Transcript Slide 1

Linguistic Theory
Lecture 8
Meaning and Grammar
A brief history
• In classical and traditional grammar not much
distinction was made between grammar and
meaning:
– Word categories (noun, verb, adjective, etc.) and
syntactic functions (subject, object, etc.) were
defined in semantic terms
• In American structuralism meaning was shunned
•
as it was not observable
In Generative grammar, meaning was considered
relevant, but separate from grammar
The generative position
• Grammaticality and ‘semanticality’ are not
the same thing:
– Colourless green ideas sleep furiously
– * Furiously sleep ideas green colourless
• Grammatical facts have semantic
consequences because semantics
‘interprets’ what syntax provides:
The generative position
Syntax
 structure 
(produces)
Semantics
(interprets)
Problems
• How do we know what are syntactic facts
and what are semantic facts?
• E.g. ‘selectional restrictions’
– Sincerity frightens John
– John frightens sincerity
• Is the oddity of the second sentence a syntactic or
semantic fact?
• It depends on your theory – we do not have
reliable intuitions about this.
How do grammar and meaning
interact?
• An early theory was that transformations
do not change meaning (Katz Postal
Hypothesis)
• This lead to the idea that meaning was
associated with Deep Structure, while
Surface Structure was associated with
phonology:
How do grammar and meaning
interact?
• Deep Structure
meaning
transformations
Surface Structure
phonology
How do grammar and meaning
interact?
• But this is problematic as there are certain
surface structure facts that influence
meaning:
– It seems to John that he is smart
– He1 seems to John [ t1 to be smart]
– John believes the king of France is bald
– The king of France1 is believed [t1 to be bald]
How do grammar and meaning
interact?
• So it seems that the situation should be:
• Deep Structure
meaning
transformations
Surface Structure
phonology
• But what is the nature of D-structure and
S-structure meaning?
Aspect of meaning associated
with D-structure
• In 1965 Chomsky proposed to account for
the oddity of ‘John frightens sincerity’
through a lexical property (=selectional
restriction):
– Frighten: sentient object
• These restrictions apply at D-structure:
– * John frightens sincerity
– * sincerity is frightened by John
Aspect of meaning associated
with D-structure
• In 1967 Fillmore proposed a limited set of
‘cases’ which play a role in determining
the interpretation of elements
accompanying verbs:
Aspect of meaning associated
with D-structure
• Agentive (perceived instigator of an action):
John broke the window
• Instrumental (object used to put into effect
•
•
•
•
an action): the stone broke the window
Dative (affected object): the window broke
Factive (resulting object): bake a cake
Locative (place at, also to): put the pen on
the table, went to London
Objective (dustbin category)
Aspect of meaning associated
with D-structure
• These roles were stated in the verb’s lexical
entry, restricting the contexts which they can
be used:
– Break: [(A) (I) D]
• John broke the window
• John broke the window with a brick
• The brick broke the window
• The window broke
• Case hierarchy – determines what will be
subject
–A>I>D
Aspect of meaning associated
with D-structure
• These ideas transformed into ‘theta
theory’ in GB syntax (1981)
– ‘theta’ = thematic = semantic relations
between predicates and their arguments (who
does what to who)
Aspect of meaning associated
with D-structure
• Theta theory operated with a set of theta roles
(agent, experiencer, goal, theme, patient, etc.)
• Theta roles are stated in a predicate’s lexical entry
but assigned to elements in certain positions
– John hit Bill
agent
patient
– But:
• No one agrees on how many or what the definitions of these
roles are
• There is disagreement over how involved in syntax they are
Aspect of meaning associated
with D-structure
• A theory of minimal involvement:
– The theta criterion
• All theta roles must be assigned to an argument
• All arguments must be assigned a theta role
– * John hit
– * John smiled Mary
(not enough arguments)
(not enough theta roles)
– Theta roles are assigned to governed
positions
Aspect of meaning associated
with D-structure
• But this theory cannot account for why
•
•
subjects tend to be agents not patients
For this you seem to need to refer to specific
theta roles
Solutions
– Grimshaw: specific theta roles play a role in
determining the ‘prominence’ of an argument
– Hale and Keyser: the lexicon stores argument
structure as sub-trees so the positions of
arguments are given in the lexicon
Aspects of meaning associated
with S-structure
• Binding relations:
– John1 seems to himself1 [t1 to be smart]
– * it seems to himself1 [that John1 is smart]
– It seems to John1 [that he1 is smart]
– * he1 seems to John1 [t1 to be smart]
• These relations seem to be established after
movement and therefore do not apply at Dstructure but at S-structure
Aspects of meaning associated
with S-structure
• Binding relations:
– But there are problems:
• [Which picture of himself1]2 did Mary say John1 thinks
Susan likes t2
– At S-structure the reflexive is not in a position
where it can be properly bound by John
• --- Mary (did) say John1 thinks Susan likes [Which
picture of himself1]
– At D-structure the reflexive is not in a position
where it can be properly bound by John
– So where are binding relations established?
Aspects of meaning associated
with S-structure
• Binding relations:
– Due to restrictions on movement, the wh-phrase
cannot move in one go, but moves to the
beginning of each clause
– wh-P1 did Mary say [t1 John believes [t1 Susan
likes t1]]]
– In one of the intermediate positions it is in the
relevant relation with John
– But these are neither D- or S-structure positions
A new level of representation
• The following sentence is ambiguous:
– Every man loves a woman
• For every man there is a woman (a different one) who he
loves
• There is a woman (the same one) and every man loves her
• In some languages these meanings are associated
with different sentences:
– Minden férfi szeret egy nőt
– Egy nőt minden férfi szeret
• It is not accident that the quantified phrase at the
beginning is interpreted like this (see English
examples too)
A new level of representation
• It has been proposed that these quantified
expressions undergo a movement to the
front of the clause and their order
determines the interpretation:
– Every man1 a woman2 [t1 loves t2]
– A woman1 every man2 [t2 loves t1]
• But where does this movement take place
and why can’t we see it (in English)
A new level of representation
• Suppose there is another level of
representation after S-structure which is
associated with semantic interpretation =
Logical Form
• Suppose S-structure feeds phonological
interpretation = Phonological Form
A new level of representation
•
D-structure
movement
S-structure
LF
PF
A new level of representation
• Any movement that takes place between
S-structure and LF will not affect PF, but
will affect meaning
• Therefore we have semantically motivated
but phonologically invisible movement
A new level of representation
• This helps to unify languages even more
– Chinese does not have (overt) wh-movement
– But the Chinese can still ask questions
– Thus we either assume that asking questions
in Chinese involves a very different process
– Or we assume that Chinese does have whmovement, but only at LF