Grammatical Relations

In its broadest sense, the term grammatical relation (or grammatical role or grammatical function) can be used to refer to almost any relationship within grammar, or at least within SYNTAX and MORPHOLOGY. In its narrowest sense, grammatical relation is a cover term for grammatical subject, object, indirect object, and the like. To understand what grammatical relations are, in this more specific sense, it will help to contrast them with intuitively similar but distinct syntactic concepts such as thematic roles, Cases, and syntactic positions.

We can see the difference between surface grammatical relations and semantic relations or THEMATIC ROLES (e.g., agent, goal, theme, etc.) by examining examples (1) and (2). In these examples, the noun phrase "that story" has the same thematic role (theme or patient) in both the active and the passive versions of the sentence; however, the grammatical relation of "that story" differs in the two sentences. In the active sentence, "that story" has the grammatical relation of object, whereas in the passive sentence, it has the grammatical relation of subject.

(1) active: This girl wrote that story.
(2) passive: That story was written by this girl.

It is not necessary, however, to appeal to derived contexts in order to distinguish surface grammatical relations and thematic roles. Although the subject of an active sentence is very often an agent, there are sentences without an agent role and in such sentences, some other thematic role such as theme, goal, or experiencer is associated with the grammatical relation of subject:

(3) The ball rolled down the hill.
(4) The woman received the letter.
(5) The boy enjoyed the ice cream.

Grammatical relations are also distinct from Cases (e.g. nominative, accusative, dative, etc.). Although the grammatical relation of subject is often associated with nominative Case, whereas the grammatical relation of object is often associated with accusative Case, there are many examples of other pairings of these grammatical relations and Cases. For example, in Icelandic some verbs take a dative subject and a nominative object:

(6) Barninu batnaši veikin.
  the.child-DAT recovered.from the.disease-NOM(*ACC)
  'The child recovered from the disease.' (Yip, Maling, and Jackendoff 1987: 223)

Hindi also has verbs that take dative subjects, but most transitive verbs in Hindi take an ergative subject and a nominative object (in the perfective aspect):

(7) Raam-ne roTii khaayii thii.
  Ram(masc.)-ERG bread(fem)-NOM eat(perf, fem) be(past, fem)
  "Ram had eaten bread." (Mahajan 1990: 73)

Thus we see that grammatical relations are distinct from Cases.

Finally, grammatical relations can also be distinguished from syntactic positions. If an object is fronted in a topicalization construction, for example, it remains an object despite the fact that it is located above the subject in the syntactic structure.

(8) a. I read that book.
  b. That book, I read.

Similarly, grammatical relations remain constant across many word orders in languages that allow scrambling. For example, in all of the word order variants of the Hindi sentence below, Ram has the grammatical relation of subject and banana has the grammatical relation of object:

(9) a. Raam-ne kelaa khaayaa.
    Ram-ERG banana-NOM ate
  'Ram ate a banana.'
  b. Raam-ne khaayaa kelaa.
  c. Kelaa raam-ne khaayaa.
  d. Kelaa khaayaa raam-ne.
  e. Khaayaa raam-ne kelaa.
  f. Khaayaa kelaa raam-ne. (Mahajan 1990: 19)

So far, we have been discussing only surface grammatical relations; that is, the grammatical relations that hold at surface structure after all movements or other grammatical relation-changing processes have occurred. However, one may also speak of deep or initial grammatical relations. In both the active sentence in (10) and the passive sentence in (11), one may say that that banana has the initial grammatical relation of object.

(11) Ram ate that banana.
(12) That banana was eaten by Ram.

Thus there is a closer association between initial grammatical relations and thematic roles than there is between surface grammatical relations and thematic roles.

Almost all linguists use the terms grammatical relation, subject, object, and the like in a descriptive sense. However, theories of grammar differ widely with respect to the question of the theoretical status of grammatical relations. The controversy centers around the question of how many formal devices of what kind the correct theory of grammar includes. Proponents of RELATIONAL GRAMMAR (Perlmutter and Postal 1977; Perlmutter 1983) maintain that grammatical relations are primitive notions in the theory of grammar and that universal generalizations about language are formulated in terms of those primitives. Others maintain that no grammatical rules make crucial reference to grammatical relations, as distinct from thematic roles, Cases, or syntactic positions and that adding grammatical relations to the theory of grammar is unnecessary and therefore undesirable (e.g., Chomsky 1981; Hoekstra 1984; Williams 1984; Bhat 1991). Between these poles, there are several middle and variant positions. Some take the position that although grammatical relations are not primitive notions of grammar, they do play an important role in grammar as derived notions (Anderson 1978). Some argue for the need for finer-grained grammatical relations, such as adding "restricted object" (Bresnan and Kanerva 1989), within the theory of LEXICAL FUNCTIONAL GRAMMAR.

Many works whose titles contain the phrase "grammatical relations" are not so much concerned with this theoretical controversy, but rather with a somewhat broader sense of grammatical relations, pertaining to how Case, agreement, and/or word order identify or distinguish subjects and objects. In this broader sense, a "theory of grammatical relations" is assumed to include a theory of Case and agreement systems that can account for all of the cross-linguistic differences that occur (see TYPOLOGY). Such work often focuses on languages or particular constructions in which the familiar associations between subjects and nominative Case or objects and accusative Case do not hold. For example, they address the question of how or why dative or ergative Case is assigned to subjects in constructions such as (6) or (7), instead of nominative Case; and why nominative Case is instead assigned to the objects in those constructions. One approach to this problem of nonprototypic associations between Cases and grammatical relations has been to propose that at some level, the prototypic association (e.g., between nominative Case and subjects) actually does hold, but that grammatical relations (or structural positions) are inverted at some level in the derivation (e.g., Harris 1976; Marantz 1984). Others maintain that assuming such a close association between Case and grammatical relations (or syntactic positions) is not correct and that there are conditions under which subjects can take other Cases, especially lexical (inherent, quirky) Cases, freeing up the nominative Case which can then be assigned to an object (e.g., Yip, Maling, and Jackendoff 1987).

Other works with "grammatical relations" in the title focus on the question of how to describe and analyze a range of constructions that appear to involve changes in grammatical relations, such as passive, causative, or applicative constructions, or on restrictions on various grammatical processes such as relativization that may be stated in terms of grammatical relations (e.g., Gary and Keenan 1977; Marantz 1984; Baker 1988).

-- Ellen Woolford

© 2003 MIT Press