Functional grammar

from Wikipedia, the free encyclopedia

The functional grammar or engl. functional grammar (FG) is a linguistic theory in the late 1970s by Simon C. Dik was developed in Amsterdam, explicitly as an alternative model to the standard model of transformational grammar of Noam Chomsky . After Dik's death in 1995, the theory was further developed mainly by his colleague Kees Hengeveld and in its current form is still very close to the original formulation. Since 2004, the theory has been further developed under the name Functional Discourse Grammar (FDG), mainly by Kees Hengeveld and Lachlan Mackenzie (Hengeveld & Mackenzie 2008).

Central assumptions about language

Dik's central assumption about language is its purpose-built character as a means of communication. Dik thus focuses on the function of language . This is the sense in which the term Functional Grammar should be seen: A linguistic model that is based on the function of language rather than on its external form. With this central assumption, Dik calls for a departure from the heuristic reduction of pragmatics that was often used in the past . In concrete terms, however, Dik's treatment of pragmatics does not mean general pragmatics in the sense of speech acts and language as action, but the area of discourse pragmatics , essentially the relationship between the information structure of a linguistic expression and its realization, for example when dealing with topic and focus (Dik 1991: 267ff.).

The grammar formalism described by Dik is pragmatic-based in this sense. From Dik's point of view, the next most important linguistic level is semantics , which, itself influenced by pragmatics, in turn has an influence on syntax . An example of such an influencing of the syntax would be an active-passive alternation, which is determined by the semantic roles of the other players in the utterance and is semantically motivated in this sense.

At the same time, FG is a formal model because it uses methods of formal semantics (such as predicate logic ) and claims to be implementable as a computer program and thus testability.

Basic properties of functional grammar

As a grammar theory, functional grammar is a general theory about the grammatical organization of natural languages. Their main characteristics are:

  • Language is primarily seen as an instrument of social interaction
  • As such, v. a. the pragmatics matter. Syntax and semantics should be analyzed within the pragmatics.
  • The hierarchy between these three levels of analysis is clear: the most important thing for functional grammar is pragmatics, subordinate to it is semantics, which in turn is subordinate to syntax. There is neither an "autonomous" syntax, nor an "autonomous" semantics.
  • The desired grammar model should satisfy the forms of adequacy described in more detail below.
  • Three different types of functional and relational terms are mainly used to structure the functional grammar:
  1. Semantic functions encode various roles that the participants can take on a subject (agent, patient, recipient, etc.)
  2. Syntactic functions present alternative perspectives on the presentation of facts (subject, object)
  3. Pragmatic functions indicate the distribution of information, which depends on the respective situation (topic, focus)
  • Linguistic expressions are first described on the basis of abstract predication patterns that are built up using predication patterns recorded in the lexicon.
  • Expression rules then determine the form and order of the constituents, which are fundamental for the respective linguistic expression. The categorical and functional properties within the predication are decisive.
  • The functional grammar avoids structure-changing transformation rules.

Ideas about grammar

Basic structure of grammar formalism

The grammar formalism of the FG essentially consists of the description of abstract expressions, the Underlying Clause Structures (UCS), which are gradually formed from predicates and terms and which are related to or generate concrete linguistic utterances through expression rules.

Structure of the underlying clause structure

Predicates

The UCS are formed from predicates and terms. Some elementary predicates and terms are part of the lexicon , others are created from these elementary predicates and terms (using predicate formation and term formation ). So the predicate for throw back would be a predicate derived from the elementary predicates for throw and back . All predicates and terms together form the fund of a language.

Predicates are expressions for properties or relations. These are predicates in the sense of predicate logic , not the grammatical relation of the predicate from the Latin school grammar . In this sense, not only verbs are predicates, but all content words of a language. So "house (x)" is just as much a predicate as "hit (x, y)".

A difference between the predicates in the FG and the classic predicate logic is the use of so-called restrictors. If predicates are put together in the FG, this is done through the use of these restrictors, written as ":", for example in the form "Japanese (x): Buddhist (x)". This can be paraphrased with "The set of x for which: x is Japanese, restricted to the set of x for which: x is Buddhist". The corresponding predicate logic form would be "Japanese (x) & Buddhist (x)", where the "&" is a predicate logic "AND". The relevant facts are the same in both cases. The difference is that the predicate logic "&" is reversible, so that "Japanese (x) & Buddhist (x)" is equivalent to "Buddhist (x) & Japanese (x)". This is not the case with the restrictors and they are thus able to grasp the difference between the linguistic utterances The Japanese Buddhist and The Buddhist Japanese (Dik 1997, Section 6.2).

Predicates are always part of a predicate frame that describes the properties of the predicate. An example of the predicate frame of a transitive verb would be: 

 throw[V](x1:<animate>(x1))Agent (x2:<concrete>(x2))Goal (x3)Direction

First the word form (throw) appears, then the part of speech (V). The following describes the argument positions of the verb. The argument in the player position with the semantic role of the agent must be animated (animate), the player affected by the facts ( Goal , generally the term patient or patient has prevailed for the role that Dik calls goal ), here the thrown The subject must be concrete and the third argument (with the semantic role of location ) is not subject to any such selection restriction.

In addition to such nuclear predicates, the optional so-called satellites can now be added, which take positions that are not specified by the predicate framework, for example for a temporal specification of the predicate with the help of words such as yesterday or soon . Dik calls such a predicate frame extended by satellites an extended predicate frame .

Terms

The second essential component of an Underlying Clause Structure (UCS), in addition to predicates, are the terms. Formally, terms are the arguments of the predicates, semantically they are expressions that reference entities (Strictly speaking, Dik (1991: 255) writes that terms instruct the addressee to identify an entity that corresponds to the profile of the term). Examples of terms would be Das Haus or The Purple Plastic Bag . There are only very few elementary terms, only proper names and personal pronouns are available as elementary terms, other terms, such as the purple plastic bag, are created from predicates. Terms are the entities that are related to each other by a predicate.

A predication containing two terms ( the garden and the dog ) would be e.g. B .: 

 present: (definite singular x1:garden [N])Location (definite singular x2:dog [N])

Levels in the UCS

Within the Underlying Clause Structure (UCS), three different levels are distinguished in the form of functions:

  • Utterance situation: level of pragmatic functions such as topic and focus.
  • Player level: level of the semantic functions such as agent and goal ( patient ).
  • Perspective level: level of the syntactic functions of subject and object.

In this sense, individual elements of a linguistic utterance take on different categories at different levels. In the sentence Peter buys ice cream, for example, Peter is agent, topic and subject at the same time , while ice cream is goal (patient), focus and object at the same time.

The linguistic expression used by the expression rules on the UCS 

 present: (definite singular x1:garden [N])Location(definite singular x2:dog [N])

can be obtained (or generated from UCS in an implementation of the formalism) is not yet clear. The UCS corresponds roughly to the statement The dog is in the garden . In a certain utterance situation (for example in a list of the obstacles to be overcome for a break-in) the following utterance would be conceivable, which also agrees with the UCS: There is the dog in the garden . This example illustrates the possibilities offered by taking the pragmatic level into account, because the characterization of the enumeration makes it possible to distinguish the two linguistic expressions in the underlying structure (Dik 1997, Section 8.7.2).

The differentiation of the level of the semantic roles , ie the players and the syntactic (grammatical) relations enables the description of syntactic alternations such as passivation without one form having to be derived from the other. In an active sentence, there is a match between subject and agent, while if subject and goal (patient) in the UCS match, this would correspond to a passive sentence.

Operators on Predications

If the terms have been inserted into the predicate frames as described , we have a predication that contains the full proposition or state of affair (SoA) of the sentence, but is not yet specified further. For this purpose, operators are now applied to the entire predication, for example in the Underlying Clause Structure (UCS) above the operator "present", which can itself be seen as a predicate with the full predication as an argument. Operators for the mode (such as interrogative or declarative ) are also inserted in this step .

This predication, which is now fully specified, is finally specified with the help of expression rules for form, order and intonation and thus related to a concrete linguistic utterance (in the description) or converted into such (in the generation).

Summary of grammar formalism

FG is therefore a monostratal model, because although a distinction is made between the Underlying Clause Structures (UCS) and the linguistic expressions and these are related to one another by expression rules, no different levels are assumed on which concrete linguistic expressions are, see above there are no syntactic derivation mechanisms. In this sense, the formation of the linguistic utterances takes place step by step within a process chain, on a single level.

The pragmatism orientation makes FG a descendent grammar model that starts from the overall situation in which an utterance is made, in contrast to an ascendant grammar model that starts from the smallest parts, e.g. from phonology to morphology to syntax .

Treatment of the data

Dik generally attaches great importance to linguistic data: "Whenever there is some overt difference between two constructions X and Y, start out on the assumption that this difference has some kind of functionality in the linguistic system" (Dik 1997, chap 1.6).

FG thus has an inductive character, since, similar to Bloomfield's descriptivism, it starts from concrete linguistic data, in contrast to a deductive model such as Generative Grammar according to Chomsky , where an ideal linguistic competence abstracted from the concrete use of language is the focus of the theory.

In the central areas of pragmatics and semantics , the FG is primarily dependent on questioning informants (elicitation) as well as consulting their own mother tongue insights ( introspection ). Other sources such as experiments or corpora cannot be used without further ado (a generation of semantic knowledge would possibly be possible through an automatic processing of corpora, e.g. to determine paradigmatic or syntagmatic relations) for the determination of semantic knowledge (e.g. for the selection restrictions in predicate frames).

To evaluate the overall model, on the other hand, corpora and thus spontaneous language data can also be used in the FG, for example to check whether utterances in corpora can be described by the FG formalism.

Claim of the model

The objective of the FG is very comprehensive, Dik (1997, chapter 1) formulates the following central question: "How does the natural language user (NLU) work?". This question clearly identifies FG as a model with a mentalistic claim.

Following the formulation of this central question, Dik identified five skills of the NLU that play essential roles in human communication:

  • linguistic capacity: ability to produce and interpret linguistic expressions.
  • epistemic capacity: ability to build and manage a knowledge base that is used for language processing.
  • logical capacity: the ability to draw conclusions from available knowledge.
  • perceptual capacity: the ability to perceive one's environment and take it into account when processing speech.
  • social capacity: ability to take the situation into account when processing language.

In addition, Dik formulates three completely different adequacy criteria of his own, alluding to the three adequacy criteria of descriptive, explanatory and observation adequacy demanded by Chomsky :

  • Pragmatic adequacy: Direct consequence of the basic assumption that language is a means of communication.
  • Psychological adequacy: Findings from psycholinguistic research on language acquisition, processing and interpretation must be taken into account.
  • Typological adequacy: The theory should be applicable to languages ​​of different typological status.

In particular, the requirement of typological openness gives the model a strongly description-oriented character, since such an openness would make it a universal description tool, as well as a universalistic claim that sees it as the goal to make generally valid statements about language as a whole, not about a specific language or To make language family.

As already mentioned in connection with the central question, the claim of psychological adequacy characterizes FG as a mentalistic model that, like generative syntax theory, seeks to be a model for human language ability, in contrast to purely application or description-oriented approaches such as HPSG .

In contrast to Chomsky's nativistic hypothesis, FG assumes that linguistic universals do not arise from innate properties, but from the necessities of human communication (in this sense, for example, the fact that all languages ​​have a distinction between functional and content words would be based on necessity relating the contents of a linguistic utterance to one another) as well as the physical and psychological constitution of the person (e.g. a restriction of the nesting depth of subordinate clauses due to the limited possibilities of human short-term memory), and can thus be characterized as a non-nativistic model.

Main task of linguistic research

The aim of research within the FG is to develop a language-independent formalism for describing language. This requires extensive adaptation of the existing formalisms to many different languages ​​(Dik 1991: 248). In this sense, the language description is a central subject of research in the FG.

Another research area results from the requirement of formalizability: The implementation of FG on a computer. Dik himself has mainly worked in this field since the 1980s. Dik's own and other implementations (e.g. Samuelsdorff 1989) use the logic-oriented programming language Prolog ( Programming in Logic ), which seemed to be particularly suitable due to its strong orientation towards predicate logic, but implementation is also possible in any other programming language. The work in this area focuses heavily on the generation and abstract representation of linguistic expressions, not on processing ( parsing ), which, like generation, is part of linguistic capacity .

In addition, the demands for pragmatic and psychological adequacy suggest a certain openness and interdisciplinary cooperation if findings from relevant subjects such as psychology and sociology are to be taken into account.

Application orientation and applicability

advantages

Dik's "Functional Grammar" seems to take into account many aspects of language that are neglected in other models but are important for a complete description of language:

  • A consideration of pragmatics , for example as described above for a list or to describe the difference between the terms Buddhist Japanese and Japanese Buddhist .
  • The central role of semantics , for example when assigning who to animate and which to inanimate players in a relative clause or when restricting the selection of arguments in predicate frames.
  • The distinction between semantic roles and grammatical relations, e.g. to describe active-passive alternation without deriving them from one another.
  • The consideration of typological peculiarities of many languages, for example in the form of the Semantic Function Hierachy (SFH) for the subjectivability of players with certain semantic roles.

weaknesses

The use of characteristic semantic primitives to restrict the selection (see semantic relation ) of certain argument positions in the Underlying Clause Structures (UCS) could in practice lead to the problems known with this semantic model, for example with relational properties such as relationships, as well as with verbs, gradual differences and Colours. The coding of the fine semantics in the lexicon, however, is generally an unsolved problem.

The procedure not to describe violations of the selection restrictions as ungrammatic, but z. Treating it as a metaphor, for example, may represent an immunization strategy, for example if no special interpretation strategy has been worked out, which in this case could reduce testability, applicability and the scientific value of the model.

Also, the resultant from the semantic orientation concentration on introspection and elicitation for data acquisition for determining selection restrictions of argument positions in predicate frame could cause problems and scientific theory reduce a value of the data obtained, elicited there and from introspection easily be misinterpreted data gathered in the predetermined question can, for example, if influencing factors that go beyond the question are not taken into account.

Also in the area of ​​the operators for the temporal specification of the predication, the universal claim and the practical requirements diverge, because the operators named by Dik at this level such as "present" and "progressive" are not universal categories, but the Underlying Clause Structures (UCS ) has the right to be coded language-independent before the expression rules are applied.

Summary characterization of the model

In summary, Simon C. Dik's “Functional Grammar” can be considered a

literature

Web links