Transcript Slide 1
Linguistic Theory
Lecture 9
Grammatical Functions
Introduction
• The notion of grammatical function
(subject, object, etc.) seems to be a basic
element of grammatical analysis. But:
– There are questions of how they are to be
defined
– There are questions of whether they are
present in all languages (universality)
– There are questions of how they are to be
identified
The classical approach
• The study of classical languages, which were
rich in morphology and allowed fairly free
word orders, and which did not distinguish
rigorously between form and meaning, lead
to a particular view of grammatical functions:
– They were associated with words
– They were semantically defined
– They were morphologically identified
Grammatical functions = words
• Because the notion of a phrase did not
become prevalent until the 1900s, syntactic
phenomena was mostly seen as facts about
words:
– Their forms
– Their meanings
– Their functions
• The subject of the sentence was therefore
defined as that word with a particular form
and meaning
Semantic definition
• Two common approaches:
– The subject is what the sentence is about
(complements ‘predicate’)
– The subject is the one who carries out the
action described by the verb
• Subject = topic
• Subject = agent
Topics
• Topics are the ideas that link a set of
sentences (discourse) as being ‘about’
something.
– Sentences form a consistent body if there is a
continued string of topics in them
• The topic is often associated with
phonological reduction
– Reduced to a pronoun
– Reduced to nothing (where allowed)
Topics
• E.g.
Two dogs are drinking in a bar. A horse
walks in (to the bar). He says: “is this
chair free?”. One dog turns to the other
(dog) and (he) says: “wow – a talking
horse!”
Topics
• Now consider the following sentences:
– Mary entered the room
– John kissed her
• In the second sentence
– The topic of the sentence is not the subject
– The subject is not a topic
• Moreover:
– It seems John is rich
• ‘it’ cannot be the topic as it is meaningless (the
sentence would be ‘about’ nothing)
Agent
• Not all subjects are agents
– Not all verbs involve actions
– Even activity verbs may have non-agent subjects
(in passive)
– There may be a connection between ‘subject’ and
theta role
• Fillmore’s Case hierarchy
– But this is not straightforward
• John fears sincerity
(experiencer – theme)
• Sincerity frightens John
(theme – experiencer)
Semantic approaches conclusion
• As is usual, semantic definitions of
syntactic phenomena are rarely
straightforward and do not yield
unproblematic results
Morphological approach
• There are two main morphological facts
associated with the subject:
– Case (morphological – not Fillmore’s)
• Subject is nominative
– Agreement
• Verb agrees with subject features
Grammatical functions and Case
• Standard assumptions:
– Subject = nominative
– Object = accusative
• But even for languages where this seems
to hold, there are problems:
– John believes she is smart
– John believes her to be smart
• Is the accusative object here?
Reasons to believe in accusative
subjects
• The similarity of
– John believes she is smart
– John believes her to be smart
• The dissimilarity of
– John believes her to be smart
– John persuaded her to be smart
• John persuaded her that she should be smart
• * John believed her that she should be smart
Reasons to believe in accusative
subjects
• Subject properties of accusative element
– John believes there to be a problem
• There is a problem
• * John persuaded there to be a problem
– John believes the cat to be out of the bag
• The cat is out of the bag
• John put the cat out of the bag
Reasons to believe in accusative
subjects
• So, not all subjects are nominative and
not all accusative elements are objects.
Further problems for Case
identification of grammatical functions
• Not all languages have Case distinctions
(are grammatical functions universal?)
• Different Case systems:
– Most European languages have the
nominative-accusative case system
• He left
1V
• He loves her
2V3
– 1 and 2 = nominative
– 3 = accusative
Further problems for Case
identification of grammatical functions
– Some languages do not do things this way
(Tsez – North Caucasian):
• ziya bik’i-s
1V
• cow go-past
– “The cow left”
• eniy-ā
ziya bišer-si
• mother-casecow feed-past
– “Mother fed the cow
• 1 and 3 zero marked case
• 2 differently marked case
23V
Further problems for Case
identification of grammatical functions
• Dilemma:
– Do we say that 1+2 are subjects in English
(common case = nominative) and 1+3 are
subjects in Tsez (for the same reason)?
– Do we say that 1+2 are subjects in both
languages but that the object of a transitive
verb is assigned the same case as the subject
of the intransitive verb in some languages?
Grammatical functions and
agreement
• Standard assumptions:
– Verb agrees with subject
– Verb does not agree with object
• But even for languages where this seems
to hold there are problems
– There is a cloud in the sky
– There are clouds in the sky
• What is the source of verb agreement?
Further problems for agreement based
identification of grammatical functions
• Not all languages have agreement
morphology (Chinese)
• Some languages (Chukchi) have too much
– the verb agrees with everything (subject
and object)
• Some languages (Tigre) have agreement
only with what would be object in other
languages (universal subject?)
The structural approach
• American structuralists
– For Indo-European
• Subject is a phrase
• The phrase which sits in a certain structural position
– But – Relativity
• You can only analyse a languages and a culture from
the point of view of that system
• Trying to impose notions from other systems onto a
language is inevitably wrong
– So, subject is not a universal notion
The structural approach
• Generative grammar
– Adopted the structuralist structural definition
of grammatical functions
– Disagreed with Relativity
– But disagreed amongst themselves as to
whether grammatical functions are basic to
the system
Subject a derived notion
• We define the subject as anything that sits
in a certain position (NP immediately
inside S):
Subject
S
NP
VP
V
NP
Subject a derived notion
• But not everything that ends up in this
position starts there: things move:
– ------ was seen Mary
– Mary was seen
• So, in this case, the notion ‘subject’ is only
•
•
established at S-structure
It would seem reasonable to assume that the
notion is always only established at Sstructure
Therefore at D-structure there is no ‘subject’
Subject a derived notion
• In mid-1980s it was argued that all
arguments start off inside VP
– So no argument is in ‘subject position’ at Dstructure
– The subject in all sentences is derived
Subject a derived notion
• E.g. Word order in Arabic:
• Arabic shows two main word orders:
– SVO
verb agrees with subject
– VSO
verb is in 3.sing. Form
– a Qara?-a al-tulaab-u
al-kutub-a
read-past the students-nom the books-acc
“the students read the books”
– b al-tulaab-u
qara?-uu
al-kutub-a
the-students-nom read-past-3pl. the books-acc
“the students read the books”
Subject a derived notion
• We can account for this if we assume:
– The verb always moves out of VP to inflection
position
– Movement to subject position is optional
– When NP moves to subject position it
becomes subject and verb agrees with it
– If it does not move, there is no subject and
the verb has ‘default’ 3.sing. form
Disagreements
• Some generative theories disagree and
claim that grammatical functions are not
derived notions, but basic building blocks
of grammar
Lexical Functional Grammar
• LFG assumes that sentences are analysed
at two levels:
– F-structure
– C-structure
• Unlike S- and D-structure, these are not
derived one from the other, but operate in
parallel
Lexical Functional Grammar
• F-structure is not a constituent structure
– More abstract assignment of elements to
functions:
• Predicate: love
• Subject: John
• Object: Mary
Lexical Functional Grammar
• C-structure is a constituent structure associated
with an F-structure by rules which map F-structure
elements onto C-structure elements:
•
S
NP2
John
VP
V1
NP3
loves
Mary
1 = predicate
2 = subject
3 = object
Lexical Functional Grammar
• This helps in dealing with non-configurational
languages, where word order is unimportant
(e.g. Warlbiri)
• witta-jarra-rlu ka-pala yalumpu wajili-pi-nyi
kurdu-jarra-rlu maliki
• small-dual-erg pres-3du that
chase-nonpast child-dual-erg dog
• “two small children are chasing that dog”
• In such languages the c-structures are very
different to English-type languages, but fstructures are similar and map onto the
relevant bits.