*!*Ghomeshi, Jila and Diane Massam. 1994. To Appear, Linguistic Analysis



tải về 302.24 Kb.
trang9/10
Chuyển đổi dữ liệu11.06.2018
Kích302.24 Kb.
#39842
1   2   3   4   5   6   7   8   9   10

it directly interacts. (see discussion pg 328.)

\bigskip


! Some data covered by reference point analysis, but not
by c-command:

\medskip


! Backward Anaphora

In 88% of examples in van Hoek's corpus, the subject


served as the antecedent, and there was a marked difference
in prominence between subject and pronoun (pronoun being
in parenthetical expression, scene setting backgrounded clause, etc.):
{As it creeps along, the Hayward fault has
plenty of opportunity to damage the works of man...[Motorland 3/89] pg 330
None of the attested examples were of the following type:
{His mother loves John (Lakoff 1968) (people vary on whether they
find this acceptable: notice that the prominence difference between
pronoun and NP is not great))

THe prominense asymmetries are expected on reference point model,


e.g. the overwhelming tendency for the antecedent to be the subject
and the rarity of sentences like \ex{0.

Cross-linguistic predictions:  tendecies shoudl be the same, but


conventionalized constructional schemas may differ, so languages
may differ in the degree to which they all backwards anaphora.

\bigskip


! Anaphora in Discourse: (see also Ariel 1988, 1990; Fox 1987a,b;
Giv\'on 1983, 1985, 1989; CHafe 1979, 1987, Tomlin 1987).

Speakers tend to use a full NP after discourse unit boundaries,


even when a pronoun would be unambiguous:
{She [Ripley] did not see the massive hand reaching
out for her form the cncealment of deep shadow.  But Jones did. He yowled.
\\
! Ripley spun, found herself facing the creater. [Alien, pg 267] cited
on page 332

\bigskip


! Point of View (Kuno 1987; Zribi-Hertz 1989; Pollard & Sag 1992)
{\%The idea that John might have cancer worried him.
(adapted from Kuno 1987) cited page 333

In \ex{0, the viewpoint and the content viewed aer only


linked by an implicit relation, not directly through a complement
chain.  The full NP indicates that ``the idea that John
might have cancer" is not formulated from John's point of view.

Point of view effect even within complement chains:


Kuno (1987) citing Haj Ross:
\eenumsentence{ *That Oscar was all alone frightened him.
That Oscar is five points ahead of the other andidates apparently pleases him.
In (b), the clausal subject is taken to be common knowledge,
and is thus accessible independently of its relationship to
the viewer.

\bigskip
! Reflexives

Prototype: reflexive pronoun is DO, and NP antecedent is subject:
(Mary saw herself).

Schematic characterization: reflexive's antecedent must be


highly accessible and conceptual adjacent to it:
i.e., directly interconnected by some relation (such as a verb's
profile).

{*John$_i$ saw him$_i$


This is said to be essentially preempted by the more entrenched (well-established, frequently occuring) reflexive pattern.

*!*Van Riemsdijk (1978a) A Case Study in Syntactic Markedness.


The Peter de Ridd Lisse. as discussed in Hornstein and Weinberg

PPs can have structure with a COMP node:  [P|| t [P| about t]], with


P|, not P|| being a bounding node: this would allow prep stranding, but since
it is internal to the PP, it does not easily capture the fact that only
(roughly) argument PPs allow extraction.  (H& W: only PPs which are within the VP).

Phrase structure rules which generate PPs with COMPS are more marked than


PS rules generating PPs without COMPS.

*!*Van Valin, Robert D Jr.  The Acquisition of Wh-Questions adn


the Mechanisms of Language Acquisition.  TO Appear in M Tomasello,
ed., The New Psychology of Language: Cognitive and FUnctional
Approaches to Language Structure.  Hillsdale, NJ.: LEA.

Discusses children's production of: medial WH (THorton 1990, 1995):


\eenumsentence{ Who do you think who is in the box?
WHo do you think who the cat chased?
What do you think what Cookie Monster likes?

Discussion of subjacency


Van Valin (1991): children construct the grammar of tehir language
based on 1) their initial
(general) cognitive endowment and 2) the evidence to which they are exposed.

Discussion of focus based on Lambrecht 1994):

In Eng, WH  is in the pre-core slot.

Wh questions are typically narrow focus in that their focus is a


single constituent represented by the wh-expression.  (Y/N questions
can be narrow or wide focus:  Is that a boy or a girl?)

Unmarked focus position:


In English:
last constituent of the core
V-final languages: position immediately before the verb.

The most marked narrow focus is on the subject


(because subj is normally topic).

Potential focus domain: part of the sentence in which a focal element may occur


Actual Focus domain: the element(s) acutally in focus in a particular utterance.

Van Valin (1993):


\begin{quote
The potential focus domain in complex sentences: A subordinate clause
may be within the potential focus domain if it is a direct daughter
of (a direct daughter of...)
the clause node which is modified by the illocutionary force operator.
{quote

Subjacency explanation:

\begin{quote
...the element questioned (the focus Np in direct Y/N question or the WH expression in Wh-quetions must be in a clause within
the potential focus domain.
{quote

[What's wrong with:


Do you like the boy who is wearing the red HAT?]

(cf * What color hat do you like the boy who is wearing).

{After you left the party, did you take M to the movies?
Yes

No (=didn't take Mary, not didn't leave the party)


No, Bill did (= BIll took M, not B left the party)
No, Susan
?No, befoore (Better: no, it was beofre we went to teh party.
No, the park (= went to the park, not after you left the park).

[Did you take a man who was wearing a red coat to teh movies?


?No, a green coat.
No, it was a green coat.

Did Max return th epapers which the secretary photocopied to the lawyer?


?No, faxed.
No they were faxed. (my suggested answers).
No, the butler (can't mean the butler photocopied them)

Does M believe that S hid hte files?


No, the photographs.

The context of adverbial clauses and restrictive RCs is normally presupposed


and consequently constructing questions with the focus in one of these
structures generates a pragmaitc contradiction.

``It has never been argued that the source of a child's knowledge of the principles governing


the interpretation of Y/N questions is anything other
than the verbal interactions in which the child is involved. pg 12

Van Valin: the constraint on Y/N interpretation is transfered to


Wh questions.

WIlson and Peters (1988): child produced violations:


\eenumsentence{ What are you cookin on a hot ?
What are we gonna go at Auntie and ?
What are we goonna look for some  with Johnnie?
Same child played the following game:

\eenumsentence{ carefiver: What did you eat?  Eggs and

child: bacon
Oh, that's a \\
child: aleph.

Gricean foundation: principle above is ultiamtely derivable from


Cooperative Principle and Quantity.

Braine (1992): discusses how something like hte theory of clause structure


in section 3 could be acquired on teh baisis of `natural logic' of cognition and positive evidence.

Stromswold (1995): early production of WH questions by 12 children in CHILDES

for who questions, subj and obj questions appear at roughly teh same time
(6 produced objects first, four proudced subj first, one at the same time).

What and which questions: seven prouduced obj questions first, four


at teh same time and only one proudced what equestions first.

Note that objects are more prototypical foci than subjects!


*!*Washio, Ryuichi. 1997. resultatives, compositionality and language variation.


Journal of East Asian Linguistics.
Looks at resultatives in Japanese, and French and English, arguing that
J and F have a subset of cases involving only verbs which lexically code
patients.   Is compatible with my account (she sent me a reprint).
Nice!

*!*Wasow,  Tom. Draft. 1994. Remarks on Complex


NP Shift. Stanford University.

About pragmatic factors:


---I agree with you that the question of why it should be
that heavy things come at the end is an interesting one.

You don't mention the psycholinguistic finding that


the (first and) last thing in a string is the best remembered.
E.g., if I tell you to remember the following string:  "7 5 6 3 4 2 1 2 9"
you're at least likely to remember 7 and 9.  This idea has been used
in the language acquisition literature--I think there might
be a reference in the Slobin 1985 volume.

So, it might be that what comes at the end isn't necessarily indefinite, but


it is what the speaker intends the listener to focus on.  Of course,
this formulation sounds odd in the face of:

 1.Ignore compLETely free will.

But at the end of a talk about free will it is somewhat unusual
to be asked to ignore it, so it COULD be that the speaker intended
to draw extra attention to "free will."

There also might be a desire to dramatize by creating tension and so


delaying a certain element until the end.  It seems like this is a possible
explanation for Clinton's:

 2. put at the center of what our administration stands for the Nat'l Service Corps.

And for:

  3. That brings to the plate Barry Bonds.

2-3 could also be motivated by the idea that the last thing is focused on.
Also contrastiveness might be a relevant factor in determining
what is focused.  Is it possible that when J. Etchemendy said (4),
he was contrasting blocks with something else?

 4. assigns to the tokens blocks

Of course, these things are hard to quantify.  It may be that you
want to stay clear of them.  But they probably could be tested by
some psycholinguistic experiments (e.g., asking speakers to provide orders
given different contexts).

--You note that relative complexity is significant when complexity of


the NP and complexity of the adjacent phrase are removed.
Is it possible that indefiniteness is significant when complexity is
removed?  That is, is indefiniteness correlated closely
with complexity?

--Example 12 seems to create a problem for any transformational account.

--Nice critique of Larson.

--I don't understand why preference rules can't be made to


account for categorical judgments.  Couldn't you weight the preference rule
27f (avoid ambiguity regarding the role of an NP) high enough
such that any violation of it would automatically render the
sentence below the threshold for grammaticality?

--The begrudge sentence is nice.

--You don't discuss the ordering of the subject and verb with respect
to each other and the verb's complements/adjuncts.  Do you have an idea of
why they fall outside the domain of your preference rules?

Thanks for giving me an early copy.   I might be interested in running some


of the psycholinguistic tests someday.  I would of course check with you first, though, to see whether you are planning, or have already done, similar
experiments.

 
Linear Precedence in English:

less oblique phrases precede more oblique phrases
phrases with fewer nodes precede phrases with
more nodes
phrases with fewer words precede phrases with more words
phrases with less stress precede phrases with more stress
avoid discontinuous phrases or collocations
avoid ambiguity regarding the role of an NP
*!*Wasow, Thomas. Remarks on Grammatical Weight. 1995.
Revised version of above. Stanford University.

``In many lngs the linear sequencing of constituents tends


to be in order of increasing size and/or complexity.":
grammatical weight. pg 1.

Tom goes through several previously proposed criteria for


heavy NP shift, including Zec and Ickelas' ``the dislocated
NP is licensedwhen it contains at lesat two phonological
phrases." and Niv's that claimed givenness in discourse
was what mattered.

Phonological constraint is counterexampled by cases cited


that were not meant to be read out loud (from Enderton!)
and by:
{Ignore COMPLETELY free will. (after a talk about
free will)

Other interesting attested cases:


{that will bring to the plate X
{Help us become one voice and express tothe owners
and players our concerns!!!
{take into account inflation
Other proposals (NP must dominate S or PP) are counterexampled
by:
{ hold hostage all foreigners from UPI corpus

``I conclude that there  are multiple factors at work in


what I have been calling weight effects.  In the words of McDonald,
et al. 1993: 196 `Word order in speech is a complex product
of factors that reflect language structures (including factors
linked to syntactic and metrical structures) and factors
that reflect the retrievability of information from
memory and general knowledge." pg 9

Three constructions are looked at:

heavy NP shift, particle movement and the dative alternation.

Looked at bring, give, hand, offer, pay, send, show, take,


teach, tell, write.

McDonald et al. 1993 found that animacy influences ordering


choices in some constructions but not in others.

He does not look at phonological or informational criteria


because these are harder to pin down from a written corpus.
``I am convinced that informatioal factorsdo play
a role in contituent ordering (though I suspect that
the notion of "givenness" is not quite the right one).
[Good topic for an experiment]

Heaviness is taken to be categorial; weight, graded.


TW comes down on the side of weight: NPs are more likely
to shift the heavier they are.

Weight effects are dependent on relative, not absolute


weight.

Weight correlates well with the number of phrasal


nodes, and the number of words.

The mean weight of the second constituent ranges from 2.5 to


4.5 times that of the first.
\subsection{Why do weight effects exist?

Many researchers have suggested that long complex elements


can cause listeners to lose track of where they are in parsing
the sentence.

Tom suggests that demands on the SPEAKER play a more important role:


saving longer constituents for later is helpful in sentence planning
and production.

He claims that it is counterintuitive that the speaker would have


already in his mind both constituents in order that he may
count up the number of words in each and compare.  And disfluencies
in speech indicate that syntactic structure is selected
during ongoing speech (cites a number of H. Clark references).

The sentence \ex{1 gives Greenwald another second or two


to check who the next batter is:
{that brings to the plate Barry Bonds

\subsubsection{Case study one: collocations vs non-collocations


TW collected 827 sentences with certain verb preposition
combinations.
He judged 53\% of the V-PP pairings to be collocations.
Of those, 54\% were heavy NP shifted.  Of the non-collocational
VPs, only 15\% were heavy NP shifted.

He then coded the collocations for whether they were transparent


or opaque (transparent collocation should be transparent
to a speaker who knows the literal meaning of each of the words).
Bring to X's attention, bring to an end/close were
considered transparent; take into account/take into
consideration were considered opaque.  47\% of the transparent
collocations exhibited HNPS; 60\% of the opaque collocations
did.  (The difference is significant p$<$ .01)

In the case of an opaque VP, HNPS serves both speaker and


hearer.  Speaker: ``having decided on the collocation,
it provides extra time and reduces memory load to
utter all of it immediately."; Hearer: ``the PP needed to assign
the appropriate interpretation to the verb occurs immediately
after the verb." pg 24.

HNPS in transparent collocations helps only the speaker.


And still the difference between transparent collocations
and non-collocations is highly significant (chi-squared test
p $<$ .001)

\subsubsection{Case study two: lexical dispositions

Assumptions: To the listener, the earlier it is possible
to predict subsequent structure, the better: fewer
possible continuations entail less memory load
[wouldn't this predict transparent collocations to
help the listener after all, then?]

The speaker would like to delay decisions as long as possible


to reduce the amount of planning neededand allow more time
to formulate and articulate thoughts.

!
So, listeners should prefer early committment,


speakers should prefer late committment.

\begin{verse

Vt: always require DO  (bring, carry, make, place, put, set, take)

Vp: can have DO, but can occur with only PP (add, build, call, draw, give, hold, leave, see, show, write

{verse

If the Listener's concerns dominate, we would expect Vt to


have higher rates of HNPS than Vp, since the HNPS version
of Vp would be ambiguous syntactically (is there going to
appear a DO or not?)

For Vt, HNPS is an instance of early committment.


(better from Listener's point of view)

For Vp, HNPS constitutes late committment (better from Speaker


point of view)

Vp's are shifted 9.3\% of the time ,whereas Vt are shifted


only 5.6\% of the time.

footnote 28:  Hawkins (1990, 1994) claims heavy Nps tend to


be first in Japanese and V final lngs.  Tom has another
story for these cases.

Particle Movement is looked at: lots of lexical variation


with make up, build up, turn out, take on never being
shifted in the corpus and look over being shifted 50\%
of the time.

Question T raises:  why does PM occur at all?  The particle


is so short.  Is it because the particle is analyzed
as an adverb (e.g. up = ``completely"
so that it's natural place is after the DO?

He suggests: more adverbial = more transparent = fewer meanings


in the dictionary = less frequent [I don't see these equivalences]
= more likely to be polysyllabic (heavier)

Found: less polysemous V+particle combinations occur with


V-NP-Particle more often.

Less frequent,the more likely the particle is to occur


after the NP, too.

The lexical variation argues against any explanation purely


in terms of tree configurations.

\noindent Questions to Tom:

\bigskip

1.  When you discuss the second finding in support of the above


interpretation concerning Vt vs Vp and HNPS, you note that
listeners prefer early commitment in general.  But couldn't that
idea be interpreted broadly to imply that listeners appreciate
knowing that a collocation is being
used as much as speakers? That is, presumably both the listener
and speaker are accessing the collocation in the mental
"constructicon."  But since the listener would like to access
as much as possible as early as possible, wouldn't s/he
prefer pulling out the entire collocation even if it is
semantically transparent?

\bigskip


2. Once you have the empirical finding that collocations tend
to shift more than non-collocations, isn't it possible that
Vp's shift more than Vt's simply because there are more collocations
among Vps than among Vt's?  I don't know if that's the case,
and maybe you did control for that.
It may not be the case, but, it worried me because it may be
likely, being motivated in the following way:
because the PP occurs directly next to the V when there is no
DO, the simple contiguity could encourage the V+PP becoming
a collocation.  And since in some of your examples, the same
basic meaning can occur either with a DO or without a DO
(e.g. give (\$) at the office the collocational
use might be relevant.)

 Then again, taking your point of view, your findings with PM


could be taken to argue against the idea that collocations are
more common with Vp verbs, since verbs that can occur intransitively
with particles were NOT more likely to shift, contrary to what
my hypothetical motivation would predict.

\bigskip


3.  Finally, the speaker's ease of production is important
in spoken speech and you do show that the effects hold
in the spoken corpus, but why do you expect them to hold
in the written corpus at all?  If the speaker has a chance
to edit his writing, then you might expect him to edit
in such a way as to make comprehension easier; at that
point, he does know what he wants to say, so ease of production
isn't an issue.  Would you
counter this by saying that the writer is basically lazy
and so little editing is done?  That might be right,
but then you'd expect correspondingly fewer cases of HNPS
in heavily edited genres, right?

*!*Wechsler, Stephen. to appear  Explaining Resultatives without Unaccusativity.


Proceedings of the 1997 Texas Linguistic Society Conference on the Syntax and
Semantics of Predication
U of Texas at Austin.

Control resultative construction: verb selects for resultative predicate:


\eenumsentence{ Robert ran clear of the car/free of the car/*exhausted
Sally painted the door red/a pale shade of red/?sticky/*beautiful/*noticeable
John hammered the metal flat/smooth/shiny/into the ground/*beautiful/*safe
*John laughed silly/off his chair ex 6: page 2

Raising resultative construction: the verb does ! not select for


the resultative phrase.
\eenumsentence{ The joggers ran their Nikes threadbare.
Olof painted herself into a corner
We laughed the speaker off of the stage
We laughed ourselves silly
We yelled ourselves hoarse. (ex 7: pg 2
W's claim: the yeller's hoarseness is not a typical goal or effect of yelling.

BUT, he notes that discourse can override the "selectional restrictions"


of  control resultatives:
\eenumsentence{ She soaps me slippery
He kissed them alive (from Levin and Hovav 1995)
Also:
\eenumsentence{ He kissed mother unconscious.

AND, he notes Green's 1972 observation that the result predicate normally


describes an objective not subjective property.  This observations explains
why many of the examples in \ex{-1 above are *'ed.

He notes (with me) that the result is predicated of the affected theme (my


patient).

L&H's DOR:


\eenumsentence{ *the dog barked hoarse
the dog barked itself hoarse
"If the dog is not semantically appropriate for predication in (11)a,
then why should it become appropriate in (11)b?" pg 4

BACKGROUND feature: encodes pragmatic info (not necessarily truth conditional).


  TELOS: result state or endpoint of an event

Paint is said to have an optional TELOS feature (but since TELOS is


tải về 302.24 Kb.

Chia sẻ với bạn bè của bạn:
1   2   3   4   5   6   7   8   9   10




Cơ sở dữ liệu được bảo vệ bởi bản quyền ©hocday.com 2024
được sử dụng cho việc quản lý

    Quê hương