Working Group on Core and Oblique Arguments and Adjuncts
Most of this chapter focuses on dependents of verbs. Some similar patterns of other parts of speech are discussed at the end of the chapter.
Most of this chapter assumes that the dependents are nominal phrases. Dependents realized as clauses are discussed at the end of the chapter.
- Core vs. oblique
- Coding strategies
- Can adjectives have core arguments?
- Can adverbs have core arguments?
- Can nouns have core arguments?
- Clausal complements
- Summary of relations
Core Arguments vs. Oblique Modifiers
The UD taxonomy is centered around the fairly clear distinction between core arguments (primarily subjects and objects) and other dependents. It does not make a distinction between adjuncts (general modifiers) and oblique arguments (arguments said to be selected by a head but not expressed as a core argument).
The Definition of Core Arguments
The core/oblique distinction is ultimately an information packaging distinction. All or nearly all languages have a basic way of expressing the one or two arguments of most verbs (intransitive and transitive verbs), and this unmarked form of argument expression is as a core argument. If additional arguments can appear that are treated similarly to these arguments, they may also be regarded as core arguments. (Some languages have no additional core arguments, while other languages allow multiple object arguments, for instance.) Status as a core argument is decoupled from the semantic roles of participants. Normally, depending on the meaning of a verb, many different semantic roles can be expressed by the same means of encoding core arguments. Nevertheless, there is a correlation: agent and patient or theme roles of predicates in their unmarked valence are normally realized as core arguments.
Syntactically, there is not a single criterion which can be used crosslinguistically to distinguish core arguments from obliques, though there are often good and useful criteria for particular languages. These include:
- Verbs usually only agree with core arguments
- Oblique arguments may usually or always appear marked by an adposition while core arguments appear as bare nominals
- Certain cases, traditionally called nominative, accusative, and absolutive typically mark core arguments
- Core arguments in many languages occupy special positions in the clause, often adjacent to the verb
- Syntactic phenomena such as being the controller of a subordinate clause argument or the target of relativization are limited to core arguments in some languages
At the end of the day, the distinction must be drawn and documented on language particular grounds. For example, many languages have certain verbs which take arguments in oblique cases such as dative or an experiencer case, but these arguments should be regarded as core arguments based on their syntactic behavior being parallel to the arguments of other transitive verbs.
Avoiding an Argument/Adjunct Distinction
Many grammatical frameworks suggest that some obliques are selected by or are arguments of a head (for instance, a source argument of from the Queen is an argument of the head receive), while other obliques are general adjuncts, which can appear with any predicate without the head selecting for them (for instance, a temporal argument such as after the holidays).
However, the argument/adjunct distinction is subtle, unclear, and frequently
argued over. For instance, syntacticians at certain times have argued for
various obliques to be arguments, while at other times arguing that they are
adjuncts, particularly for certain semantic roles such as oblique instruments
or sources. We take the distinction to be sufficiently subtle (and its
existence as a categorical distinction sufficiently questionable) that the
best practical solution is to eliminate it. Nevertheless, if the distinction
is available in a treebank that is being converted to UD, it can be preserved
using subtypes of dependency relations: obl:arg is used for oblique
arguments, and bare obl then denotes adjuncts.
The core-oblique distinction is generally accepted in language typology as being both more relevant and easier to apply cross-linguistically than the argument-adjunct distinction. See, for example:
- Avery D. Andrews. 2007. The Major Functions of the Noun Phrase. In Timothy Shopen (ed.) Language Typology and Syntactic Description: Clause Structure (2nd ed), Cambridge University Press, Cambridge, United Kingdom, pp. 132-223. (1st edition, 1985.)
- Sandra A. Thompson. 1997. Discourse Motivations for the Core-Oblique Distinction as a Language Universal. In Akio Kamio (ed.) Directions in Functional Linguistics. Benjamins, Amsterdam, the Netherlands, pp. 59-82.
Coding Strategies
English
In English, nominal core arguments are bare noun phrases (that is, without preposition). Oblique arguments and nominal adjuncts are prepositional phrases. There is one exception: a bare nominal may be used as a temporal adjunct expressing duration:
- He works the whole week.
In an unmarked declarative sentence, the core argument preceding the verb is the subject, and if there is another core argument following the verb, it is the object. A finite verb agrees in person and number with its subject:
- The boy eats one apple.
- The boy eats many apples.
- The boys eat one apple.
- The boys eat many apples.
If the arguments are realized as personal pronouns, the subject is in the nominative form (I, he, she, we, they) and the object is in the accusative (me, him, her, us, them). Nouns do not inflect for case in English.
Transitive clauses (those that have an object) can be passivized, which means:
- Active verb form is replaced by passive (finite auxiliary + participle).
- Former object becomes subject.
- Former subject either disappears or becomes an oblique argument.
- One apple is eaten (by the boy(s)).
- Many apples are eaten (by the boy(s)).
The inability to passivize of He works the whole week is an argument in support of the claim that the clause is intransitive and the whole week is an adjunct rather than an object. However, this test is not sufficient because there is a small set of verbs that have objects but do not passivize:
- John has a new car. (*A new car is had by John.)
- Friday does not suit me. (*I am not suited by Friday.)
Therefore, durational temporal adjuncts have to be stated as an exception, and this is the one case where the argument/adjunct distinction cannot be avoided in UD.
- He works the whole week.
On the other hand, even temporal noun phrases can be objects if the verb requires an object:
- He spent the whole week in Oslo.
Joakim: I think word order tests can be used to distinguish these two cases (instead of appealing to the arg/adj distinction):
- John worked the whole week in Paris. John worked in Paris the whole week.
- John spent the whole week in Paris. *John spent in Paris the whole week. </span>
Dan: In English, yes. But the same adjunct appears in Spanish and Czech too.
Some English verbs allow two objects (i.e., two core arguments following the verb):
- Peter gave Kate a book.
- Tom teaches me mathematics.
The traditional approach outside UD is to call the first object indirect and the second object direct; it is often defined in terms of semantic roles, saying that the recipient is the indirect object. UD avoids referring to semantic roles and says instead that indirect object is a core argument of the verb that is not its subject or direct object; the (direct) object is then defined as “the second most core argument of a verb after the subject.” In the above examples, the recipient (Kate, me) is arguably more core than the theme (book, mathematics) because the recipient can be promoted in passivization while the theme cannot:
- Kate was given a book by Peter. (*A book was given Kate by Peter.) Nathan: “A book was given Kate” sounds unusual or archaic but not completely ungrammatical to me. You can find examples on the web of “advice was given him/her”, for example. Some COCA examples: “how do you feel about John Poindexter and the sentence that was given him today?” (1990 spoken), “But when the rooms had been shown and the paper was given him to sign, Frisch hesitated, clutching the pen.” (1990 fiction). Not many COCA examples with by-phrases—only found this one from a 2012 interview with Sarah Palin, who is known for her colorful use of English: “And he got a standing ovation at the end of the speech, and that was given him by those who paid attention and stayed to the end, if you will, and heard what he had to say.” No examples in EWT of a passive verb with
iobj, though. - I am taught mathematics by Tom. (*Mathematics is taught me by Tom.)
The second object can be promoted only if both the subject and the first object are recoded as oblique arguments (in the case of to give) or the first object is removed (in the case of to teach):
- A book was given by Peter to Kate.
- Mathematics is taught by Tom. (*Mathematics is taught by Tom to me.)
Therefore, the UD v2 guidelines actually require that the second object
is labeled iobj. Yet in the current English data (UD 2.1), it is still
the first object that is labeled iobj, and the second object is obj.
This has to be resolved by either modifying the guidelines, or the data
(diverging from what people traditionally understand under indirect object).
Joakim: I don’t think this is specified by the v2 guidelines. I tried to argue for it on GitHub, based on syntactic criteria, but people pushed back saying that it is convenient for NLP applications to have the same relation for the recipient role in both cases. This needs to be resolved.
TO DO: How do we explain that an oblique nominal is annotated as subject in subordinate infinitival clause, as in It was impossible for him to attend the meeting. Some authors (I do not remember where exactly I saw it) say that it is quite common that core arguments are coded differently in subordinate clauses. But can we prove that the for-phrase is indeed a subject? It is not sufficient to say that it has the same semantic role as the corresponding bare nominal in a main finite clause. If it was sufficient, then the by-phrases in passive clauses would be subjects too!
Joakim: We definitely need to allow different coding in subordinate clauses. That is the rule in languages like Latin and Turkish. But I think we should save this for the section on subordinate clauses.
Spanish
The behavior of Spanish core arguments is somewhat similar to English but there are differences. Like in English, it is typical for a core argument to be a bare nominal without preposition. However, prepositions are not completely excluded. If the object is a person, it is marked by the preposition a:
- Vimos a alguien. “We saw somebody.”
The subject’s person and number is cross-referenced by verbal inflection. Spanish is a pro-drop language, meaning that the subject can be omitted if it is a personal pronoun.
- (Yo) bebo vino. “I drink wine.”
- (Tú) bebes vino. “You drink wine.” (singular)
- El hombre bebe vino. “The man drinks wine.”
- (Nosotros) bebemos vino. “We drink wine.”
- (Vosotros) bebéis vino. “You drink wine.” (plural)
- Los hombres beben vino. “The men drink wine.”
If the arguments are realized as personal pronouns, the subject is in the nominative form (yo, tú, él, nosotros, vosotros, ellos) and the object is in the accusative (me, te, lo, nos, os, los). Nouns do not inflect for case in Spanish.
If both core arguments are present and if they are realized as full noun phrases, the prototypical word order is the same as in English: the subject precedes the verb and the object follows it. However, if the object is realized as a pronominal clitic (and if the verb is finite indicative), the object precedes the verb.
- El hombre lo bebe. “The man drinks it.”
Transitive clauses (those that have an object) can be passivized.
- El vino es bebido por los hombres. “The wine is drunk by the men.”
- Las bebidas son bebidas por los hombres. “The drinks are drunk by the men.”
Both coding strategies that are used for core arguments can also appear with adjuncts. Bare nominal adjuncts are rare, the exception being durational temporal adjuncts. In contrast, the preposition a can be used with various directional and temporal adjuncts.
- Él trabaja toda la semana. “He works the whole week.”
- Subiremos al tren a las cinco. “We will be boarding the train at five.”
Neither toda la semana nor al tren or a las cinco can be promoted via passivization.
Some Spanish verbs allow two objects:
- Pedro dio un libro a Isabel. “Pedro gave Isabel a book.”
- Pedro le dio un libro a Isabel. “Pedro gave Isabel a book.”
- Pedro le dio un libro. “Pedro gave her a book.”
- Santiago me enseña las matemáticas. “Santiago teaches me mathematics.”
One of the objects typically corresponds to the recipient semantic role and it most likely refers to a person, therefore it will be marked by the preposition a (if it is realized as a full noun phrase). If it is represented by a pronominal clitic, it will be in the dative form (identical with the accusative except for the third person, which is le, les, and does not distinguish gender). It is not uncommon that both the noun phrase and the clitic are present, as in Pedro le dio un libro a Isabel. (this is known as “clitic doubling”).
The object that is marked by the preposition a or by the dative case of the pronoun (i.e., the one with the recipient role) is labeled as an indirect object; the unmarked/accusative object is direct. This is in line with the UD v2 guidelines: if the clause is passivized, the direct object is promoted to the subject relation, while the indirect object stays untouched. Hence the indirect object is less core than the direct object.
- Un libro fue dado a Isabel por Pedro. “A book was given to Isabel by Pedro.”
Joakim: Why do we treat the recipient as core in Spanish but not in English when they both have prepositions? Is this because prepositions are found with direct objects in Spanish but not in English? And what would we then do in French, where the recipient takes a prepositions but where prepositions are never used with prepositions? Same as English? It would also be possible (at least in French) to argue for a mixed system, where full noun phrases are oblique (because they take a preposition) but pronouns are core (because they have dative case).
Czech
Classification of verbal arguments in Czech depends on case morphology. There are certain anomalies of the case system when the argument is a quantified phrase (with a cardinal number or a pronominal quantifier, the head noun may have different case than the entire phrase). We exclude quantified phrases from the following overview.
The coding strategy most typical for Czech core arguments is bare noun phrase in nominative or accusative. Some authors claim that core arguments are not marked for case. This is not true and Czech is one of the counter-examples. The nominative can be considered unmarked in the vague sense that it is the default case that is used if there are no external factors requiring another case. However, it is not unmarked in the morphological sense: many nouns must use suffixes to form the nominative (and the same holds for the accusative).
The nominative argument is the subject, the accusative is object. The subject’s person, number, and sometimes also gender and animacy are cross-referenced by verbal inflection. Czech is a pro-drop language, meaning that the subject can be omitted if it is a personal pronoun.
- Snědl jsem jablko. “I ate an apple.” (masculine)
- Snědla jsem jablko. “I ate an apple.” (feminine)
- Snědl jsi jablko. “You ate an apple.” (masculine singular)
- Snědla jsi jablko. “You ate an apple.” (feminine singular)
- Chlapec snědl jablko. “The boy ate an apple.”
- Snědl jablko. “He ate an apple.”
- Snědla jablko. “She ate an apple.”
- Snědlo jablko. “It ate an apple.”
- Snědli jsme jablko. “We ate an apple.” (masculine animate)
- Snědly jsme jablko. “We ate an apple.” (feminine)
- Snědli jste jablko. “You ate an apple.” (masculine animate plural)
- Snědly jste jablko. “You ate an apple.” (feminine plural)
- Snědli jablko. “They ate an apple.” (masculine animate)
- Snědly jablko. “They ate an apple.” (feminine or masculine inanimate)
- Snědla jablko. “They ate an apple.” (neuter)
Czech word order is free and while the SVO order is preferred by default, other permutations are possible and may be required to distinguish topic and focus.
- Chlapec snědl jablko. “The/a boy ate an apple.”
- Jablko snědl chlapec. “It was a/the boy who ate the apple.”
- Chlapec jablko snědl. “What the boy did to the apple was that he ate it.”
- Jablko chlapec snědl. “As for the apple, what the boy did to it was that he ate it.”
- Snědl chlapec jablko. “A boy ate an apple.”
- Snědl jablko chlapec. “A boy ate an apple.” (Most unusual order but not impossible.)
Transitive clauses (those that have an accusative object) can be passivized.
- Jablko bylo snědeno chlapcem. “The apple was eaten by the boy.”
- Jablka byla snědena chlapcem. “The apples were eaten by the boy.”
There are a few verbs with accusative objects that cannot passivize, for instance:
- Chlapec dostal čokoládu. “The boy got chocolate.” (*Čokoláda byla dostána chlapcem.)
- Chlapec má čokoládu. “The boy has chocolate.” (*Čokoláda je mána chlapcem.)
It is not guaranteed that a bare accusative nominal is a core argument. It can also be a durational temporal adjunct, as in:
- Pracuje celý týden. “He works the whole week.”
The nominal celý týden cannot be promoted to subject via passivization, which supports the claim that it is not an object; however, this test is not sufficient because of the exceptional verbs like dostat “to get” and mít “to have”. Therefore, durational temporal adjuncts have to be stated as an exception and the argument/adjunct distinction cannot be avoided in this case.
Many two-argument verbs in Czech specify the second argument as a bare noun phrase in a case other than accusative. Whether these arguments are core arguments is a point of disagreement among different authors. The current (UD 2.1) approach in UD for Czech and several similar Indo-European languages is to analyze them as core.
Some of the verbs resemble prototypical transitive verbs in that their arguments have the roles of agent and patient. For instance, the verb pomoci “to help” takes a dative argument:
- Zuzka pomohla Martinovi s domácím úkolem. “Zuzka helped Martin with his homework.”
The clause can be passivized. However, the dative argument is not promoted by the passivization to the subject relation. It stays in dative and the verb does not cross-reference its person, number or gender. Instead, verbal morphology switches to the default singular neuter agreement with unexpressed subject. This suggests that the dative argument is less core-like (if at all) than a standard accusative object.
- Martinovi bylo pomoženo s domácím úkolem. “Martin was helped with his homework.” (*Martin byl pomožen s domácím úkolem.)
If dative arguments are core objects, then we may want to distinguish them from benefactive adjuncts that are also encoded as bare nominals in the dative (that is, if we acknowledge that benefactives should be adjuncts):
- Zuzka Martinovi udělala večeři. “Zuzka prepared a dinner for Martin.”
- Zuzka Martinovi koupila večeři. “Zuzka bought a dinner for Martin.”
- Zuzka Martinovi objednala večeři. “Zuzka ordered a dinner for Martin.”
- Zuzka Martinovi snědla večeři. “Zuzka ate Martin’s dinner.”
Joakim: If benefactives and datives have the same encoding, perhaps they should both be regarded as oblique (to facilitate annotation).
Similarly, there are objects realized as bare genitives and even instrumentals, with the same passivization pattern as datives. Instrumental objects have to be distinguished from demoted oblique agents in passive constructions, and from instrumental adjuncts.
- Novináři musí dbát zásad objektivity. “Journalists must observe principles of objectivity.” (genitive)
- Musí být dbáno zásad objektivity. “Principles of objectivity must be observed.”
- Karel hýbal nábytkem. “Karel moved furniture.” (instrumental)
- Nábytkem bylo hýbáno. “The furniture has been moved.”
The Czech grammar also recognizes prepositional objects but we do not consider them core (which means they are not objects in UD). In fact, the definition of object in the Czech grammar is identical with argument and leads to the argument/adjunct distinction, disfavored in UD. Most adjuncts are prepositional phrases, thus the bulk of the decisions would have to be done for prepositional phrases; that is the main reason why bare datives are analyzed as core objects while prepositional phrases are not. It is worth noting that the neuter singular passivization that is described above for bare datives, genitives and instrumentals is also available for prepositional phrases and even for some intransitive verbs, although such constructions are rare:
- Spoléhali na ředitelovo rozhodnutí. “They relied on the director’s decision.”
- Na ředitelovo rozhodnutí bylo spoléháno. “It was relied on the director’s decision.”
- Ředitel rozhodl v pátek. “The director decided on Friday.”
- V pátek bylo rozhodnuto. “It was decided on Friday.”
Some Czech verbs allow two objects. Typically, one object is accusative and the other is dative; nevertheless, some other combinations are possible, too.
- Petr dal Katce knihu. “Petr gave Katka a book.” (dative + accusative)
- Tomáš mě učí matematiku. “Tomáš teaches me mathematics.” (accusative + accusative)
- Muž vyhrožoval sousedovi smrtí. “A man threatened his neighbor with death.” (dative + instrumental)
The dative-accusative construction can be passivized so that the accusative object is promoted to subject, the dative object stays as it is, and the former subject disappears (or, rarely, is transformed to instrumental). It is thus confirmed that the dative object is less core than the accusative, hence the dative should be labeled as indirect.
- Kniha byla dána Katce (Petrem). “The book was given to Katka (by Petr).”
The verb učit “to teach” is special in that it allows two accusatives: one representing the theme (mathematics) and the other the recipient (me). Either one can be omitted and then we have a normal transitive clause with an accusative object that can be promoted via passivization. However, if both arguments are present in the active clause, it is not possible to promote one of them and leave the other untouched; the grammar does not tolerate a bare accusative argument in a passive clause. Also note that there is a slightly archaic alternative where the theme (not the recipient!) takes the dative form. Here the passive might in theory be available but it still sounds clumsy.
- Tomáš učí matematiku. “Tomáš teaches mathematics.”
- Matematika je učena Tomášem. “Mathematics is taught by Tomáš.”
- Tomáš mě učí. “Tomáš teaches me.”
- (Já) jsem učen Tomášem. “I am taught by Tomáš.”
- Tomáš mě učí matematiku. “Tomáš teaches me mathematics.”
- (*Jsem učen matematiku Tomášem. *Matematika je učena mě Tomášem.)
- Tomáš mě učí matematice. “Tomáš teaches me mathematics.” (mathematics in dative)
- (?Jsem učen matematice (Tomášem).) “I am taught mathematics (by Tomáš).”
There thus does not seem to be any evidence that one of the accusatives is more core than the other. We have an example of a clause with two objects, neither of which is indirect.
Finally, in the dative-instrumental construction, the passivization follows rules similar to clauses with one non-accusative object: former subject disappears but the objects stay untouched. Moreover, if the subject is not removed but transformed to an instrumental argument, it is likely that the original instrumental argument will be removed instead. Yet it is not completely ungrammatical to keep them both (see the example below; but it is highly preferred that the two instrumental arguments are not adjacent).
- Muž vyhrožoval sousedovi smrtí. “A man threatened his neighbor with death.”
- Sousedovi bylo vyhrožováno smrtí. “The neighbor was threatened with death.”
- Sousedovi bylo vyhrožováno mužem. “The neighbor was threatened by the man.”
- Mužem bylo sousedovi vyhrožováno smrtí. “The neighbor was threatened by the man with death.”
The dative-instrumental construction is one where both objects are clearly less core than accusative objects, but none of the two is more or less core than the other. Therefore, none of them can be labeled as indirect.
- Petr dal Katce knihu. “Petr gave Katka a book.”
- Kniha byla dána Katce Petrem. “The book was given to Katka by Petr.”
- Tomáš mě učí matematiku. “Tomáš teaches me mathematics.”
Basque
See also: Fernando Zúñiga, Beatriz Fernández (draft 26.6.2014): Grammatical relations in Basque
In Basque, like in Czech, nominal case morphology is essential for recognition of core argument relations. However, instead of nominative-accusative, the core pair of cases in Basque is ergative-absolutive. Most two-argument verbs have one argument in the ergative and the other in the absolutive case (but see below for other possibilities). The ergative argument is labeled as subject, the absolutive argument is object. With primary transitive verbs, the ergative argument corresponds to the agent and the absolutive argument to the patient.
- Ekaitzak itsasontzia hondoratu du. “The storm has sunk the ship.”
The single argument of intransitive verbs takes mostly the absolutive but sometimes the ergative form. It is labeled as subject in both cases.
- Gizona hil da. “The man has died.” (absolutive)
- Urak irakin du. “The water has boiled.” (ergative)
The third core case is the dative. Arguments in all three core cases are cross-referenced on finite verbs. Thanks to cross-referencing, the arguments can be omitted if they are just personal pronouns.
Some two-argument verbs take dative+absolutive, instead of ergative+absolutive:
- (Niri) ardoa gustatzen zait. “I like wine.” (lit. “To-me the-wine pleasing is.”)
In the current data (UD 2.1), the dative argument seems to be labeled iobj
even in dative-absolutive constructions. However, Zúñiga and Fernández (2014)
write that the dative encodes the A function in such constructions; that
would mean that it should be nsubj.
Some two-argument verbs take ergative+dative, instead of ergative+absolutive:
- Irakasleak haserre begiratu die ikasleei. “The teacher has looked angrily at the students.”
In the current data (UD 2.1), the dative argument seems to be labeled iobj
even in ergative-dative constructions. However, Zúñiga and Fernández (2014)
write that the dative encodes the P function in such constructions; that
would mean that it should be obj.
There are verbs that take all three core arguments. In such constructions,
the ergative encodes the A function, absolutive the P function (also T function = theme-like),
and dative the G function (goal). In terms of dependency relations, it seems
reasonable to label the ergative as nsubj, absolutive as obj and dative
as iobj just to distinguish them and to acknowledge that absolutives and
ergatives are more frequent than datives; though the grammar does not seem to
target the absolutive argument in specific rules, which would make it
more core-like than the dative.
- (Nik) (zuri) liburua eman dizut. “I have given you a book.”
- Zezenak saihetsa pitzatu zidan. “The bull cracked my rib.” (ergative argument: 3 singular; absolutive argument: 3 singular; dative argument: 1 singular)
Basque does not have a canonical passive construction, although there are constructions that have been called passive by some authors in the past.
While the passive, if it existed, would decrease the valency of the verb, there is another operation that increases the valency: the causative:
- Active verb form is replaced by causative.
- Former subject becomes dative object and denotes the causee.
- A new ergative subject appears and denotes the causer.
Analogically to the passive, it is recommended that the relations with modified interpretation (if overtly present) are labeled as subtypes: the causer as nsubj:caus and the true agent as iobj:agent.
- (Guk) arazo hau ikusi genuen. “We have seen this problem.” (active)
- (Horiek) (guri) arazo hau ikuserazi digute. “They have made us see this problem.” (causative)
Joakim: I think it might be better to use “nsubj:caus” and “iobj:caus”, indicating that these are arguments of a causative and therefore do not have the expected role but without specifying the role (cause and agent). This is more parallel with “nsubj:pass”, where we just indicate that it is the subject of a passive and therefore has an unexpected role but without specifying what the role is.
The fact that the causative is available for dative-absolutive verbs supports the claim that the dative argument indeed is the subject in active voice of those verbs.
- Zopa izugarri gustatzen zaio mutilari. “The boy likes the soup a lot.”
- Goseak zopa hori izugarri gustatuerazi zion mutilari. “Hunger made the boy like that soup a lot.”
Causative is only marginally accepted with ditransitive verbs, apparently due to the marginal acceptance of two dative-marked arguments in the same clause.
Yidiɲ
See also:
- Avery D. Andrews: The major functions of the noun phrase (2007). In Timothy Shopen (ed.): Language Typology and Syntactic Description, second edition, volume I: Clause Structure. Pp. 193-197. Cambridge University Press. ISBN 978-0-521-58156-1.
- R. M. W. Dixon: A Grammar of Yidiny (1977). Cambridge: Cambridge University Press. ISBN 978-0-521-21462-9.
Yidiɲ (Pama-Nyungan, Australia) has a combination of the ergative-absolutive system (similar to Basque) and the nominative-accusative system (similar to Czech). The former pair is typical for nouns, the latter for pronouns.
- Ŋayu maŋga:ɲ. “I laughed.” (nominative)
- Buɲa maŋga:ɲ. “The woman laughed.” (absolutive)
- Ŋaɲaɲ buɲa:ŋ wuɹa:ɲ. “The woman slapped me.” (ergative-accusative)
- Ŋayu buɲa wuɹa:ɲ. “I slapped the woman.” (nominative-absolutive)
- Waguɖaŋgu guda:ga wawa:l. “The man saw the dog.” (ergative-absolutive)
There is a type of subordinate clause that is called dative subordinate (datsub) because its head verb takes a subjunctive suffix similar to the dative suffix of nouns. The clause may be analyzed as a relative clause dependent on a noun phrase in the main clause. The coreferential noun phrase in the subordinate clause may be elided. There is a near requirement (with some exceptions) that the coreferential nominal has S/P function in both clauses, i.e., it is the subject of an intransitive predicate or the object of a transitive predicate:
- Ŋayu maŋga:ɲ (ŋaɲaɲ) buɲa:n wuɹa:ɲunda. “I, (who) was slapped by the woman, laughed.”
- Ŋaɲaɲ buɲa:ŋ wuɹa:ɲ (ŋayu) maŋga:ɲunda. “I, (who) was lauging, was slapped by the woman.”
If we want to use this construction for a nominal with A function (i.e., the subject of a transitive predicate) in either of the two clauses, we must first transform the clause to antipassive, which means:
- Active verb form is replaced by antipassive.
- Former object is transformed to the dative or locative case and becomes oblique argument.
- The subject is still subject, but now of an intransitive predicate. Its function changed from A to S.
Analogically to the passive, we may want to use a relation subtype for the demoted object: obl:patient.
- Waguɖaŋgu guda:ga wawa:l. “The man saw the dog.” (active)
- Wagu:ɖa gudaganda wawa:ɖiɲu. “The man saw the dog.” (antipassive)
- Ŋayu buɲa wuɹa:ɲ. “I slapped the woman.” (active)
- Ŋayu buɲa:nda wuɹa:ɖiɲu. “I slapped the woman.” (antipassive)
It is now possible to combine clauses so that the subject of an originally transitive (now intransitive) predicate is coreferential:
- Ŋayu maŋga:ɲ (ŋayu) buɲa:nda wuɹa:ɖiɲunda. “I, (who) was slapping the woman, laughed.”
- Ŋayu buɲa:nda wuɹa:ɖiɲu (ŋayu) maŋgaɲunda. “I, (who) was laughing, slapped the woman.”
Andrews notes that besides the relative clause analysis, these sentences can be also interpreted adverbially. The core-oblique distinction and the antipassive stay the same but the UD trees change:
- Ŋayu maŋga:ɲ (ŋayu) buɲa:nda wuɹa:ɖiɲunda. “I laughed while slapping the woman.”
- Ŋayu buɲa:nda wuɹa:ɖiɲu (ŋayu) maŋgaɲunda. “I slapped the woman while laughing.”
Tagalog
Tagalog is a representative of a subset of Austronesian languages that have been described as symmetric-voice languages. There is an agent voice, where the most core argument corresponds to the agent role, and a patient voice, where the most core argument corresponds to the patient. It is tempting to reuse our labels of active and passive for these voices. There are at least two important differences from what these terms mean in Indo-European languages, though. First, the two voices in Tagalog are more symmetric and it is hard to say that the active is default, unmarked, and the passive is marked. The passive is not morphologically more complex than the passive. Both the agent and the patient regularly appear in both voices. They are coded differently, but both are coded as core, not oblique arguments. Both “active” and “passive” clauses thus can be transitive.
The second difference is that there are more voices than the agent-oriented and the patient-oriented one. There is also a voice that fronts a locational argument, for example.
The arguments are marked by function words that could be analyzed as either prepositions, or case-bearing determiners. Although adpositions are often associated with oblique arguments and adjuncts, we have seen that it is not a universal rule. Spanish marks a human direct object with the preposition a; Slavic and other languages have prepositional object that, despite not being considered core in UD, seem to work the same way as other non-accusative (but preposition-less, and currently core) arguments; and in Japanese, all arguments are marked by postpositions, including the subject and the direct object. On the other hand, if the nominal markers in Tagalog are determiners, then Tagalog somewhat resembles modern German, where nominal inflection has been greatly reduced and the varying forms of determiners are often the only clue about the case of a noun phrase. Categorizing the Tagalog markers as determiners seems to be more favored in the literature.
The most core-like argument (also called pivot) is marked by the determiner ang. The other core argument (if any) is marked by the determiner ng (pronounced nang). A different set of determiners is used with proper nouns. No determiners are used with personal pronouns, but the pronouns themselves inflect for case.
There have been disputes about whether the pivot is subject and whether
Tagalog has a subject at all. Nevertheless, there does not seem to be any
better solution than to simply identify it with the subject in UD.
EDIT: Further discussion is needed. Andrews cites
others and says (below examples 124) that both ang and ng arguments are
core, but the actor is more subject-like regardless of whether it is pivot or
not. That would mean that we should use the nsubj relation for the ang
argument of agent-voice verbs, the ng argument of patient-voice verbs,
and we would have to select one of two ng arguments in other voices,
depending on which argument semantically fits the agent role.
It seems appropriate to mark the determiners and the personal pronouns with
the Case feature: the pivot with nominative, and the other core argument
with accusative.
- Nagsalita ka. “You spoke.”
- Natalisod ka. “You tripped.”
- Natalisod ang babae. “The woman tripped.”
- Natalisod si Pedro. “Pedro tripped.”
Transitive (two-argument) sentences have two core arguments, one in the
nominative (pronoun or determiner) and the other in the accusative. We may
want to use relation subtypes for the arguments in the patient voice (similarly
to what we do in the English passive): nsubj:pass for the subject and
obj:agent (not obl:agent!) for the object. However, the subtypes should
not be taken to suggest that the agent voice is default and unmarked.
- Naglilinis siya ng bahay. “She cleans the house.” (agent voice)
- Nililinis niya ang bahay. “She cleans the house. / The house is cleaned by her.” (patient voice)
Locative, directional and benefactive arguments are normally coded as oblique. However, there are additional voices where these arguments become subjects. One of the reasons why an argument is fronted as the subject is that the subject is understood as the topic of the sentence. The ang determiner implies definiteness (while ng-marked arguments can be definite or indefinite). Here is an example of a sentence in four different voices:
- Magaalis ang babae ng bigas sa sako para sa bata. “The woman will take some rice out of a/the sack for a/the child.” (agent voice)
- Aalisin ng babae ang bigas sa sako para sa bata. “A/the woman will take the rice out of a/the sack for a/the child.” (patient voice)
- Aalisan ng babae ng bigas ang sako para sa bata. “A/the woman will take some rice out of the sack for a/the child.” (locative voice)
- Ipagaalis ng babae ng bigas sa sako ang bata. “A/the woman will take some rice out of a/the sack for the child.” (benefactive voice)
Since the agent and patient stay core arguments even in the locative and benefactive voices, we actually have a ditransitive clause with three core arguments. In contrast, the verbs of giving, which are typical representatives of ditransitive predicates in other languages, form a standard transitive clause in the agent and patient voices, as the recipient is coded as a directioal (locative) oblique dependent.
- Nagbigay ang lalaki ng libro sa babae. “The man gave a book to the woman.” (agent voice)
Plains Cree
See also:
- Arok Elessar Wolvengrey (2011): Semantic and pragmatic functions in Plains Cree syntax (PhD thesis). Utrecht: LOT. ISBN 978-94-6093-051-5.
The Algonquian (North American) language Plains Cree is similar to Basque in its cross-referencing of both the subject and the object by verbal inflection. It is also very dissimilar to Basque due to its almost complete lack of case marking on the nouns. Yet the marking of the argument on the verbal head is sufficient to allow for a relatively free word order (unlike English).
Like in many other languages where person and number of an argument is cross-referenced by the verb, the argument does not need to appear as a separate word if it would be just a pronoun.
- Niwīcihānānak. “We help them.”
- Niwīcihikonānak. “They help us.”
The two verb forms in the two previous examples differ in Voice. The first example is in the direct voice, where the argument higher in a language-specific obliqueness hierarchy is more agent-like and the other argument is more patient-like. First person arguments are higher than third person arguments, therefore the agent is “we” and the patient is “they”.
In the second example, the argument-marking morphemes have not changed but the verb is in the inverse voice where the argument lower in the hierarchy (“they”) is the agent and the higher argument (“we”) is the patient.
In a parallel to the active-passive voice system of English, we can postulate
that the argument higher in the hierarchy is the subject (i.e., the most core
argument), and the lower argument is the object.
And like with English passives, the voice has to be considered when one wants
to map the grammatical relations to semantic roles.
In the two previous examples, the arguments are not overtly present and the
dependency relations are not visible. However, if there are overt arguments
in an inverse voice clause, we may again want to use relation subtypes to
signal that the default interpretation does not apply.
TBD: Should we use nsubj:pass and obj:agent
as in symmetric-voice languages? Or something else? Or should we even replace
nsubj:pass in other languages by something more neutral?
Alternatively, we could make the direct-inverse voice system special in that
we resolve the roles already in the relation labeling, i.e., higher-ranking
argument will be labeled nsubj in direct voice and obj in inverse voice,
and lower-ranking argument will be obj in direct voice and nsubj in
inverse voice. That is, if there are not some typical subject tests known
from other languages that will contradict our proposal.
At present I do not know anything about subjecthood tests applied to Plains
Cree.
- Nikī-wīcihāw Cān. “I helped John.”
- Nikī-wīcihāwak Cān ēkwa Mēriy. “I helped John and Mary.”
If two third-person arguments are involved, one of them is considered proximate (more topical, higher in the obliqueness hierarchy) and the other is considered obviative (less topical, lower in the obliqueness hierarchy). The obviative noun is marked morphologically by the suffix -a. Note that Plains Cree distinguishes singular and plural Number for proximate but not for obviative third person arguments.
Current UD data (as of UD 2.1) does not include languages with direct-inverse
voice systems and proximate-obviative division of third-person arguments.
Joakim: But I think similar cases are discussed
with reference to UD in this paper on Arapaho.
The guidelines define a special value for the “fourth person”, which could be
used for obviative arguments (see Person). This value is inspired by the
Unimorph project.
However, the Unimorph specification seems to favor using Person=3 with an
additional subfeature for proximate and obviative status:
“In some languages, a fourth person category is used to describe an otherwise
third-person referent that is differentiated from other third-person referents
by a switch-reference-like distinction … or, more commonly, by a distinction in
obviation status… For the purposes of morphological distinctions, these fourth
person categories may call for dedicated verbal morphology. While in some
cases their meaning can be captured by third person (3) plus switch-reference
features or features marking pragmatic voice distinctions (such as the
proximate (PRX) and obviative (OBV)), we include a fourth person category with
the feature 4 to allow for identification of a fourth person category when the
semantic distinctions are complicated or not strictly inflectional in nature.”
A possible alternative would be to revoke the “fourth person” and define
a separate feature Obviation with the values Prx and Obv. There are two
places where the feature could be used: as a verbal feature (cross-referencing
the argument) and as a nominal feature (marking the argument). Especially in
nominal inflection it seems preferable to use a new feature rather than assert
that the noun inflects for person (with values 3 and 4 only).
Yet another option would be to include the nominal inflection under Case
marking, i.e., a noun with the obviative morpheme would be Case=Acc and
the unmarked nouns would be Case=Nom. However, such an extension could be
perceived as stretching the usual definition of case too far. The main oddity
would be that it only applies to third-person referents and only if two such
referents compete as arguments of the same verb. If one argument of the verb
is first-person and the other third-person, both will stay in the “nominative”
regardless of their semantic roles. Therefore, defining a new feature,
separate from Case, is probably a better and less confusing solution.
- Cāniy kī-wīcihēw Mērīwa. “Johnny helped Mary.” (direct voice)
- Cānīwa kī-wīcihēw Mēriy. “Mary helped Johnny.” (direct voice)
- Cāniy kī-wīcihik Mērīwa. “Mary helped Johnny. / Johnny was helped by Mary.” (inverse voice)
- Cānīwa kī-wīcihik Mēriy. “Johnny helped Mary. / Mary was helped by Johnny.” (inverse voice)
Ditransitive verbs (that is, verbs with three arguments) still cross-reference only two arguments on the verb. It is always the recipient-like argument that is cross-referenced, while the theme stays outside. Hence the recipient (goal) is more core than the theme, and deserves to be labeled obj. The question is whether the theme is still a core argument (and thus labeled iobj) or it is already oblique. Allowing it the core status seems to be at a similar level of tolerance as with the non-accusative objects in languages like Czech. The themes in Plains Cree are not indexed on the verb, though their own surface form (bare nominal) is not different from the core arguments.
- Nikī-miyāw anima masinahikan. “I gave him/her that book.” (direct voice)
- Nikī-miyik anima masinahikan. “He/she gave me that book.” (inverse voice)
- Nikī-kiskinwahamawāw nēhiyawēwin. “I taught him/her Cree.” (direct voice)
- Nikī-kiskinwahamāk nēhiyawēwin. “He/she taught me Cree.” (inverse voice)
In the above example, the verb stem used is for animate objects, while masinahikan “book” is inanimate. That is a proof that the 3rd person singular cross-reference on the verb does not refer to the book but to an animate recipient that is not overtly represented in the sentence.
Even though Plains Cree does not use morphological cases to distinguish agents
from patients, there is a form of nouns that can be classified as the locative
case (Case=Loc). It clearly marks the noun as oblique and unable to be
cross-referenced by verbal inflection. The other, bare nominal forms, could
then be tagged as nominative (Case=Nom), or, optionally, left with the Case
feature empty.
- Akocikanihk nikī-ahāwak nitastisak. “I put my mitts on the shelf.” (mitts are grammatically animate, despite the semantics)
Adpositions may accompany the locative, too:
- Nikī-nīhtināw akocikanihk ohci. “I took him/her/it down from the shelf.”
TO DO: Discuss somewhere the layered features that are needed when multiple arguments are cross-referenced by verbal morphology. Try to standardize the layer names.
Can Adjectives Have Core Arguments?
Under certain circumstances, yes. One possibility is that the adjective is used as a non-verbal predicate, possibly with a copula. Then it usually has a subject child (nsubj or csubj).
[en] Mary has been very kind to us.
Participles
may be tagged as either verbs or adjectives.
In both cases they retain certain features that are normally associated with
verbs. If a participle is derived from a transitive verb and tagged ADJ,
we have an adjective with object:
[cs] Řidič opravující auto musí mít reflexní vestu. “A driver repairing a car must wear a reflective vest.”
It is unusual for a non-participial adjective to have a core object. One English example is worth, as in
[en] It is worth $10.
(At least this analysis occurs in UD English EWT 2.2. The putative object resembles objects of verbs in that it is obligatory and it is a bare nominal, without a preposition. However, it is still not treated as an object of a verb, and the clause is not transitive. There is no passivization pattern that would promote the object ($10) to the subject position, and remove the original subject (it).)
Attributively used adjectives (regardless whether they are participles or not) do not have subjects.
Related issues:
Can Adverbs Have Core Arguments?
Under certain circumstances, yes. One possibility is that the adverb is used as a non-verbal predicate (typically indicating a location), possibly with a copula. Then it usually has a subject child (nsubj or csubj).
[en] We can be outside.
Converbs
may be tagged as either verbs or adverbs.
In both cases they retain certain features that are normally associated with
verbs. If a converb is derived from a transitive verb and tagged ADV,
we have an adverb with object.
Being non-finite forms, converbs do not have subjects.
There are currently no examples of adverbs that are not converbs and have objects.
Can Nouns Have Core Arguments?
No, with one exception. The noun may be used as a non-verbal predicate, possibly with a copula. Then it usually has a subject child (nsubj or csubj).
[en] John is a teacher.
In this case, the subject John is not part of the same nominal phrase as teacher. It is the subject of the entire clause, whose predicate is teacher.
UD makes a clear distinction between modifiers of a nominal on one side,
and everything else on the other side. Even verbal nouns,
if they are tagged NOUN and not VERB, must have their modifiers attached via
relations reserved for modifiers of nouns.
For example, in [en] to take action, action is an object of to take (obj(take, action)).
If the phrase is nominalized, the coding of the argument becomes oblique (requiring the
preposition of) but the relation is neither obj nor obl; it is nmod:
[en] Taking of any action is prohibited.
Clausal Complements
So far we were dealing with nominal arguments. But sometimes an argument is realized as a clause. If it occurs in subject position, it is labeled as clausal subject.
TO DO: What does “subject position” mean in free word order languages? What other subject properties can be tested with clauses?
Otherwise it may be a clausal complement.
Unfortunately, the term is somewhat misleading. UD does not distinguish arguments from adjuncts
(in other words, complements from modifiers), at least not at the universal level.
What it does distinguish is core arguments from oblique arguments + adjuncts.
It is thus not sufficient that the clause fills a slot defined by the valency of the matrix verb,
i.e., that it is in some sense obligatory. We must require more. We must require that the clause
corresponds to a core argument, that is, obj (or iobj, at least in theory).
If it corresponds only to an oblique
argument (or an adjunct), it should be labeled as adverbial clause.
Consequently, clausal complements can depend on verbs but not on adjectives or adverbs
because the latter two normally do not have core objects.
If an adjective or adverb is modified by a clause, that clause must be advcl.
Even if it fills an obligatory slot in the adjective/adverb’s valency frame.
(Analogically, if a clause complements a noun, it must be labeled as adnominal clause.)
Even clauses that alternate with direct nominal objects (and thus are labeled ccomp)
often do not work the same way as their nominal counterparts. For instance, English
direct objects can be passivized, as in:
- Tell me the secret.
- The secret will be told (to me).
- I will be told the secret.
However, a clausal complement of the same verb does not passivize the same way. It needs an expletive subject instead:
- Tell me that you will come.
- It will be told (to me) that you will come.
- I will be told that you will come.
Note that UD does not distinguish direct and indirect clausal complements.
At present it is assumed that ccomp alternates with obj (direct object).
TO DO: Find Polish LFG examples where there are two ccomps in one sentence, one corresponding to obj and the other to iobj.
Summary of Relations
- nsubj — nominal subject, corresponding to the S function in intransitive clauses and to the A function in transitive clauses.
If some of the
nsubjsubtypes are used in the language, then the barensubjdenotes the remaining situations not covered by the subtypes.- nsubj:pass — nominal subject in a passive clause
- nsubj:caus — nominal subject in a causative clause
- obj — nominal object, corresponding to the P function in transitive clauses.
- iobj — nominal indirect object, one of two objects in ditransitive clauses, the one that is “less core-like” of the two.
If the two objects are equal in there core-likelihood, none of them is
iobj. If some of theiobjsubtypes are used in the language, then the bareiobjdenotes the remaining situations not covered by the subtypes.- iobj:agent — true agent in a causative clause, if encoded as an indirect object.
- obl — nominal oblique argument or adjunct.
If some of the
oblsubtypes are used in the language, then the bareobldenotes the remaining situations not covered by the subtypes.
TO DO: ccomp, xcomp