Discourse cohesion at C2

A text is more than a sequence of sentences. What turns a sequence into a text — what gives it texture, in Halliday and Hasan’s term — is cohesion: the network of explicit linguistic ties that connect each sentence to what came before and what comes after. Halliday and Hasan’s Cohesion in English (1976) maps the system in five categories: reference (pronouns, demonstratives, comparatives), substitution (one, do, so), ellipsis (zero-substitution), conjunction (and, but, however, therefore), and lexical cohesion (repetition, synonymy, hyponymy, collocation).

At C1 you handled most of these as separate phenomena. At C2 the move is integrative: you see cohesion as a single system whose density and choice of devices vary by register, and you start reading long texts the way an editor reads them — tracking each referent through every sentence, noticing where the chain breaks. You also start writing this way: every sentence in your draft is asked what does this connect back to, and what does it set up next?

For a Russian-speaking C2 learner, the systematic gap is in reference tracking and lexical cohesion. Russian pronoun systems work differently — Russian licenses null subjects, uses demonstratives more freely, and tolerates referent ambiguity that English does not. Russian academic prose also tolerates more repetition than English, where lexical variation under hyponymic or synonymic cohesion is the norm. This lesson maps the system and trains the production habits.

Reference — anaphora and cataphora

Reference ties point to entities mentioned elsewhere. Anaphora points backward (the referent has been introduced); cataphora points forward (the referent is about to be introduced); exophora points outside the text (deictic this room, that man over there).

Anaphora — the workhorse

English anaphora is mostly pronominal: he, she, it, they, this, that, these, those, the + repeated noun (the anaphoric definite).

A man walked into the bar. He ordered a beer.

Here he anaphorically refers to a man. The anaphoric tie is the cohesive device.

Anaphoric chains in long texts

In a paragraph, a single referent may take a chain of forms:

Sarah Chen founded the company in 2019. She had previously led product at Stripe, where the engineer built three of the early payment APIs. The Stripe alum brought a Silicon Valley sensibility, but Chen also leaned into something Stripe didn’t have: a Pacific-Northwest ethos of slow, deliberate growth.

The chain: Sarah Chen → she → the engineer → the Stripe alum → Chen. The chain alternates between pronoun and lexical repeats (epithet, role-name, surname) to avoid both monotony and reader confusion. This is the C2 move Russian-trained writers often miss — they either over-repeat the proper noun (reads as wooden) or over-pronominalize across long stretches (reads as ambiguous).

Cataphora — the literary anticipator

Cataphora moves the pronoun before its referent:

Even as he stepped onto the stage, the senator knew the speech was a mistake.

Cataphora is marked — it draws attention to itself. Use it for emphasis, not casually. American journalism uses cataphora in lede sentences for hook effect:

She’d never told anyone before. But this afternoon, in a windowless office in Cambridge, Dr. Maria Sandoval would say the words out loud.

The she in sentence one is cataphoric to Dr. Maria Sandoval in sentence two. The reader’s mild puzzlement is the hook.

Demonstrative reference — this, that, these, those

Demonstratives anchor reference deictically in time, space, or discourse position.

• This — proximal; signals the just-mentioned or the upcoming topic. This argument has problems.
• That — distal; signals already-handled or distanced topic. That argument, however, is exhausted.

A robust C2 distinction: this/that used pronominally refers to whole propositions or events, not single nouns:

The CEO resigned at 4pm. This caught the board off guard.

This refers to the CEO resigned at 4pm, the entire prior proposition. Russian это works similarly but English uses pronominal this a touch more sparingly — overuse reads as breathless.

Substitution — one, ones, do, so, not

Substitution replaces a phrase with a dummy form. Three classes:

Nominal substitution: one, ones

I’ll have a coffee. — A small one or a large one?

One substitutes for coffee. Cohesive tie back to the introduced noun.

Verbal substitution: do, do so

She finished the report. He should do so too.

Do so substitutes for finish the report. Formal register prefers do so; casual register uses do (it) or do too.

Clausal substitution: so, not

Is he coming? — I think so. / I think not.

So / not substitutes for the entire embedded clause he is coming.

The C2 task: deploy these to avoid repetition. Russian rarely needs the substitution because Russian licenses ellipsis more freely; English speakers reach for one, do, so in slots where Russian writes nothing.

Ellipsis — silence that means something

Ellipsis is substitution by zero — omitting a phrase that the reader recovers from context.

Nominal ellipsis

I bought two books — a long [ø] and a short [ø]. (= a long one, a short one)

Verbal ellipsis

She’ll arrive at six and he [ø] at seven. (= he will arrive at seven)

Clausal ellipsis

Will she come? — She might [ø]. (= she might come)

English ellipsis is constrained — you cannot ellipt anything the reader cannot recover. Russian ellipsis is more permissive (Russian licenses null subjects and verb-elided constructions Russian English speakers carry over). The classic Russian carry-over: dropping the auxiliary in an English answer where English wants it kept.

Russian-direct: — Are you coming? — Going.

English-correct: — Are you coming? — I am. (auxiliary retained; main verb ellipted)

The English rule: ellipt the lexical verb, retain the auxiliary. I am, I will, I have, I would — these answer-shapes are obligatory.

Conjunction — explicit logical ties

Conjunction signals the logical relation between two stretches of text. Halliday & Hasan distinguish four types:

• Additive: and, also, in addition, furthermore, moreover, what’s more
• Adversative: but, yet, however, on the other hand, nevertheless, still
• Causal: because, since, therefore, thus, hence, consequently, as a result, so
• Temporal: then, next, after that, meanwhile, subsequently, finally, in the end

C2 register differentiation:

Mixing registers (Moreover, we’re broke) reads as awkward. Match conjunction to register.

Lexical cohesion — the most underused C2 device

Lexical cohesion ties sentences through the choice of related words rather than grammatical reference. Five subtypes:

Repetition

The founder wrote the original code. The founder also wrote the documentation.

Pure repetition. English tolerates less of this than Russian — overuse reads as wooden in English.

Synonymy / near-synonymy

The founder wrote the original code. The architect of the system also wrote the documentation.

The second mention varies the lexical form. Architect of the system is near-synonymous with founder in context. This is the C2 move: epithets and role-names that vary the referent.

Hyponymy / hyperonymy (subset / superset)

We bought three plants — a monstera, a ficus, and a pothos.

Monstera, ficus, pothos are hyponyms of plants. Hyponymy creates cohesion via the kind-of relation.

The monstera, the ficus, the pothos — every plant in the place needs water.

Reverse direction: specific instances followed by the hyperonym. This is the C2 generalization move common in American journalism.

Meronymy (part-of)

The car was a wreck. The engine was seized, the transmission dead, the interior rotted.

Engine, transmission, interior are meronyms of car. The text coheres because the parts evoke the whole.

Collocation

The trial was a circus. The judge glared, the defense stalled, the jury dozed, the verdict surprised no one.

The cohesion is not via any single ref-tie but via judicial scene collocation: trial, judge, defense, jury, verdict belong together by association. American long-form journalism uses collocational cohesion heavily.

Tracking referents in long texts — the C2 reading skill

Read the following passage from a New Yorker-style profile (composed in the house style):

Mira Patel grew up in a four-bedroom split-level in Sunnyvale, the only daughter of two software engineers. The Patels had emigrated from Hyderabad in 1992; her parents met on a Sun Microsystems shuttle and were married within the year. The marriage produced Mira and, four years later, a son, Sanjay. The siblings were close, but it was the older one who, by twelve, had decided she was going to be an artist.

Trace the chains:

• Mira: Mira Patel → the only daughter → her parents (possessor) → Mira → the older one → she.
• Parents: two software engineers → the Patels → her parents → the marriage (predicate-cohesion).
• Family unit: split-level (mero) → the marriage → the siblings (collocational).

Eight referents tracked across four sentences via five different devices. This is the C2 prose reader’s job: notice every tie, never lose the thread.

Building cohesion in your own writing — the C2 production drill

The drill: write a five-sentence paragraph on any topic. After drafting, edit each sentence with the question what does this sentence tie back to? and the parallel question what does this set up for the next sentence? Mark each tie explicitly:

Sentence 1: introduces X.

Sentence 2: anaphoric reference to X (pronoun or definite). Conjunction (additive / adversative / causal) to S1.

Sentence 3: introduces Y as a hyponym/meronym of X. Conjunction to S2.

Sentence 4: lexical cohesion via collocation with X and Y. Demonstrative reference (this/that) to S2 proposition.

Sentence 5: cataphoric to the next paragraph’s topic. Conjunction (temporal or summative) to the whole paragraph.

This is editorial work, not invention. A C2 writer drafts intuitively and then audits cohesion the way a programmer audits dependencies.

Common cohesion pitfalls in Russian-to-English transfer

Dropped auxiliary in elliptical answers

Russian — Придёшь? — Приду. maps to English — Coming? — I am. not — Coming? — Come. Retain the auxiliary; ellipt the lexical verb.

Demonstrative this used where the is wanted

Russian это covers more semantic space than English this. Russian-trained writers produce This argument has been made before where English wants The argument has been made before — the anaphoric definite. This is marked; the is unmarked.

Pronoun ambiguity from null-subject habits

Russian licenses null subjects: Пришёл, увидел, победил. English forces overt subjects: He came, he saw, he conquered. Russian writers carry this over and produce English with too few pronouns; the reader loses the chain.

Over-repetition where lexical variation is wanted

Russian academic prose repeats the same noun more than English does. The factor X causes Y. The factor X also influences Z. The factor X is therefore important. English wants The factor X causes Y. It also influences Z. This makes it important. — pronoun + demonstrative.

Missing conjunctions on relation shifts

Russian often relies on context to signal causal or adversative relations; English wants explicit therefore, however, but, because. Bare juxtaposition reads as logic-gap.

Cohesion in spoken vs written register

The cohesion devices map differently onto spoken and written English. The C2 speaker controls both surfaces.

Spoken cohesion

• Heavy use of demonstratives (this, that, these, those) as discourse-deictic pronouns.
• Heavy pronominalization — he, she, it, they carry chains through long stretches.
• Substitution via one, do, so, not is common.
• Ellipsis is freer in speech than in writing — I will, I have, I might answer-shapes ellipt the lexical verb.
• Conjunction is mostly and, but, so — formal conjunctions (moreover, hence, therefore) rarely surface in speech.
• Lexical variation is moderate — speakers don’t have time to optimize lexical choice.

Written cohesion

• Anaphoric definite (the X) is common where speech might use a pronoun, because the writer needs to reduce ambiguity for the reader.
• Demonstrative pronouns this, that used for whole-proposition anaphora are particularly common in academic writing.
• Substitution is less frequent — writers prefer explicit referents.
• Ellipsis is controlled — only patterns clearly recoverable from context.
• Conjunction span is wider — formal conjunctions surface freely.
• Lexical variation is heavy — synonyms, near-synonyms, epithets cycle to avoid repetition.

Real US example — spoken cohesion in a podcast

And the funny thing was, she’s known about it for months. They told her in March. And she was the one who decided to roll with it anyway. So that’s where we are now.

She, they, her, the one, it, anyway, so that — six cohesion devices, all light. Pronominalization carries the work.

Real US example — written cohesion in a New Yorker piece

Chen had founded the company in 2019. The founder, then twenty-eight, had spent the previous five years at Stripe; her co-founder, Diaz, had come from a smaller payments startup in Brooklyn. The pair had met at a conference in Lisbon and, by their accounts, decided in a single afternoon that they would build something together.

Chen, the founder, her co-founder, Diaz, the pair, they, their — eight cohesion devices using anaphoric definite (the founder, the pair), proper noun anaphora (Chen, Diaz), possessives (her, their), and pronoun (they). The lexical variation is the texture.

AmE vs BrE in cohesion

• Demonstrative density: AmE journalism uses this and that pronominally slightly more than BrE; BrE leans on the latter, the former.
• Connective register: BrE retains whilst, amongst, betwixt; AmE uses while, among, between. Mixing reads as foreign.
• Numbered list cohesion: AmE academic prose often uses First…, Second…, Third…; BrE prefers Firstly…, Secondly…, Thirdly… AmE writers find firstly slightly off.
• Lexical variation: AmE editorial style (Strunk & White, the New York Times manual) actively encourages varying the referent across mentions. BrE academic style tolerates more repetition.

Functional view — cohesion device map

Common Russian-speaker mistakes

1. Pronominal under-use: nominal repetition where English wants pronoun. Sarah said Sarah would arrive at six → Sarah said she would arrive at six.
2. Demonstrative over-use: this and that used as anaphoric pronouns where the simple the + N or pronoun is wanted. The CEO resigned. This caused panic (correct) vs The CEO resigned. This CEO had been in office for ten years (wrong — should be He had been).
3. Auxiliary dropped in elliptical answers: — Are you coming? — Coming → — Are you coming? — I am. Auxiliary stays.
4. No conjunction on relation shifts: bare juxtaposition reads as logic-gap. Insert however, therefore, in fact, but explicitly.
5. Wooden repetition of proper nouns and full noun phrases: vary via near-synonym, role-name, or epithet. Tim Cook → the Apple CEO → Cook → the executive.
6. Mismatched register conjunctions: moreover in casual text or plus in academic. Match.
7. Null subjects from Russian: Came in, sat down, started talking (Russian-style) → He came in, sat down, and started talking. English keeps the subject in the first clause; ellipts in coordinated continuation.

Summary

• Halliday & Hasan’s five cohesion categories: reference, substitution, ellipsis, conjunction, lexical cohesion.
• Reference covers personal pronouns, demonstratives, anaphoric definites. C2 prose alternates these across long chains to avoid both monotony and ambiguity.
• English ellipsis retains the auxiliary, ellipts the lexical verb in answers: I am, I will, I have.
• Conjunction must match register: moreover/furthermore (academic) vs also/plus (casual) vs however/that said (mid-formal).
• Lexical cohesion via synonymy, hyponymy, meronymy, and collocation gives C2 prose its texture. Russian-trained writers under-use lexical variation.
• Editing for cohesion is editorial labor — drafting first, auditing second. Every sentence should answer what does this tie back to?
• AmE uses this/that pronominally slightly more than BrE; both varieties prefer lexical variation over repetition.

Next lesson: Register mastery at C2 — reading the room in any register, switching for humor and sincerity, recovery after slip, and deliberate register-mixing for effect.