Generative Grammar - Wikipedia
Generative Grammar - Wikipedia
grammar
Generative grammar is a t heoret ical approach in linguist ics t hat regards grammar as a domain-
specific syst em of rules t hat generat es all and only t he grammat ical sent ences of a given
language. In light of povert y of t he st imulus argument s, grammar is regarded as being part ly
innat e, t he innat e port ion of t he syst em being referred t o as universal grammar. The generat ive
approach has focused on t he st udy of synt ax while addressing ot her aspect s of language
including semant ics, morphology, phonology, and psycholinguist ics.[1][2]
As a research t radit ion, generat ive grammar began in t he lat e 1950s wit h t he work of Noam
Chomsky.[3] However, it s root s include earlier st ruct uralist approaches such as glossemat ics.[4]
Early versions of Chomsky's approach t o synt ax were called t ransformat ional grammar, wit h
subsequent variant s known as t he government and binding t heory and t he minimalist program.[5][6]
Recent work in generat ive-inspired biolinguist ics has proposed t hat universal grammar consist s
solely of synt act ic recursion, and t hat it arose recent ly in humans as t he result of a random
genet ic mut at ion.[7]
Frameworks
There are a number of different approaches t o generat ive grammar. Common t o all is t he effort
t o come up wit h a set of rules or principles t hat formally defines every one of t he members of
t he set of well-formed expressions of a nat ural language. The t erm generative grammar has been
associat ed wit h at least t he following schools of linguist ics:
Milit ary funding t o generat ivist research was influent ial t o it s early success in t he 1960s.[12]
Generat ive grammar has been under development since t he mid-1950s, and has undergone many
changes in t he t ypes of rules and represent at ions t hat are used t o predict grammat icalit y. In
t racing t he hist orical development of ideas wit hin generat ive grammar, it is useful t o refer t o t he
various st ages in t he development of t he t heory:
A core aspect of st andard t heory is t he dist inct ion bet ween t wo different represent at ions of a
sent ence, called deep st ruct ure and surface st ruct ure. The t wo represent at ions are linked t o
each ot her by t ransformat ional grammar.
syntactic constraints
generalized phrase structures (X-bar
theory)
Context-free grammars
Generat ive grammars can be described and compared wit h t he aid of t he Chomsky hierarchy
(proposed by Chomsky in t he 1950s). This set s out a series of t ypes of formal grammars wit h
increasing expressive power. Among t he simplest t ypes are t he regular grammars (t ype 3);
Chomsky argues t hat t hese are not adequat e as models for human language, because of t he
allowance of t he cent er-embedding of st rings wit hin st rings, in all nat ural human languages.
At a higher level of complexit y are t he cont ext -free grammars (t ype 2). The derivat ion of a
sent ence by such a grammar can be depict ed as a derivat ion t ree. Linguist s working wit hin
generat ive grammar oft en view such t rees as a primary object of st udy. According t o t his view, a
sent ence is not merely a st ring of words. Inst ead, adjacent words are combined int o
constituents, which can t hen be furt her combined wit h ot her words or const it uent s t o creat e a
hierarchical t ree-st ruct ure.
The derivat ion of a simple t ree-st ruct ure for t he sent ence "t he dog at e t he bone" proceeds as
follows. The det erminer the and noun dog combine t o creat e t he noun phrase the dog. A second
noun phrase the bone is creat ed wit h det erminer the and noun bone. The verb ate combines wit h
t he second noun phrase, the bone, t o creat e t he verb phrase ate the bone. Finally, t he first noun
phrase, the dog, combines wit h t he verb phrase, ate the bone, t o complet e t he sent ence: the dog
ate the bone. The following t ree diagram illust rat es t his derivat ion and t he result ing st ruct ure:
Such a t ree diagram is also called a phrase marker. They can be represent ed more convenient ly in
t ext form, (t hough t he result is less easy t o read); in t his format t he above sent ence would be
rendered as:
[S [NP [D The ] [N dog ] ] [VP [V at e ] [NP [D t he ] [N bone ] ] ] ]
Chomsky has argued t hat phrase st ruct ure grammars are also inadequat e for describing nat ural
languages, and formulat ed t he more complex syst em of t ransformat ional grammar.[14]
Evidentiality
Some linguist s such as Geoffrey Pullum have quest ioned t he empirical basis of povert y of t he
st imulus argument s, which mot ivat e t he crucial generat ive not ion of universal grammar.[15]
Linguist ic st udies had been made t o prove t hat children have innat e knowledge of grammar t hat
t hey could not have learned. For example, it was shown t hat a child acquiring English knows how
t o different iat e bet ween t he place of t he verb in main clauses from t he place of t he verb in
relat ive clauses. In t he experiment , children were asked t o t urn a declarat ive sent ence wit h a
relat ive clause int o an int errogat ive sent ence. Against t he expect at ions of t he researchers, t he
children did not move t he verb in t he relat ive clause t o it s sent ence init ial posit ion, but t o t he
main clause init ial posit ion, as is grammat ical.[16] Crit ics however point ed out t hat t his was not an
evidence for t he povert y of t he st imulus because t he underlying st ruct ures t hat children were
proved t o be able t o manipulat e were act ually highly common in children's lit erat ure and everyday
language.[15] This led t o a heat ed debat e which result ed in an increasing split bet ween generat ive
linguist s and applied linguist ics in t he early 2000s.[17][18]
The sentence from the study which shows that it is not the verb in the relative clause, but the verb
in the main clause that raises to the head C°.[19]
Recent argument s have been made t hat t he success of large language models undermine key
claims of generat ive synt ax because t hey are based on markedly different assumpt ions,
including gradient probabilit y and memorized const ruct ions, and out -perform generat ive t heories
bot h in synt act ic st ruct ure and in int egrat ion wit h cognit ion and neuroscience.[20]
Generat ive-inspired biolinguist ics has not uncovered any part icular genes responsible for
language. While some hopes were raised at t he discovery of t he FOXP2 gene,[21][22] t here is not
enough support for t he idea t hat it is 't he grammar gene' or t hat it had much t o do wit h t he
relat ively recent emergence of synt act ical speech.[23]
Generat ivist s also claim t hat language is placed inside it s own mind module and t hat t here is no
int eract ion bet ween first -language processing and ot her t ypes of informat ion processing, such
as mat hemat ics.[24][a] This claim is not based on research or t he general scient ific underst anding
of how t he brain works.[25][26]
Music
Generat ive grammar has been used in music t heory and analysis since t he 1980s.[27][28] The most
well-known approaches were developed by Mark St eedman[29] as well as Fred Lerdahl and Ray
Jackendoff,[30] who formalized and ext ended ideas from Schenkerian analysis.[31] More recent ly,
such early generat ive approaches t o music were furt her developed and ext ended by various
scholars.[32] [33][34][35][36]
See also
Cognitive linguistics
Cognitive revolution
Digital infinity
Formal grammar
Functional theories of grammar
Generative lexicon
Generative metrics
Generative principle
Generative semantics
Generative systems
Linguistic competence
Parsing
Phrase structure rules
Syntactic Structures
Notes
References