Grammatical relations

From KneeQuickie
Jump to: navigation, search
Grammatical Relations

The main grammatical relation is "Subject"; all other grammatical relations are called "objects". If there are more than one object, the main one may be called "primary object", or it may be called "direct object", depending both on the language, and on the linguist writing about it. The other object(s) may be called "secondary object(s)", or it(they) may be called "indirect object(s)", once again depending both on the type of language, and on the school of linguist writing about it. Apparently no language allows more than four grammatical relations. If a language allows four, then two of them are "secondary objects" or "indirect objects". A clause may not have more than one of the same kind of grammatical relation, so if a language allows a clause to have two secondary or indirect objects, they will nevertheless be of different kinds.


Mapping Theory

An MAP is a Morphosyntactically-Licensed Argument Position. MAPs are closely related to GRs, but may not be considered identical by every linguist (just as some linguists distinguish grammatical functions or syntactic relations or syntactic functions from GRs). MAPs are direct, core terms. They are distinguished from oblique arguments (and also from adjuncts) by agreement and/or case-marking and/or word-order. In particular,

  • Only direct core terms must be agreed with by the verb; verbs never agree with obliques nor adjuncts. (But verbs may not have to agree with all direct core terms, either; depending on the language, the verb may not agree with any, or may agree with at most one, or at most two, etc.)
  • In languages with cases, direct core terms have "grammatical" or "structural" cases. A "structural" case is different from a "semantic" case; "semantic" cases convey the noun's semantic or thematic role, while "structural" cases can mark nouns with a (usually wide) variety of semantic roles; but the verb (usually by voice-marking) determines which semantic role each structurally-cased noun occupies. Nouns marked with semantic cases, OTOH, consistently have one of a few semantic roles, not varying from one verb to another.
  • Especially if the verb doesn't have to agree with a direct core term, and/or it isn't marked with a structural case, (and, frequently, even if the verb does have to agree with it or it is marked by a structural case), a direct core term may have to come in a fixed word-order relative to the verb and to the other direct core terms. Obliques and adjuncts, OTOH, usually may be distributed fairly freely relative to each other; sometimes or in some languages they may also be distributed fairly freely relative to the core, so long as they don't interfere with the core-internal word-order of the members of the core.

For most of the below-listed "syntactic privileges" of Subjects and "correlates" of Absolutives, there are some languages where they are also shared by some of the other MAPs in that language. For instance, sometimes MAPs in addition to the subject, or in addition to the subject and the direct object, can antecede reflexives; and/or can "raise"; and/or float quantifiers.

Subject

Keenan’s “Subject Properties List”

Keenan, Edward L. (1976) "Towards a Universal Definition of 'Subject'", in Charles N. Li (editor) Subject and Topic, New York: Academic Press (pp. 303-333)

How to Use the List

Keenan suggested that any pronoun or noun-phrase that had more of the properties on this list than it didn't have, could be called "a subject" of its clause; and any pronoun or NP that had more of these properties than any other pronoun or NP in its clause, could be called "the subject" of its clause. (I need to insert a reference to support this.)

The List

  • Autonomy Properties
    • Independent Existence
    • Indispensability
    • Autonomous Reference
      • Possible Controller of Stipulated Coreference
        • Can Control Reflexive Pronouns
        • Possible Controller of Co-Referential Deletions and Pronominalizations
      • Possible Controller of Switch-Reference Indicator
      • Control Verb Agreement If Anything Does
      • Easiest to Stipulate Co-Reference Across Clause Boundaries
        • Reflexive Pronouns of Verbs of Thinking in Sentential Complements which are Bound by NPs in the Matrix Clause Can Occur as Subjects in the Complement Clause
        • NPs which Can be Co-Referentially Deleted in Sentential Complements when Co-Referential with Matrix NPs Always Include Subjects
        • NPs which Can be Co-Referentially Deleted Across Co-Ordinate Conjunctions Include Subjects
        • NPs which Can be Co-Referentially Deleted Under Verb Serialization Usually Include Subjects
      • Absolute Reference
      • Presupposed Reference
      • Metaphoric Idioms
      • Topic
      • "Highly Referential" NPs (e.g. Personal Pronouns, Proper Nouns, and Demonstratives), Can Always Occur As Subjects
      • Subjects Are the Most Natural Targets of "Advancement" Transformations.
      • Subjects Have Wider Logical Scope than Non-Subjects.
      • Subjects Usually Occur Earlier than Any Other Noun Phrase in Basic Sentences.
      • Subjects Can Be Relativized, Questioned, and Cleft.
      • Possessors of Subjects Can Be Relativized, Questioned, and Cleft.
      • A Relativized Subject is Rarely a Personal Pronoun; a Relativized Personal Pronoun is Rarely a Subject.
      • Subjects Can Undergo Raising.
      • Subjects Can Be Expressed by Independent and/or Emphatic Pronouns which Can Be Conjoined with Full NPs.
      • Subjects Can Launch Floating Quantifiers.
  • Case Marking Properties
    • If Any NPs Are Not Case-Marked, the Subjects of Intransitive Basic Sentences Are Not Case-Marked.
    • Subjects Change Their Case Marking Under Causativization.
    • Subjects Change Their Case Marking Under Action Nominalizations.
  • Semantic Role
    • The Semantic Role of the Subject is Predictable from the Form of the Main Verb.
    • If There Is An Agent of a Basic Sentence the Subject will Normally be the Agent.
    • The Addressee of Imperatives is Normally the Subject.
    • Subjects of Basic Sentences are Usually in the Same Position, Case-Marking, and Verb-Agreements as Causers of Causative Sentences.
  • Immediate Dominance



The Notion of “Subject”

Absolutive

Keenan’s “Correlates of the Absolutive vs Ergative Distinction”

Keenan, E.L. 1984. Semantic correlates of the ergative/absolutive distinction. Linguistics 22, 197-223.

  • Bondedness to the Verb. This is the opposite of "Immediate dominance" in the Subject Properties List above. Absolutives tend to be the "siblings" of verbs.
    • Existence Dependency. The clause may indicate a change in the existence-status of an absolutive noun-phrase. That is, it may denote that the absolutive is coming into existence (or being created), or, that it is going out of existence (or being destroyed). This is the opposite of the "independent existence" property on the Subject Properties List.
    • Multiple Senses. The meaning of a verb can vary according to the nature of its absolutive argument in many ways, some unpredictable. On the contrary, the meaning of a transitive verb is likely not to vary according to the nature of its transitive subject; or, if it does, it will vary in few ways, all of them quite predictable.
    • Selectional Restrictions and Verbal Classifiers: "As noted by Moravcsik, predicates in a language may impose highly specific selectional restrictions on absolutive arguments, but typically impose only weak and rather general restrictions (such as humanness or animacy or concreteness) on transitive subjects." "Given an arbitray predicate in English we may infer more about the semantic nature of its absolutive argument than, if the predicate is transitive, we can infer about its transitive subject." "A morphologically more regular, but semantically similar, phenomonon, is the 'object' classifiers on verbs in various Amerindian languages. For instance Navaho has six; applied to transitive verbs they indicate that the direct object is a round solid object, or a long slender object, or a wool-like mass, or a mud-like mass, etc.; applied to intransitive verbs they indicate the same about the intranstive subject. But there are no such classifiers to indicate the properties of transitive subjects."
    • Noun Incorporation. If a noun is incorporated into a verb, it is much, much likelier that it is an absolutive argument that is so incorporated, than that a transitive subject is so incorporated. Also, if the transitive subject is incorporated, it is "less incorporated" than an absolutive would be.
  • Thematic Role
    • Patient. Absolutives are often Patients; their existence state is affected by the action expressed by the predicate. Transitive subjects rarely are patients in the same sense.
    • Theme. The set of arguments whose path of movement is specified by the Goal and Source locatives of verbs of motion, always includes the Absolutive arguments. The path of movement of a transitive subject cannot be so specified unless the path of the absolutive argument is also specified.
  • Control Phenomena
    • Control of Predicated Adjectives. Adjectives within predicates are normally understood to refer to the absolutive argument. If they are understood to modify the transitive subject instead, this is marked and exceptional.
    • Control of Predicate Infinitives. It is overwhelmingly more often the absolutive, rather than the transitive subject, which "controls" the subject position of any infinitive within a predicate.
Primary Object

Any grammatical relation that isn't a subject is called an object. For some languages ("primative" languages), the main object is called a "primary object" by all linguists. For other languages ("directive" languages), some call the main object a "direct object", but others still call it a "primary object".

Is There a Universally accepted Notion of “Primary Object”?

One of the problems here is that a "primary object" in a primative language is somewhat different from a directive language's "direct" (or primary) object. Another is that in a syntactically ergative language, where the absolutive argument deserves to usually be considered the syntactic subject, the ergative argument seems the natural candidate for "primary object". Apart from those considerations, we might say (please note this is my own suggestion -- I'm not sure how many professional linguists would like it, nor how much eldin raigmore 00:22, 3 March 2009 (UTC));

  • Any object that has more "Absolutive Correlates" than it lacks, and also has more "Absolutive Correlates" than any other object (any other grammatical relation except the subject), might deserve to be called the "Direct Object". As a "tie-breaker", (that is, if it has almost as many "Absolutive Correlates" as it lacks, but not more, and at least as many as any other object, but not more than all), if it has more "Subject Properties" than any other object (any other grammatical relation except the subject), it should be called "the Direct Object".
  • Any object that has more "Subject Properties" than it lacks, and also has more "Subject Properties" than any other object (any other grammatical relation except the subject), might deserve to be called the "Primary Object". As a "tie-breaker", (that is, if it has almost as many "Subject Properties" as it lacks, but not more, and at least as many as any other object, but not more than all of them), if it has more "Absolutive Correlates" than any other object (any other grammatical relation except the subject), it should be called "the Primary Object".

Keenan's "Subject Properties List" Revised for Primary and/or Direct Objects

  • Autonomy Properties
    • Independent Existence: Not a correlate of absolutives, so doesn't apply to Direct Objects; however it does apply to Primary Objects.
    • Indispensability: The Primary or Direct Object may be even less dispensible than the Subject.
    • Autonomous Reference
      • Possible Controller of Stipulated Coreference
        • Can Control Reflexive Pronouns: Necessarily second to the subject, if at all. Depends on whether the language's reflexive pronouns can cross clause-boundaries (in which case they can co-refer only with subjects), or can co-refer with objects (in which case the co-referring noun-phrase must be a clausemate higher in the Noun Phrase Hierarchy than the reflexive pronoun).
        • Possible Controller of Co-Referential Deletions and Pronominalizations: I'm not sure about this yet.
      • Possible Controller of Switch-Reference Indicator: There are three kinds of switch-reference indicators; same vs different subject, same vs different object, and same vs different location. Most commonly the "same" referent is the subject of either the marked clause or of the reference clause. If instead it is an object of both clauses, it is likelier to be the primary or direct object of each than to be a secondary or indirect object of both.
      • Control Verb Agreement If Anything Does: Second to subjects, primary objects are likeliest to control verb agreement. Some languages with poly-personal agreement whose grammar is mostly "directive", nevertheless have a "primative" agreement system; the verb agrees with the subject and with the primary object (whether or not that is also the direct object). But some "directive" languages with poly-personal agreement have their verbs mostly agreeing with the subject and with the direct object.
      • Easiest to Stipulate Co-Reference Across Clause Boundaries: For the most part, primary objects or direct objects must be second on each of the following lists, after the subject.
        • Reflexive Pronouns of Verbs of Thinking in Sentential Complements which are Bound by NPs in the Matrix Clause Can Occur as Subjects in the Complement Clause: Or as primary or direct objects, second only to subject.
        • NPs which Can be Co-Referentially Deleted in Sentential Complements when Co-Referential with Matrix NPs Always Include Subjects: And if they include anything else, they include primary or direct objects.
        • NPs which Can be Co-Referentially Deleted Across Co-Ordinate Conjuntions Include Subjects: And if they include anything else, they include primary or direct objects.
        • NPs which Can be Co-Referentially Deleted Under Verb Serialization Usually Include Subjects: And if they include anything else, they usually include primary or direct objects.
        • But note that primary or direct objects are likelier to co-refer to the subjects of infinitives which are embedded in the predicate.
      • Absolute Reference: More a property of primary objects than of direct objects. Second place to subjects.
      • Presupposed Reference: More a property of primary objects than of direct objects. Second place to subjects.
      • Metaphoric Idioms: I don't see how these apply to primary or direct objects.
      • Topic: New topics are usually introduced as primary or direct objects, the first time they occur.
      • "Highly Referential" NPs (e.g. Personal Pronouns, Proper Nouns, and Demonstratives), Can Always Occur As Subjects: And, to a lesser degree, as primary objects. But highly non-referential NPs are typically direct objects.
      • Subjects Are the Most Natural Targets of "Advancement" Transformations:
        • and in "applicative" languages, primary, or direct, objects, are the other most natural targets.
        • and in languages with ditransitive verbs, "dative movement", which is essentially applicativization that promotes the secondary or indirect object to the primary-or-direct-object "slot", naturally "targets" the primary or direct object.
        • Direct Objects, or Primary Objects, are always "promotable" to Subject by a language's Voice system ("passivization"), if anything is. In some languages, no single transformation can "promote" anything else (other than the Primary or the Direct Object, whichever the language has) to Subject; in such languages, in order to promote anything else to Subject, first it has to be promoted to Primary or Direct Object by some "applicativization" transformation.
      • Subjects Have Wider Logical Scope than Non-Subjects: I don't yet see how to revise this one to apply it to objects.
      • Subjects Usually Occur Earlier than Any Other Noun Phrase in Basic Sentences: But primary or direct objects usually occur closer to the verb than any other noun phrase in basic sentences.
      • Subjects Can Be Relativized, Questioned, and Cleft. Ordinarily the Noun-Phrase Accessibility Hierarchy governs this; Direct Objects, or Primary Objects, are second after Subjects on this hierarchy, so, if anything besides Subjects can be Relativized or Questioned or Cleft, Primary or Direct Objects can.
      • Possessors of Subjects Can Be Relativized, Questioned, and Cleft. Ordinarily the Noun-Phrase Accessibility Hierarchy governs this; Direct Objects, or Primary Objects, are second after Subjects on this hierarchy, so, if the possessors of anything besides Subjects can be Relativized or Questioned or Cleft, then the possessors of Primary or Direct Objects can be.
      • A Relativized Subject is Rarely a Personal Pronoun; a Relativized Personal Pronoun is Rarely a Subject. I don't know about objects.
      • Subjects Can Undergo Raising. Generally, they are raised to primary or direct objects.
      • Subjects Can Be Expressed by Independent and/or Emphatic Pronouns which Can Be Conjoined with Full NPs. I don't know about objects.
      • Subjects Can Launch Floating Quantifiers. I don't know about objects.
  • Case Marking Properties
    • If Any NPs Are Not Case-Marked, the Subjects of Intransitive Basic Sentences Are Not Case-Marked. Obviously this applies to absolutives as easily as to subjects; it may depend on the morphosyntactic alignment of the language.
    • Subjects Change Their Case Marking Under Causativization. I don't think this can be modified to apply to objects.
    • Subjects Change Their Case Marking Under Action Nominalizations. I don't think this can be modified to apply to objects.
  • Semantic Role
    • The Semantic Role of the Subject is Predictable from the Form of the Main Verb. Especially in languages with applicatives, the same is likely of the Primary Object or the Direct Object.
      • If There Is An Agent of a Basic Sentence the Subject will Normally be the Agent. Depending on the morphosyntactic alignment of the language, it may be that the Absolutive is usually the Subject, and, if there is an Agent, it is usually the Primary Object.
      • The Addressee of Imperatives is Normally the Subject. Normally it isn't an Object.
      • Subjects of Basic Sentences are Usually in the Same Position, Case-Marking, and Verb-Agreements as Causers of Causative Sentences. Primary Objects, on the other hand, are likelier to correspond to Causees ("Agents of Effect"). I'm not sure about Direct Objects.
  • Immediate Dominance: Exactly the reverse of the case for Subjects. Primary Objects and/or Direct Objects are probably the closest to the verb; the node just above the verb is probably also the node just above the Absolutive participant, which, if it isn't the Subject, is probably the Primary Object or the Direct Object.


Do Primary Objects Occur in Nearly Every Language?

Most languages seem to have either two or three grammatical relations. Many seem to have one and only one -- the Subject. Possibly more have two than have three. Few seem to have four, and none seem to have more than four. There is a great deal of controversy over whether or not a few languages have no grammatical relations at all. But there is widespread agreement that nearly all have at least one; and that the languages which, controversially, don't have any, uncontroversially, don't have more than one.

Secondary Object

Languages that Have Secondary Objects vs. Languages that Don't

In most transformations that involve the promotion or demotion of a participant, the participant is promoted to or demoted to the language's lowest MAP on the Accessibility Hierarchy.

If a language has secondary or indirect objects, then that is the language's third MAP or third GR. If the language case-marks it, its case will be structural rather than semantic; if that's true that proves it is the third GR. Possibly trivalent verbs (and higher-valent verbs, if there are any) will be required to agree with it; if that's true that also proves it is the third GR. Otherwise, (and, possibly, in addition) it may be required to have a fixed word-order with respect to the verb, the subject, and the primary or direct object.

But there are additional tests. Languages with three GRs tend to be "indirect-object oriented". This means they tend to have "dative applicatives" rather than applicatives; and that causativizing a monotransitive clause will result in the causee being demoted to secondary object rather than to an oblique. Such languages are likelier to have a "dative" MSA than a "dechticaetiative" MSA.

If a language's suspected "indirect objects" are not secondary objects after all, then they are an oblique argument, not one of the language's GRs. If the language case-marks it, its case will be semantic rather than structural; it can have only one of a few similar semantic roles, (perhaps "Recipient, Beneficiary, or Causee"), no matter which verb, or what form of the verb, is involved. Possibly trivalent verbs (and higher-valent verbs, if there are any) will never be required to agree with more than two participants. The "indirect object" may have fairly free word-order compared to other oblique arguments and to adjuncts, and, possibly to the core as a whole (provided it doesn't interfere with the relative word-order of the subject, verb, and primary or direct object).

As for the additional tests: Languages with two GRs tend to be "direct-object oriented". This means they tend to have applicatives rather than "dative applicatives"; and that causativizing a monotransitive clause will result in the causee being demoted to an oblique case which is probably not the same as the "indirect object"'s case. Also, such languages are likelier to have a "dechticaetiative" MSA than a "dative" MSA.


One Language’s Secondary Object May Differ from Another’s

Dechticaetiative vs. Dative Languages

In ditransitive clauses, a language with "dative" morphosyntactic alignment will treat the Theme the way it treats the Patient in monotransitive clauses (with regard to Agreement, and/or Case, and/or Word-Order); but the Recipient will receive a different, "Dative" treatment.

In ditransitive clauses, a language with "dechticaetiative" morphosyntactic alignment will treat the Recipient the way it treats the Patient in monotransitive clauses (with regard to Agreement, and/or Case, and/or Word-Order); but the Theme will receive a different, "Dechticaetiative" treatment.

Most dechticatieative languages are "direct-object oriented"; and most have only two GRs (and/or only two MAPs, if that's different). Conversely most direct-object oriented languages, and most languages with just two GRs, are either dechticaetiative, or partially dechticaetiative, or have a tendency towards partial dechticaetiativity.

However, some linguists think some languages are Dechticaetiative in alignment, but do have three GRs, including secondary objects; these linguists think that in these languages, the Theme in ditransitive clauses is usually a secondary object rather than an oblique argument.

Most dative languages are "indirect-object oriented"; and most have at least three GRs (and/or three MAPs, if that's different). Conversely most indirect-object oriented languages, and most languages with three or more GRs, are either dative, or partially dative, or have a tendency towards partial "dativity" (if that's a word).

However, some linguists think some languages are Dative in alignment, but do have only two GRs; they think these languages' "indirect objects" are more oblique arguments than proper secondary objects; these linguists think that in these languages, the Recipient in ditransitive clauses is usually an oblique argument rather than a secondary object.



The Noun-Phrase Accessibility Hierarchy

Comrie & Keenan came up with this hierarchy;

 Subject > Direct Object > Indirect Object > Oblique Argument > Possessor > Object of Comparison

Keenan, Edward L. & Comrie, Bernard 1977, "Noun phrase accessibility and universal grammar." Linguistic Inquiry 8(1):63-99.

Suitably modified to fit the language's morphosyntactic alignment, it successfully predicts, in languages which have relative clauses, which NP of an RC can be relativized; also, it successfully predicts which NPs require a different relativization strategy.

Comrie & Keenan initially proposed that this hierarchy also governed which position the causee (the agent-of-effect) would be demoted to in a morphologically causativized clause (assuming the agent was the subject before causativization). In fact, though, later it was discovered that only a minority of languages work that way, although they do include several well-known and well-studied major languages with very many speakers.

The lower in the accessibility hierarchy an MAP is, the likelier it is to be doubled. Direct core terms are never tripled, but many languages allow obliques to be tripled. In some languages indirect objects may be doubled. Some linguists advise one another to be careful of accepting that a clause has two direct objects; but some linguists are fairly sure that some clauses in some languages do have two DOs. Clauses that were formerly analyzed as having two subjects are now analyzed as having a topic - comment structure in which the comment is a clause with only one subject.eldin raigmore 00:09, 3 March 2009 (UTC)