argument, the paragraph-level ordering needs work before you touch word-level cohesion.
Circle every pronoun and check its antecedent. A “this,” “it,” or “these” with no clear, single antecedent in the previous sentence is a common cohesion failure, especially after a sentence containing more than one plausible referent.
Track key terms for unsignalled synonym drift. If a construct is called “the intervention” in one paragraph and “the treatment condition” in the next with no indication they’re the same thing, a reader may reasonably wonder if you mean two different things.
Check that every conjunction is doing real logical work. “However” should signal an actual contrast, not just serve as a paragraph-opening habit; misused connectives actively mislead readers about the relationship between ideas.
Read the section aloud, or have a colleague unfamiliar with the work read it once through without interruption. Places where a listener stumbles, re-reads, or asks “wait, what does that refer to?” are reliable indicators of a cohesion or coherence break that silent self-review tends to miss, because a writer already knows what they meant.

How this relates to hedging and signposting

Cohesion, coherence, hedging, and signposting are frequently discussed together under the broader label of metadiscourse (following Ken Hyland’s framework, also referenced in CASRAI’s signposting guide), but they solve different problems and are diagnosed and fixed differently:

Cohesion/coherence (this guide) is about whether the underlying ideas connect logically and are linked by the right grammatical devices — a structural property of the argument itself.
Hedging is about calibrating how confidently each individual claim is stated, independent of how well that claim connects to its neighbours.
Signposting is about explicit, self-referential orientation phrases (“this section addresses…”) layered on top of an already-coherent structure — signposting cannot fix a genuinely incoherent argument, it can only announce structure that already exists.

A manuscript can fail on any one of these independently of the others: a well-signposted section can still be internally incoherent, and a coherent, cohesive paragraph can still overclaim if the hedging is off. Editing for all three separately, rather than treating “flow” as one undifferentiated problem, tends to produce more targeted revisions.

Frequently asked questions

What is the simplest way to explain the difference between cohesion and coherence?

Cohesion is about the ties between sentences (the words and grammar that connect them); coherence is about whether the ideas, taken as a whole, make logical sense. Text can have one without the other.

Are cohesion and coherence the same as using transition words?

No. Transition words (conjunction, in Halliday and Hasan’s terms) are one of five cohesion devices, alongside reference, substitution, ellipsis, and lexical cohesion. Adding transition words to an incoherently ordered paragraph does not fix the underlying problem — it can even make it worse by asserting a logical relationship (“therefore,” “as a result”) that the ideas don’t actually support.

How can I check the coherence of my own paper?

Read only the opening sentence of each paragraph in a section, in order. If that sequence reads as a reasonable outline of the section’s argument, the paragraph-level ordering is likely coherent; if it reads as an arbitrary list, the ideas need to be regrouped or reordered before any sentence-level editing.

Can a paragraph be too cohesive?

Yes. Overusing explicit connectives (“however,” “furthermore,” “therefore”) on nearly every sentence can make prose feel mechanical and can misrepresent the actual logical relationships between ideas. Skilled academic writing typically relies more on reference and lexical cohesion, which are less intrusive, and reserves explicit conjunctions for genuine logical pivots.

Do peer reviewers actually comment on cohesion and coherence?

Reviewers rarely use the terms “cohesion” or “coherence” directly, but comments like “hard to follow,” “unclear what this refers to,” “this paragraph seems to jump between two ideas,” or “the logic here isn’t clear” are, functionally, cohesion or coherence critiques, and are among the most common structural feedback research writers receive.