Simple explanation
What is the difference between cohesion and coherence? Cohesion concerns the linguistic connections that hold parts of a text together. These include reference, repetition, lexical relationships, substitution, ellipsis, conjunctions, and discourse markers. Coherence is broader: it concerns whether the ideas form a logical, interpretable whole. A text may contain many words such as however, therefore, and moreover and still lack coherence if its argument develops unpredictably or its claims are poorly connected. Local cohesion and global coherence Local cohesion operates between nearby sentences and clauses. A sentence should normally connect to information that has already been introduced before adding something new. Global coherence operates across the whole text. Each paragraph should contribute to a larger purpose, and the relationship between sections should remain visible even when the discussion becomes complex or draws on several sources. Why multi-source writing is more demanding A single-source explanation can often follow one relatively stable line of thought. Multi-source writing requires the writer to manage overlapping claims, conflicting evidence, methodological differences, and varying degrees of certainty. At C2 level, sources should not simply appear one after another. They should be grouped according to the role they play in the writer's argument. The given-to-new principle One powerful way to improve information flow is to move from familiar information to new information. A sentence can begin by referring to something already established and then extend it. For example, after introducing a decline in commuting, the next sentence might begin with This reduction in daily travel before explaining a new environmental consequence. Synthesis rather than source-by-source summary Summary reports what individual sources say. Synthesis explains how several sources relate to the same question. A strong C2 paragraph may begin with a claim, introduce evidence from one source, qualify it with another, and then use a third source to explain why the findings differ. The paragraph remains controlled by the idea rather than by the sequence of authors. Cohesion without excessive linking words Advanced writing does not require a connector at the beginning of every sentence. Cohesion may be created through lexical repetition, reference chains, parallel structures, consistent terminology, and logical sequencing. A discourse marker should clarify a genuine relationship, not merely make the text sound more formal. Maintaining coherence across long sections In a long text, readers need periodic signals that remind them how individual points relate to the central argument. Topic sentences, mini-syntheses, section transitions, and repeated key concepts help maintain orientation. Each new section should both connect backward to what has been established and move the argument forward.
Grammar formula
Given-to-new information flow
given information + clear reference + new informationRemote work has altered patterns of collaboration. This shift has also changed how managers evaluate productivity.
Agreement across sources
source A claim + alignment marker + source B evidence + synthesisMorgan links flexible schedules to higher retention. In line with this, Chen reports lower turnover among hybrid teams, suggesting that autonomy may influence long-term employee commitment.
Contrast across sources
source A claim + contrast marker + source B counter-evidence + interpretationLee presents remote work as largely beneficial. By contrast, Ahmed identifies substantial coordination costs, indicating that its effects depend partly on task interdependence.
Qualified claim
main claim + concessive limitation + conditionThe evidence broadly supports flexible working arrangements, although the benefits appear less consistent in highly collaborative roles.
Cross-paragraph transition
reference to previous point + new analytical directionThis productivity-focused account, however, overlooks the social consequences of prolonged remote work.
Mini-synthesis
multiple findings + interpretive summary + broader implicationTaken together, these findings suggest that flexibility is beneficial only when organisational support is sufficient.
Core rules
- 1
Organise paragraphs around claims, not sources
A paragraph should normally have one dominant analytical focus. Evidence from different sources should support, challenge, refine, or contextualise that focus rather than appearing as separate mini-summaries.
- 2
Make reference chains explicit
Pronouns and demonstratives such as this, these, such, and they must have clear antecedents. In dense writing, a noun phrase such as this methodological limitation or this apparent contradiction is often clearer than a bare this.
- 3
Name the relationship between sources
Do not merely report that different authors say different things. Show whether one source supports, challenges, qualifies, extends, or contextualises another.
- 4
Use repetition strategically
Repeating a central technical term can strengthen coherence. Avoid replacing every repeated term with a synonym if the synonym changes the meaning or weakens conceptual consistency.
- 5
Build conceptual transitions between paragraphs
A new paragraph should not feel detached from the previous one. Its opening should show whether it extends, narrows, contrasts with, or reinterprets the previous discussion.
- 6
Match discourse markers to the real logical relation
Use addition markers for addition, contrast markers for contrast, causal markers for cause, and concessive markers for qualification. An inaccurate connector damages coherence even when the grammar is correct.
- 7
Preserve the writer's analytical voice
Sources provide evidence, but the writer controls the argument. Topic sentences, comparisons, qualifications, and synthesis statements should show how the evidence is being interpreted.
- 8
Signal major shifts in argumentative direction
When moving from evidence to limitation, from consensus to disagreement, or from explanation to evaluation, provide enough signalling for the reader to recognise the shift immediately.
Useful examples
Several studies associate remote work with greater employee autonomy. This increased control over working conditions may partly explain the higher job satisfaction reported in distributed teams.
Several studies connect remote work with more employee autonomy, and that greater control may help explain higher job satisfaction.
This increased control creates a precise reference chain from autonomy to a new interpretation.
Morgan attributes the decline in staff turnover to flexible scheduling, while Chen emphasises improved managerial trust. Taken together, these findings suggest that retention may depend on several interconnected aspects of workplace autonomy.
Morgan focuses on flexible schedules and Chen on managerial trust, but together their findings suggest that several forms of autonomy may improve retention.
Taken together synthesises two sources instead of leaving them as separate reports.
Early research treated digital collaboration primarily as a technological challenge. More recent work, however, has shifted attention towards organisational culture and informal communication.
Earlier studies focused mainly on technology, whereas newer research gives more attention to culture and informal communication.
However marks a genuine shift in research emphasis.
The two surveys appear to reach similar conclusions, but this apparent agreement masks an important methodological difference: one measures perceived productivity, whereas the other records completed tasks.
The surveys seem to agree, but their similarity hides a key difference in how productivity is measured.
This apparent agreement refers to an entire proposition rather than a single noun.
Cost reduction is frequently presented as a major advantage of automation. That argument becomes less straightforward, however, once training, maintenance and transition costs are included.
Automation is often said to reduce costs, but the claim becomes more complicated when additional expenses are considered.
That argument provides a clear backward reference before the claim is qualified.
Although Patel reports a substantial productivity gain, the study focuses exclusively on experienced employees. This limitation makes it difficult to generalise the findings to newly recruited staff.
Patel reports higher productivity, but because the study only examined experienced workers, the findings may not apply to new employees.
This limitation names exactly what is being carried forward into the next sentence.
Environmental benefits are evident at the level of daily commuting. A broader assessment, however, must also consider household energy consumption, data-centre demand and changes in residential patterns.
Reduced commuting may bring environmental benefits, but a complete assessment must also include energy use, data centres, and residential change.
The second sentence deliberately broadens the analytical frame.
Both authors acknowledge that access to digital infrastructure affects educational outcomes. Where they differ is in the mechanism they propose: Lewis emphasises connectivity, whereas Rahman focuses on digital literacy.
Both authors agree that digital infrastructure matters, but they disagree about why: one stresses connectivity and the other digital literacy.
The passage establishes common ground before identifying the precise disagreement.
The policy may therefore reduce short-term congestion without addressing the underlying demand for private transport. This distinction is crucial when evaluating whether the intervention produces temporary relief or structural change.
The policy may reduce traffic temporarily without changing demand for private transport, and that difference matters when judging its long-term impact.
This distinction converts the previous analytical contrast into an evaluation criterion.
One explanation is that employees value the freedom to choose where they work. A complementary explanation concerns time: eliminating long commutes may improve both well-being and perceived control over the working day.
Employees may value freedom of location, but another compatible explanation is that removing long commutes improves time control and well-being.
A complementary explanation signals addition without implying contradiction.
The literature does not support a simple conclusion that online learning is either superior or inferior to classroom instruction. Rather, outcomes vary according to learner autonomy, course design and the type of interaction required.
Research does not show that online learning is simply better or worse; outcomes depend on learner autonomy, course design, and interaction.
Rather replaces an oversimplified binary conclusion with a more precise interpretation.
Taken as a whole, the evidence points to a conditional rather than universal benefit. Flexible working appears most effective when employees have sufficient autonomy, appropriate technology and clearly defined coordination practices.
Overall, the evidence suggests that flexible work is beneficial under certain conditions rather than in every situation.
Taken as a whole synthesises a larger body of evidence.
Common mistakes
- ❌ Smith supports remote work. Moreover, Jones disagrees with remote work. ✅ Smith presents evidence in favour of remote work. By contrast, Jones identifies several disadvantages. Moreover signals addition, not opposition. The connector must reflect the actual logical relationship.
- ❌ Remote work can reduce commuting. This can also affect cities, companies and people in many ways. ✅ Remote work can reduce commuting. This reduction in daily travel may affect urban transport demand, office occupancy and residential choices. A bare this is too vague in dense writing. The noun phrase makes the reference precise.
- ❌ Smith argues that productivity rises. Jones argues that communication becomes harder. Brown argues that workers feel isolated. ✅ Although some studies associate remote work with higher productivity, others identify communication difficulties and social isolation, suggesting that performance gains may involve organisational trade-offs. The incorrect version lists sources independently. The corrected version synthesises them around one analytical claim.
- ❌ Remote work improves flexibility. Furthermore, it may create isolation. Furthermore, managers may struggle to coordinate teams. ✅ Remote work improves flexibility, but these gains may be accompanied by greater social isolation and coordination difficulties. Furthermore suggests compatible addition, but the later points qualify the initial benefit.
- ❌ The study used only 25 participants. However, the sample was small. ✅ The study used only 25 participants. This small sample limits the extent to which the findings can be generalised. The incorrect version merely restates the same fact. The corrected version develops its analytical significance.
- ❌ There are several benefits of AI in education. On the other hand, personalised feedback is one of them. ✅ There are several potential benefits of AI in education. One of the most significant is the possibility of providing personalised feedback at scale. On the other hand introduces contrast, not an example or specification.
- ❌ The results were inconsistent. This shows the method is useless. ✅ The results were inconsistent. This variability suggests that the method may be sensitive to contextual or methodological factors. The original conclusion is stronger than the evidence supports. Advanced writing should match the level of certainty to the available evidence.
Learning tips
- Group sources by theme, agreement, disagreement, limitation, or methodological difference before writing.
- If every paragraph discusses only one source, you are probably summarising rather than synthesising.
- After drafting a paragraph, check whether you can state its central claim in one sentence.
- Use explicit reference phrases such as this discrepancy, this pattern, or such an interpretation when a bare pronoun could be ambiguous.
- Do not replace technical terms with synonyms purely to avoid repetition.
- Choose a discourse marker only after identifying the real logical relation between the statements.
- End complex sections with a brief synthesis that explains what the evidence collectively contributes to the argument.
- Check whether the opening sentence of each paragraph both connects to the previous discussion and introduces a new analytical step.
Practise this topic
Interactive exercises for this topic are coming soon.
PlannedRelated grammar
Related dictionary words
Frequently asked questions
Are cohesion and coherence the same thing?
No. Cohesion refers to linguistic connections between parts of a text, whereas coherence refers to the logical unity and interpretability of the text as a whole. A text can contain many cohesive devices and still be poorly organised.
How many linking words should I use in a C2 essay?
There is no ideal number. Use them only when they clarify a real logical relationship. At C2 level, cohesion should also come from reference, lexical continuity, information structure, and logical sequencing.
How do I combine several sources in one paragraph?
Begin with the paragraph's main claim, then introduce sources according to their relationship to that claim. One source may support it, another may qualify it, and a third may explain a discrepancy. Finish with a short synthesis where appropriate.
Is repetition always bad in academic writing?
No. Strategic repetition of key technical terms can strengthen coherence. Replacing every repeated word with a synonym may actually reduce precision or make reference chains harder to follow.
When is this too vague as a reference word?
It is too vague when the reader could reasonably connect it to more than one idea in the preceding sentence. In that case, use a fuller expression such as this discrepancy, this decline, or this methodological limitation.
What is the difference between summary and synthesis?
A summary mainly reports what one or more sources say. A synthesis integrates information from several sources to develop, compare, qualify, or evaluate a larger argument.
How can I improve coherence between paragraphs?
Make the opening of each paragraph connect explicitly to the previous stage of the argument while introducing a new analytical direction. The transition should show whether the next paragraph extends, contrasts with, or reframes what came before.
Should every paragraph begin with however, moreover, or therefore?
No. Advanced writing benefits from varied transitions. Logical sequencing, reference chains, topic continuity, and sentence structure can often create coherence without an explicit connector.