🗣️ 29. Ellipsis, word order and information structure

Ukrainian word order is free in grammar and bound in meaning. Case endings and agreement say who did what to whom, so Брат читає книжку (Brat chytaie knyzhku), Книжку читає брат and Читає брат книжку all mean 'the brother is reading the book'; what the order says is which part is given (the theme, what the sentence is about) and which is new (the rheme, what is said about it), and the neutral sentence stress falls on the last word. Grammars describe a normal order (subject – verb – object – adverbials) from which a speaker departs to connect with the context or to emphasise; narrative and presentational sentences put the verb first; and folk poetry and song keep orders that prose has given up. The same freedom lets Ukrainian leave out what the context supplies: subjects, objects, verbs, whole predicates. Present-tense бути 'be' is normally absent, and the 2019 orthography regulates the dash that marks a zero copula or an omitted member. This chapter describes the orders and their meanings, focus and topic devices, the zero copula, the kinds of ellipsis, the delta of song and folk poetry over prose, with treebank counts, and what Mova does with them.

Contents:

1. Free order, fixed meaning
2. The normal order
3. Theme, rheme and sentence stress
4. Postverbal subjects and preverbal objects: counts
5. Verb-first sentences
6. Topics, dislocation and the resumptive pronoun
7. Focus: stress, particles and the це-cleft
8. Order inside the clause: clitics, particles, short before long
9. Order inside the noun phrase
10. The zero copula and the dash
11. Ellipsis: subjects and objects
12. Ellipsis of the verb: gapping, answers, verbless predicates
13. Asyndeton
14. Song and folk poetry: a delta over prose
15. Counts in the treebanks
16. Quick reference: in UD
17. How Mova handles it
18. Sources

Related chapters: the clause and its members in 21. Clause structure; phrases in 22. Phrases; questions in 26. Questions; voice as an information-structure choice in 28. Voice and the passive; subject drop and reference in 35. Reference and coreference; register and song in 37. Register and variation; discourse particles in 38. Dialogue and discourse markers; intonation in 3. Syllable, stress and intonation.

1. Free order, fixed meaning

1.1 The six orders of a transitive clause

With a subject, a verb and an object that distinguish nominative from accusative, all six orders are grammatical. Each answers a different implicit question:

The two most common are SVO and OVS, and their difference corresponds to English active versus passive or a cleft: Книжку читає брат is 'The book is being read by the brother' or 'It's the brother who is reading the book', with no passive and no cleft needed. Smal-Stockyj (1913, § 62) puts it for German readers: "the word order is free, but not indifferent".

1.2 When case does not decide

When subject and object have the same form (neuter nouns, many plurals, indeclinable nouns: Мати любить дитя; Колесо зачепило дерево), the default reading is SVO: the first noun is the subject. Agreement can decide where case cannot: Дитя любить мати (the verb agrees with either, but мати is the expected agent) is ambiguous out of context; Колеса зачепили дерево is decided by the plural verb. Speakers avoid the ambiguous OVS in writing.

2. The normal order

Smal-Stockyj (1913, § 62) states the rules of the "normal" (unmarked) order:

1. Question words, relative words and conjunctions open their clause.
2. The negation stands directly before the word it negates (chapter 27).
3. Prepositions precede their noun phrase; attributes stand next to their noun.
4. The subject with its attributes precedes the predicate; objects and adverbials follow it — "as in the modern languages of the West".
5. Of two objects, the dative comes before the accusative; of several objects or adverbials of one predicate, the shorter usually comes before the longer.
6. After a sentence-initial object or adverbial, the verb usually precedes the subject: В однім городі жив купець "In a certain town there lived a merchant".
7. Weak personal and reflexive pronouns attach to the first word of the clause — a conjunction, a question word — "roughly as in German": Як собі хто постелить, так ся виспить "As one makes one's bed, so one sleeps" — with the reflexive ся detached before the verb, as in the Galician standard of his time and in south-western speech today; the modern standard writes так і виспиться.

He adds that German habits do not apply: there is no obligatory inversion in questions and no verb-final order in subordinate clauses. Departures from the normal order are made "to emphasise a word by its unusual place, or to connect with what was said, by putting first the part of the new sentence that was already mentioned".

Rule 5 shows as a tendency in the treebanks, overridden by length and novelty. Among postverbal noun objects of one verb, the dative precedes the accusative in 33 and follows it in 32 cases in IU; in ParlaMint the accusative comes first in 88 of 98 cases (UD 2.18, all splits, 2026-11-06). The parliamentary figure is a formula of the chair, Надаю слово народному депутату … 'I give the floor to people's deputy …', in which the long and new dative (title and name) goes last: rule 5's second half and the theme–rheme principle together beat the case order.

3. Theme, rheme and sentence stress

3.1 Given before new

The unmarked Ukrainian sentence moves from what is known to what is new. In a text this produces chains: the rheme of one sentence becomes the theme of the next, or a constant theme runs through several.

У селі жив старий коваль. Коваль мав трьох синів. Найстаршого звали Іваном. 'In the village there lived an old smith. The smith had three sons. The eldest was called Ivan.'

The first sentence introduces the smith at the end (verb before subject, rule 6); the second takes him as theme; the third takes найстаршого (from трьох синів) as theme and puts the name last.

3.2 Neutral and emphatic stress

In the neutral order the main sentence stress falls on the last word of the clause, which is the rheme. An order can be kept and the stress moved to another word for contrast or emphasis: Брат читає книжку with the stress on брат means 'It's the brother (not the sister) who's reading the book'. Emphatic stress is the spoken counterpart of the OVS order; writing, which has no stress marks, prefers to move the word. Intonation is described in chapter 3.

3.3 Articles by word order

Ukrainian has no articles, and order does part of their work: a noun at the end, as rheme, tends to be new and indefinite (У кімнату ввійшла жінка 'A woman entered the room'); a noun at the beginning, as theme, tends to be given and definite (Жінка ввійшла в кімнату 'The woman entered the room'). Demonstratives (цей, той) and один, якийсь make the distinction explicit when needed.

4. Postverbal subjects and preverbal objects: counts

Positions of nominal subjects (nsubj of a VERB) and direct objects (obj of a VERB), UD 2.18, all splits, 2026-11-06:

Pronoun subjects are almost always preverbal (they are themes by nature), so the noun subjects are the informative figure: one third of noun subjects follow the verb in the written treebank, and more than two fifths in parliamentary speech. The spoken figure is raised by presentational and existential sentences of the chair and the speakers (Слово має…, Надійшла пропозиція…, Виступає…), all with the new participant last. Preverbal objects are less frequent and are themes or contrastive foci (Цей закон ми підтримаємо 'This law we will support').

5. Verb-first sentences

The verb comes first in several kinds of sentence:

Smal-Stockyj (1913, § 62) singles out verb-first order in narrative: Шуміла вода по камінню, цвірінькали горобці в лозах, а в серці бідного Миколи клубилося горе "The water roared over the stones, the sparrows chirped in the willows, and grief gathered in poor Mykola's heart" (Ukrainian text read from the OCR of the 1913 edition, spelling modernised) — two verb-first clauses describing the scene, and a third with an adverbial first and the subject last.

6. Topics, dislocation and the resumptive pronoun

6.1 Fronting

Any constituent can be moved to the front as a topic without a pronoun copy: Цю книжку я вже читав 'This book I've already read'; Про це ми поговоримо завтра 'We'll talk about that tomorrow'; Грошей у мене немає 'Money, I don't have'. This is the normal way to make a non-subject the theme; it is neutral in all styles.

6.2 Dislocation with a pronoun

Speech also uses left dislocation: a noun phrase is set off at the start and resumed by a pronoun in the clause: Такі закони, вони є не, знаєте, неґрунтовні… 'Such laws, they are, you know, ill-founded…' (ParlaMint-UA_2022-01-25-m0.u262.p1.lang1.s6); І, я думаю, ці результати вони сьогодні обертаються до нас 'And, I think, these results, they are turning back to us today' (ParlaMint-UA_2004-11-27-m0.u23.p26.lang1.s4). And right dislocation, the afterthought: Коли воно закінчиться, це слідство? 'When will it end, this investigation?' (ParlaMint-UA_2003-10-14-m0.u47.p2.lang1.s11); Ми робимо вибіркові перевірки, наш комітет 'We do spot checks, our committee' (ParlaMint-UA_2003-10-14-m0.u139.p2.lang1.s2).

Dislocation is a feature of unplanned speech; edited prose avoids it. UD marks the dislocated element as dislocated: 19 in IU against 49 in ParlaMint (all splits, 2026-11-06), despite the similar sizes of the two treebanks.

6.3 А as a topic shifter

The conjunction а introduces a contrastive or new topic: Я працюю, а він відпочиває 'I work, and he rests'; А тепер про бюджет 'And now about the budget' (chapter 38).

7. Focus: stress, particles and the це-cleft

7.1 Focus particles

The particle stands directly before the focused word (тільки він, навіть діти), or, for ж, же, -то, таки, after it. The orthography writes -то, -таки with a hyphen when they emphasise one word and apart when another particle intervenes (2019, § 44.3).

7.2 The це-cleft

Ukrainian marks an exhaustive focus by putting це 'this, it' before the focused constituent at the start of the clause: Це я тобі дзвонив 'It was me who called you'; Це ви порушили нашу домовленість 'It was you who broke our agreement'. Це here does not agree and is not the subject (the verb agrees with я, ви); it is a focus particle. The construction is frequent in speech and accepted in writing. UD marks such це as expl in some sentences and as a subject of a copular structure in others: the expl relation has це as its commonest lemma (89 of 116 in IU, 131 of 135 in ParlaMint, all splits, 2026-11-06), although not every expl це is a cleft (it also introduces a predicate after a dash: Поезія — це завжди неповторність, §10).

7.3 Contrastive negation

The corrective не X, а Y focuses X and Y (chapter 27, §2): Не я, а він.

8. Order inside the clause: clitics, particles, short before long

• Weak pronouns tend to the second position, after the first stressed word or a conjunction: Я тобі казав rather than Я казав тобі when тобі is not in focus; Дай мені його 'Give it to me'. The tendency is weaker than in the South and West Slavic languages; a pronoun at the end is stressed and focused (Я казав тобі! 'I told you!').
• The conditional particle б/би follows the first word or the verb: Я б пішов, Пішов би я, Якби я знав.
• Delta, south-western speech: the reflexive ся is a mobile clitic that can stand before the verb, after the first word of the clause (як ся маєш?, так ся виспить), where the standard has the postfix -ся fixed to the verb (як маєшся? is not used; the standard greeting is як справи? or як ся маєш? as a fixed phrase). The 1913 Galician grammar of Smal-Stockyj still writes it this way (§2, rule 7).
• The particles ж/же, бо (in the emphatic sense 'come on'), ну, от cluster at the start (chapter 38).
• Short before long (Smal-Stockyj § 62.2): a pronoun or a single word precedes a long phrase; long, heavy constituents and clauses go to the end, where they are also rhematic.
• Adverbials of time and place often open the sentence as scene-setting themes (Учора в Києві відбулася нарада), with the verb before the subject (rule 6).

9. Order inside the noun phrase

The treebank counts show how fixed the adjective position is in prose and speech alike: amod precedes its head in 9,282 and follows it in 464 cases in IU (5%), 8,806 and 438 in ParlaMint (5%) (all splits, 2026-11-06). The postposed adjective is therefore a strong marker of style — poetry, song, address, terminology — not a free variant.

The approximative inversion of numeral and noun (років п'ять, хвилин за десять 'in about ten minutes') is grammaticalised: the order itself means 'about'.

10. The zero copula and the dash

10.1 The present tense

The present tense of бути is normally not expressed in nominal predicates: Мій батько — інженер 'My father is an engineer'; Київ — столиця України; Вона вчителька; Мені холодно. The form є is used for existence and possession (У мене є брат 'I have a brother'; Є питання 'There is a question'), for emphasis (Він і є той лікар 'He is the very doctor'), in definitions and bookish style, and with an instrumental predicate in official language (Він є членом комісії 'He is a member of the committee'). The last construction is frequent: with є as copula and a noun predicate, the instrumental outnumbers the nominative 67 : 7 in IU and 56 : 27 in ParlaMint (UD 2.18, all splits, 2026-11-06). The purist syntax opposed the predicative instrumental (Smerechynskyi 1932, ch. IX); the modern norm accepts both cases after the past and future copula, where the instrumental also dominates (IU 122 : 56, ParlaMint 37 : 26 instrumental : nominative). See chapter 21.

10.2 The dash

The 2019 orthography (§ 161, I) regulates the dash in the simple sentence:

1. Between subject and nominal predicate, both nouns (or cardinal numerals) in the nominative, with a zero copula: Кит — ссавець; Київ — столиця України; П'ять і п'ять — десять; especially with a contrast: Мій брат — лікар, а сестра — учителька.
2. No dash, usually: when the predicate is emphasised less than usual (Мій батько інженер); when the subject is це (Це наша хата); when an adverbial or other member precedes the predicate (Діти завжди діти); when the predicate precedes the subject (Хороша-таки штука життя); when the subject is a personal pronoun (Я син простого лісоруба); before a negated noun predicate (Серце не камінь), unless the contrast не … а is stressed (Мій батько — не інженер, а майстер); before a comparison with як, мов, наче (Життя як казка), unless emphasised (Вода — як скло).
3. Between subject and predicate when one or both are infinitives: Життя прожити — не поле перейти.
4. Before це, оце, то, ось, значить introducing the predicate: Поезія — це завжди неповторність.
5. In place of an omitted member, mostly the predicate: На одній ниві пшениця, на другій — жито; Останні дні — знову в Парижі; and in headings and slogans: Книга — поштою; Мир — народам! The dash may be left out where no pause is wanted: Мотря вибігла з хати. Мелашка за нею.

The dash is thus the written sign of a pause in the intonation, and the rules list where a pause is expected. In the treebanks, of nominal predicates (noun, adjective, pronoun or proper noun heads with an nsubj), 427 have a copula and 751 none in IU, 223 and 778 in ParlaMint; a dash before the predicate appears in 170 of the IU and 90 of the ParlaMint zero-copula clauses (all splits, 2026-11-06). The lower share in ParlaMint reflects transcription practice as much as speech.

11. Ellipsis: subjects and objects

11.1 Subjects

A subject pronoun can be omitted when the verb ending identifies it and the referent is clear: Прийду завтра 'I'll come tomorrow'; Знаєш, що сталося? 'Do you know what happened?'. In the present and future the ending carries person and number; in the past it carries gender and number only (прийшов, прийшла, прийшли), so first and second person subjects are kept more often in the past (Я прийшов, Ти прийшов), and context does the rest. The details of zero subjects and their antecedents, including the south-western dialects that drop past-tense subjects more and mark person on the verb by clitics, are in chapter 35. In UD a dropped subject leaves no node.

11.2 Objects

An object given in the context is often omitted, especially in answers and with verbs of perception, knowledge and giving: Бачив? — Бачив. 'Did you see (it)? — I did'; Дай! 'Give (it)!'; Знаю 'I know (it)'. English needs it; Ukrainian does not. The omitted object leaves no node in UD either.

12. Ellipsis of the verb: gapping, answers, verbless predicates

12.1 Gapping and coordination

In coordinated clauses the repeated verb is omitted in the second, and the gap is marked in writing by a dash where there is a pause: Я пішов додому, а він — у кіно 'I went home and he to the cinema'; На одній ниві пшениця, на другій — жито (2019 orthography, § 161). A short spoken example with the predicate omitted: А ми повинні знати, хто на нашій стороні зараз, а хто ні! 'And we must know who is on our side now and who isn't!' (NSDC_UA_28_Feb2014-152).

12.2 Fragments and answers

Answers repeat only the focus: Коли ти прийдеш? — Завтра.; Хто це зробив? — Я. Vote results in the Rada are fragments of a fixed form: За — 58 'In favour: 58' (ParlaMint-UA_2022-01-25-m0.u124.p4.lang1.s1), with the verb проголосувало understood.

12.3 Verbless predicates of motion and speech

Speech omits verbs of motion and saying when the adverbial or the quotation is enough:

• Я — додому. 'I'm off home.' Ми в магазин. 'We're going to the shop.' Ти куди? 'Where are you off to?'
• А він мені: «Іди звідси!» 'And he (says) to me: "Get out!"'
• Тепер — до справи. 'Now, to business.'
• Що ж, до зустрічі. 'Well, till we meet.'

The type with a motion adverbial is a regular construction, not a slip: the direction is enough to imply the verb.

12.4 In UD

UD represents gapping by promoting one remnant to head and attaching the others as orphan. The relation is much more frequent in speech: 256 orphan in IU, 582 in ParlaMint (all splits, 2026-11-06). In ParlaMint the promoted heads are mostly nouns (269), proper nouns (87) and prepositions (87, from the vote formula За — 58 and its variants), with verbs (51) and adjectives (31) after them.

13. Asyndeton

Ukrainian joins clauses and words without conjunctions more freely than English, and Smerechynskyi (1932, ch. XXVIII) devotes a chapter to the "stylistic weight" of asyndeton in the folk language: nouns (хліб-сіль, батько-мати), adjectives (добрий-предобрий), verbs (сидить-думає, ходить-блукає), and clauses (Прийшов, побачив, переміг). The verbal type, two verbs in sequence for one event or a duration (сидить-думає 'sits and thinks'), is written with a hyphen and is common in folk tales and colloquial narrative. Clausal asyndeton without a dash or with a colon expresses cause, condition or result: Не прийде — не треба 'If he won't come, never mind'; Сонце зійшло — стало тепло. These are parataxis in UD and are discussed with coordination in chapter 24.

14. Song and folk poetry: a delta over prose

Folk song and the poetry that follows it (Shevchenko and his successors) keep orders and ellipses that prose has narrowed or lost. As with the other styles of this manual, they are best described as a delta over the literary standard: the grammar is the same, and a small set of rules is switched on or off. The examples below are from public-domain folk songs (as cited by Smerechynskyi 1932) and from Shevchenko (d. 1861); no copyrighted lyrics are quoted.

Two properties of the song style matter for any parser or measurement. First, the unit is the line: a line is often a clause or a phrase, and the parallel lines of a stanza share material that only one of them states (the ellipsis row). A sentence segmenter trained on prose punctuation has no cue to follow. Second, the rhyme and the metre choose among grammatical variants: postposed adjectives, uncontracted endings and verb-final -но/-то forms are all grammatical options in the language that the song selects much more often; they are not errors and they are not archaisms to the listener, who learned them in childhood. This is the sense in which song is its own style of speech with its own grammar (chapter 37); surzhyk and dialect forms in popular song are discussed there.

15. Counts in the treebanks

All counts: UD 2.18, Ukrainian IU (CC BY-NC-SA, counts only) and ParlaMint (CC BY-SA), all splits, syntactic words, awk over the CoNLL-U files, 2026-11-06.

16. Quick reference: in UD

17. How Mova handles it

17.1 Roles independent of order

The Ukrainian fact extractor (kg/src/uk.rs) assigns roles by case and agreement, not by position: the subject is the nominative that agrees with the predicate (fn agrees: person and number in the present and future, gender and number in the past), objects are chosen by case and government. Word order is used only to break ties: among several nominative candidates, the one before the predicate is taken (before.len() == 1). Measured with kg parse (CPU, worktree at commit ac9a23149, VESUM, empty proper-name list; 2026-11-06):

So Mova reads who did what to whom correctly across orders, and it does not misread OVS as SVO when the cases are distinct. What it discards is the information structure: the three sentences become one fact, and the theme–rheme difference that the order carried is lost. The fact keeps the word index of each member (Member.word), so the order could be recovered, but nothing reads it as theme and rheme.

The particle rule for це implements the cleft of §7.2: "Це ви порушили…: це alongside another candidate is a particle" (comment and code in fn verbal), added after the development gold standard graded this sentence wrong (examples/gold-v1.tsv, verdict bad:це-particle).

17.2 Zero copula and ellipsis

Verbless sentences go to fn nominal_sentence, which accepts four patterns: a dash after the first word (Київ — столиця України.), це first (Це кіт.), two words (Кіт білий.), or three or more words ending in a nominative noun with the first word a noun (Ти мій друг.). A dash together with a finite verb rejects the sentence (dash with verb), which excludes gapping by design. Measured (same run):

17.3 The generator chooses an order

The generator (kg/src/textgen.rs) orders the chunks of a fact (predicate, subject, objects, adverbials) by a Markov chain over chunk classes, trained on the parsed sentences, and finds the best permutation by dynamic programming over subsets. It therefore produces the most frequent order for a fact's classes (usually subject – verb – object), not the order of the source. The round-trip gate (kg — the Mova knowledge graph, "Generator") classifies each regenerated held-out sentence as verbatim, same words in a different order, or different words; the middle class is exactly the information-structure loss described above.

17.4 Not yet

• No theme–rheme or focus annotation, and no use of order, це, or focus particles to mark them in facts.
• No reading of dislocation, gapping, fragment answers, or verbless motion predicates.
• No generation of a requested order (for example, OVS for a given patient), which would be the Ukrainian counterpart of choosing a passive in English.
• No song-style segmentation by line or handling of postposed adjectives beyond what the agreement rules happen to catch.
• No Ukrainian dependency parser or UD baseline (no Ukrainian model in tools/mova-parse; langseg is Latin-only; no row in What every language gets, 2026-11-06); the pragmatic labels of prag/src/schema.rs are read for English only.

18. Sources

• Smal-Stockyj, Stephan. 1913. Ruthenische Grammatik. Berlin/Leipzig: Göschen. § 62 (word order: fixed positions; the normal order; inversion after an initial object or adverbial; weak pronouns; verb-first narrative; no inversion in questions, no verb-final subordinate clauses). stockyj-1913-ruthenische-grammatik (public domain).
• Smerechynskyi, Serhii. 1932. Нариси з української синтакси. Ch. IX (predicative instrumental and nominative), ch. XXIII (paratactic constructions and the folk-song examples quoted in §14), ch. XXVIII (asyndeton). smerechynskyi-1932-narysy-z-ukrainskoi-syntaksy (public domain in UA/EU).
• Український правопис. 2019. § 44.3 (hyphen with -то, -таки), § 161 (the dash in the simple sentence: zero copula, infinitives, це, omitted members, headings). ukrainskyi-pravopys-2019 (official document).
• Shevchenko, Taras. Причинна (1837) and Садок вишневий коло хати (1847): first lines quoted (public domain).
• Shvedova, Maria, Arsenii Lukashevskyi, and Andriy Rysin. 2025. Developing a Universal Dependencies Treebank for Ukrainian Parliamentary Speech. UNLP 2025. shvedova-2025-developing-universal-dependencies-treebank-ukrainian (CC BY 4.0): the parliamentary treebank as a record of spoken syntax.
• UD Ukrainian ParlaMint 2.18 (CC BY-SA 4.0), including the NSDC 2014 transcript: sentences cited by sent_id; counts. UD Ukrainian IU 2.18 (CC BY-NC-SA 4.0): counts only. All 2026-11-06.
• Mova: kg/src/uk.rs, kg/src/textgen.rs, kg — the Mova knowledge graph, kg/examples/gold-v1.tsv, prag/src/schema.rs, What every language gets.

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/languages/uk/29-ellipsis-word-order-information-structure/