Transcript Lecture 3

Linguistic Theory
Lecture 3
Movement
A brief history of movement
• Movements as ‘special rules’ proposed to capture
facts that phrase structure rules cannot
• Movements catch on – anything you can do, I can
do better (with a movement)
• Too many movements – loss of explanation
• Restrictions on movements – fewer movements to
account for the same amount of linguistic
phenomena
• Movements become more general
• There is only one movement: move anything
anywhere
Why do we need
transformations?
• Q: what is the connection between the
following two structures?:
S
S
NP
John
VP
V
NP
loves Mary
• A: nothing
NP Aux
John may
VP
V
NP
read Det N
a book
• Q: what is the connection between the
following two structures?:
S
S
NP
John
VP
V
NP
loves Mary
NP Aux
VP
Mary was
V
loved
• A: quite a lot
• But if all these structures are produced by
the same kinds of rules (PS rules) they
should all be as unconnected as the first
two
• Therefore we need grammatical rules
which are able to relate structures to each
other
• A transformation takes a structure and
changes it to form another structure
• Thus transformations are capable of
relating structures
• E.g. a ‘passive’ transformation could take
a structure such as:
S
NP
John
VP
V
NP
loves Mary
• and ...
• Delete its subject:
S
NP
VP
V
NP
loves Mary
• Move the object to subject position:
S
NP
VP
Mary
V
loves
• Insert the passive auxiliary:
S
NP Aux
VP
Mary was
V
loves
• Insert the passive morpheme:
S
NP Aux
VP
Mary was
V
loved
• The passive structure is related to the
active because it was formed from it
• A technical problem:
– which structures are basic and which are
formed from transformations?
• There is no real reason to think of any
structure as more basic than any other
• So the relation between structures must
be more indirect
• Suppose all structures start off with a
more abstract underlying form which is
then ‘transformed’ into the structure we
actually see:
underlying form
transformations
surface form
• Originally these were called Deep
Structure and Surface Structure
• We can relate different surface structures
by transforming them from the same deep
structures:
Deep Structure
active
transformations
passive
transformations
active
Surface Structure
passive
Surface Structure
• This also allows us to capture certain
ambiguities:
Deep Structure
transformations
Deep Structure
transformations
Surface Structure
• Two Deep Structures could be operated
on by different transformations to produce
the same Surface Structure
• E.g.
– the bomber’s destruction
• the bomber destroyed something
• something destroyed the bomber
the bomber destroyed ---
--- destroyed the bomber
nominalisation
nominalisation
destruction the bomber
movement
the bomber destruction
genitive insertion
the bomber’s destruction
Things go wild with
transformations!
• What you can do with a transformation:
– move things about
– insert new things
– delete old things
– change structure
• So what can’t you do?
– In principle, any two structures could be
related
• Mathematically, a transformational
grammar is equivalent to an Unrestricted
Rewrite System
– this is the most powerful kind of grammar
which can generate any kind of language
– so we can’t use it to explain why human
languages are as they are
• Transformations are specific to particular
structures in particular languages – the
passive transformation for English is not
the same as the passive transformation for
French
– so in principle languages could vary infinitely
– this is the same as Linguistic Relativity
• Children would have to learn the
particulars of each transformation of their
language
– they must guess which transformations are
used from hearing surface forms
– but if there is no limit to how a transformation
might be, this would be impossible without
direct instruction
– no one directly instructs children in language
learning
– so they shouldn’t be able to learn language
• Deletion transformations are particularly
problematic:
– presumably, the hearers job is to recover the
Deep Structure on hearing a surface form
– this would seem to involve applying
transformations backwards: putting moved
items into their original positions and getting
rid of inserted material
– but it would also involve the recovery of
deleted material, and it is not clear how that
can be done
• For example, suppose the underlying structure of
a passive sentence were:
S
NP
VP
John
V
NP
loves Mary
• Applying the passive transformation we delete the
subject, move the object and insert auxiliary and
morpheme to get:
– Mary was loved
• It is easy to see how we can move the subject
back into object position and remove the auxiliary
and morpheme, but how do we know to insert
‘John’?
• We might say that we do not need to recover
the exact element that was deleted as this
was something like ‘someone’
S
NP
someone
VP
V
NP
loves Mary
• But in other cases, this is not how we recover
missing information:
– John wants to go to the party but Mary doesn’t
• want to go to the party
• * do something
What syntacticians did next
• The solution to all these problems was to
impose restrictions on transformations
– the emphasis turned from what could be done
with transformations to what they couldn’t do
– for example, given that deletion causes so
many problems, it was suggested that we
should not make use of deletion rules
transformations + restrictions =
generalisation
• Restricting transformations meant that
more phenomena had to be handled with
less grammatical devices
• So the remaining grammatical devices had
to be more general than previously
• A good example of how restrictions lead to
generalisation is the passive
Having got rid of deletion, the D-structure
for the passive could no longer be claimed
to be the same as for the active as passives
appear to have no underlying subjects:
• Active D-structure
–
S
• Passive D-structure
–
NP Aux
VP
John may V
NP
see Mary
S
NP
Aux
VP
--- may be V
NP
seen Mary
• The active and passive are still related by
what they share in common: ‘Mary’ is the
object of ‘love’
• The only transformation needed now is to
move the object to subject position
• Passive S-structure
–
S
NP
Aux
VP
Mary may be V
seen
NP
• We can simply state the passive
transformation thus:
– move the object of a passive verb into the
empty subject position
• However, there are passives which do not
involve the movement of an object:
– John believed [Mary to be nice]
– Mary was believed [--- to be nice]
• So we should not restrict the moved NP to
the object position:
– Move an NP into the empty subject position of
a passive verb
• This seems too general as not all NPs can
move to this position:
– John believes [Mary is nice]
– * Mary was believed [--- is nice]
• However, it seems that no NP can move
out of a finite clause, so this is an
independent restriction:
– it seems [Mary is nice]
– Mary seems [--- to be nice]
– * Mary seems [ --- is nice]
• In fact, the similarities between these
observations indicate that we should
extend the movement rule to
constructions which are not even passives:
– --- was believed [Mary to be nice]
– Mary was believe [--- to be nice]
– --- seems [Mary to be nice]
– Mary seems [--- to be nice]
• So the rule becomes:
– move an NP into a vacant subject position
• However:
– the fact that the subject position has to be
vacant follows from general principles:
• one position cannot be filled by two elements
• deletion is not allowed
– the fact that the NP must move to a subject
position also follows from general principles:
• the subject position has to be filled by something
in all sentences:
– it seems [Mary is nice]
– * seems [Mary is nice]
• this is not true of object position:
– Mary smiled
– * Mary smiled it
• We can therefore generalise our rule to:
– Move an NP
• In other words, this simply states that it is
possible for NPs to move
• The details of actual movements fall out
from general principles
• There is no time to demonstrate it, but the
same generalisation happened to all other
movements
• In the end, all we needed was one
movement rule:
– move anything anywhere
• This is not as chaotic as it sounds
– it is a statement that movements are possible
– the details of actual movements all fall out
from general principles independent of the
movements themselves