A brief historic overview of Syntax & Early stages in ... · Generative Grammar Syntax and limits of Phrase Structures TransformationalGrammar X-bar theory Outline 1 A brief historic

A brief historic overviewGenerative Grammar

A brief historic overview of Syntax&

Early stages in Transformational SyntaxSyntactic Theory

Winter Semester 2009/2010

Antske Fokkens

Department of Computational LinguisticsSaarland University

20 October 2009

Antske Fokkens Syntax — History 1 / 56


Outline

1 A brief historic overviewGrammar in the early daysTowards Modern Linguistics

2 Generative GrammarSyntax and limits of Phrase StructuresTransformational GrammarX-bar theory



Grammar in the early daysTowards Modern Linguistics

Outline






Early work on Grammar

There is a long tradition of describing language’s structure:

In most cases, language was analyzed so that classicaltexts could be read

Grammar described archaic forms of language




Examples of early grammarians and linguistic work

India: Pan. ini (estimated 4th century B.C.)

China: Erya (author unknown) (3rd century B.C.)

Greece: Dionysius Thrax (2nd century B.C.), ApolloniusDyscolus (2nd century A.D.)

Rome: Donatus (4th century A.D.), Priscian (6th centuryA.D.)

France: Lancelot et al (1660) Grammaire générale etraisonnée (Port Royal)




Pan. ini’s grammar

Sanskrit grammar, said to be short and complete

Includes topics as syntax, morphology, phonology andpragmaticsEspecially known for the As. t.adhyayı:

describes algorithms that can be applied to lexical items toform wordssystematic and highly technicalfocus on brevity: difficult to read

Pan. ini is said to have influenced the foundations of manyaspects of modern linguistics:

Structuralism (Ferdinand de Saussure and LeonardBloomfield)Generative grammar (Noam Chomsky)Optimality theory




Diachronic Linguistics

Discovery of Sanskrit and its obvious resemblance to Latinand Greek lead to development of comparative linguistics

Originally mostly guided towards languages with historicrecords

Interest in other languages stimulated researchers todescribe language

Gradual shift of interest: from prescriptive to descriptivegrammars




Ferdinand de Saussure (1857 - 1913)

Sanskrit scholarHis course notes were published posthumously by hisstudents in cours de linguistique générale (1916)

Turned the attention from diachronic linguistics tosynchronic linguisticsFormulated the arbitrariness of signIntroduces the terms “langage”, “langue” and “parole”




Ferdinand de Saussure (cont)

Langage, Langue and paroleLangage is the faculty of speech: it is heterogeneousconsisting of physic, physiological and psychological factsA Langue is a homogeneous system of symbols that maybe mapped to meaning, it is a social product , exterior ofindividualsParole is the act of using language, it is also here wherepsychology comes into play

Saussure’s work is seen as the starting point of’structuralism’, introducing “syntagmatic analysis”: whatelements can occur in which context: what does itcontribute to the meaning?




Towards modern syntax

Structuralism (20-30ies, Bloomfield), distributionalism(50ies Hockett, Harris)

Categorial Grammar (30ies, Ajdukiewicz)

Dependency Grammar (30ies, Tesnière)



Syntax and limits of Phrase StructuresTransformational GrammarX-bar theory

Outline






Chomsky’s Syntactic Structures

Main task for linguist: separate grammatical strings fromungrammatical stringsTwo issues:

How to define grammatical strings?Corpus or statistical methods: fail because of creativecharacter of languageGrammaticality cannot be determined by ’meaningfulness’Proposed method: native speaker judgments

What kind of system can describe all grammatical strings ofa natural language? It must

1 consist of a finite set of rules2 be descriptively adequate3 be explanatory




Level of formal grammar

Easy to show: English is not a finite state grammarCompare (after Chomsky (1957)):

(i) If S1, then S2 .(ii) Either S3 , or S4.(iii) If either S3 , or S4, then S2 .(iv) *If S1, or S2 .

Phrase Structure Grammar?




Phrase Structure Grammar

Chomsky on Phrase Structure Grammar:

Not flawed in the same way a finite state grammar isThere are probably languages that cannot be described bya PSG

Later shown to be (most likely) true for Dutch, and definitelyfor Swiss German

If English can be described by a PSG, remains to be seen

There are, however, other grounds to consider PSGsinadequate to describe natural language...




Adequacy of linguistic theory

How to test whether a linguistic theory is adequate?

1 Can it account for the data?

2 Can it account for data in a straight-forward way, or will itlead to extreme (implausible) complexity?

3 Can the same system be used to construct grammars forall natural languages?




Limits of Phrase Structure Grammar

Phrase Structure Grammar may be able to generate allgrammatical strings, but it cannot capture regularities inrelations between expressions

Coordination:1 The topic of the lecture is syntax2 The topic of the book is syntax3 The topic of the lecture and of the book is syntax

Passivization:1 Noam Chomsky wrote Syntactic Structures2 Syntactic Structures was written (by Noam Chomsky)




Three levels of morpho-syntactic representation

Phrase Structure Grammar: D-structure

Transformations: S-structure

Morpho-phonemics: Final output




Transformations

How to capture grammatical phenomena such as agreement,coordination, passivization?

Main idea: spilt syntactic structures in a deep structure(d-structure) and surface structure (s-structure)

Phrase Structures create deep-structures

Transformations apply to deep-structures deriving asurface structure→ sentences and their passives have the same d-structure




Information in Syntactic Structures

In addition to how the sentence can be composed insmaller parts, we want to know how these parts relate toeach otherIn syntactic structures such information comes from:

1 Definitions of grammatical functions2 The lexicon3 Features on categories




Grammatical functions and Grammatical Categories

Grammatical functions (subject, object, predicate) aredefined in relation to the deep structure (Standard Theory):

Subject-of-S [NP, S]Object-of-V [NP, VP]Predicate-of-S [VP, S]

Syntactic properties are generally represented by(boolean) features, e.g.

N: [+N, -V]V: [-N,+V]A:[+N,+V]




Subcategorization and lexical insertion

Lexical items come with a subcategorization frame.E.g.:

love: [V;–NP]smile: [V;–]rely: [V;–PP]think: [V;–S’]

Lexical Insertion Rule (Ouhalla (1994): p.50):Insert lexical item X under terminal node Y, where Ycorresponds to the categorial features of X, and YP correspondsto the subcategorisation properties X.

Based on Ouhalla (1994) p.45–50




Transformations: Passivization

Passivization: optional

Structural analysis:

NP – Aux – V – NP

Structural change:

X1 – X2 – X3 – X4

↓

X4 – X2 + be + en – X3 – by + X1

(Chomsky (1957: p112))




PSG and Transformation: tense

Starting with PS-rule: S → NP Aux VP

Consider the following examples:

The boy watched the movieThe boy will watch the movieThe boy doesn’t watch the movieThe boy didn’t watch the movie, but his friend didWatch the movie, she wondered whether the boy will.

Tense seems to be part of ’Aux’ rather than VP:S → NP Aux VPAux → Tense (Modal) (Neg)

based on Ouhalla (1994)




PSG and Transformation: tense (cont)

The structure of the boy watched the movie is:NP – tense – V – NP

The tense marker thus precedes the verb watch in thed-structure.How can we make sure that tense will be marked on themain verb in spell-out?

1 Apply a transformation moving V to Aux?

S-Structure: [[NPThe boy][Aux watchi -ed][VP _i the movie]]

2 Apply a transformation moving tense to V?

S-Structure: [[NPThe boy] [Aux _i ][VPwatch -edi the movie]]





Evidence for moving tense

Adverbs can precede or follow a VP in English:(i) The boy cleverly avoided Bill.(ii) The boy avoided Bill cleverly.(iii) The boy will cleverly avoid Bill.

If V moves to Aux, the verb precedes the VP on the surfaceAdverbs should be able to follow the verb, but(iv) *The boy avoided cleverly Bill.

The conjugated verb thus remains in situ, and tense mustmove to the VP, if there is no modal: ’affix hopping’





Affix-hopping

We also want to account for the following:1 Auxiliary verbs do move to Aux: Mary was often happy2 Tense does not move to VP when VP dominates an

auxiliary: *The boy have watched the movie3 Affixes cannot ’hop’ over negation: *The boy (do) not

watched the movie

Affix hopping: Move Tense (from Aux) to V, providedi) Aux does not dominate a Modal or Negationii) V has the feature specification [-AUX] (i.e. is not an

auxiliary)iii) VP does not dominate a V with feature specification [+AUX]

based on Ouhalla (1994), definition p. 98 (before subject-aux inversion)




Transformational grammar: initial stages

Standard Theory: interpretation from d-structure

Extended Standard Theory: interpretation from d-structure,s-structure and possibly final derived structure

Trace theory: when transformations move elementsaround, these elements leave a trace:→ semantics can be interpreted from s-structure only




Language Model

D-structure

S-structure

Logical Form (LF) Phonetic Form (PF)




Assumptions in transformational syntax

There is a difference in competence and performance, i.e.between what the speaker knows of the language and how(s)he uses it

Children can learn a complex system such as language soeasily, because the basis is innate: we are born alreadyhaving a Universal Grammar (UG) in our mind

Descriptive adequacy: describe language as known by thespeaker (according to competence)

Explanatory adequacy: plausibility of the analysisdepending on whether it is easily learnable given our UG




Contributions to Syntax

Syntax was placed in the centre of linguistic researchAims of syntax go beyond description:

Attention for the (more) formal side of syntaxAttention for psychological aspect of grammar

This lead to more systematic research on linguistic data:native speaker judgments, distinction between grammaticalfrom ungrammatical




Some remarks

Transformational syntax states thatGrammar (PSG + transformation) must be able to generateall expressions that are part of the languageA speaker must have access to s-structure and d-structure(in Standard Theory) to interpret an expression

Because of such remarks, many take transformationalgrammar as a language production model: This is notnecessarily the case

The first aim of the transformational approach is to studyhow language works as a system that can easily belearned by children




Some more remarks

Because a language production/interpretation model is notthe aim of transformational grammar, the model is not themost suitable for such approaches

Despite its efforts towards formal description, the detailsare often not formal enough for computational approaches→ how does Lexical Insertion work really?

This also applies (perhaps in somewhat lesser extends) toX Theory, Government and Binding and Minimalism




X-bar theory: motivations

X-bar theory was developed in the seventies to designphrase structures in a more theoretically sound wayIt ended up addressing several issues:

1 stronger generalization than previously used PSG2 introducing a structural difference between complements

and modifiers3 removing a redundancy between lexical contribution and

the contribution of PS-rules (mentioned by Ouhalla 1994)




Redundancy

Redundancy: the items that may form a VP is determinedboth by the subcategorization properties of the verb, andby the Phrase-Structure rules.

Is it possible to use only one of the two?

We can use only subcategorization, but then thisinformation must be present at all levels

Projection Principle:“Representations at each syntactic level (i.e., L.F., and D- andS-structure) are projected from the lexicon, in that the observethe subcategorization properties of lexical items.”

Chomsky (1981) p. 29




Generalization

Can we define phrase structure rules in a way thatcaptures cross-linguistic properties of syntactic structures?

Can we define phrase structure rules in a way that allowsto capture commonalities in structure within a language(e.g. subject of a sentence or an NP in English)?

Can we define phrase structure rules in a way thatdistinguishes complements from adjuncts?




X-bar Theory: definitions

We can generalize PS-rules as follows:

XP → ...X...

We say that XP is the maximal projection of X

In X-theory X is an obligatory element on the left-hand sideof the rule. It is called the head of the maximal projection.

The maximal projection XP and its head X are different barlevels of X




The X-bar Convention I

X Convention: a “theory of syntactic categories”There are three major claims:

1 There is a set of syntactic features in UG defining possiblelexical categories. A language selects the lexical categoriesit uses from UG (in much the same way as it selectsphonemes)

2 Each lexical category X defines supercategoriesX’,X”,...,Xk . Xn and Xn−1 are related through the followingPS-rule:

X n→ ...X n−1...

The head of Xn may be defined as either Xn−1 or lexicalcategory X




The X-bar Convention II

3 Grammatical formatives are defined as feature complexesand a prime notation:

2

6

4

αF1

βF2

...

3

7

5

i

e.g. V’:2

6

6

6

4

+Subj+Object+Comp...

3

7

7

7

5

’ N’:2

6

6

6

4

+Subj−Object+Comp...

3

7

7

7

5

’

based on Jackendoff (1977)




Bar-levels and properties

How many bar-levels does each category have?This is an empirical question: how many are needed toaccurately describe language?For this overview, we follow Jackendoff (1977) and supposethree bar-levels for each category: X ′, X ′′ and X ′′′

Lexical categories are of type X , maximal projections X ′′′,for most categories this is XP (for V this is S)




The PS-rule’s canonical form

We suppose that elements appearing left or right of X n−1

are either major categories or specified grammaticformatives (such as tense)

The canonical form of the X PS-rule is then:

Xn → (C1)...(Cj) – Xn−1 – (Cj+1)...(Ck ),and for all Ci either Ci = Y′′′ for some lexical category Y, orCi is a specified grammatical formative.

Jackendoff (1977: p.36)

Language specific rules determine on what side of Xdifferent elements may appear




Parallelism between structures (example)

Important idea in X Theory: if there are parallel relationsacross categories, these categories must be syntacticallyparallel in respect to the relationFor instance: the subject of a sentence (V”’) and thesubject of an NP (N”’):

1 John has proved the theorem2 John’s proofs of the theorem





Two (old) proposed structures

S

N”

N’

N

John

V”

Specv

T

Pres

have en

V’

V

prove

N”

the theorem

N”

SpecN

Preart

Several

of Poss

N”

N’

N

John

’s

N’

N

proofs

P”

of the theorem

Chomsky’s analysis presented by Jackendoff (1977: p. 38)




A uniform structure for subjects: step 1

Assumption: several of is not the specifier, but part of ahigher NP: N”

N’

N or Q

Several

of N”

SpecN

Poss

N”

N’

N

John

’s

N’

N

proofs

P”

of the theorem

Adapted from Jackendoff (1977: p.40)





Assumption 2: note that the ’s always occurs with subjects of NPs, alsoin cases where the subject moved there (consider the city’s destructionby the enemy )→ ’s is inserted at the last moment: N”

SpecN

N”

N’

N

John

N’

N

proofs

P”

of the theorem





There is no category ’Spec’: both SpecN and SpecV can be removed:

S

N”

N’

N

John

T

Pres

have en V’

V

prove

N”

the theorem

N”

N”

N’

N

John

N’

N

proofs

P”

of the theorem

Jackendoff (1977: p.40-41)




Final step: three bar-levels

There are only two bar levels so far: we add a bar-level one to N and V:

S

N”’

John

V”

T

Pres

have en V’

V

prove

N”’

the theorem

N”’

N”’

John

N”

N’

N

proofs

P”’

of the theorem

from Jackendoff (1977: p.41)




Uniform Subject Structures, Concluding remarks

In English, the grammatical relation ’subject-of’ can now bedefined as:

[N”’,[+ Subj]]

For motivation of why three bar levels would be preferable,see Jackendoff (1977)




Arguments versus Modifiers

In principle, complements can attach to X’ or X”

Bar-level and structure are also used to distinguishcomplements that are functional arguments from othermodifiers

In English, all functional arguments (except the subject)immediately follow their head

If a head strictly subcategorizes for a term, then itcombines with X’

Other complements combine with X”

In practice, the criterion of Xn→ Xn−1 is somewhat weakened, so that some complements may combine with X”’




X’ vs X” complements: an example

N”’

Det”’

the

N”

N’

N

King

P”’

of England

P”’

from England





Some remarks on X-bar Theory

X-bar theory is a module of grammar concerned with thePhrase Structure of grammar

It has been widely adopted in syntactic theory

X-bar structure is still used in (some versions of) GB andMinimalism

References to it are also found in purely computationallinguistic work that are not necessarily focusing onsyntactic analysis




Concluding remarks

In this lecture, we have seen:

That diachronic linguistic research lead to descriptivelinguistics

That transformational syntax emerged from a need toimprove on structural approaches

A (somewhat simplified) overview of Standard Theory,including examples of transformations in English

Fundamental ideas in transformational syntaxA (simplified) overview of X Theory

What are its aims?What does an X conform structure look like?




What to retain from this lecture?

Chomsky’s ideas on syntactic research:What is the aim of syntactic research? I.e. what are theaims of transformational grammars?Chomsky’s assumptions concerning innateness ofgrammar and grammaticality

The basic architecture of the language model assumed inthe transformational approach (d-structure, s-structure, PF,LF)The principles of X Theory:

what phrase structures look like in Xwhat different bar-levels may represent in X




What you do not need to know (for this lecture)

Specific names and dates from the historic overview

How to formalize transformationsDetails of examples presented in this lecture, i.e.

Passivization in transformational syntaxAffix-hoppingThe exact structure of subject-head, functional argumentsand other complements in X

The exact motivation of particular analyses presentedhere: most were highly simplified, and would requiresubstantial additional reading




Some presupposed knowledge

Please make sure you are familiar (and comfortable) with thefollowing concepts:

Constituency

Phrase Structure Grammar

Subcategorization

If not the following sources may be of help:

Judith Köhne’s slides on the preparatory course web-page

Sag, Wasow and Bender (2003) (First two chapters)

Ouhalla (1994) (Chapter 2)




Bibliography I

Bloomfield, Leonard (1933). Language. New York: Henry Holt.

Duszkowski, W. and W. Marciszewski and J. van Benthem (eds) (1988).Categorial Grammar. Amsterdam: John Benjamins.

Chomsky, Noam (1957). Syntactic Structures. Mouton, The Hague.

Chomsky, Noam (1965). Aspects of the Theory of Syntax. CambridgeMassachusetts: MIT Press.

Chomsky, Noam (1981). Lectures on Government and Binding. Berlin:Mouton.

Haegeman, Liliane (1991). Introduction to Government and BindingTheory. Oxford: Blackwell Publishers.

Jackendoff, Ray (1977). X Syntax: A Study of Phrase Structure.Cambridge Massachusetts: MIT Press.

Ouhalla, Jamal (1994). Introducing Transformational Grammar. NewYork: Oxford University Press.




Bibliography II

Sag, Ivan A., Thomas Wasow and Emily M. Bender (2003). SyntacticTheory. A Formal Introduction. Palo Alto: CSLI Publications.

Saussure, Ferdinand (1916). Cours de linguistique générale. Eds.Charles Bally and Albert Sechehaye. Paris: Payot.


Documents

A brief historic overview of Syntax & Early stages in ... · Generative Grammar Syntax and limits of Phrase Structures TransformationalGrammar X-bar theory Outline 1 A brief historic