Grammatical relations

Grammatical relations are not universal linguistic phenomena in RRG. To posit grammatical relations in a language, one must obtain evidence of restricted neutralization. That is, a semantic or pragmatic relation is neutralized for syntactic purpose, and the neutralization must be restricted for actors and undergoers instead of other arguments. Restricted neutralization can be attested in two aspects. One is the coding property such as verbal agreement. Take English for example. A a verb agrees with a third person actor in a declarative active construction as in (2.6a). A third person undergoer also renders this mechanism in (2.6b). A similar phenomenon is found in the active and passive sentences in (2.7). The constant pattern of agreement is insensitive to the semantic contrast but sensitive to the syntax. Moreover, the neutralization is restricted because (i) the actors and undergoers (i.e. neutralization) trigger the verbal agreement and (ii) only certain semantic roles: non-third person actors and undergoers (i.e. restriction) do not trigger such mechanism.

(2.6) English

a. The hound bites the boar. (Actor agreement)

b. The hound dies. (Undergoer agreement)

(2.7) English (Van Valin & LaPolla 1997:251)

a. John kills the ducklings. (Actor agreement) b. The ducklings are killed by John. (Undergoer agreement)

Restricted neutralization can also be attested in the behavioral property such as control of the missing argument in a complex construction. The want-construction in English is a clear example as shown in (2.8). First, the omitted arguments in the infinitival cores are irrelevant to the contrast of semantic roles, which indicates the neutralization. Second, the missing arguments are restricted to actors (2.8a-b) and undergoers (2.8c-d) instead of other arguments. This is evidence of neutralization because the choice of the missing argument cannot be stated in purely semantic terms (Van Valin 2005:252). The crucial point here is (2.8e) whereby the missing argument cannot be undergoer when the second core is the active construction. This is evidence of restriction.

(2.8) English (my examples)

a. Chrisi wants ____i to drink a beer. [Actor of trans. V]

b. Chrisi wants ____i to sing in the park. [Actor of intrans. V]

c. Chrisi wants ____i to be stronger. [Undergoer of intrans. V]

d. Chrisi wants to ____i be interviewed by the journalist.

[Undergoer of trans. V,

passive construction]

e.*Chrisi wants the journalist interviews ____i. [Undergoer of trans. V, active construction]

RRG considers that grammatical relations are construction-specific phenomena, since restricted neutralization found in one construction may not be identical in another construction within a language. One piece of evidence can be drawn from English relative clauses. As shown in (2.9), the heads of English relative clauses display neutralization. Relative pronouns can virtually carry any semantic roles because the heads function as AGENT, PATIENT, RECIPIENT, LOCATION and SOURCE. This pattern, different from the want-construction in (2.8), indicates the non-restrictive selection of semantic roles. Accordingly, English relative clauses do not involve grammatical relations.

(2.9) English (Van Valin 1997:253)

Mary talked to the man (a) who [AGENT] bought the house down the street.

(b) who [PATIENT] the dog bit.

(c) to whom [RECIPIENT] Bill sold the house.

Mary looked at the box (d) in which [LOCATION] the jewelry was kept.

(e) out of which [SOURCE] the jewelry had been taken.

RRG considers that plenty of languages exhibit grammatical relations, but these languages do not exhibit identical pattern of grammatical relations. That is, the notion of subject will not be exactly the same in languages that exhibit grammatical relations.

That is to say, the patterns of restricted neutralization vary from languages to languages. Van Valin & LaPolla (1997) use English, Warlpiri, Enga, and Dyirbal to

elaborate this claim. These four languages exhibit different patterns of restricted neutralization in terms of subjecthood. Table 2.3 summaries this part.

Table 2.3 Restricted neutralization of semantic roles (Van Valin & LaPolla 1997:269)

Intransitive verbs

Transitive verb

Grammatical relations

‘Subjects’

Warlpiri, Enga Yes no yes [S²², A]

English Yes yes yes [S, A, d-s]

Dyirbal Yes yes yes [S, U, d-s]

Acehnese No no no d.n.a.

Warlpiri and Enga exhibit restricted neutralization in intransitive verbs but not in transitive verbs. The single argument of an intransitive verb is reserved for either actor or undergoer. By contrast, it is only the actor that acts as the subject of a transitive verb or the missing argument of linked verb as in the want-construction.

English and Dyirbal behave alike, since both transitive and intransitive verbs exhibit restricted neutralization. Both languages allow using syntactic mechanism to derive actor or undergoer to be the subject, such as the passive or antipassive constructions.

A crucial point lies in Acehnese. This language does not apply to the notion of

‘subject’. Neither intransitive nor transitive verbs exhibit restricted neutralization because the selection of ‘subject’ is semantically sensitive but not driven by syntax.

22 The abbreviation S refers to the single argument of a predicate. A refers to actor. The abbreviation

Constructions such as the active clauses only allow actor to be the ‘subject’, while constructions such as possessor-raising²³ only allow the undergoer to be the ‘subject’.

To reiterate RRG’s claim of grammatical relations, any study on this issue must consider two fundamental aspects. One is to recognize that they are not linguistically universal. The other is that grammatical relations are not identical across languages.

The cross-linguistic diversity of grammatical relations reveals the difficulty to apply the traditional notions ‘subject’ and ‘object’ to all languages. Since this topic is not the central issue of this dissertation, I use theoretically more neutral terms such as nominative arguments and genitive arguments, instead of more sensitive terms like

‘subject’ or ‘direct object’ in my dissertation.