Elamur: **Mova** · [Parse](https://mova.elamur.ai/free/en/parse/) · [Dictionary](https://md.dictionary.elamur.ai/free/en/) · [Translate](https://md.translate.elamur.ai/free/en/) · [Research](https://md.research.elamur.ai/free/en/) · [Groupants](https://md.groupants.elamur.ai/free/en/)

[Mova](https://md.mova.elamur.ai/free/en/) · [this page in HTML](https://mova.elamur.ai/free/en/docs/languages/uk/25-relative-clauses/)

[home](https://md.mova.elamur.ai/free/en/) / [docs](https://md.mova.elamur.ai/free/en/docs/) / [languages](https://mova.elamur.ai/free/en/docs/languages/) / [uk](https://md.mova.elamur.ai/free/en/docs/languages/uk/) / 25. Relative clauses

# 25. Relative clauses

**On this page**

1. [1. Overview: the relative words](#1-overview-the-relative-words)
2. [2. *Який*: agreement and case](#2-який-agreement-and-case)
3. [3. Prepositions, possessors and pied-piping](#3-prepositions-possessors-and-pied-piping)
4. [4. *Котрий*](#4-котрий)
5. [5. Invariable *що* and the resumptive pronoun](#5-invariable-що-and-the-resumptive-pronoun)
6. [6. *Чий*](#6-чий)
7. [7. Relative adverbs: *де*, *куди*, *звідки*, *коли*](#7-relative-adverbs-де-куди-звідки-коли)
8. [8. Free relatives and correlatives: *той, хто*; *те, що*](#8-free-relatives-and-correlatives-той-хто-те-що)
9. [9. The paratactic relative of folk speech](#9-the-paratactic-relative-of-folk-speech)
10. [10. Mood in relative clauses: *який би*](#10-mood-in-relative-clauses-який-би)
11. [11. Restrictive and non-restrictive clauses; punctuation](#11-restrictive-and-non-restrictive-clauses-punctuation)
12. [12. Participial phrases as relative alternatives](#12-participial-phrases-as-relative-alternatives)
13. [13. Counts in the treebanks](#13-counts-in-the-treebanks)
14. [14. Quick reference: relative clauses in UD](#14-quick-reference-relative-clauses-in-ud)
15. [15. How Mova handles it](#15-how-mova-handles-it)
16. [16. Sources](#16-sources)

A Ukrainian relative clause follows the noun it modifies, is set off by commas, and opens with a relative word. The common written relative is the inflected pronoun *який* (*yakyi*) 'which, who', which takes its gender and number from the antecedent and its case from its role inside the clause. Beside it stand the invariable particle *що* (*shcho*) 'that', which leaves the case to a resumptive personal pronoun when the gap is oblique, the older and regionally coloured *котрий* (*kotryi*), the possessive *чий* (*chyi*) 'whose', the adverbs *де*, *куди*, *звідки*, *коли* ('where, whither, whence, when'), and the free relatives *хто* and *що* with or without a demonstrative correlative (*той, хто…*; *те, що…*). Folk speech adds a paratactic way of relativising, by a coordinated clause with a personal pronoun, which the purist syntax of the 1920s–30s recommended in place of *який* and *котрий*. Participial phrases are the compact alternative: passive participles in *-ний*/*-тий* are common, active present participles in *-учий*/*-ючий* are restricted. This chapter describes the forms, their agreement and case, the choice between them in the treebanks, and what Mova does with them today.

Contents:

1. Overview: the relative words
2. *Який*: agreement and case
3. Prepositions, possessors and pied-piping
4. *Котрий*
5. Invariable *що* and the resumptive pronoun
6. *Чий*
7. Relative adverbs: *де*, *куди*, *звідки*, *коли*
8. Free relatives and correlatives: *той, хто*; *те, що*
9. The paratactic relative of folk speech
10. Mood in relative clauses: *який би*
11. Restrictive and non-restrictive clauses; punctuation
12. Participial phrases as relative alternatives
13. Counts in the treebanks
14. Quick reference: relative clauses in UD
15. How Mova handles it
16. Sources

Related chapters: the pronoun paradigms in [12. Pronouns, determiners and numerals](https://md.mova.elamur.ai/free/en/docs/languages/uk/12-pronouns-determiners-numerals/); participles in [13. Verbs](https://md.mova.elamur.ai/free/en/docs/languages/uk/13-verbs/) and [28. Voice and the passive](https://md.mova.elamur.ai/free/en/docs/languages/uk/28-voice-and-passive/); complement clauses with *що* (the conjunction) in [23. Complementation](https://md.mova.elamur.ai/free/en/docs/languages/uk/23-complementation/); other subordinate clauses in [24. Coordination and subordination](https://md.mova.elamur.ai/free/en/docs/languages/uk/24-coordination-subordination/); interrogative uses of the same words in [26. Questions](https://md.mova.elamur.ai/free/en/docs/languages/uk/26-questions/); register in [37. Register and variation](https://md.mova.elamur.ai/free/en/docs/languages/uk/37-register-and-variation/).

---

## 1. Overview: the relative words

| relative word | form | antecedent | what it encodes | example |
| --- | --- | --- | --- | --- |
| *який, яка, яке, які* | adjectival declension | any noun or pronoun | gender and number of the antecedent; case of its own role | *книжка, яку я читаю* 'the book (that) I am reading' |
| *котрий, котра, котре, котрі* | adjectival declension | any noun | same as *який* | *людина, котра прийшла* 'the person who came' |
| *що* | invariable | any noun; *той*, *те*, *все* | nothing; role shown by a resumptive pronoun if oblique | *хлопець, що з ним я говорив* 'the boy I talked with' |
| *чий, чия, чиє, чиї* | adjectival declension | a person (usually) | the possessor; agrees with the possessed noun | *поет, чиї вірші ми читали* 'the poet whose poems we read' |
| *де, куди, звідки, коли* | adverbs | places, times | location, goal, source, time | *місто, де я народився* 'the town where I was born' |
| *хто* | pronoun, declined | *той*, *кожен*, *всі*, none | an animate referent | *той, хто прийшов* 'the one who came' |
| *що* (pronoun) | pronoun, declined | *те*, *все*, *щось*, none | an inanimate referent | *те, чого ми боялися* 'what we were afraid of' |

The relative word stands at the beginning of its clause, after a preposition if one governs it (§3). Smal-Stockyj (1913, § 62) states the rule for the whole family: question words, relative words and conjunctions open their clause, and the clause otherwise keeps the free order of a main clause; the verb is not sent to the end as in German.

Two homonymies matter for parsing. *Що* is a relative particle (§5), a relative and interrogative pronoun (*що* 'what', declined *чого, чому, що, чим, чому*), and the complementiser 'that' of [chapter 23](https://md.mova.elamur.ai/free/en/docs/languages/uk/23-complementation/). *Який* is also an interrogative (*Який це будинок?* 'Which house is this?') and an exclamative (*Який гарний день!* 'What a fine day!'); *коли* is also 'if' in older and colloquial usage.

---

## 2. *Який*: agreement and case

### 2.1 The rule

*Який* (*yakyi*) declines like a hard adjective. It agrees with its antecedent in **gender and number** and takes the **case** that its function inside the relative clause requires:

| antecedent | role in the clause | form | example |
| --- | --- | --- | --- |
| *хлопець* m. sg. | subject | nom. *який* | *хлопець, який співає* 'the boy who sings' |
| *дівчина* f. sg. | direct object | acc. *яку* | *дівчина, яку ти бачив* 'the girl you saw' |
| *село* n. sg. | locative after *у* | loc. *якому* | *село, у якому я виріс* 'the village I grew up in' |
| *друзі* pl. | dative object | dat. *яким* | *друзі, яким я написав* 'the friends I wrote to' |
| *справа* f. sg. | object of a genitive verb | gen. *якої* | *справа, якої він боїться* 'the matter he is afraid of' |
| *ліс* m. sg. | instrumental | instr. *яким* | *ліс, яким ми йшли* 'the forest we walked through' |

The antecedent's case plays no role: *Я бачу хлопця* (gen.-acc.), *який співає* (nom.). The relative is a pronoun that refers back; it is not an adjective attached to the noun.

### 2.2 Animacy in the accusative

In the accusative singular masculine and in the accusative plural, *який* follows the animacy of its antecedent, as adjectives do (see [11. Nouns](https://md.mova.elamur.ai/free/en/docs/languages/uk/11-nouns/)):

- *будинок, який я купив* 'the house I bought' (inanimate: acc. = nom.);
- *брат, якого я зустрів* 'the brother I met' (animate: acc. = gen.);
- *листи, які я отримав* 'the letters I received'; *студенти, яких я навчаю* 'the students I teach'.

### 2.3 Agreement with special antecedents

- **A coordinated antecedent** takes the plural: *батько й мати, які нас чекали* 'father and mother, who were waiting for us'.
- **A polite *Ви*** referring to one person: the relative is plural, as is the verb of *Ви*: *Ви, які стільки зробили…* 'You, who have done so much…' (the pattern is that of polite plural agreement; see [32. Politeness and hedging](https://md.mova.elamur.ai/free/en/docs/languages/uk/32-politeness-and-hedging/)).
- **A noun of common gender** or a feminine role noun referring to a woman: the relative follows the referent, as the predicate does: *моя колега, яка…*; *лікарка, яка…*.
- **Indeclinable nouns** take the gender assigned by the grammar: *кафе, яке…* (neuter), *таксі, яке…* (neuter).

### 2.4 Linear order inside the clause

After the relative, the clause has free order. The subject often follows the verb when it is the new information of the clause: *книжка, яку написав мій дід* 'the book my grandfather wrote' is as natural as *книжка, яку мій дід написав*; the first puts *мій дід* in the rhematic end position ([chapter 29](https://md.mova.elamur.ai/free/en/docs/languages/uk/29-ellipsis-word-order-information-structure/)).

---

## 3. Prepositions, possessors and pied-piping

### 3.1 Prepositions go with the relative

A preposition always precedes its relative; there is no stranding of the English kind (*the man I talked to*):

- *людина, з якою я працюю* 'the person I work with';
- *питання, про яке ми говорили* 'the question we talked about';
- *стіл, на якому лежить книжка* 'the table on which the book lies';
- *причина, через яку він пішов* 'the reason why he left'.

The locative of *який* after *у/в* and *на* is *якому/якім* (m., n.), *якій* (f.), *яких* (pl.): *у якому* and the shorter *у якім* are both standard; *якім* is the older, more literary form.

### 3.2 The possessive genitive

When the relative is the possessor of a noun inside the clause, the genitive *якого/якої/яких* follows that noun, and the whole noun phrase moves to the front of the clause:

- *письменник, книжки якого я читав* 'the writer whose books I read';
- *країна, столиця якої — Київ* 'the country whose capital is Kyiv';
- *будинок, на даху якого гніздяться лелеки* 'the house on whose roof storks nest'.

The genitive of *який* may not precede its head (*✗ якого книжки*). The alternative is the possessive relative *чий* (§6), which does precede: *письменник, чиї книжки я читав*.

### 3.3 Relatives inside infinitives and complements

Ukrainian allows the relative to belong to an embedded infinitive: *книжка, яку я хочу прочитати* 'the book I want to read'; *людина, з якою варто поговорити* 'a person it's worth talking to'. Extraction out of a finite *що*-clause is avoided in writing (*✗ книжка, яку я думаю, що він прочитав*); speech repairs it with a resumptive or by rephrasing (*книжка, яку, я думаю, він прочитав*, with *я думаю* as a parenthetical).

---

## 4. *Котрий*

*Котрий* (*kotryi*) declines and agrees like *який*. As a relative it competes with *який* in the same positions: *людина, котра прийшла*; *будинок, у котрому я живу*. It also has meanings *який* lacks: 'which (of several)', ordinal 'which number' (*Котра година?* 'What time is it?'; *котрий раз* 'for the umpteenth time'), and an indefinite in *котрийсь* 'some one or other', *декотрий* 'some'.

The modern written language prefers *який*; *котрий* is noticeably more frequent in western, especially Galician, usage and in older literary texts, and in speech it is a stylistic variant rather than a marked form. Counts (relative uses, PronType=Rel, all splits, UD 2.18, 2026-11-06): *котрий* 32 against *який* 625 in IU, 4 against 980 in ParlaMint. In the 2024 Leipzig news sample (1 million sentences), all forms of *котрий* occur 747 times against 131,417 forms of *який* (all uses, not only relative), so *котрий* is about half a percent of the pair in news.

A ParlaMint example (2022) where a speaker uses it: *програма термодернізації, котра є на кілька мільярдів гривень* 'a thermal-modernisation programme that amounts to several billion hryvnias' (ParlaMint-UA_2022-01-25-m0.u234.p3.lang1.s3).

### 4.1 The purist view and the codified norm

Smerechynskyi (1932, ch. X) argued that both *котрий* and *який* used as pure relative connectors were foreign to the folk language (he relates them to Polish *który, jaki* and Russian *который, какой*), that the folk language's own relative was the particle *що* (§5) or a paratactic construction (§9), and he cites Syniavskyi's remark that the "purity" of a writer could be measured by how rarely he used them. He also notes that the literary language of his time already used them widely. The modern standard has followed usage: *який* is the unmarked written relative, recommended in school grammar, and *що* is its stylistic alternative. This is a clear case where the codified norm caught up with how people write, and the purist claim survives as a stylistic preference for *що* in narrative and in poetry.

---

## 5. Invariable *що* and the resumptive pronoun

### 5.1 The construction

*Що* (*shcho*) as a relative does not inflect and says nothing about the antecedent. When the gap it leaves is a **subject** or an **inanimate direct object**, nothing else is needed:

- *люди, що прийшли* 'the people who came' (subject);
- *лист, що я написав* 'the letter (that) I wrote' (object, inanimate).

When the gap is **oblique** (a prepositional phrase, a dative, an instrumental, a genitive object or possessor) or an **animate object**, a personal pronoun in the required case marks it inside the clause:

| role of the gap | with *що* + resumptive | with *який* |
| --- | --- | --- |
| animate direct object | *дівчина, що її я бачив* | *дівчина, яку я бачив* |
| prepositional object | *хлопець, що з ним я говорив* | *хлопець, з яким я говорив* |
| dative | *друг, що йому я писав* | *друг, якому я писав* |
| locative | *хата, що в ній я виріс* | *хата, в якій я виріс* |
| possessor | *сусід, що в нього хата біла* | *сусід, у якого хата біла* |

The resumptive is a clitic-like weak pronoun, typically placed right after *що* or before the verb (*що його*, *що з ним*). It is the only marker of case and of gender: *село, що в ньому* (neuter) versus *вулиця, що на ній* (feminine).

### 5.2 Style and history

Smerechynskyi (1932, ch. XXIII, II A–B) presents *що* as the "purely relative particle" of the folk language and gives many folk and literary examples, including *що* for place and time (*хата, що в ній*, *день, що того дня*). Grammars of his period (Smal-Stockyj 1913, Simovych 1919) list *що* beside *який/котрий* without restriction. Today *що* with a nominative or inanimate accusative gap is neutral in all styles, frequent in fiction and folk song, and common in official formulas (*особи, що перебувають…* 'persons who are…'). With a resumptive pronoun it is literary-poetic or colloquial and is rare in official prose, which prefers *який*.

### 5.3 How the treebanks annotate it

Both UD Ukrainian treebanks tag relative *що* as **SCONJ** with relation **mark** inside the `acl:relcl` clause, even when it fills the subject gap; the gap is not represented. So *депутатів, що записалися* 'the deputies who have signed up' has `mark(записалися, що)` and no `nsubj`. ParlaMint has the formula in every session: *На табло прізвища народних депутатів, що записалися* 'On the board, the names of the deputies who have signed up' (ParlaMint-UA_2003-10-14-m0.u71.p1.lang1.s3). Where *що* is a declined pronoun (*те, що відбувається* in some sentences, *те, чого боялися*), it is PRON with PronType=Rel and a core relation; the two treebanks are not fully consistent in choosing between the analyses for nominative *що* after *те*.

Counts (UD 2.18, all splits, 2026-11-06; `acl:relcl` clauses whose head has a direct dependent *що* with relation `mark`):

| treebank | `acl:relcl` | with *що* as `mark` | of these, with a 3rd-person personal pronoun among the clause head's dependents |
| --- | --- | --- | --- |
| IU (fiction, news, law, web…) | 1,156 | 415 (36%) | 47 |
| ParlaMint (parliamentary speech) | 1,170 | 153 (13%) | 3 |

The pronoun count is an upper bound on resumptives (a 3rd-person pronoun dependent may be an ordinary argument). The contrast is a genre effect: IU contains much fiction, where *що* is a narrative choice, while the parliamentary transcripts, though spoken, use the written-official *який*.

### 5.4 *Що* in fixed official formulas

The pattern *визнати X таким, що…* 'declare X to be such that…' is common in legal and parliamentary language: *визнати вибори Президента такими, що не відповідають* 'to declare the presidential election as not complying' (ParlaMint-UA_2004-11-27-m0.u99.p2.lang1.s3). It replaces an active participle (*невідповідними*, or the restricted *✗ невідповідаючими*), see §12.

---

## 6. *Чий*

*Чий* (*chyi*) 'whose' declines like a soft possessive adjective (*чий, чия, чиє, чиї*; gen. *чийого, чиєї*) and **agrees with the possessed noun**, not with the possessor: *поет, чия збірка вийшла* 'the poet whose collection came out' (*чия* f. with *збірка*); *мати, чий син загинув* 'the mother whose son died' (*чий* m. with *син*). It precedes its noun. It is literary and less frequent than *якого/якої* + noun (§3.2): the treebanks have 3 relative *чий* in IU, none counted in ParlaMint. The sentence-level uses of *чий* are interrogative (*Чия це книжка?* 'Whose book is this?') and in proverbs (*Чий хліб їси, того й пісню співай* 'Whose bread you eat, his song you sing', a correlative free relative, §8).

---

## 7. Relative adverbs: *де*, *куди*, *звідки*, *коли*

A place or time antecedent can be relativised by an adverb instead of a preposition and *який*:

| adverb | meaning | example | equivalent with *який* |
| --- | --- | --- | --- |
| *де* | where (location) | *місто, де я народився* | *місто, у якому я народився* |
| *куди* | whither (goal) | *село, куди ми їдемо* | *село, до якого ми їдемо* |
| *звідки* | whence (source) | *край, звідки він родом* | *край, з якого він родом* |
| *коли* | when | *день, коли все почалося* | *день, у який все почалося* |
| *поки*, *доки* | while, as long as | *час, поки ми чекали* | — |

*Де* is also used for abstract "places" — situations, documents, events: *рішення судів, де окремі акти визнані актами тероризму* 'court decisions in which some acts are recognised as acts of terrorism' (ParlaMint-UA_2022-01-25-m0.u115.p3.lang1.s2). This extended *де* is colloquial and is criticised in style guides, which prefer *у яких*; in the transcripts it is frequent. Smerechynskyi (1932, ch. XXIII, II В) regarded the relative adverbs as less natural in folk speech than *що* with a demonstrative adverb (*там, що…*; *тоді, що…*), but in the modern language they are standard.

Counts of relative adverbs as direct dependents of `acl:relcl` heads (UD 2.18, 2026-11-06): IU *де* 67, *коли* 21, *куди* 2, *звідки* 1, *поки* 1; ParlaMint *де* 35, *коли* 22, *поки* 7.

---

## 8. Free relatives and correlatives: *той, хто*; *те, що*

### 8.1 Headed by a demonstrative

The commonest "free" relatives have a demonstrative head *той* (m.), *та* (f.), *те* (n.), *ті* (pl.) or a quantifier *кожен*, *всі*, *все*, *усе*, *щось*, and the relative pronoun *хто* (animate) or *що* (inanimate):

- *той, хто прийшов перший* 'the one who came first';
- *усі, хто хоче* 'everyone who wants';
- *те, що ти сказав* 'what you said';
- *все, чого ми навчилися* 'all that we have learned' (*чого*: genitive governed by *навчитися*).

The demonstrative carries the case of the matrix role, the relative the case of the embedded role: *Я вдячний тим, хто допоміг* (*тим* dat. of *вдячний*; *хто* nom.); *Розкажи про те, що бачив* (*те* acc. after *про*; *що* acc.).

Agreement with *хто*: the verb after *хто* is singular (masculine in the past) even when *хто* refers to a plural set: *ті, хто прийшов*. Plural agreement (*ті, хто прийшли*) is common in speech and is accepted as a variant. The treebanks show both, with the singular dominant: in `acl:relcl` clauses with *хто* as subject under a plural head, the verb is singular in 13 and plural in 2 cases (IU), singular in 17 and plural in 6 (ParlaMint). A spoken example with the plural and no comma: *Я даю тим людям хто записались* 'I give (the floor) to those people who have signed up' (ParlaMint-UA_2003-10-14-m0.u156.p1.lang1.s2).

In IU the commonest heads of `acl:relcl` are *той* (54) and *те* (32); in ParlaMint *те* (37) comes after the nouns *людина* (43) and *питання* (42).

### 8.2 Without a head

Headless relatives stand in argument positions, especially in proverbs, and are typically **correlative**: the relative clause comes first and a demonstrative picks it up in the main clause:

- *Хто рано встає, тому Бог дає* 'God gives to him who rises early' (*хто* nom. — *тому* dat.);
- *Що посієш, те й пожнеш* 'As you sow, so shall you reap';
- *Де тонко, там і рветься* 'Where it's thin, there it tears';
- *Хто не працює, той не їсть.*

The demonstrative may be dropped if its case is the same as the relative's (*Хто шукає, (той) знайде*). In UD, a headless *хто*-clause in subject position is `csubj` or, when the treebank treats the relative as the head, `nsubj` with an `acl:relcl` — practice differs between the treebanks.

### 8.3 Indefinite-concessive free relatives

With *б*/*би* or the particle *не* the free relative becomes 'whoever, whatever': *хто б не прийшов* 'whoever comes'; *що б ти не казав* 'whatever you say'; *куди не глянь* 'wherever you look'. The *не* here is not a negation (cf. Smal-Stockyj 1913, § 58, on *ні* in the same function: *куди оком ні глянеш* 'wherever you look'); both *не* and *ні* occur, and modern usage prefers *не*. See [chapter 27](https://md.mova.elamur.ai/free/en/docs/languages/uk/27-negation/), §9.

---

## 9. The paratactic relative of folk speech

Folk narrative often expresses what a written text would put in a relative clause by a **coordinated or juxtaposed clause** with a personal or demonstrative pronoun. Smerechynskyi (1932, ch. XXIII, I) classifies the types, with many examples from folk tales:

| type | example (folk tale pattern) | written equivalent |
| --- | --- | --- |
| *а* + pronoun | *Жила собі баба вдова, а в неї був син.* | *…вдова, у якої був син* |
| *і* + pronoun | *Був такий музикант, і він так гарно грав…* | *…музикант, який так гарно грав* |
| pronoun, no conjunction | *Був собі купець, він усе їздив по морі.* | *…купець, який усе їздив* |
| *і* + verb, no pronoun | *Була баба і мала одну дочку.* | *…баба, яка мала одну дочку* |
| juxtaposition, no pronoun | *Бачили дівчину, від криниці воду несла.* | *…дівчину, що несла воду* |
| demonstrative | *…де є двір, а в тому дворі живе баба* | *…двір, у якому живе баба* |

These patterns are alive in storytelling and conversation. A UD tree analyses them as coordination or `parataxis`, not as relative clauses; the referential link is coreference ([chapter 35](https://md.mova.elamur.ai/free/en/docs/languages/uk/35-reference-and-coreference/)). For a reader of folk texts, they are the commonest "relative clauses" there are.

---

## 10. Mood in relative clauses: *який би*

A relative clause takes the conditional-subjunctive particle *б*/*би* (with the past form of the verb) when the antecedent is non-specific — wished for, sought, denied or questioned:

- *Шукаю людину, яка б могла допомогти* 'I'm looking for someone who could help';
- *Немає такої книжки, якої б він не прочитав* 'There is no book he hasn't read';
- *Чи є тут хтось, хто б знав дорогу?* 'Is there anyone here who knows the way?'

Smal-Stockyj (1913, § 58) lists "restrictive relative clauses" among the uses of the *би*-form, with a question of the same kind (*is there among you at least one person who would tell the truth?*). With a specific antecedent the indicative is used: *Я знайшов людину, яка може допомогти*. The double negative of the second example is the normal Ukrainian way to say "every": see [chapter 27](https://md.mova.elamur.ai/free/en/docs/languages/uk/27-negation/), §4.

---

## 11. Restrictive and non-restrictive clauses; punctuation

Ukrainian does not distinguish restrictive from non-restrictive relatives by the relative word or by punctuation: **both are set off by commas** (2019 orthography, § 158, on separating the parts of a complex sentence; a clause inside the main clause is enclosed by two commas). *Студенти, які склали іспит, отримали стипендію* is either 'The students who passed the exam got a grant' or 'The students, who passed the exam, got a grant'. The difference is made by:

- intonation (a non-restrictive clause is a separate intonation unit with a lower pitch);
- a demonstrative on the antecedent, which forces the restrictive reading: *ті студенти, які…*; *та книжка, що…*;
- *такий … який/що* for a characterising clause: *такий будинок, який нам потрібен* 'the kind of house we need';
- context: a proper-name antecedent is normally non-restrictive (*Київ, який…*).

Commas are sometimes omitted in speech transcripts and on the web (the ParlaMint example in §8.1), not in edited prose.

---

## 12. Participial phrases as relative alternatives

### 12.1 Passive participles

The passive participles in *-ний*, *-тий* (*написаний*, *збудований*, *відкритий*) and, from *-ти*-stems, *-ений* (*зроблений*), are the common compact alternative to a relative clause: *лист, написаний учора* = *лист, який написали вчора* 'the letter written yesterday'. A participial phrase after its noun is separated by commas; a short one may precede (*учора написаний лист*). Their formation and the passive constructions are in [chapter 28](https://md.mova.elamur.ai/free/en/docs/languages/uk/28-voice-and-passive/). In the treebanks they are frequent: 1,239 passive participles (`VerbForm=Part|Voice=Pass`) in IU, 812 of them `amod` and 46 `acl:relcl`; 914 in ParlaMint, 431 `amod`, 91 `acl:relcl` (an `acl:relcl` label on a participle marks a participle phrase treated as a reduced relative).

### 12.2 Active participles: restricted

Active present participles in *-учий*/*-ючий*, *-ачий*/*-ячий* exist (*квітучий* 'blooming', *палаючий* 'burning', *виконуючий*), but the standard restricts them. Many have become adjectives (*квітучий сад* 'a blooming garden', *блискучий* 'brilliant', *лежачий* 'lying', *сидячий* 'sedentary'), and the productive use as a verb form with objects (*✗ студент, виконуючий завдання*) is considered non-normative and a calque of Russian; the norm uses a relative clause (*студент, який виконує завдання*) or a noun (*виконувач обов'язків* 'acting', in place of *виконуючий обов'язки*). The purist literature of the 1920s–30s argued against the related adverbs of the type *чаруюче* "charmingly" (Smerechynskyi 1932, ch. XXII, 2), and the restriction on the participles became the school rule. Usage has not followed the rule here: Starko & Rysin (2025, `starko-2025-dictionary-derussification`) report that imperfective active participles are the largest group of Russianisms flagged in the VESUM dictionary, nearly 3,500 items, and that these forms occur more than 730,000 times (about 410 per million words) in the GRAC reference corpus; the replacements their tools offer are, first, a *що*-clause with a third-person verb (*що блокує* for *блокуючий* "blocking") and, second, a deverbal adjective in *-льний*, *-ливий* (*збиральний*, *обтяжливий*). This is one of the places where the codified norm and measured usage disagree, and the tools side with the norm. Counts confirm how marginal they are in edited text: 81 active participles (`Voice=Act`) in IU against 1,239 passive; 80 in ParlaMint against 914. Active past participles of intransitive perfective verbs (*змарнілий* 'emaciated', *почорнілий* 'blackened', *зів'ялий* 'withered') are normal, and are adjectives in function.

### 12.3 Adverbial participles

A clause in *-ючи*/*-вши* (*читаючи* 'reading', *прочитавши* 'having read') is not a relative alternative but an adverbial; it shares the matrix subject ([chapter 24](https://md.mova.elamur.ai/free/en/docs/languages/uk/24-coordination-subordination/)). Smal-Stockyj (1913, § 57) lists the old use of these forms where a relative or temporal clause would stand.

---

## 13. Counts in the treebanks

All counts: UD 2.18, Ukrainian IU (122K tokens; fiction, news, law, letters, web, Wikipedia; CC BY-NC-SA, counts only) and ParlaMint (109K tokens; Rada transcripts 2003–2022 and one 2014 National Security Council meeting; CC BY-SA 4.0), all splits, syntactic words, made with `awk` over the CoNLL-U files on 2026-11-06.

**The relative word of each `acl:relcl` clause** (first direct dependent with PronType=Rel, or *що* as `mark`):

| relative | IU | ParlaMint |
| --- | --- | --- |
| *який* | 497 | 808 |
| *що* (`mark`) | 415 | 153 |
| *де* | 67 | 35 |
| *хто* | 38 | 35 |
| *котрий* | 31 | 4 |
| *коли* | 21 | 22 |
| *що* (PRON) | 8 | 11 |
| other (*чий, куди, поки, звідки, як, щоб…*) | 7 | 39 |
| none found at the head (relative nested in a phrase, e.g. *батько якого*) | 72 | 63 |
| total | 1,156 | 1,170 |

The ParlaMint "other" row includes clauses introduced by *як*, *щоб*, *якщо* that the annotators also labelled `acl:relcl` (11, 8 and 2), a sign that the speech data stretches the category.

**Case of relative *який*** (all PronType=Rel tokens of *який*):

| case | IU | ParlaMint |
| --- | --- | --- |
| nominative | 309 | 637 |
| accusative | 131 | 204 |
| genitive | 85 | 60 |
| instrumental | 45 | 35 |
| locative | 43 | 40 |
| dative | 12 | 4 |

Subjects dominate, as in most languages (the accessibility hierarchy); in the parliamentary speech two thirds of all *який* are subjects.

**Position**: the relative clause follows its head in 1,151 of 1,156 IU cases and in all 1,170 ParlaMint cases. The five IU exceptions all have the head *той*: correlative structures with the clause first (§8.2).

**All PronType=Rel tokens** (any relation): IU *який* 625, *коли* 170, *що* 158, *як* 103, *де* 103, *хто* 85, *котрий* 32, *поки* 15, *куди* 15, *скільки* 11; ParlaMint *який* 980, *що* 128, *хто* 111, *коли* 106, *як* 53, *де* 50, *поки* 18, *чому* 16, *котрий* 4.

---

## 14. Quick reference: relative clauses in UD

| construction | example | analysis |
| --- | --- | --- |
| *який* subject | *хлопець, який співає* | `acl:relcl(хлопець, співає)`, `nsubj(співає, який)`; *який* DET (IU: all 625; ParlaMint: 779 DET, 201 PRON) with `PronType=Rel` and agreeing features |
| *який* with preposition | *село, у якому* | `obl(виріс, якому)`, `case(якому, у)` |
| possessive genitive | *письменник, книжки якого* | `nmod(книжки, якого)`; the relative is not a direct dependent of the clause head |
| invariable *що* | *люди, що прийшли* | `mark(прийшли, що)`, SCONJ; no subject node |
| *що* + resumptive | *хлопець, що з ним я говорив* | `mark(говорив, що)`, `obl(говорив, ним)`; coreference not marked |
| relative adverb | *місто, де я народився* | `advmod(народився, де)`, ADV `PronType=Rel` |
| headed free relative | *той, хто прийшов* | `acl:relcl(той, прийшов)`, `nsubj(прийшов, хто)` |
| participial phrase | *лист, написаний учора* | `amod(лист, написаний)` or `acl` / `acl:relcl`; `VerbForm=Part |
| paratactic relative | *Жила баба, а в неї був син.* | `conj` or `parataxis`; no relative |

(The part of speech of relative *який* differs: IU tags all 625 as DET; ParlaMint 779 as DET and 201 as PRON. The feature `PronType=Rel` is constant and is what the counts above use.)

---

## 15. How Mova handles it

### 15.1 The Ukrainian fact extractor

Mova's only Ukrainian sentence analyser is the conservative fact extractor in `kg` ([`kg/src/uk.rs`](https://md.mova.elamur.ai/free/en/docs/kg/), described in [kg — the Mova knowledge graph](https://md.mova.elamur.ai/free/en/docs/kg/)). It turns a **simple** sentence into one fact (predicate, subject, objects by case, adverbials, attributes), with morphology from VESUM ([`kg/src/morph.rs`](https://md.mova.elamur.ai/free/en/docs/kg/)) and a compact grammar code per word ([`kg/src/gram.rs`](https://md.mova.elamur.ai/free/en/docs/kg/)). It does not parse relative clauses by design:

- Any comma rejects the sentence (`Reject::Punct`), and Ukrainian relative clauses are always comma-delimited in edited text.
- The list `CONJ` in `uk.rs` contains *що*, *де*, *куди*, *коли*, *як* and the relative-interrogative words *хто*, *який/яка/яке/які*, *чий/чия/чиє/чиї*, *котрий*; a sentence containing one of them in any form listed is rejected as `conjunction`. Inflected forms of *який* other than the four nominatives (*яку*, *якого*, *якому*…) are not on the list.

Measured with `kg parse` (CPU, worktree at commit `ac9a23149`, VESUM `dict_corp_vis.txt.bz2`, an empty proper-name list; 2026-11-06):

| sentence | result | comment |
| --- | --- | --- |
| *Я бачив дім, який збудував батько.* | rejected: punctuation `,` | as designed |
| *Книжка яку я читаю цікава.* (comma omitted) | rejected: not covered by rules | *яку* is not on the conjunction list; the rules then find two finite-verb readings they cannot combine |
| *Хто рано встає тому Бог подає.* | rejected: conjunction | *хто* on the list |
| *Це хата де я народився.* | rejected: conjunction | *де* on the list |
| *Біля хати росте яблуня з великими яблуками.* | parsed: `predicate→рости`, `subject→яблуня`, `adverbial:біля+v_rod→хата`, `adverbial:з+v_oru→яблуко`, `attribute→великий` | a prepositional modifier, the "relative-free" alternative, is read; it is attached to the verb, not to *яблуня* |

So no relative clause is ever turned into a fact, and none is turned into a wrong one: the gate is strict. The cost is coverage: in the Tatoeba run that built the graph, sentences with commas and conjunctions are the largest rejection classes ([kg — the Mova knowledge graph](https://md.mova.elamur.ai/free/en/docs/kg/), "Rejections").

### 15.2 The generator

The graph-to-text generator ([`kg/src/textgen.rs`](https://md.mova.elamur.ai/free/en/docs/kg/)) produces one simple sentence per fact. It never produces a relative clause, and there is no operation that would fold two facts sharing a participant into a head and a relative.

### 15.3 Not yet

- No Ukrainian dependency parser: `tools/mova-parse` has no Ukrainian model, and `langseg` segments Latin script only, so Cyrillic Ukrainian text is not routed to any parser. No Ukrainian UD baseline is recorded in [What every language gets](https://md.mova.elamur.ai/free/en/docs/languages/PACKAGE/) as of 2026-11-06.
- No splitting of complex sentences into simple ones (a relative clause would become a second fact with the antecedent as its member).
- No resolution of *що* + resumptive pronoun, of the possessive genitive *якого*, or of the paratactic relative (§9), which needs coreference ([chapter 35](https://md.mova.elamur.ai/free/en/docs/languages/uk/35-reference-and-coreference/)).
- No register signal from the choice of *що*, *котрий* or *який*, though §§4–5 show it is a style marker.
- The pragmatic labels of `prag/` ([`prag/src/schema.rs`](https://md.mova.elamur.ai/free/en/docs/prag/)) are produced only for English.

---

## 16. Sources

- Smerechynskyi, Serhii. 1932. *Нариси з української синтакси (у зв'язку з фразеологією та стилістикою)*. Kharkiv: Radianska shkola. Ch. X («Которий (котрий)», «який», «що»), ch. XXII.2 (adverbs of the type *чаруюче*), ch. XXIII (paratactic and hypotactic relative constructions; *що* as relative particle). `smerechynskyi-1932-narysy-z-ukrainskoi-syntaksy` (public domain in UA/EU).
- Smal-Stockyj, Stephan. 1913. *Ruthenische Grammatik*. Berlin/Leipzig: Göschen. § 57 (active participles), § 58 (particles; *би* in restrictive relative clauses; *ні* after question words), § 62 (word order: relative words first). `stockyj-1913-ruthenische-grammatik` (public domain).
- *Український правопис*. 2019. Kyiv: Naukova dumka (official text). § 158 (comma between the parts of a complex sentence). `ukrainskyi-pravopys-2019` (official document, not protected by copyright).
- Starko, Vasyl, and Andriy Rysin. 2025. An Electronic Ukrainian Dictionary as a Derussification Tool. eLex 2025. `starko-2025-dictionary-derussification` (CC BY-SA 4.0; paraphrased, not copied).
- 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).
- UD Ukrainian ParlaMint 2.18 (CC BY-SA 4.0): sentences cited by `sent_id`; counts. UD Ukrainian IU 2.18 (CC BY-NC-SA 4.0): counts only. Leipzig Corpora Collection, Ukrainian news 2024, 1M sentences (CC BY; Goldhahn, Eckart & Quasthoff 2012): word counts. All counted 2026-11-06.
- Mova: [`kg/src/uk.rs`](https://md.mova.elamur.ai/free/en/docs/kg/), [`kg/src/morph.rs`](https://md.mova.elamur.ai/free/en/docs/kg/), [`kg/src/gram.rs`](https://md.mova.elamur.ai/free/en/docs/kg/), [`kg/src/textgen.rs`](https://md.mova.elamur.ai/free/en/docs/kg/), [kg — the Mova knowledge graph](https://md.mova.elamur.ai/free/en/docs/kg/), [`prag/src/schema.rs`](https://md.mova.elamur.ai/free/en/docs/prag/), [What every language gets](https://md.mova.elamur.ai/free/en/docs/languages/PACKAGE/).
- [← previous 24. Coordination and subordination](https://md.mova.elamur.ai/free/en/docs/languages/uk/24-coordination-subordination/)
- [next → 26. Questions](https://md.mova.elamur.ai/free/en/docs/languages/uk/26-questions/)
- [↑ back to Ukrainian: a reference manual for Mova](https://md.mova.elamur.ai/free/en/docs/languages/uk/)