🗣️ 36. Discourse relations

A text is more than a list of sentences: one clause gives the cause of another, the result, a contrast, an example, a condition, the next event. Ukrainian marks these discourse relations with conjunctions (бо, тому що, хоча, але; chapter 24), with connective adverbs and parenthetical words (отже, тому, проте, однак, зате, натомість, по-перше, тобто, зокрема), with correlative pairs (хоч … але, якщо … то), and — more than English — with nothing at all: two clauses side by side, joined by intonation in speech and by a comma, colon or dash in writing. This chapter maps the relations to their Ukrainian markers, gives their frequencies and registers in news, mixed prose and parliamentary speech (counted on 2026-11-06), describes the punctuation that signals implicit relations, the folk preference for parataxis that Smerechynskyi documented, the variation across styles and varieties, and what Mova does with Ukrainian discourse today, which is nothing beyond rejecting most of it.

Contents

1. Means of marking a relation
2. The relations and their markers
3. Frequency and register
4. Position and punctuation of connectives
5. Implicit relations: asyndeton
6. Parataxis in the folk language
7. Time, aspect and the order of telling
8. Correlatives and doubled marking
9. Attribution
10. Varieties and styles as deltas
11. Quick reference: discourse relations in UD
12. How Mova handles it
13. Sources

Related chapters: conjunctions, subordinate clauses and their frequencies in 24. Coordination and subordination; anaphoric це, через це in 35. Reference and coreference; implicature of order and of connectives in 33. Implicature and presupposition; discourse markers in dialogue in 38. Dialogue and discourse markers; aspect in 13. Verbs; punctuation in 7. Punctuation, capitals, abbreviations and numbers; register in 37. Register and variation.

Conventions as in chapter 20: examples in Cyrillic italics, romanized by the national system (2010) on first use and glossed; sentences from UD Ukrainian-ParlaMint (CC BY-SA 4.0) quoted with their sent_id; UD Ukrainian-IU (CC BY-NC-SA 4.0) and Brown-UK (CC BY-NC-SA 4.0) used for counts only. Corpus counts are case-insensitive whole-word matches (grep -oiw, 2026-11-06) in the Leipzig ukr_news_2024 sample (1 M sentences, about 14.2 M words, CC BY) and in Brown-UK (1,037,038 words, mixed genres including fiction); treebank counts are UD 2.18, all splits, awk. Per-million figures are rounded.

1. Means of marking a relation

The conjunctions are described in chapter 24; this chapter treats them as markers of relations and adds the rest.

2. The relations and their markers

The table follows the senses of the Penn Discourse Treebank (PDTB) — temporal, contingency, comparison, expansion — and lists the Ukrainian markers of each, with the style each belongs to. "Neutral" means usable in any register.

Several markers serve more than one relation. Водночас is temporal ('at the same time') and contrastive ('at the same time, however'); а is additive, contrastive and, with negation, corrective (не я, а він); коли and як are temporal and conditional; тому is a result adverb on its own and part of the reason conjunction тому що; нарешті closes a list and marks the last event of a sequence.

3. Frequency and register

3.1 Three corpora

Per million words, 2026-11-06 (raw counts in parentheses):

(The Leipzig sample is about fourteen times larger than Brown-UK; its per-million figures are the steadier.) News prefers the bookish reason markers оскільки and тому що and the instance marker зокрема, the workhorse of reported politics (зокрема, йдеться про…). Brown-UK, with fiction, essays and science, uses бо almost three times as often, and the reflective connectives отже, тож, проте, однак, тобто two to six times as often: news reports events and quotes; argumentative and narrative prose builds chains of reasoning. Зате is rare in news and ten times more frequent in Brown-UK: it belongs to the narrator's and the conversationalist's 'but at least'.

3.2 Speech and writing in the treebanks

Clause-linking in the two treebanks (UD 2.18, 2026-11-06; IU 99,495 words, ParlaMint 89,858):

Parliamentary speech subordinates less (858 adverbial clauses against 1,618) and juxtaposes more (2,031 parataxis); it states results with the plain adverb тому ('that's why') more than three times as often as writing (273 vs 83), and enumerates arguments with по-перше, по-друге. The conjunction counts by mark are in chapter 24, § 5.2: якщо and щоб lead in parliament, як, бо and the converbs in writing.

The two treebanks also annotate the same words differently: ParlaMint treats тобто mostly as cc (26), IU as mark (24); IU attaches отже as discourse or mark, ParlaMint as parataxis or discourse. A model trained on one and tested on the other loses points on connectives for reasons of convention, not of grammar.

3.3 Так як, позаяк and the purist tradition

The causal так як ('since, as') is a long-standing target of Ukrainian normative writing. Simovych (1919, simovych-1919-hramatyka-ukrainskoi-movy-dlia, § 374, note) already wrote that the educated Ukrainian's так як is "an obvious Muscovitism" (from Russian так как) to be avoided, and that the bookish позаяк, used by intellectuals for тому що, means something else where it is native — in the Pidhiria it means 'insofar as' — and should not be used either. A century later the Leipzig news sample has 85 так як (most of them the comparative так, як 'just as', not causal) against 5,913 оскільки and 6,026 бо; Brown-UK has 4. The causal так як survives in speech and in some official writing (ParlaMint: 5 sequences так як, against 56 тому що); in edited prose the standard alternatives have won. Позаяк survives as a rare bookish or stylised connective (26 in news).

4. Position and punctuation of connectives

4.1 Conjunction or parenthetical word

Several connectives are conjunctions at the start of a clause and parenthetical words inside it. The 2019 orthography (ukrainskyi-pravopys-2019, § 158, item 11, note 1 в) lists among the parenthetical words that "indicate the connection of thoughts and the order of presentation" по-перше, по-друге; з одного боку, з другого боку; наприклад, так, відповідно, отже, таким чином, словом, виходить, значить, зрештою, нарешті; до речі, зокрема, крім того, навпаки, and проте and однак "inside, not at the beginning of the sentence, where these words are conjunctions". The punctuation follows the function:

The treebanks reflect the double nature: отже is mark 10 times and discourse 17 times in IU.

4.2 Splitting the compound conjunctions

The compound reason conjunctions тому що and через те що can be split: the anaphoric part stays in the main clause and is stressed, the comma moves before що: Я прийшов тому, що ти просив 'I came because you asked' (the reason is the focus); Я прийшов, тому що ти просив (the reason is added). Brown-UK writes the split form тому, що three times as often as тому що (321 vs 99); the news sample about equally (3,407 vs 3,291), but the comma form also includes the correlative of other constructions (завдяки тому, що; дорівнює тому, що). The split is a focus device of the written standard; in the treebanks ParlaMint has 56 adjacent тому що and 38 тому, що sequences.

5. Implicit relations: asyndeton

5.1 The punctuation of implicit relations

When no connective is present, the reader infers the relation from the content, the order and, in writing, the punctuation. The 2019 orthography gives each mark its relations in an asyndetic complex sentence:

The orthography adds that a dash between clauses "conveys a sharper juxtaposition of their content" than the more usual comma. A colon thus reads as бо, a dash as то, тому, і раптом or але, depending on content.

5.2 What the treebanks show

The punctuation directly before a parataxis clause that follows its head (2026-11-06):

Writing distributes the relations over the full set of marks; the parliamentary transcript, punctuated by stenographers, reduces almost everything to the comma (1,163 of about 1,430), so the relation that the speaker's intonation carried is lost in the written record. A parser or a relation classifier trained on IU and run on transcripts sees colons and dashes that are not there.

6. Parataxis in the folk language

Smerechynskyi (1932, smerechynskyi-1932-narysy-z-ukrainskoi-syntaksy, chs. XXVIII–XXIX) argued that parataxis — clauses and words placed side by side, with or without a conjunction — is "the greatest and most characteristic feature of the Ukrainian folk language", running through it "from the connection of single words to the connection of whole sentences". He described asyndeton as "a progressive fact" in the development of the language: the omission of the conjunctions that convey logical shades, and their replacement by intonation, stress and pauses, gives concision and force and, in words he quotes from Bulakhovskyi, "provokes heightened activity of thought, which itself joins separate sentences and thoughts with logical bridges". He classified the asyndetic complex sentences of folk narrative and song by the relation they express:

He set the paratactic style against the hypotactic style of the bookish language, in which "the thought is broken up among many shades of dependence, mostly artificial and formal", and listed the "high positive features" of parataxis: imagery, concision, energy, economy of attention, naturalness. The modern reader should take the evaluation as a programme of the 1920s purist and stylistic school; the description holds: folk narrative and song prefer juxtaposition, and the reader of them must infer most relations.

Simovych (1919, § 360) gives the same in school-grammar terms: clauses of a compound sentence "are each independent, but by thought, by content they connect", and the relation — conjunctive, contrastive, causal, consequential — is recognised either from the content alone (Тихесенько вітер віє, степи, лани мріють, Shevchenko) or from the conjunction (Умер співець, але живуть у серця слова безсмертні).

7. Time, aspect and the order of telling

Aspect carries much of the temporal structure that English marks with connectives and tenses:

A chain of perfective past verbs without connectives (§ 6) is read as a sequence of events in the order of telling; a reordering needs a marker (перед тим, до того, раніше, the pluperfect-like був пішов of older and dialect texts).

Mixing tenses in narration. Smerechynskyi (1932, ch. XIV) showed that folk narrative combines past, present and future forms within one sentence of a story in the past: Як коли прийшли туди, кладе хліб на стіл 'When they came there, (he) puts the bread on the table'; Внесла дівка рушник на тарілці, підходить до батька 'The girl brought in a towel on a plate, goes up to her father'. The present makes the event vivid, "as if someone looks through binoculars and tells the others what he sees". For a reader the tense switch does not signal a new time; the relation is still succession.

8. Correlatives and doubled marking

Ukrainian often marks one relation twice, in both clauses — a pattern English mostly avoids (Although …, but … is ungrammatical in English):

The correlative то after a conditional or temporal clause is frequent in speech: ParlaMint has 86 то as mark, IU 22 as cc. Smerechynskyi (1932, ch. XXV) gives a chapter to the folk correlatives так — як, як — так, хоч — хоч, що — що, чи — чи.

9. Attribution

Reported speech and thought are discourse relations too (PDTB Attribution): the content is someone's. Ukrainian marks them with:

• a verb of saying or thinking and що, ніби, нібито, мовляв: Він сказав, що прийде; Він відмовився: мовляв, зайнятий 'He refused — said he was busy' (мовляв 'as he put it', distancing);
• parenthetical каже, кажуть, за словами …, як повідомляє …, на думку …: Злість, кажуть, сатані сестриця (Kotliarevskyi, quoted by Simovych 1919, § 347) 'Malice, they say, is Satan's sister';
• the particles of reported or doubted content нібито, начебто 'allegedly' (news uses them to distance itself from a claim);
• in folk speech, direct speech after що without a shift of person (chapter 35, § 7; Smerechynskyi 1932, ch. XVII).

In the Leipzig news sample мовляв occurs 551 times (39 per million) and за словами is a standard news formula; in Brown-UK мовляв has 78 per million.

10. Varieties and styles as deltas

Each variety or style is listed only by what it changes in the discourse marking of the literary standard.

11. Quick reference: discourse relations in UD

UD annotates the syntactic attachment of connectives, not the relation itself:

None of the Ukrainian works in the research library describes a corpus with discourse-relation annotation (PDTB- or RST-style), and Mova holds none (2026-11-06).

12. How Mova handles it

Mova's discourse relations are links between frames — cause, next, quote, the conditional flag — set by the frame reader from English cues (because, so, then, after, if) on trees of the English parser (prag/src/frames.rs, set_cause, set_next; Pragmatics: frames — who did what to whom, why, and what came next, Links); the story layer and the QUD stack (world/src/story.rs, world/src/qud.rs) work on those frames. None of this reads Ukrainian: there is no Ukrainian parser model in mova-parse, langseg has a Latin-script table only, and the cue lists are English.

The Ukrainian fact extractor kg (kg/src/uk.rs, kg — the Mova knowledge graph) takes simple sentences only. Its conjunction list CONJ contains most clause linkers — і, й, та, а, але, або, чи, що, щоб, коли, якщо, бо, ніж, ані, зате, проте, адже, хоча, хоч, поки, доки, як, наче, ніби, мов, неначе, тому, аніж, and the interrogative-relative words — and any sentence containing one is rejected. It also rejects any comma and any dash in a sentence with a verb ("a dash alongside a verb means two parts"). What passes are single clauses with a connective adverb that is not on the list, which becomes an adverbial or particle member without a relation.

Measured on 2026-11-06 (kg parse, CPU, worktree commit ac9a23149, VESUM, empty name list):

12.1 Not yet

• No Ukrainian discourse relations of any kind: no cause, next or quote links between Ukrainian frames.
• No reading of connective adverbs and parenthetical words as relations; fixed connectives (до того ж, крім того, таким чином) are not treated as units.
• No reading of the colon and the dash as relation markers; no complex sentences.
• No Ukrainian discourse-annotated corpus for measurement.

13. Sources

• Smerechynskyi, Serhii. 1932. Нариси з української синтакси (у зв'язку з фразеологією та стилістикою). Ch. XIV (mixing tenses in narration), ch. XVII (indirect speech), ch. XXV (correlative pairs), ch. XXVIII (asyndeton; B: asyndeton of clauses, types 1–15), ch. XXIX (parataxis and its "high positive features"). smerechynskyi-1932-narysy-z-ukrainskoi-syntaksy (public domain in UA/EU)
• Simovych, Vasyl. 1919. Граматика української мови для самонавчання та в допомогу шкільній науці. §§ 347 (parenthetical words), 360 (compound sentences by content), 374 with note (causal clauses; так як, позаяк). simovych-1919-hramatyka-ukrainskoi-movy-dlia (public domain)
• Український правопис. 2019. § 43 (compound conjunctions written as one word), § 158 item 11 with notes 1–2 (parenthetical connectives; проте, однак initial vs medial), § 159 (semicolon), § 160 item 1 (colon in asyndetic sentences: content, cause), § 161 II item 1 (dash in asyndetic sentences: time, condition, consequence, concession). ukrainskyi-pravopys-2019 (official text, not under copyright)
• Universal Dependencies 2.18: UD_Ukrainian-IU (CC BY-NC-SA 4.0; counts only), UD_Ukrainian-ParlaMint (CC BY-SA 4.0); counted 2026-11-06, all splits.
• Leipzig Corpora Collection, ukr_news_2024 (1 M sentences; Goldhahn, Eckart & Quasthoff 2012; CC BY), leipzig-news-2024; Brown-UK (Starko & Rysin; CC BY-NC-SA 4.0; counts only), brown-uk; counted with grep -oiw 2026-11-06.
• Shevchenko, Kotliarevskyi, Nechui-Levytskyi, Franko and folk texts, as quoted in the grammars above: public domain.
• Mova: kg/src/uk.rs (CONJ, the punctuation gate in fn parse), kg — the Mova knowledge graph, prag/src/frames.rs, Pragmatics: frames — who did what to whom, why, and what came next, world/src/story.rs, world/src/qud.rs, mova-parse.

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/languages/uk/36-discourse-relations/