Transcript SYNTAX

SYNTAX
Wu Yaqing
[email protected]
Office hours: Tuesday14:30-16:30 or
by appointment
 "Language is the most massive and inclusive
art we know, a mountainous and anonymous
work of unconscious generations."
 Edward Sapir Language (1921)
 Writing is clearly an optional accessory; the real
engine of verbal communication is a spoken
language we acquired as children. (Language
Instinct, Steven Pinker 1994)
Why syntax?
 徐烈炯（2001）：“句法学占据整个语言学的核心地位。
欧美国家较好的大学一般都有一个语言学系，每个系都至
少开设一、两门句法必修课。只教社会语言学等边沿学科
的大概不能算语言学系。有不少语言学系只教句法学和
音系学，不教或者很少教其他分支学科，却仍然可以成
为国际上公认的最好的语言学系，所以句法学是每个学语
言学的学生必读的。”
Syntax, in its most general sense, is the
study of the structure of sentences in
natural language. In this course, we will
approach syntax from the perspective of
generative grammar, as pioneered
through the work of Noam Chomsky, and
developed over the past fifty years.
Course Aims & Objectives
The aims of this course are three-fold.
 First, to understand the nature of language as viewed
from the generative perspective, to sharpen students
awareness that natural language is structure-dependent,
and to understand the sort of insight about language this
perspective affords.
 Second, to understand certain general principles or
laws of language which can be elucidated from the
generative perspective by the empirical methods.
Notes
The material in the course primarily will be
drawn from English and Chinese, but
occasional reference will be made to other
languages( although no prior knowledge is
presupposed).
Upon completion of the course, students will
possess:
 a. conceptual framework for syntactic
analysis;
 b.the ability to observe and describe
syntactic data;
 c.the ability to represent syntactic
structures formally and functionally
Requirement
(i) Regular class participation.
(ii) Active class participation and
presentation.
(iii) There will be one course paper. Rules
for writing the course paper will be
explained in classes.
Assessments
Grades will be determined on the basis of
class participation, exercises and course
paper. The course paper will count for
70% of the grade, with the remainder
determined by the exercises and class
participation.
Important notes
 Class lectures will cover the topic described in the
topical syllabus. The readings in the text are
supplementary to material presented in class. While
class material will correspond in large part to material in
the text, the content of what is presented in class will
diverge at points. Morever, there will be material
presented in class which goes beyond that covered in
the text, and material presented in the text which is not
covered in class. Anything in the text which you do not
fully understand should be discussed with the teacher,
either in section or by appointment.
Textbooks
(1) Vivian Cook & Mark Newson.
Chomsky’s Universal Grammar:An
Introduction.外研社
(2) 徐烈炯《生成语法理论》上海外语教育出版社
;
(3) 宋国明《句法理论概要》中国社会科学出版社
;
(4) Andrew Radford’s Transformational Syntax.外
研社
introduction to Government and binding theory.
How hard is Chinese?
Some things about Chinese are hard,
some are easy.
 Hard things:
 Chinese shares very little vocabulary with European languages, so
speakers of these languages have to work harder than if they were learning
 another European language. And even though Chinese shares vocabulary
with several Asian languages (especially Korean, Japanese, and
Vietnamese),this shared vocabulary is often difficult to recognize.
 The writing system is definitely hard to learn, though there is
nothing conceptually difficult about it; there is just a lot to
memorize.
 Chinese is a tone language--that is, different pitch patterns do
not just add emotional color, as in English; they actually
distinguish one word from another. How much of a problem
this is depends a lot on the individual student: students with a
good ear do not necessarily find this a difficulty.
 Easy things:
 Unlike many European languages, Chinese has no irregular
verbs or noun plurals to learn, because words have only a
single form, with no suffixes for tense, number, case, etc.
(There are some particles which work somewhat like tense
endings, but they always take the same form, no matter what
they are added to.)
 Chinese speakers are usually tolerant of a foreigner's
mistakes--perhaps because so many Chinese themselves speak
standard Mandarin Chinese as a second language.
 Chinese is everywhere.
The nature of universal grammar
 It is a theory of human languages (including meaning
and sounds) or rather a theory about human mind,
human knowledge
 Behavior /performance(the actual use of language in
concrete situations) vs. competence/capacity (the
speaker/hearer’s knowledge of his language)
 Knowledge of language: what is the concept of
knowledge? Where is language?汉语/苗语
 Practitioners: practicing linguists or psychologists. Vs.
philosophers and cognitive scientists: ‘reflective’ part of
the field. (NL words and technical terms)
 Polyglot?
Plato’s three levels of learning
 Different levels of knowledge:
(1) knowledge learned from senses, which is not
wisdom. e.g. Fire is hot.
(2) Wisdom, which is more than that kind of
knowledge obtained from sensing objects and
their qualities. Scientists.
(3) Highest level of abstraction.
 Tension between science and philosophy
 Science cannot replace philosophy: the latter is
argument about things of which we are ignorant.
 language research as empirical science?
 Linguistic intuitions are known to be subject to individual
variation
 ？我放了书在桌子上。
 Language data are often messy and do not lend themselves to
elegant generalizations in terms of exceptionless laws.
 Are linguistc genealization culturally bound? Do all languages
have the grammatical category of subjects of a sentence? Do
all languages have complementizers or clause introducers like
‘that’?
Object of inquiry
internalized language (IL), externalized language (EL)




Why is I-Language our primary concern?
IL is a system represented in the mind/brain of a particular individual.
what is our cognitive shift?
The shift of focus from dubious concept of EL to the significant notion
of IL was a crucial step in early generative grammar.
 “Linguistics is the study of IL, and the basis for attaining this
knowledge”.
 The history of linguistics shows a move from an E-language to an ILanguage approach.
 tubalarasa: blank slate.
 operant conditioning vs biological endowment: creative aspect
of language use
 All organisms have special subsystems that
lead them to deal with their environment in
special ways.Some of these subsystems are
called “mental” or “cognitive”.
 The development of cognitive systems, like
others, is influenced by the environment but
the general course is genetically determined.
 Changes of nutrition, for example, can have a
dramatic effect on development, but will not change
a human embryo to a bee or a mouse, and the same
holds for cognitive development.
 The evidence is strong that among the human
cognitive systems is a faculty of language(FL), apart
from severe pathology, FL is close to uniform for
humans.
Idealization
 Linguistic theory is concerned primarily with an ideal
speaker-listener, in a completely homogeneous speech
community, who knows its language perfectly and is
unaffected by such grammatically irrelevant conditions as
memory limitations, distractions, shifts of attention and
interest, and errors (random or characteristic) in applying his
knowledge of the language in actual performance. ( Chomsky,
1965, 3)
 I-language research aims to represent this
mental state, the speaker’s knowledge of
the language, not the sentences they
have produced, it explore the mind rather
than the environment.
 ‘E-language, if it exists at all, is derivative,
remote from mechanisms and of no
particular empirical significance, perhaps
none at all’. (Chomsky, 1991a, p10)
Epistemology of inquiry
 It tries to provide a solution to Plato’s problems
 Where is the knowledge of language? (some sort of
mental state, arrangement of the stuff in the brain)
 How is it acquired? (the acquisition of
language/knowledge-how do we acquire it? How does it
arise in the individual?)
 ·How is it put in use? (how do we use language/
knowledge?)
Russell Problem and Orwell problem are
derived from Plato problems:
 How is it that with so little experience we
know so many things?
 Orwell problem: How is it that with so rich
experience, we know so little?
 An input-output discrepancy
 It is a theory of language acquisition
 It is explicit like any other science, a science in its very
real sense, a science as opposed to art.
 It is a formal theory: explicit as science (physics,
chemistry)
 It is a formal tool measuring human mind. It is a
modular system.
 It is an approximation to scientific truth
 It is an international enterprise
 It is explanatory as opposed to descriptive
 Its task is not to offer a descriptive theory (grammar)
of English or Chinese.
 It's a theory of universal grammar
The “facts”
Child:
Nobody don’t like me.
Mother:
No, say “Nobody likes me.”
Child:
Nobody don’t like me.
Mother:
No, say “Nobody likes me.”
Child:
Nobody don’t like me.
Mother:
No, say “Nobody likes me.”
Child:
Nobody don’t like me.
[dialogue repeated five more times]
Mother:
Now listen carefully, say “Nobody likes me.”
Child:
Oh! Nobody don’t likeS me.
(McNeill, 1966)
The intuitive problem




The child makes an error.
The adult may correct or identify the error.
But the child ignores these corrections.
So, how does the child learn to stop
making the error?
Recovery from Overgeneralization
 u-shaped curve: went - goed - went




child must stop saying:
“goed”
“unsqueeze”
“deliver the library the book”
Argument from Poverty of Stimulus





Noisy input
Incomplete input
Ignoring correction
Not enough feedback
Unclear referentially
 It is an invitation for an answer to the
simplest question like: how come that we
human beings can speak whereas animal
cannot. Is the human brain programmed
to learn language? In what ways are
languages similar and how can they differ?
(1)John knows English or John speaks
English. What does that means?
(2) What has happened from John’s not
being able to speak English to John’s
being able to speak English? What do
we do with language?
 Tension between facts and theory: tug of war
A tension between facts and theory: a
geographical discovery by Eratosthenes.
 Conceptual necessity and empirical necessity
 Why do linguistic theory matter in language
data
 How one conceives data of language
determine the data we use
 Theory affect what we perceive the data
 Data description will be very difficult without
theory;
 Instrumental technique are also theory
dependent
 Theory often leads to reanalysis of familiar
data
 Counterexamples do not refute a theory
Idealization
Linguistic theory is concerned primarily with an
ideal speaker-listener, in a completely
homogeneous speech community, who knows
its language perfectly and is unaffected by such
grammatically irrelevant conditions as memory
limitations, distractions, shifts of attention and
interest, and errors (random or characteristic) in
applying his knowledge of the language in
actual performance. ( Chomsky, 1965, 3)
 Citation of exceptions is in itself of very little
interest. Counterexamples to a grammatical
rule are of interest only if they lead to the
construction of a new grammar of even greater
generality or if they show some underlying
principle is fallacious or mis-formulated.
Otherwise, citation of counterexamples is
beside the point. (P.ix. Chomsky and
Halle ,1968)
 Counter examples may signal the presence of
other factors rather than the falisity of the
hypothesis. They lead to refinement of the
theory rather than refute the theory
 “Scientists fail to reject paradigms when faced
with the anomalies or counter instances. They
could not do so and still remain scientists.”
(Kuhn, p.78)
 One’s conception of language determines the type
and range of data one would admit as relevant. The
theory of language may affect how data are perceived
and conceptualized. What counts as evidence for a
particular theory is not sth that can be decided uopn
independent of the theory content. Theories are not
generally given up in the face of counterexamples,
rather counterexample often contribute to the
refinement of theories by drawing attention to
additional factors.
Universal grammar and parameter
 Principles that hold of all languages are said to
be part of UG, which results from the biological
endowment or innateness hypothesis.
 UG must be geared to any human language and
not to just one, and it is not sufficient to enable
one to speak a language. The native language
is that spoken by the Child’s immediate
environment.
 Variation between different languages and
parameter: WH-movement Polish/English/
Chinese. Parametric variation.
A tension between UG and Parameter
⑴UG contains a set of absolute universals,
notions and principles which do not vary
from one language to the next.
⑵There are language-specific properties
which are not fully determined by UG
but which vary cross-linguistically.
The solution to this conundrum is to say
that UG lays down the basic principle,
but that each language is free to pick and
choose the parameters along which
those principles are realized.
Language learning and language
acquisition
 Triggering experience Language
X→UG(with parameters)→Core grammar
Language X
 What are learned? New words and less
usual constructions. Core grammar and
marked periphery of the grammar (the
exceptional or marked patterns of the
language, which are acquired later)
A child’s language acquisition task
Parameter setting
+ new words learning
+ social or cultural conventions learning
The nature of universal grammar
 Structural dependency
 A principle common to the syntax of all
languages.
 Definition: operations on sentences such
as movement require a knowledge of the
structural relationships of the words rather
than their sequence.
 Structural Dependency
1.The man who is first in line is coming.
2.*Is the man who __ first in line is coming?
3.Is the man who is first in line ___ coming?
This only applies to non-parameterized
aspects of language.
More cases
 *Who did John believe the man that kissed ___
arrived?
 Who did John believe __ kissed his buddy?
 *What did you stand between the wall and __?
 *What did you see a happy ___?
 阿Q[从小][便学会浇花]。
 阿Q[从小便][学会浇花]。（蔡维天）
 无鸡鸭也可，无鱼肉也可，蔬菜一碟足矣。
 无鸡，鸭也可，无鱼，肉也可，蔬菜一碟，足矣。
No need for positive evidence
Chomsky: “A person might go through much or all
of his life without ever having been exposed to
relevant evidence, but he will nevertheless
unerringly employ the structure-dependent
generalization, on the first relevant occasion.”
Hornstein and Lightfoot “People attain knowledge
of the structure of their language for which no
evidence is available in the data to which they
are exposed as children.”






Susan must leave. Must Susan leave?
Bill is sleeping. Is Bill sleeping?
1.Invert the first two words in the (a) sentences.
Mary read the book. *Read Mary the book?
The man left. *Man the left?
2.Move the Aux to the sentence-initial position.







Mary has been sleeping.
*Been Mary has sleeping?
3. Move the first Aux to sentence-initial position.
The man who is here can swim.
*Is the man who here can swim?
Did Mary read the book?
Did the man leave?
 Crain and Nakayama (1987)did relevant
experiments involving elicited production
of yes-no questions, which clear
demonstrated children do not make this
kind of mistake.
 non-structural dependent error.
 Why don’t children make such error?
 4.Move the first aux after the subject to
the front.
Constituent tests






Pronominalization
John left. He left. (pronoun stands for a noun?)
The man left. *The he left?
The man in the room left. *He in the room left?
The old man left. *Old he left?
Topicalization and coordination
The head parameter:a particular language consistently
has the heads on the same side of the complements in all its
phrases, whether head-first or head-final.




SVO languages vs. SOV languages
how to determine the head?
How to set the head parameter?
The relative position of heads and complements
needs to be specified only once for all the
phrases in a given languages.
Three adequacies
 Observational adequacy: Facts are
accounted for
 Descriptive adequacy: also generalization
 Explanatory adequacy: a goal hard to
achieve
Methodology: Contrastive study
An extremely powerful theoretical tool to know what a
grammatical sentence is: Basic techniques in
contrastive study
(1)Use one’s intuition;
(2) Make contrast in a minimal pair; formulate a rule which
can account for the well-formed /ill-formed sentences as
well.
 ·Approximation method
 First approximation, 2nd ….
 ·how to represent syntactic structure: distinction
between metalanguage and object language
 ‘Snow is white’ is true iff snow is white.
 HUDA contains four letters.
The Unit of Syntactic Analysis
Sentence parsing :immediate constituents and
ultimate constituents. Structural hierarchy
 Constituency: If a category X is dominated by
another category Y, then X is a constituent of Y.
 Dominance/ mother node precedence/sister
node
 Horizontal relation: precedence
 Vertical relation: dominance
 Phrase structure rule /rewrite rule
 Labeled brackets vs. tree diagram
what are Spoonerisms?
 Spoonerisms are phrases, sentences, or words in
language with swapped sounds. Usually this happens
by accident, particularly if you're speaking fast.
Come and wook out of the lindow is an example.
 Of course, there are many millions of possible
Spoonerisms, but those which are of most interest
(mainly for their amusement value) are the ones in
which the Spoonerism makes sense as well as the
original phrase. Go and shake a tower and a wellboiled icicle illustrate this well (go and take a
shower, a well-oiled bicycle).
 The name Spoonerism comes from the Reverend
William Archibald Spooner who is reputed to have
been particularly prone to making this type of verbal
slip.
 有主持人把“多好的孩子们哪”说成“多孩的好
子们那”
 “移动联通小灵通的用户请发短信至”说成“移
动联通的小用户们请……”
 “现在我们电线连话在现场的记者”
 “在打击车匪路霸的行动中，有他的身影；
在拐卖妇女儿童的行动中，有他的身
影……
 “法院开庭”说成“法庭开院”。
 “据新华社消息,今天凌晨,伊拉克军队已经
成功地切断了科威特的两条输卵管道。”
“迅雷不及掩耳盗铃之势”
"在上周刚举行了一场别开婚面的生礼。"
 韩乔生：您最近真是来者不善罢甘休，一路刀砍
斧劈头盖脸，像赵子龙七进七出污泥而不染。王
朔：这算什么呀，哥们不过是小试牛，我要是拿
出真本事，他们还不全军覆。
 韩乔生：有人说你疯了，还说你病得不轻于鸿毛。
 王朔：我装疯卖就可以把他们丫的搞定，到底是
谁病入膏了？
 韩乔生：可是大家对您颇有微词不达意。
 王朔：我是流氓我怕谁，我不学无，我醉生梦，
我无可救，我胡言乱。
 韩乔生：可是有人说你炒作，说你绑架媒体无完
肤。
 王朔：我就炒作了，我就是让他们看看什么叫货
真价的炒作，我明修栈，暗渡陈，声东击，得陇
望，兵不厌，瞒天过，你瞧瞧他们丫的炒作，故
弄玄，低三下，黔驴技，卑鄙无……
 韩乔生：那您真是炒作专家长里短。
 王朔：我这刚略施小，他们就一塌糊了。所以他
们嫉贤妒，对我恶语相。韩乔生：木秀于林志玲，
风必催芝华士，众口铄金喜善，积毁销骨肉皮，
枪打出头鸟儿问答，出头的椽子先烂醉如泥。
 王朔：他们就是想看我丑态百，那我就顺水推，
你看他们开始如坐针，忍无可，自卫还了吧。
韩乔生：这叫置之死地而后快乐大本营。
 王朔：我就讨厌这帮人动不动就同仇敌，千军万
我都见过了，还怕他们这帮乌合之。
 韩乔生：那是，您所向披靡靡之音。
 王朔：其实我慈眉善，心地善，助人为，行侠仗，
八荣八，我都占全了。
Garden path sentences







The horse raced past the barn fell.
When Fred eats food gets thrown.
Mary gave the child the dog bit a bandaid.
I convinced her children are noisy.
Helen is expecting tomorrow to be a bad day.
I know the words to that song don't rhyme.
She told me a little white lie will come back to
haunt me.
 Until the police arrest the drug dealers control
the street.
 The dog that I had really loved bones.
 That Jill is never here hurts.
 The man who whistles tunes pianos.
 The old man the boat.
 The cotton clothing is made of grows in Mississippi.
 Have the students who failed the exam take the
supplementary.
 Every woman that admires a man that paints likes
Monet.
 The raft floated down the river sank.
 We painted the wall with cracks.
Projection principle(PP)
 Lexical information must be syntactically
represented.
 Subcategorization and selectional
restriction
 0-place predicate, 1-place… 2-place…
 Extended projection principle (EPP)
 When verbs appear in phrases or
sentences, they rarely come alone.
Mastering the use of verbs requires
knowledge of the company they ask for.
2.Words and Phrases: Syntactic category
and its distribution
 Mental lexicon and categorial specification
of a word.
 Constituency tests/criteria
 Displacement, insertion, deletion
coordination, replacement with a pro-form
 他不但骂了你，而且打了我。（coordination）
 不但他骂了，而且你打了我。(coordination)
 也许你打了我，你也许打了我，他打了也许我。
(insertion)
 是他打了你。他是打了你。他打了是你
(displacement)
 他不应该打我，你也不应该。 (Deletion)
 打人，你不应该。(movement)
 你打了我，他也这么做了。(substitution with
pro-form)
 Jeeves will meet his employer at the castle.
 Miss Marple will read the letters in the garden
shed this afternoon.
 3. predicates and arguments
 3.1 Subcategorization. adjuncts.
Transitive, ditransitive and intransitive.
Types if universals
 UG does not insist all the languages are
the same.The variation introduced through
parameters allow universals to be all but
undetectable in particular languages.
 Greenbergian Universals
 Accessibility hierarchy
 All languages have relative clauses in which the
subject of the relative clause is related to the
noun.
 subject object  indirect objectobject of
preposition  Genitiveobject of comparison.
 All languages start at the left of the hierarchy
and have subject relative clauses, no language
can avoid this sequence.
 The accessibility Hierarchy is datadriven,an implicational universals.
 No compelling reason within UG theory
why this should be the case.
 UG are theory-driven. They may not be
breached but they need not be present.
The language faculty
 Language knowledge is independent of
other aspects of the mind.
 Human Mind is divided into separate
compartments,or modules, each
responsible for some aspect of mental life.
Language faculty, mathematics, vision,
logic.
Cognitive theories
 Mind is a single unitary system,
connectionism
Principles and rules
 “what we know is not a rule system in the
conventional sense. In fact, it might be that the
notion of the rule in this sense…has no status in
linguistic theory”
 (Chomsky, 1986,P151)
 Rules are artifacts of the interaction between
the principles and the lexicon. The information
states in rules should be interpreted as general
principles that affect all rules rather than a
property of individual rules.
3.2 Argument Structure and
Thematic Structure
primitive property: a property which does not follow from
anything else.
 Argument: An argument is the phrase that denote the thing that
has some property or is involved in such relationship.
 Two kinds of arguments: 1) the subject, whose presence in a
sentence is for the most part independent of the particular verb. 2)
There are the arguments associated with a particular
verb./sentential argument
 One-place predicate: stumble, dither, run; two-place predicates:
imitate, see;
 three-place predicates: give, put.
 The sets of arguments selected by the verb, including the subject
argument, constitutes its argument structure.
 Distinction between subcategorization
frames and argument structure/implicit
argument: buy, envious;
 Argument of other lexical categories:
adjectives and nouns.
Thematic role /theta role/θ-role
The semantic role played by an argument in relation to its
predicate:
 agent/actor: instigator of some action
 theme/patient: entity undergoing the effect of some
action (Mary fell over)
 recipient/possessor/benefactive: entity receiving
/possessing some entity (John got Mary a present)
 experiencer: entity experiencing some psychological
state (John felt happy)
 goal: entity towards which something moves (John went
home)
 source, location, etc.
Thematic/theta grid
Theta Criterion:
1) Each argument is assigned one and only
one theta role.
2) Each theta role is assigned to one and
only one argument.
The projection principle
Lexical information is syntactically
represented.
Lexical category and functional
category (comp, det, Infl)
着想、操心、关心、说事、过不去、抬杠
Complementizer(Comp)
 It denotes a particular category of clauseintroducing word such as that/if /for,
 It is also used to denote the pre-subject position
in clauses which is typically occupied by a
complementizer like that/if /for, but which can
also be occupied by an inverted Aux in
sentences like Can you help? It has the function
of introducing the S-complement of the verbs.
 Clausal arguments/ small clause
 Expletives:it as a slot-filler,
expletive/pleonastic it.
 Extraposition. There in existential
sentence.
 Neither Auxiliaries nor the copula be
assign thematic roles.
 Extended Projection Principle: refers to
the general property of sentence that they
must have subjects.
 The choice of the object affects the
thematic role of the subject while the
choice of the subject argument does not
affect the role of the object.
 Direct theta-marking (internal argument:
complement)/ Indirect theta-marking
which is compositionally determined by
the whole verb+complement structure.
(external argument: subject).
 Phrase Structure: X-theory
 Linear relationship and hierarchical
relationship
 Endocentric and exodocentric
Garden path sentences





The horse raced past the barn fell.
When Fred eats food gets thrown.
Mary gave the child the dog bit a bandaid.
I convinced her children are noisy.
Until the police arrest the drug dealers control
the street.





The dog that I had really loved bones.
The man who whistles tunes pianos.
The old man the boat.
The cotton clothing is made of grows in Mississippi.
Every woman that admires a man that paints likes
Monet.
 The raft floated down the river sank.
 We painted the wall with cracks.
Dominance
 The vertical relations between the elements in a
tree. (immediate dominance)
 Node A dominates node B iff A is higher up in
the tree than B and if you can trace a line from A
to B going only downwards.
 Precedence: the horizontal relations between
the elements in a tree. (Non-branching/
branching)
 Node A precedes node B iff A is to the left of B
and A does not dominate B or B does not
dominate A. Sisterhood.
 Governor (the element which
governs )/governee (the element that is
governed)
 Government (i)A governs B if A is a governor;
(ii) A and B are sisters;
Governors are heads.
 If X is a head and it governs Y then X headgoverns Y. All the constituents governed by a
node constitute the governing domain of that
node. When a V governs an element and
assigns an internal theta role to it we say that it
theta-governs this element.
 The structure of Phrases: ? the common
properties of VP, NP, AP, PP
Substitution is structure-determined:
Only constituents can be substituted for by
an element.
Projection
A projection is a constituent which is
an expansion of a head word.
e.g. an NP such as students of
linguistics is a projection of its head
noun students.
 A minimal projection (zero projection) is a
constituent which is not a projection of
some other constituent: hence heads are
minimal projection.
 An intermediate projection is a constituent
larger than a word, but small than a
phrase.
 A maximal projection is a constituent
which is not contained with any larger
constituent with the same head.
 He is proud of you (MaxP).
 He is proud (MiniP&MaxP).
 Complements combined with X to form X’
projections, adjuncts combined with X’ to
form X’ projection, The specifier combines
with the topmost X’ to form the maximal
projection. X’’=XP for convenience’ sake.
 The relative order of the constituents is
not universally fixed.
 The differences between the internal
structures of the phrases. V& P take
complement, N & Adj do not.NP AdjP, PP
are preceded by a Spec. not for VP.
The principles of X-bar Theory will be part
of UG, which is innate. Very little data will
suffice to allow the child to fix the ordering
constraints of the language he is learning.
The Structure of Sentence: ?
Agreement
Two words are said to agree in respect of
some grammatical features if they have
the same value for the relevant
features.AgrS /AgrO
 Italian and French AGR are stronger than English
and Chinese AGR.
 Infinitival clause: How is agreement realized?
 No tense marking or agreement in infinitives and the to
in infinitives is assumed to correspond to the verb
inflection. I didn’t expect Poirot to abandon the
investigation.
 The structure of IP
 I’’→Spec;I’ I’ →I;VP.
 Subject is not a primitive notion in GB theory.
 INFL dominates the inflectional morphology of the verb,
affixes and infinitival to, which are not independent
lexical categories or words, and the aspectual auxiliaries
have and be and the modals.
 INFL is a non-lexical head, for it does not dominate
open class lexical head.
 S’ is a projection of C: Whether the same analysis can
be extended to S’?
 The Comp (that, if, whether, for) determines the type of
clause(IP).
 That and if take a finite clause as their complement;
 for selects an infinitival one and whether selects either
type of clause.
 Head to head movement: Movement of a word from one
head position to another.
 Speaker A: Honey-buns, there’s sth I wanted to ask you.
 Speaker B: What, sweetie-pie?
 Speaker A: If you will marry me.
 Speaker B: (pretending not to hear): What d’you say,
darlin’?
 Speaker A: Will you marry me?
 Rationale for the proposals that INFL goes to
the position of C, which was occupied by the
Comp in subordinate clauses, and that WHwords move to [Spec, CP].When will you marry
me?
 The structure of CP: C’’→Spec; C’; C’ →C;IP
 All syntactic structure is built on the basis of the
X’-format.
 Small clause poses a question to the analysis
Structural relation
 Specifier: The grammatical function fulfilled by certain
types of constituent which (in English) precede the head
of their containing phrase. What did John do?
 Specifier-head relation (agreement): a relation between
a head and its specifier. E.g. We may say that the
Subject-Aux agreement involves a spec-head relation in
sentences like He has gone, since has is the head of
the clause and agrees with the specifier he.(also found
in 64a,67,69)
 Parametric variation in the morphological realization of
this agreement. Overt/covert
 C-command: Node A c-command node B iff (a) A does
not dominate B and B does not dominate A; (b) the first
branching node dominating A also dominates B.
 Government: A governs B iff (i) A is a governor; and (ii) A
c-commands B and B c-commands A.
 C-Command : a metaphor. To use a simple train station
metaphor, one node X c-command another node Y if
you can get from X to Y by taking a northbound train
from X, getting off at the first stop, and then taking a
southbound train to Y.
M-command: What is it for？Quit his job in the autumn/
leave in the autumn
The notion of c-command and government have been
reformulated in terms of maximal projection
C-command (2):
 A c-commands B iff A does not dominate B and every X
that dominates A also dominates B.
 It is strict c-command when X is equated with the first
branching node. If not, X is a MP.
 Government (3): A governs B iff A m-command
B and no barrier intervenes between A and B.
 Maximal projection are barriers to
government.Governors are heads.
 Binary branching vs. Ternary branching why is
the former preferred? More constrained for
reasons of economy and elegance/aesthetically
more satisfying./easier for processing and
acquisition p130.A matter of learnability.
 Syntactic features: What are syntactic primitives?
Distinctive features in phonlogy. Lexical
category as complexes of syntactic features.
Chapter 3 Case Theory
Formal properties of overt NP
 Abstract case vs. Morphological case: the former is an
universal property, the latter varies cross-liguistically.
Lists offer no insight in the phenomena that are listed.
 English has three cases-nominative,
objective/accusative and genitive. Personal pronoun has
three cases (I, me, my respectively), and NP inflected
only for genitive case (John’s).
 English has a full-fledged system of abstract case.The
degree of overt morphological realization of abstract
case varies parametrically from L to L.
 complements of prep and verb: case is
assigned under government, by the governors
(heads).Not all verbs assign cases to their
complement and NP/AdjPcannot either.
 What is responsible for the nominative cases in
the subject position? [+Tense]/[+AGR]
 Barrier for government move towards him
/Poirot attacked him.PP is a barrier for
government by V and vp is a barrier for
government by I.
 In order to exclude multiple governors and to define
maximal projection as barriers, we introduced the notion
minimality into the definition of government.
 A governs B iff (i) A is agovernor; (ii) Am-commands B;
(iii) minimality is respected.
 Minimality is defined as: A governs B iff there is no
node Z such that (I) Z is a potential governor for B;(ii) Z
m-commands B;(iii) Z does not m-command A.
 Each head determines a domain of influence into which
outside heads cannot govern: the maximal projection of
the head is the governing domain.
 How do we account for the Accusative case of
the subject NP of the infinitival clause?
 [For him to attack him] would be surprising.
p154.
 *[Him to attack Bill] would be illegal. *I prefer
very much [him to go now].
 The sentences can be saved by the insertion of
for as the complementizer of the non-finite
clause or by the omission of the overt subject of
the infinitival clause.
 Case Filter (Why do NPs need cases?)
 Every overt NP must be assigned abstract case.
 Exceptional case marking (ECM): Objective
subjects in the infinitive clause (I believe [him to
be honest]) are said to carry exceptional case
(its case being checked by the preceding verb).
 The ‘exceptionality’ is related to the fact that
maximal projections normally constitute
barriers for case assignment from the
outside.
 Small clause AP is not a barrier for an outside governor.
 What is a barrier? Maximal projection such as VP,CP
and PP are barriers, infinitival IP and small clauses are
not.
 John opened the door. The wind opened the door. A
chisel opened the door. *John and a chisel opened the
door.
 A: My dog’s got no nose. B: Oh, really. How does he
smell? A: Awful.
 NP is Nominative when in Spec of AgrSP, where AgrS is
[+Agr]
 Adjectives and nouns : They are not case
assigners in English. Of-insertion can save the
ungrammatical sentence.
 Failure of Of- insertion and two types of case
assignment: Structure case assignment and
inherent case assignment. The former is merely
subject to structural requirements and is blind to
thematic relations, and the latter is dependent
on two conditions: (i) theta-role assignment and
(ii) government.
 Inherent case condition: If A is an inherent case
assigner, then A assigns case to an NP if and only if A
theta-marks the NP. P165 Inherent case in German
 Adjacency requirement: It is a structural requirement
which predicts that case assigners must not be
separated from the NPs which they case mark by
intervening material between them. This requirement
can also be interpreted as the following: a condition
requiring that two expressions must be immediately
adjacent in order for some operation to apply. e.g. to
can only contract onto want (forming wanna) if the two
are immediately adjacent. (42e)
 Passivization and argument structure
 Why agent role is not assigned to an NP in an A-position.
It is absorbed by the passive morphology on the verb.
Passivized verbs loses the ability to assign structural
ACCUSATIVE case to its complement.
Properties of passivization and the connection between the
properties .
(1)the verb morphology is affected;
(2)the external theta role is absorbed;
(3)the structural case of the verb is absorbed;
(4)the NP which is assigned the internal theta role of the
passive verb moves to a position where it can be
assigned case.
(5)the movement of the NP is obligatory in vies of the case
filter.
(6)the movement of the NP is allowed because the subject
position is empty.
 Visibility requirement: In order to be
recognized as an argument of some
predicate an NP must be made visible.
Invisible NPs cannot be assigned a theta
role. NPs are licensed by virtue of their
case properties.
 Chain: a set of one or more constituents
comprising an expression and any traces
associated with it.
Theta criterion
 Each argument A appears in a chain
containing a unique visible theta position P,
and each theta position P is visible in a
chain containing a unique argument A.
 c.f. Each argument is assigned one and
only one theta role and each theta role is
assigned to one and only one argument.
ASSIGNMENT (to be due next weekday)
(1)P182. Exercise 2. (4), (5),(8)
(2)P184. Exercise 5. (1), (2), (8)
(3) Why is ‘ECM’ impossible in the following
contexts?
a.John tried Mary to leave.
b. It seemed John to be nice guy.
c. I assured you John to be a nice guy.
Chapter four
 Anaphoric and overt NPs
The interpretation of NP and the binding
theory (hence GB Theory)
 Informally, binding is a type of referential dependency,
whereby the reference of the bound element is exactly
the reference of the antecedent. Binding is a semantic
relation. To say that one constituent X binds ( or serve
as the binder for) another constituent Y is to say that X
determines the semantic properties of Y. e.g . (pronoun
of laziness)
 Johni wants PROi to leave. John binds PRO.
 Every dogi loves itsi owner.
 The preliminary working assumption is that all aspects
of binding are determined in the syntactic representation;
the semantic binding relation on this view is then
parasitic on the syntactic one.
 The device of indexing . the index of an NP correlates
with what the NP refers to. Co-indexation.
 Reference is a relationship between part of a sentence
and the external world. Referential dependence and
coreference are two different relations, as can be seen
in this sentence:
 [No contestanti ] thought that shei would win.
Three types of NPs should be distinguished
(1)reflexives and reciprocals (each other)are anaphors.(It must have
an antecedent in the same CL)
(2)Pronominal expressions.(It cannot have an antecedent in the same
clause)
(3)R-expressions. (It must be free, i.e. not bound )
We refer to the referential dependency of (i) (ii) as anaphora.

Johni likes him*i,j. Johni (Mary*i,*j)likes himselfi,*j. Johni
believes that Mary likes himi.

*Johni believes that Mary likes himselfi.

*Johni likes Johni. *Hei likes Johni. *Johni thinks that Mary likes
Johni. *Johni can’t stand Johni’s teacher. *Hei can’t stand Johni’s
teacher.

Pragmatic factors should be excluded from our consideration.
Context-independent.
 More on reflexives: Reflexives must be bound, locally
bound (clause-mate condition), by the antecedent.(1)
 Binding : A binds B iff (i) A C-command B; (ii) A and B
are co-indexed.
 The domain of reflexive binding and the revision of the
Principle (1) p200.
 A reflexive must be bound inside a clause that contains
it and its governor.(2)
*Poiroti believes [Miss Marple’s description of himselfi]
Poirot believes [any description of himself].
 A reflexive must be bound in the minimal domain
containing it, its governor and a subject.(3)
*Poiroti believes [CP that [IP himselfi is the best].
*Maryi expects[IP us to like herselfi].
Maryi expect [IPus to like ourselvesi].
 Two objections to the tensed clause proposal:
epistemological objection and empirical
objection.
 What is subject? Subjects of clauses, tensed or
infinitival, and the subjects of NPs.
 Distinction between subjects of finite clauses
and those of non-finite ones and NPs. The AGR
of finite clauses is SUBJECT.
An anaphor cannot be bound if
(1)it is marked with NOMINATIVE CASE;
(2)it is in a domain that has a subject and its
antecedent is outside of the domain.
 张先生的朋友批评了自己。张先生的阴谋反而害了自己。（xu’s
article in 1997 and the literatures cited in the
bibliography）
More on reflexives evidence from Chinese: Coreferentiality.
Disjoint referentaility.
(1)他吹嘘自己 *自己吹嘘他 为了自己的利益，谁也阻挡不了他。
(2)张先生批评自己。 张先生批评他。张先生承认自己错了。张先生承
认他错了。
(3)张先生的朋友批评了自己。 张先生的阴谋反而害了自己。
(4)张先生的爸爸i的阴谋被自己i的朋友识破了。张先生i的爸爸j的钱
被自己i/j的朋友偷走了。
(5)小张不肯告诉小李自己的分数。老师不肯告诉学生自己的分数。
(6)这件事告诉他自己以前的想法不一定对。(subject orientation)
(7)小张说小李只关心自己。小张说小李不关心自己。(pragmatic
consideration)
(8)小张说小李知道小王不喜欢自己（long distance
binding ）Minimal clause effect/Maximal clause effect)
我想老奶奶i还不知道卖鱼的骗了自己i。
(9)小张j知道小李i不喜欢自己i/j。小张知道我i不喜欢自己i。
(Blocking effect)
(10) 我j知道小张i不喜欢自己i/j。小张知道小李i不喜欢他
自己i。
(11) 张三j认为李四i不喜欢自己i/j.
 如果“自己”是反身代词，按‘约束论”它不应该受“张三”的约束，如
果“张三”是人称代词，按“约束论”它不应该受“李四”的约束。
 三种解决方法：1）认为“约束论”理论解释力不足，提出了
generalized binding theory.2）认为“约束论”的理论解释力可以
进一步挖掘的提出了反身代词随INFL抽象移动。3）认为汉语、朝语
等语言的反身代词本身不同于英文反身代词的看法，这些语言中的反
身代词具有人称代词和反身代词的双重属性，这样可以解释例
（11）。这种属性称为 “logographoricity”. Open question.
Approximation.(1st approximation, 2nd approximation, …)
 对待例外的态度。西方现代理论十分欢迎“例外”和“反
例”，认为与已有规则相矛盾的“例外”和“反例”是推动
理论发展的动力，是进行理论逼近的依据。不为维护已有
理论把它们拒之门外，而是主动寻找它们，从中找出理论
的不足，从而追求理论的完美。
立论——例外/反例n→理论n——例外/反例n+1→理论
n+1…
 Anaphors: Anaphors include reflexives and reciprocals.
An anaphor must be bound in the minimal domain
containing it, its governor and an accessible
subject/SUBJECT, where the italicized part can be
replaced with the governing domain.
 Pronouns: In many major respects, ordinary pronouns
are the opposite of anaphors.
First, unlike anaphors, they don’t actually need
antecedents, although they can have them.
He wrote an opera. ( he :contextually given male
individual )
Phil said he wrote an opera. (he could be either Phil or
some contextually given male individual )
Second, the antecedent of a pronoun does not have to ccommand that pronoun.
Johnnyi’s managerj exploited himi.(himselfj).
Third, the antecedent of a pronoun can be quite far away
from the pronoun, and it cannot be too close to that
pronoun.
 To summarize, the constraint on pronoun is that they
must be free in a particular syntactic domain.
Pronouns must be free wherever anaphors must be
bound. Pronoun must be free in its governing domain.
This is sometimes called ‘Disjoint Reference Condition’.
 R-expressions: they don’t require antecedents of any
type, they have their own semantic content. They are illformed in the contexts where anaphors are well-formed.
*Hei/Briani hates Briani.
 A non-c-commanding co-indexed R-E is allowed. Hisi
manager exploited Johnnyi.
 R-E differs from pronouns in that they cannot be bound
even in the contexts where pronouns can be bound. Rexpressions must be free everywhere.
 J.Ross说：“如果你真的想了解语言中所发生的情况，别
把眼光放在诸如The farmer kills the duckling之类的句子
上。使人感兴趣的现象出现在（合格与不合格句的）边
缘地带，只有同冗长且复杂得不可思议的句子打交道时才
会看到。”
 What books did you order John to ask Bill to persuade
his friend to stop reading?
 The man who the boy who the students recognized
pointed out is a friend of mine.
Problems remain in the binding theory
 Implicit arguments: an open question.
 Possessive pronouns and Anaphors: evidence from E
and other language.
 NP types and features: NPs as feature complex
 Syntactic primitives and feature decomposition
 Reciprocal and reflexives: [+anaphor,pronominal].Pronouns: [-anaphor, +pronominal]
 R-expressions: [-anaphor, -pronominal]. ? [+anaphor,
+pronominal]: the last NP.
 The last NP is subject to contradictory requirements: it
must be at the same time be bound and free in its GC.
What is the way out? No GC. An element might lack a
GC if it does not have a governor. If an overt NP lacks a
governor then this NP will not be case-marked and will
be ruled out by virtue of the case filter. No overt NP but
covert NP PRO meets such requirement.
 The binding theory can be used as evidence for the
presence of PRO.
*Sidi permitted his childrenj [(PROj) to mutilate himselfi]the
cat/ themselvesj/each otherj.
Sidi promised his children [PROi to mutilate himselfi] the
cat/*themselvesj/ *each otherj
PRO is a pronominal anaphor, it cannot have a GC, it is
neither governed nor case-marked.
Circularity and I-within-I filter
 Agr is introduced to avoid the apparent paradox.
a.Maryi thinks that [IP [a beautiful picture of herselfi]j [Agrj
[is hanging on the wall of the post office.
b. *[John and Susan]i think [ that [IP each otheri [Agri is
intelligent]]].
 Agr is not accessible if changing the index of Agr and
the index of the subject to the index of the anaphor
would yield a situation in which the anaphor and the NP
that contains it have the same index.
 The condition described above in which
the anaphor an the NP that contains it
have the same index is called the Iwithin-I Condition. Consider a concrete
example of how it is supposed to work. In
(a), a beautiful picture of herself and Agr
are coindexed with j. Let us change the
index to the index of the anaphor. …[IP [a
beautiful picture of herselfi]i Agri…]
 We now have a situation in which herself
is contained in an NP with the same index.
Thus Agr is not an accessible subject, and
IP is not the GC for herself. Hence herself
can have an antecedent outside of the
clause containing the NP that dominates it.
Assignment ( to be due next Thursday)




Exercise 2 from the textbook
Exercise 3. From sentence 9-15.
Exercise 6.
Exercise 9.
Chapter 5
Non-overt Categories: PRO and Control
Empty category
A category which is covert (i.e. which is silent or null and
hence has no overt phonetic form). Empty categories
includes traces, the null pronoun PRO and pro, the null
generic determiner,etc.
 We would like [ you to stay].
 We would like [to stay]
 We don’t want [anyone to upset them].
 We don’t want [to upset them]
The module of the grammar which regulates the distribution and the
interpretation of PRO is called control theory.
 Control: relationship between an empty PRO subject and its
antecedent.
 Verbs which allow an infinitive complement with a PRO
subject are said to function as control verbs.
Why should we posit PRO?
 Part of the motivation for positing PRO is semantic in
nature. In traditional grammar it is claimed that
subjectless infinitive clauses have an understood or
implicit subject- and positing a PRO subject in such
clauses is one way of capturing the relevant intuition.
 Another motivation is syntactic evidence which comes
from the syntax of reflexive anaphor.








They want [ John to help himself].
*They want [John to help themselves].
John want [PRO to prove himself]. Locality Requirement.
They want [ their son to become a millionaire/ *millionaires]
He wants [his sons to become millionaire/* a millionaire]
They want [ PRO to become millionaires/* a millionaire]
He want [PRO to become a millionaire /* millionaires]
Predicate nominals in copular constructions have to agree with the
subject of their own clause
 Features of PRO: semantically understood, syntactically
represented but implicit, phonetically null, empty, covert.
1.[CP[IP PRO to abandon the plan]] would be regrettable.
PRO[+pronominal]
2.Poirot was glad [CP[IP PRO to abandon the investigation]].
PRO[+anaphor]
 PRO cannot be used as a direct object, nor as the subject of finite
clauses, and not every infinitival construction allows PRO as its
subject.
PRO differs from overt NP in that PRO must occurs in
an ungoverned, hence caseless position. That explains
why there is no governing category for PRO, which is
[+anaphor][+Pronominal]. PRO is in complementary
distribution with overt NPs.
PRO theorem: PRO must be ungoverned.
PRO as the subject of infinitival clauses, gerundival
clauses and small clauses.
Obligatory control and optional control
Subject control and object control
C-command and obligatory control
The controller: Argument Control
Control Patterns
 (i) Pro as the subject of a complement clause.
Pro must be controlled by an NP when it is used as a
subject. Arbitrary control must be illegal. When the verb
of the matrix clause is [+WH], different properties of the
Pro obtain: either subject control or arbitrary control is
possible. See examples in 41 on p259.
A controller must be present in the case of obligatory
control. Certain verbs in English may take arguments
optionally rather than obligatorily. Lead, promise
 (ii)PRO as the subject of adjunct clause is also
obligatorily controlled.
 (iii)PRO in subject clause is not obligatorily controlled.
 (iv)Passivization and control




partial control (PC)
Johni wanted [PRO to gather in the library at 6]
Control shift (CS)
Johni asked his mother [PROi to stay up late on New Year’s
Eve]
 Split antecedent control (SAC)
 Johni asked his motherJ [PROi +j to stay up late together on
New Year’s Eve]
 How to establish the right division of the labour between
syntax and semantics
Chapter 6
Transformation: NP Movement




Passivation/ theta-role / parallelism between active
construction and passive construction
Surface structure and deep structure: the Dstructure encodes the predicate-argument
relations and the thematic properties of the
sentence, and the S-structure representation
accounts for the surface ordering of the
constituents.
Underlying order [IP e[I’ [VP [V’ [NP]]by NP]]]
Coindexation / derived subject / base position/
base generated
Question
 scope of questions: yes-no question/ echo question/ whquestion or constituent question: root question/direct
question
 embeded / indirect yes-no question and wh-question
 head to head movement in yes-no question
 no movement is involved in echo question
 wh-movement : WH-element is moved to the [Spec, NP]
 Three types of movements: wh-movement, head
movement and NP movement: “move-α”
 Target of movement (landing site)/ the element that is
moved.
 NP Movement: in passive construction,
raising verbs and adjective
 Raising verb/adjective: it denotes a word
like seem/likely or certain whose subject
are raised out of subject position in a
complement clause to become subject of
the seem / likely or certain clause.
 Trace: A trace is an empty category left
behind (as a result of movement) in each
position out of which a constituent moves,
indicated by t. The moved element is
called the antecedent of the trace.
 Trace of movement is based on the
discussion of the projection principle(EPP),
the theta-theory and the case theory.
 Local: An operation is local only if it
operates within a highly restricted domain,
crossing no more than one intervening
phrasal boundary. E.g. Agreement
typically involved a local relation between
the head and specifier of a given type of
phrase. Clause-mate condition on
agreement.
Some properties of NP-movement
 properties of A-chains: passivization and
raising are mutually exclusive.p290.
 C-command
ASSIGNMENT(to be due next week)
 Exercises 3
 Exercises 5.
Unaccusative
An unaccusative predicate is a verb like
come which allows a post-verbal subject
( as in From inside the house came a cry
of anguish).passive verbs, raising verbs
and verbs used in existential constructions
are unaccusative verbs. Unaccusative
verbs fail to assign accusative case and
lack an external theta-role. Passsive,
ergative and middle verb
Ergative
This term originally applied to languages like
Basque in which the complement of a transitive
verb and the subject of an intransitive verb are
assigned the same case. However, by extension,
it has become to be used to denote verbs like
break which occur both in structures like
Someone broke the window and in structures
like The window broke, where the window
seems to play the same thematic role in both
types of sentences, in spite of being the
complement of broke in one sentence and the
subject of broke in the other.
Unergative
An unergative verb is a verb like groan in
a sentence such as He was groaning
which has an agent subject but seems to
have no object.
Middle
Construction in which the subject is the
patient and the verb is active and
intransitive.Construction which is usually
non-eventive. e.g. The boat sinks easily.
The car drives fast./c.f. The vase broke .
The door opened.
Burzio’s generalization
(i) A verb which lacks an external argument
fails to assign ACCUSATIVE case (Burzio,
1986)
(ii) A verb which fails to assign
ACCUSATIVE case fails to theta-mark an
external argument.(1986)
Structure-preserving principle
Syntactic structures must satisfy X’-theory
at all levels of representation.
Transformations are structure-preserving.
D-structure representations are subject to the thetacriterion: all syntactic arguments of the predicates must
be realized.
According to the structure preserving principles,
movement leaves trace since positions created at Dstructure must be preserved. Traces of movement form
a chain with their antecedent.
The EPP is a principle regulating syntactic structure
which applies at all levels of syntactic representation. It
imposes that the [NP,IP] position be generated, but it
does not imposes that the position be filled by overt
elements, and it does not require that the [NP,IP]
position be filled by arguments either.
Structural case is assigned at S-structure, case is not checked at Dstructure.
At which level the BT applies: Principle B and C apply to S-structure
configurations. The evidence that Principle A can be fulfilled at Dstructure is controversial.
The feature composition of NP-traces. NP-traces are not Rexpression, for they are co-indexed with a C-commanding NP in an
A-position. NP-traces are not Pronouns for they are bound in their
GC. NP-traces are not PROs for they are governed .
NP-traces are like anaphors, they must be bound in their GC.
Antecedent: An expression which is referred to by a pronoun , a
anaphor or a trace of some kind. John cut himself shaving. He is
someone whom we respect. Mary seems to be happy. P323.
Reference for Further Reading
 宋国明《句法理论概要》第9章。第3-5节。
 吴益民（1998）双重投射假设和及物性变
异，《现代外语》第3期，30-45页。
Chapter 7
WH-Movement
Traditional TG grammars are obsessive
wh-paranoiacs.
What is WH-movement?
The target of WH-movement: WHphrases
The moved WH-element may be both an
argument of the verb or an adjunct. In a
language like English, WH-movement
occurs in WH-question, relative clauses
and a number of other construction. Whphrases may take the head position of the
moved phrase or the specifier position of
the moved phrase.
Prepositional-stranding
A stranded (orphaned) preposition is one
which has been separated from its
complement by movement of the
complement. e.g. Who were you talking to?
By extension, in a sentence such as They
have all left, the quantifier all could be
said to have been stranded, since it is
separated from the pronoun they which it
quantifies. P-stranding is restricted and
subject to the cross-linguistic variation.
Pied-piping
A process by which a moved constituent
drags one or more other constituents
along with it when it moves. e.g. To whom
were you talking?
The landing site of WH-movement
 Short vs. Long movement
In short movement, the Wh- phrase moves to
the initial position of the CP that contains its Dstructure position, the [Spec,CP] position.e.g.
[CP whati willj [IP you tj do ti ]]
In languages like English, a WH-phrase may
move out of its clause.e.g.
[CP whati willj [IP John tj claim [CP (tj) that [IP
you did ti ]]]]
 There are also instance of long movement. In
which a wh-Phrase moves out of its clause in a
single step, for there is no landing site in the
lower clause.
e.g.
a. Which professor did you go to MIT [in order to
work with t]?
b.This is the cake that John left the party [without
even trying t]?
c.John is the only person that I didn’t know
[ whether or not to believe t]?
One property that distinguishes root
questions from embedded questions is
that in the latter the INFL does not move.
A chain is an ordered set of nodes <α1…, αi,… αn>
such that
1, all the nodes are co-indexed and
2, for every pair<αi, αi+1>,αi c-commandsαi+1.

No landing site for more than one wh-phrase.
E.g.
1, To John , which book did you give? (*To whom which
books did you give?)
2, On the table, how many books did you put?(*Where how
many book…?) (How many… when?)
3, Tomorrow, where do you plan to go? (*When where
do…) (cf. Where do you plan to go when?)

 Multiple WH-movement is allowed in
Slavic languages. Superority Condition(伍
雅清，1999)
 Russian: Kto cto kogda skazal? (who what
when said, who said what when?)
 Bulgarian: Koj kogo vizda? (who whom
sees, who sees whom?)
 Czech: kdo koho videl? (who whom saw,
who saw whom?)
Wh-movement and substitution: a nonfilled [Spec, CP] can receive phrasal
constituents of any syntactic category, the
wh-phrase fills an unoccupied position.
Doubly filled COMP filter
When an overt WH-phrase occupies the
Spec of some CP the head of that CP
must not dominate an overt
complementizer.
Adjunct
An adjunct denotes an optional constituent
typically used to specify e.g. the time,
location or manner in which an event
takes place. E.g. We had a drink in the
pub.
Adjunction
A process by which one word/phrase is
adjoined to another to form a larger
word/phrase.
E.g.
1, He shouldn’t go.
2, You know that such behavior we cannot
tolerate.
 Exclusion and the metaphor of balcony: a position
created by adjunction is like a balcony.
 Phrases can only be adjoined to maximal projections
and adjunction can only be to non-argument.
 Whether WH-movement may include adjunction is a
matter of parametric variation.Polish/English
 A comparison between NP-movement and Whmovement
 NP-movement: A-position, A-chain, A-antecedent, [headi,
ti] [+case,-case].
 WH-movement: A’-position, A’-chain, A’-antecedent,
[headi, ti] [-case, +case].
D-linked wh-items
Those wh-phrases whose range of
felicitous answers is limited by a set of
objects already referred to in the
discourse or salient in the context of
utterance. A typical example of a D-linked
wh-phrases is a which-phrase.
 D-linking of wh-phrases cancels
Superiority effects in
English.(Pesetsky,1987).
a. *What did who read?
b. Which book did which student read?
 Yes-no questions have the following
denotations.
a. Does John smoke?
b. p [p[p=smoke’*(j)p=smoke’(j)*]]
(where ‘P’ is a variable over propositions)
 It represents the unit set containing either
the proposition that John smokes or the
proposition that John does not smoke,
whichever is true.
 Wh-questions have the following denotations
a. Who likes John?
b. p x[person (x)  pp[p=like’*( x,j)]]
 The denotations is taken to be the set of true
propositions expressed by sentences of the
forms ‘x likes John’, where x is a person.
Karttunen treats interrogative noun phrases as if
they were indefinite noun phrases, making use
of an existential operator.
Extraction of complement vs.
specifier
 ?Which paperi does John wonder ti
whether Bill will read ti?
 *Who does John wonder whether will write
a paper on Tagalog?
Subextraction of complement vs.
specifier
 About which wari does John think that the
president proposed a referendumi?
 *Of which countryi does John think that
the presidenti proposed a referendum?
 Arguments for postulating WH-traces: One
motivation is analogy: NP-movement leaves
traces. The second argument comes from the
theta-theory, internal theta-role is assigned to its
trace.
 Subject-verb agreement is a local process, and
is based on a clause-mate condition. On top of
that, Principle B of the Binding Theory need whtrace proposal.
 The trace of WH-movement is case-marked,
allowing case to be realized on the antecedent.
 Vacuous Movement: Movement
transformations whose effects cannot be
observed are referred to as instances of
vacuous movement. [CP Whoi [IP ti will
come]]?
 Asymmetry between the object WHphrase extraction and the subject WHphrase extraction.
 That-trace filter: The sequence of an
overt COMP followed by a trace is
ungrammatical. This filter is languagespecific.
 Ross’s Island Constraints: Indirect
question and complex NPs are islands for
Wh-movement.
Bounding Theory
It is a sub-component of the grammar
which defines the boundaries for
movement and thus determine how far an
element can be moved.
 Subjacency Condition: Movement cannot
cross more than one bounding node at a time,
where the bounding nodes are IP and NP.
[ Barriers] A is subjacent to B if there is no more
than ONE barrier between A and B. This
condition also apply to NP movement.
 The bounding nodes are subject to parametric
variation. In English IP and NP are bounding
nodes; in Italian CP and NP are bounding node.
A movement analysis is not appropriate
for left dislocation, using subjacency
condition as a diagnostic; relative clauses,
heavy NP shift and PP-extraposition from
NP are derived by means of whmovement.
 The resumptive pronoun strategy is a way of
avoiding subjacency violation.
 Wh-trace has the feature composition [-anaphor,
-pronmominal].
 Crossover: A wh-movement cannot move a
constituent across a co-indexed pronoun.
 Leftness condition: A variable cannot be the
antecedent of a pronoun to its left.
 Weak crossover: A pronoun c-commands the
trace of a constituent that appears to the left of it
and that contain its antecedent.
Reference for further reading
 吕叔湘（1985）疑问 • 否定 • 肯定，〈中国语文〉
第4期：241-250页。
 石定栩 （1999）疑问句研究，载于徐烈炯主编
《共性与个性》，北京语言文化大学出版社。
 伍雅清 （1999）特殊疑问句研究，〈现代外语〉
第1期：75-107页。
 Shi Dingxu (1994) The Nature of Chinese Whquestions. Natural Language and Linguistic
Theory 12:301-333.
Chapter 8
 An Inventory of Empty
Category
Three null elements: PRO, WH-trace and
NP-trace
 D-structure null elements with thematic
roles could only be PRO, traces being an
S-structure Phenomenon by definition.
Traces are governed, PRO cannot be
governed. NP traces must have an
antecedent whereas PRO need not have
one. NP trace and WH-trace share the
feature [-Pronominal], PRO has the
feature [+anaphor, +pronominal].
An inventory of the main properties that distinguish NPtraces from WH-traces. (P 402)
 Null elements must be formally licensed, they should be
governed in a special way, i.e. they must be properly
governed, i.e. theta government or antecedent
government.
 Empty Category Principle (ECP)
 Traces must be properly governed. A properly governs B
iff A theta-governs B or antecedent-governed B.
 Violation of both Subjacency and ECP will result in a
marked decrease of acceptability
 Non-overt subject: the pro-drop parameter
 Pro: it is a covert nominative-case
pronoun (known informally as little pro)
which represents the understood subject
of a finite clause.
◘ NP trace [+anaphor,-pronominal], WH-trace [-anaphor, pronominal], PRO [+anaphor, +pronominal]. The forth
type [-anaphor, +pronominal]
◘ There ought to be a subject position on the basis of the
EPP.
◘ The non-overt subject is not ungoverned, it is not a PRO
for it violate the PRO theorem.
◘ It is not a trace for there is no antecedent for it.
 The non-overt subject is the forth type non-overt NP we
have been looking for: non-overt pronoun
 The subject pronoun is only overtly stressed when it is
emphasized.
 Pro subjects are not a universal property
of all human languages: Pro-drop
language: Italian, Chinese? The crosslinguistic variation is referred to as prodrop parameter, which mainly depends on
the overt morphological realization of the
verb inflection.
Licensing of Pro
It is governed by a head Xoy; Let X be the
licensing head of an occurrence of pro;
then pro has the grammatical specification
of the features on X co-indexed with it.
 Q: Is pro possible either in languages with
rich agreement or no agreement at all?
 Pro-drop parameter and language
acquisition: negative setting for the prodrop parameter; overt evidence to the
contrary will fix the positive choice for Xo.
Assignment ( to be due next Thursday)
 Exercise 1 (p387)
 Exercise 7 (p390)
 Exercise 3 (437)
Reference for further reading
 黄衍（1992）汉语的空范畴，《中国语文》第5期，383-393页。
 沈阳（1996）现代汉语空语类研究，济南：山东教育出版社。
 徐烈炯（1994）与空语类有关的一些汉语语法现象，《中国语文》
第5期，321-329页。
 Huang, C.-T. James (1984) On the Distribution and Reference of
Empty Pronoun. Linguistic

Inquiry15:531-574.
 Huang, C.-T. James (1987) Remarks on Empty category in Chinese.
Linguistic Inquiry18:321-337
 Xu,L.J. (1986) Free Empty category. Linguistic Inquiry 17:75-93.
 Syntax Handout 17 for M.A. Program in Linguistics 26/10/1999
Wu Yaqing
Chapter 8
An Inventory of Empty Category
Non-overt Antecedents of wh-movement
 I know the man that Lord Emsworth will
invite e.
 This is the man that Lord Emsworth
claims that he will invite e.
 Complementizer Contraction / Que-qui
rule
 Empty operator: An operator that
undergoes Move a to [Spec, CP] but lacks
phonetic form.
 Two incidents of empty operator: infinitival
relatives and infinitival adjuncts
 Parasitic gap: A parasitic gap construction has two gaps
related to an earlier phrase, the second gap is ‘licensed’
by a trace in the sentence.
 Which letteri did I file ti without reading ei? / 你借了多少
钱没有还？
 Which booki did you lose ti before reading ei?
Cf.
ti=your watch
 John is a person whoi even people that like ei insult ti.
me
 Whoi does the fact that John dislikes ei disturb ti?
you
 Whoi do good friends of ei gossip about ti?
you
The parasitic gaps are in positions from
which there cannot be extractions. A
parasitic gap is possible only when there
is a true gap.
Assignment (to be due next week)
 Exercises 3
 Exercise 4 (p. 437)
Chapter 9
LF and PF
 LF: level of syntactic representation that
maps into semantic interpretation.
 PF : phonetic representation of a
sentence.
 Operator: It denotes (e.g.) interrogative and negative
expressions which have the syntactic properties that
they trigger aux inversion.( WHAT have you done?
Nothing would I ever do to upset anyone.)
 Variable: the category variable is used to denote ‘any
category of head which you care to choose’. The feature
variable [a ] is used to represent ‘any value for the
relevant feature which you care to choose’
 Quantifier: A quantifier is a special type of determiner
used to denote quantity. Typical quantifiers include the
universal quantifiers all/both, the distributive quantifiers
each /every and partitive quantifiers some /any, etc.
Move a and LF
The LF representations of wh-sentences in Japanese
and Chinese indicate that wh-phrases are like operators
and variables.
PF
D-structure move a
S-structure
QR, WH-raising
move a
LF
Wh-raisng: the movement of a wh-element to adjoin to an
already moved wh-element, it is another instantiation of
wh-movement to derive LF from S-structure, it is another
instantiation of move a.
The level of application of wh-movement is another
instance of parametric variation. English in S-structure
/Chinese in LF.
 John came and John left.(conjunction reduction)
=John came and left.
 Someone came and someone left. ?= Someone came
and left. (Conjunction reduction)
 John wants to be happy. (Equi NP deletion)
=John wants John to be happy
 Everyone wants to be happy. ?= Everyone wants
everyone to be happy.
 例如有甲乙丙三个人，第一句话的意思是甲要甲幸福，乙
要乙幸福，丙要丙幸福。他们都不一定要其他两个人也幸
福，而第三句中是三个人中每个人都要三个人都幸福。
They speak two languages = Two
languages are spoken by them.
 Everyone in the room speaks two
languages.(indefinite)
 Two languages are spoken by everyone
in the room. (definite)
ECP effects are observed at LF.
 Subject-object asymmetry / complement-adjunct
asymmetry
 Gamma-marking
 Traces of adjuncts will be subject to the ECP
only at LF level and not at S-S. Argument, at SS.
 Data: that-trace effect is not observed in
sentence like the following:
 Why do you think that he left early? (p.456)
 Why is gamma-marking introduced ?
 Traces of adjuncts will be subject to the ECP at
S-structure only at the level and NOT at sstructure.
 S-structure -----s-structure. LF rechecks
argument and adjunct trace, the trace left by
movement. LF is concerned only with the
meaning, that can be deleted at LF. An element
that does not contribute to the logico-semantic
representation can be deleted at LF.
 Affectα: moveαand deleteα
 ECP must be observed at LF at the latest.
Reference for further reading
 Aoun & Li. 1989. Scope and Constituency. LI
 Chierchia, G. 1992-3. Questions with
Quantifiers. Natural Language Semantics.
1:181-234
 Hornstein, N. 1995. Logical Form: From GB to
Minimalism. Blackwell.
 伍雅清，（1999） 英汉语量词辖域的歧义研究
综述，《当代语言学》第4期。
Structural dependency
 Operations on sentences such as
movement require a knowledge of the
structural relationships of the words rather
than their linear sequences.
Eg.
 Is the man who is tall John?
 *Is the man who tall is John?
(Chomsky ,1980)
The head parameter
 a parameter of syntax concerning the position of heads
within phrases for example nouns in NPs, Verbs in VPs,
etc. A particular language consistently has the heads on
the same side of the complements in all its phrases,
whether head-first or head-last.
E wa kabe ni kakatte imasu
Picture wall on is hanging (The picture is hanging on the
wall)
The projection principle
The properties of lexical items project onto the syntax of
the sentence.
 ‘Lexical structure must be represented categorically at
every syntactic level (Chomsky, 1986,P84)
E-Language and I-Language
 ‘the shift of focus from the dubious
concept of E-Language toe the significant
notion of I-language was a crucial step in
early generative grammar’ (Chomsky,
1991a, P10)
 ‘linguistics is the study of I-Language, and
the basis for attaining this knowledge’
(Chomsky, 1987)
 I language research aims to represent this
mental state, the speaker’s knowledge of
the language, not the sentences they
have produced, it explore the mind rather
than the environment.
 ‘E-language, if it exists at all, is derivative,
remote from mechanisms and of no
particular empirical significance, perhaps
none at all’. (Chomsky, 1991a, p10)
 niwa wa soko desu
garden there is (the garden is there)
 niwa wa doko desu ka?
Garden where is (where is the garden?
 Bahasa Melayu
 Dia nak pergi ke Kuala Lumpurkah?
He is going to Kuala Lumper?( is he going
to KL)
 Though ‘GB theory’ is a common label for
this model of syntax, Chomsky himself
finds it misleading because it gives undue
prominence to the two elements of GB,
whose status ‘was not fundamentally
different from others that entered into the
discussion or others that did not’(1987) .
Hence the label of “PP theory’ has come
to be seen as closer to its essence.




Johni asked to[PROi go home ].
John asked Maryi [PROi to go home].
John persuaded Maryi to [PROi go home ].
John i promised Mary to [PROi go home ].
 Minimal Distance Principle (Rosenbaum,
1967, present formulation given in
Hornstein 1999)
 A is the controller of PRO iff A Ccommands PRO and for all B which
different from A that c-commands PRO, B
c-commands A.
 This principle incorrectly predicts that
promise is the object control verb.
 The better scenario should be
…Dpi…[PROi to VP]
 where the co-indexed DP is the lowerest
DP in the matrix clause that c-commands
the [Spec, infinitival IP]
 partial control (PC)
 Johni wanted [PRO to gather in the library
at 6]
 Control shift (CS)
Johni asked his mother [PROi to stay up
late on New Year’s Eve]
 Split antecedent control (SAC)
Johni asked his motherJ [PROi +j to stay
up late together on New Year’s Eve]
 How to establish the right division of the
labour between syntax and semantics
 Move  move something anywhere.
Move  is tightly constrained.
 The theory of movement explores the
restrictions that human languages actually
place on movement.
 Only certain elements may be moved.
Only moved to certain locations, and not
more than a certain distance.
 Chain:
An alternative way of expressing movement is
as a chain consisting of links between a moved
element  and its trace (a, t)
 The shift of focus from the dubious concept of
E-language to the significant notion of Ilanguage was a crucial step in early generative
grammar (Chomsky,1991,P10 linguistics and
adjacent field: a personal view. In A.Kasher (ed.)
The Chomsky Turn. Oxford, Blackwell.5-23.
 Plagiarism comes from plagiarius, the Latin word for
kidnapper. To plagiarize means to present another
person’s language or ideas as your own---to give the
impression you have written or thought sth yourself
when you have actually taken it from someone else.
 Global plagiarism, is stealing your speech entirely
from another source and passing it off as your own.
The most blatant and unforgivable kind of plagiarism,
it is grossly unethical.
 Patchwork plagiarism occurs when a speaker pilfers
from two or three sources.
 Incremental plagiarism, occurs when the speaker fails
to give credit for particular parts---increments—of the
speech that are borrowed from other people. The most
important of these increments are quotations and
paraphrases.
Introduction to syntax
Instructor Yatsin, Wu
e-mail: [email protected]
tele:0731-8673062
 Syntax, in its most general senses, is the study of the structure of
sentences in natural language. In this course, we will approach
syntax from the perspective of generative transformational grammar,
as pioneered through the work of Noam Chomsky, and developed
over the past 50 years. Our goals are three fold. First, to
understand the nature of language as viewed from the structural
perspective, and to understand the sort of insight about language
this perspective affords. Second, to understand the nature and
application of certain empirical methods to theoretical hypotheses
within linguistics. And third, to understand certain general principles
or laws of language which can be elucidated from the structural
perspective by the empirical methods. The material in this course
primarily will be drawn from English and Chinese, but occasional
reference will be made to other languages( although no prior
knowledge is presupposed).
 Requirements:
(i) there are five homework assignments. They are
attached. You will be told in class when they are
to be done.(usually over a weekend). Written
answers will be due at the beginning of the next
class session, and will be graded.
(ii) there will be two exams: the midterm covering
material from the first half of the course, and the
final covering material from the second half. The
exams are attached. The exams will be takehome.; rules for taking the exams will be
explained in class.
 Grades will be determined on the basis of class
participation, exercises and exams. The
midterm will count 35%of the grade, the final will
be 45%(=80%), with the remainder determined
by the exercises and class participation.
 Each student must sign up for a weekly
discussion section led by the TA. The sections
will primarily review material presented in class.
 No drop cards will be signed after the third week
of classes.
Important note
Class lectures will cover the topics described in
the topical syllabus. The readings in the text are
supplementary to material presented in class.
While class material will correspond in large part
to material covered in the text. The content of
what is presented in class may diverge at points.
Moreever, there will be material presented in
class which goes beyond that covered in the
text, and material presented in the text which is
not covered in class. Anything in the text which
you do not fully understand should be discussed
with the TA, either in section or by appointment.
Minimalist Program:
An Introduction
1, Substantive categories and functional
Categories;
2, Functional Categories: C, T, v;
3, Core Functional Categories; C (expressing
force/mood), T(tense/event structure), and v
(the light verb head of transitive construction);
4, All CFCs nay have -features (obligatory for T,
v), which constitute the core of the systems of
(structural) Case agreement and “dislocation”
(Move).
5. Three kinds of features;
(1)V features (motivate head movement),
e.g. Tense;
(2)D/N features (NP movement) e.g. Case;
(3)Wh features (Wh-movement) e.g. [+wh];
5, Feature checking can only occur in
LOCAL relations:
 Spec – Head agreement
 Head/head
Feature checking occurs between a
lexical item with inflectional () features
and a functional head.
6, Interpretable features:
[+Interpretable]: (1) agreement on DP’s,
(2) Tense, (3) Q/Wh, (4) Categorial
Feature
[-Interpretable]: (1) agreement on
verbs, (2) all case features.
 Full Interpretation.
 The interface level may consist only of
legitimate LF (PF) objects.
 Legitimate Object:
 One whose formal features have been checked
(& erased)
 If the principle of Full Interpretation is not met by
LF then the derivation is said to
 CRASH. If it is met then it is said to
CONVERGE.
7, Merge and Move:
 {XP {(EA) H YP}}
 Pure Merge:
 Merge that is not part of Move:
 Argument structure:
Merge to Move
 Inclusiveness condition: introducing no new
elements but only rearranging those of the domain.
 Cycles:
 All the three components are cyclic;
 The worse case : the three cycles are independent
 The best case is that there is a single cycle only:
 and  apply to units constructed by NS, and the three
components od decivation of <PHON,SEM> proceed
cyclically in parallel.
L contains operations that transfer each unit to  and ;
In the best case, these apply at the same stage of cycle.
Assume so. Then there is an operation TRANSFER,
applying to the narrow-syntactic derivation DNS
 TRANSFER hands DNS over to  and .
 In this connection, there is NO LF: rather the
computation maps LA to <PHION, SEM> pieceby-piece cyclically. There are, therefore, no LF
properties and on interpretation of LF, strictly
speaking, though  and  interpret units that are
part of something like LF in a non-cyclic
conception.
 The mapping to  is called Spell-Out, S-O
 Remove from NS all features that do not reach to SEM
(Uriagereka (1999))
 Phases: CP and vP, but crucially not TP.
 When a phase is transferred to , it is
converted to PHON.
  proceed in parallel with NS derivation;
  is greatly simplified if it can “forget about”
what has been transferd to it at earlier
phases, other wise the advantages of cyclic
computation are lost;
 SM=sensorimotor; C-I= Conceptualintensional
 Sorting out the effect of general principles
(physical, chemical,mathematical), interface
conditions;
 Interaction of three factors:
1, individual experience (Primary linguistic
data: PLD), which selects among the options
allowed by S0
2, S0 itself, a product of evolution;
3, General peoperties of organic system;
 The last line of each derivation is a pair
<PHON,SEM>, where PHON is accessed by
SM and SEM by C-I;
 IC (Interface conditions) must be strong enough
to allow sufficient diversity of “legible”
expressions in SEM interface.
 “each derivation is “failure proof”: there must be
a way to extend it to convergent derivation”
 “…..to eliminate comparison of derivation,
backtracing and look-ahead, and “non-local”
operation generally”
 So determines the set {F} of properties
(“features”) available for languages;
 Each L makes a one-time selection of subset
[F] of {F} and a one-time assembly of
elements of [F] as its lexicon, which we take
to be a classical “list of options”,
 For each derivation D, L makes a one-time
selection of elements from LEX that will be
accessed in D: a lexical array LA(a
numeration if elements of LEX are accessed
more than once)
L has three components:
1, narrow syntax (NS) maps LA to a derivation
DNS
2, the phonological component  maps DNS to
PHON;
3, the semantic component  maps DNS to
SEM;
  is assumed to be uniform in all L;
 NS is also assumed to be uniform in all L;
(if parameters are restricted to LEX;
  is highly variable among Ls.
 Inclusiveness condition: introducing no
new elements but only rearranging those
of the domain.
 Cycles:
 All the three components are cyclic;
 The worse case : the three cycles are independent
 The best case is that there is a single cycle only:
  and  apply to units constructed by NS, and the three
components od decivation of <PHON,SEM> proceed
cyclically in parallel.
 L contains operations that transfer each unit to  and ;
 In the best case, these apply at the same stage of cycle.
Assume so. Then there is an operation TRANSFER,
applying to the narrow-syntactic derivation DNS
TRANSFER hands DNS over to  and .
Connection, there is NO LF: rather the
computation maps LA to <PHION, SEM>
piece-by-piece cyclically. There are,
therefore, no LF properties and on
interpretation of LF, strictly speaking,
though  and  interpret units that are part
of something like LF in a non-cyclic
conception.
 The mapping to  is called Spell-Out, SO
 Remove from NS all features that do not reach to SEM
(Uriagereka (1999))
 Phases: CP and vP, but crucially not TP.
 When a phase is transferred to , it is
converted to PHON.
  proceed in parallel with NS derivation;
  is greatly simplified if it can “forget about”
what has been transferd to it at earlier
phases, other wise the advantages of cyclic
computation are lost;
 FL with two interfaces: SM (senserimotor) and C-I
(conceptual-itensional) which enter into thought and
action.
 IC (interface condition) If language is to be usable at all,
its design must satisfy IC (computational efficiency)
(Galilean intuition of perfection of nature)
 Three acquisition conditions:
 (1) unexplained elements of So
 (2) IC (the principled part of So)
 (3) general properties (external to S0)
 but: (1) is empty (SMT: strong minimalist thesis)
 L generates a set of derivations.
 The last line of derivation D is a pair of <PHON, SEM>,
PHON for SM, SEM for C-I
 D converges if PHON and SEM satisfy IC, otherwise it
crashes at one or other interface.
 IC must be strong enough to allow sufficient diversity of
“legible” expressions as SEM interface;
 Each D must be “failure-proof”: there must be a way to
extend it to a convergent derivation.
 Comparison of derivations , backtracing and look-ahead
and “non-local operations are eilinated.
 So determines {F} [ a set of features properties
available for Ls]
 Each L selects [F] from {F} (2) assembly the elements
in [F] to form LEX
 LEX = “list of exceptions”
 For each D, L makes a one-time selection of elements
of LEX to be accessed in D:
 LA (a lexical array) (a numeration if elements of LEX
are accessed more than once)
 L has 3 components:
 (1)NS (narrow syntax) maps LA to a derivation DNS;
 (2) phonological component , which maps DNS to
PHON

(3) semantic component , which maps
DNS to SEM;
 is uniform for all Ls
NS is also uniform for all Ls, if
parameters are restricted to LEX
 is highly variable among Ls.
Mapping satisfy inclusiveness condition,
introducing no new elements but only arranging
those of the domain
This strong condition is true of NS, not true of ,
nor of .
  introduces only elements that are in [F] (though
typically not in LEX)
 new elements introduced by  never enter NS and are
accordingly not in [F]
If inclusiveness hold for NS, it introduces no features, even
of [F].
 All the three components are cyclic:
(optimal)
 In the worse case: the three cycles are
independent;
 In the best case, there is a single cycle
only.
 Assuming this to true,
 and  apply to units constructed by NS
cyclically in parallel;
 Operation: TRANSFER:
 TRANSFR hands over DNS to  and .
 Spell-Out (S-O) : the mapping to ;
 PH: Phases: TRANSFER units: operation,
mapping units:
 CP and vP, crucially not TP.
 When a phase is transferred to , it is
converted to PHON.
  proceeds in parallel with NS derivation,
F is generally simplified if it can “forget
about” what has been transfer to it at
earlier phases.
 S-O (spell-out) must be able to spell-out
PH in full, or root clauses will never
spelled out. But we know S-O cannot be
required to spell-out PH in full, or
displacement would never be possible.
 (5) PH = [ [ H ]]
 -H : the edge of PH (The elements of edge may
(sometimes must) raisie)
  must be spelled out at PH, but the edge
 this allows foe head-raising, raising of head-internal
subject to SPEC-T and
“escape hatch” for successive cyclic
movement through the edge.
 Merge: (which is required in a recursive system)
1. Two units already constructed in NS
2. Yields the relation Є of membership, (contain)
3. C-command is derivational determined, =
(sister of contain) functions at SEM (e.g. for
biding theory), but perhaps not with NS
Computation relies on a head-SPEC relation R (H,
SPEC), called m-command
 Now, there is no such relation;
 There is R(SPEC, H), namely c-command;
 No R(LB, H), since H is not in the minimal
search space for LB (unless LB=SPEC)
 Operations are driven by labels, then
there can be no general SPEC-head
relation at all.
 Computation driven by label will keep to
its domain, the category that guarantees
minimal search, in accord with SMT.
 Complement: the first element merged to
a head
 Later ones are SPEC;
 In the best case, there should be no
stipulations on the number of SPECs
 H and three elements K,L,M
 SO1 {M, {L, {H, K}}} free merge
 SO2 {M, {H’ {L, {H, K}}}} only twice: new head
H’
 SO3 {M, H’’ {L, H’ {H, K}}}}} only once: two
new heads: H’ and H’’
 Assuming that no stipulated restrictions on
Merge, and no projections or other
violations of inclusiveness, keeping to the
bare phrase structure”.
 Search of  satisfies some locality condition:
least embedding; closest under c-command;
 Merge satisfies the extension condition, with
zero search;
 One possibility is  is completely unchanged;
 Another is that  is as close as possible to the
head that is label of  so than any SPEC of 
becomes a higher SPEC (“tucking in”)




PHON indicate temporal order;
SEM involves only hierarchy not order;
Left-right distinction is required;
Uniformly left?
 C-I requires that SEM expresses a variety
of semantic properties, at least including
argument structure, scopal and discourserelated properties (new/old information,
specificity, etc)
 NS derivation has to provide the basis for
assignment of order at SM interface, and
for the multiplicity of semantic properties;
 Order:
 In the worst case, it is construction specific;
 A better possibility: it is fixed once for all for L
 Kayne (1994) order reflects hierarchy
 Reduce the multiplicity to duality:
argument structure and everything else;
IC imposes order on PHON and duality of
semantic interpretation at SEM, with no
interaction between -PHON and -SEM.
 NS is based on the free operation Merge;
 SMT entails that Merge of  and  is
unconstrained, either internal or external.
 Under external Merge,  and  are separate
objects;
 Under internal Merge, one is part of the other;
 Displacement from within  is the edge of ,
yielding a new SPEC;
 Argument structure is associated with external
Merge (base structure) , everything else with
internal Merge (derived Merge)
THE ECONOMY CONDITION OF
PROCRASTINATE
Ok, but how do we account for language specific variation.
e.g. French verbs move, but English ones appear not to?
1) STRONG agreement/Tense features force OVERT
movement (before Spellout)
(one way to encode this: unchecked strong features
result in an illegitimate PF object, but unchecked weak
features are irrelevant at LF)
2) Two views of Strength: 1) Morphological =Rohrbacher
(wrongo wrongo)
3) abstract = everyone else (yay!)
4) The Economy Principle of PROCRASTINATE:
Don’t move until the end of the derivation.
5) English verb movement Must check T/A,
but features weak.
Moves in LF component therefore meets
Procrastinate.
French verb movement: Must check
T/Agr overtly (strong features) VIOLATES
procrastinate.
6) In short, Strong features can override
Procrastinate.
Another Example:
7) English Wh-movement:
a) *you want to eat what
b) What do you want to eat?
8) Chinese Wh-movement
a) ni xiang chi sheme
you want eat what
"what do you want to eat?"
b) *sheme ni xiang chi
what you want eat
"what do you want to eat?“
9) English: STRONG Wh-features (overt movement violates
procrastinate)
Chinese: WEAK wh-features (covert movement, doesn’t violate
procrastinate)
BOTH languages meet the inviolable principle of F.I.
 Another example:
Assumption: Nominative case features are
checked in Spec, AgrS
Accusative case features are checked in
Spec, AgrO
10) English embedded clauses:
John thinks [ that I have read the book]
*John thinks [that I have the book read]
11) German embedded clauses:
… [dass Ich das buch gelesen hat]
*….[dass Ich gelesen hat das buch]
12) English WEAK accusative case features,
therefore NP doesn’t move overtly to spec
AgrOP, meets procrastinate and FI.
 German STRONG accusative case
features, therefore NP does move. Meets
FI, but violates procrastinate.
 THE ECONOMY CONDITION OF
SHORTEST MOVE
1. Replaces relativized minimality. Make the shortest move
possible. (evaluated in terms of skipped projections)
2).Wh-islands
a) [CP1 [IP you wonder [CP2 [IP who loves what ]
b) both CPs have STRONG wh-features to check, both
Wh-words have a wh-feature to check.
c) Scenario 1: Who moves to spec, CP2 (meets SM)
What moves to spec, CP1 (FAILS SM)
d) Scenario 2 What moves to spec, CP2, then CP1
(meets SM)
Who moves to spec, CP1 (meets SM, fails FI) etc.
3) Superiority (can’t be accounted for using ECP)
 [CP [ you persuaded John [ PRO to read aspects]
 [CP what did [you persuade John [ PRO to read ti]
 [CP who did [you persuade t [ PRO to read Aspects ]
 *[CP What did you persuade who to read t ]
(this is not a WH island --- who is not in the spec of the
lower CP, it is the object of persuade (it is in an A
position, not an A’ position, so it isn’t a potential
intervening governor)
 4) Violation of SMC results in a HMC violation in heads
 5) Violation of SMC with NPs could explain the
ungrammaticality of sentences like
 *Him hit he. (but this might also be a FI violation
because the wrong case is being checked).
 Showing that the SMC is VIOLABLE is tricky. In fact, it’s
near impossible.
 That’s why in later versions of minimalism (eg chapter 4)
Shortest move is replaced by a non-violable condition
on movement (called the minimal link condition (MLC).
(GREED)
The economy condition of Greed says “move only
for your own reasons”. This is just wrong. so
we’ll ignore it
 *Johni seems that ti left
 *Who does John leave because he likes_?
 Why did John leave?
 Who did John think that Bill claimed that Mary
suspected that everybody liked_?
 The search for rigorous formulation in linguistics
has a much more serious motivation than mere
concern for logical niceties or the desire to
purify well-established methods of linguistic
analysis. By pushing a precise but inadequate
formulation to an unacceptable conclusion, we
can often expose the exact source of this
inadequacy, and consequently, gain a deeper
understanding of the linguistic data. (preface to
Chomksy, 1957)
 The DP hypothesis resolves what was a
theoretical inconsistency between the
treatment of noun phrases and clause.
That is, according to this approach, nouns,
like verbs, project to a functional category.
Morphological evidence for DP
Yup’ik, a Central Alaskan Eskimo
language, in Yup’ik, both verb and its
subject are marked with matching ergative
case, expressed via an identical
agreement suffix (-t in this instance),as
illustrated in (1). Similarly, a noun and its
possessor are marked for agreement and
the morpheme involved (i.e. –t) matches
that found in the clause. (data from Abney
1987:(24) 39)
(1)a. angute-t kiputa-a-t.
man-Erg(PL)buy-OM-SM
‘the men bought it’
b.angute-t kuiga -t.
man-Erg (PL ) river-SM
‘the men’s river’
 Hungarian, a nominative/accusative language,
also exhibits identical agreement affixes on
nouns and verbs. In this language, case is
expressed on the possessor and the head noun
agrees with the possessor in person and
number.
 Turkish also displayed DP-internal agreement,
the possessor displayed genitive case and the
head noun agrees in number and gender with
possessor.
Syntactic evidence for DP
 Rome destroyed Carthage.
 Rome’s destruction of Carthage.
 Nouns, like verbs, may take both internal
and external arguments.
 There is compelling evidence from binding
and control phenomena that the
arguments in the nominal domain are
hierarchical arranged in a manner parallel
to that in the clausal domain.
 Floating quantifier is another piece of
evidence.
 All the girls received the grades.
 The girls all received the grades.
 Semantic properties of noun phrases indicate
that the argument/non-argument distinction is
relevant ,that non-arguments (I.e., nominal
predicates) correspond to NP and argument to
DP or alternatively, NPs are non-referential and
DPs are referential (Stowell, 1989), noun phrase
arguments to be saturated and bear thematic
roles and noun phrase predicates to be
unsaturated and lack theta-roles. General
speaking, an article may serve to “convert” a
predicate NP into an argument DP.
 Gianni é medico (*che…)
(Italian)
 John is (a) doctor (that…)
 Gianni é un medico (*che…)
 John is a doctor (that…)
 Predicate NP seem to be adjectival in
nature. (Mandelbaum,1994)
 According to Szabolcsi’s proposal, an NP
cannot, on its own, serve as an argument
because it is not introduced by a
subordinator, which may take the form of,
among other things, the definite article.
 Longobardi(1994) observes that certain articleless
nominal expressions may nevertheless function as
arguments. In many European languages, for example,
plural and mass nouns may function as arguments,
subject to parametric variation in distribution and
interpretation. it is natural to assume, as Longobardi
does, that these nominal expressions are (DP)
arguments introduced by a determiner devoid of lexical
content. Another argument forming strategy is raising
the N-head to D. This strategy may form DP-arguments
from articles nominal expressions involving proper
names.
 Propositions and clausal arguments correspond
to IP and DP. Analogously under a DP analysis,
nominal predicates correspond to NP and
nominal arguments to DP. Another advantage of
the DP-analysis is that it provides a functional
head that encodes semantic features of
determiner elements. Some of the features
claimed to be encoded in D are definiteness,
specificity, referentiality, and deixis.
 Longobardi(1994) has provided
independent evidence for N-to D
movement in Romance languages. He
argues that proper names raise to the Dposition, the locus of referentiality.
 The following example he cited support
the idea that proper names not introduced
with a definite article must raised to D in
the overt syntax in a language like Italian.
a.[IIDP mio Gianni] ha finalmente telefonato.
Italian
b.*[DP mio Gianni] ha finalmente telefonato.
 [DP Gianni mio] ha finalmente telefonato.
 (the) my John has finally called.
 ‘my John has finally called’.
Mannen
Man-the
The man
Det store huset
The big house-the
The big house
Swedish
Double definiteness
Relative clause as a support for the parallel treatment
between clause and DP
 [DP D CP]
 [DPThe [CP[NP picture of John ]i[C’ that Bill saw[ei]]]]]
 John bought the picture of himself that Bill saw. (John,
Bill as antecedents)
 Idiom chunk
 To take advantage of
 To make headway
 The advantage that he took…
 The headway that we made…
 Number in DPs, like tense in the clause,
plays an integral role in the interpretation
and legitimacy of a noun phrase.
 In English gender is restricted to singular
pronominal forms, in many languages
gender may be expressed on nouns,
determiner elements, adjectives, and/or
other modifiers.
el niño pequeño.
Spanish
The-MascSg child-MascSg small-MascSg
The small child
la niña pequeña.
The-FemSg child-FemSg small-FemSg
The small child.
Types of functional phrase.
 IP consists of two features: it has the strange property of being
doubly headed. (C&L 1993, P.530)
 Separation between AGRP &TP, both with the normal X-bar structure.
whether TP dominates AGRP or vice versa. A parameter of variation
between a VSO language such as Berber, where AGR is inside
Tense, and French, where AGR is outside Tense.
 in some language the verb and the object agree as well as the verb
and subject.
Agreement with subject and agreement with object.
 negative element in the sentence also needs to have a separate
NegP. Modal phrases, aspect phrases, and passive phrases.
DP ([DP the[NP picture of Bill] ]) [DP we [NP men]]
[DP[SpecJohn[D’s [NP picture of Bill]]]
The contrast between lexical and functional categories










Functional phrases
--closed class of heads
--dependent phonologically
phonologically
usually unstressed
have a single complement
argument
not an argument
inseparable argument
argument
no descriptive content, not
content, linked to…
linked to ‘real’ world
have grammatical features
grammatical features
linked to parameter
parameters.
lexical phrases
---open class of head
---independent
potential stressed
have one or more
separable
descriptive
do not have
not linked to
  criterion violation: John left the room
angry.