E002-IntroGB.pdf

Loading ...
Global Do...
News & Politics
Black, Cheryl A. (send email c/o Andy_Black@sil.org)
A step-by-step introduction to the Government and Binding theory of syntax
http://www.sil.org/americas/mexico/ling/E002-IntroGB.pdf [February 1999]
 1999 Summer Institute of Linguistics.  All rights reserved.  May be freely reproduced for nonprofit or educational use,
    provided that it is not modified in any way.
A step-by-step introduction to
the Government and Binding theory of syntax
Cheryl A. Black
SIL - Mexico Branch and University of North Dakota
Designed for use as a textbook, this paper provides a data-motivated,
step-by-step introduction to the main tenets of Government and Binding theory
of syntax as developed by Noam Chomsky.  An overview of the theory is
followed by sections on subcategorization, X-bar theory, movement, semantic
roles, case theory, and binding theory.  A number of the more recent additions
proposed by other linguists, such as the DP hypothesis and the split structure
for Infl, are covered in the final section.  The data used is primarily English,
but also includes Abaza, Chichewa, Eskimo, French, Hixkaryana, Mohawk,
Tiwa, Tzotzil, Urubú, and Zapotec.
1. Introduction
This text is designed to provide a data-motivated, stepwise introduction to the main tenets
of the Government and Binding (GB) theory of Syntax, which was developed mainly by
Chomsky (1981, 1982, 1986).  The content of the articles is excerpted from the Syntax II
course I taught at the Summer Institute of Linguistics at the University of North Dakota (SIL-
UND), which was in turn based upon graduate courses I took from Dr. Sandra Chung at the
University of California at Santa Cruz.  After presenting a condensed version of this material to
the staff at UND, I was encouraged to write up the material in a series of articles for Notes on
Linguistics to make the material available to a larger audience.  These articles (Black, May 96 -
Nov. 97) are combined and slightly revised here, primarily for use as a textbook.  I hope that
new students and other readers will find this method of presentation as interesting and helpful
as the past students and staff at SIL-UND have.  The basic understanding of the theory gained
should enable the reader to have much improved comprehension when reading theoretical
articles and to see how formal linguistic analysis could be helpful to their own program.
The main topics covered are:
1. Overview of Government and Binding Theory (section 2)
2. Subcategorization and X-bar theory for lexical phrases (sections 3 and 4)
3. Extending X-bar theory to sentences and clauses (section 5)
A step-by-step introduction to GB
November 1998
2
4. X-bar theory applied to languages with other word orders (section 6)
5. Constraints on movement (section 7)
6. Semantic roles (section 8)
7. Case theory (section 9)
8. Binding theory (section 10)
9. Recent updates:  VSO/OSV word orders (section 11.1), additional
functional projections (section 11.2), and the morphology/syntax
interface (section 11.3)
The majority of the data used throughout will be in English to allow the reader to learn the
theory without the hindrance of working with an unfamiliar language at the same time.  By
section 6, however, the reader should be able to begin applying the theory to the language he
or she is working with, which is the goal.
2. Overview of Government and Binding theory
GB assumes that a large portion of the grammar of any particular language is common to
all languages, and is therefore part of Universal Grammar.  The GB view is that Universal
Grammar can be broken down into two main components:  levels of representation and a
system of constraints.
GB assumes a derivational model consisting of four levels of representation, as
diagrammed in (1).  The lexicon lists the idiosyncratic properties of lexical items which
constitute the atomic units of the syntax.  These properties include what arguments the item
subcategorizes for, etc.  Lexical items are combined together at D-structure (underlying
structure).  D-structure is mapped into S-structure, which is the syntactic representation that
most closely reflects the surface order of the sentence.  S-structure is not directly interpreted
itself, but is factored into Phonological Form (PF) and Logical Form (LF).  PF is the interface
with the Phonology where shapes, sounds, and groupings of items are directly represented.  LF
is the interface with the Semantics.  Predication relationships and the scope of quantifiers and
operators of various kinds are explicitly represented in the phrase structure at LF.
(1)
Lexicon
D-structure
Move-^
S-structure
stylistic and
LF Move-^
phonological rules
PF
LF
These levels are related to one another by rules (noted in italics in (1)).  A single movement
rule, Move-^, maps between D-structure and S-structure and a similar rule maps S-structure
into LF.  Move-^ is stated as a simple rule basically allowing anything to move anywhere, since
the system of constraints (which will be introduced throughout this text) is responsible for
A step-by-step introduction to GB
November 1998 
3
correctly restricting this movement.  Stylistic and other phonological rules are assumed to take
place at PF.  This text will be limited to the D-structure and S-structure levels of
representation.
3. Constituent structure and subcategorization
A word, such as a noun, verb, adjective or preposition is a lexical category.  In structural
terms, they are called heads.  Phrases are meaningful groupings of words built up from the
lexical category of the same type that they contain.  Examples of phrases are:  NP, VP, AP
(=AdjectiveP), and PP.  But the particular head is choosy about what can combine with it to
form a phrase.
VP examples:1
(2) 
died / *died the corpse / *died to Sue about politics
(3) 
relied on Max / *relied  / *relied from Max / *relied to Max
(4) 
dismembered the corpse / *dismembered
(5) 
talked (to Sue) (about politics) / *talked that the economy is poor
(6) 
read (the book) (to John) / read that the economy is poor
(7) 
supplied the Iraqis (with arms) / *supplied
(8) 
told Sylvia (that it is raining)
(9) 
revealed (to John) that problems exist / revealed the answer (to John)
A complement is a phrase that a lexical category takes or selects.  Which complements are
taken by a particular verb is an arbitrary property of that verb:  in (2) died cannot take any
complements; in (3) relied must have a PP complement with on as the preposition; in (4)
dismembered must take an NP complement; in (5) talked can take an optional PP complement
with to as the preposition and/or an optional PP complement where the preposition is about,
etc.
We can represent these complement selection requirements in subcategorization frames, as
shown in (10), where the square brackets delimit the phrase and the environment bar indicates
the position of the lexical head.  Required complements are simply listed, whereas optional
complements are enclosed in parentheses.  Finally, in cases where a complement with a
particular head is subcategorized for, the head is listed as a feature on the complement (as for
rely and talk).
(10) 
die, 
V, 
[ _ ]
rely, 
V, 
[ _ PP[on] ]
dismember, V, 
[ _ NP ]
talk, 
V, 
[ _ (PP[to])  (PP[about]) ]
                                               
1A * before a word, phrase or sentence indicates ungrammaticality.  Our grammar must not only generate all
grammatical sentences in the language, but also correctly rule out all ungrammatical sentences.  The latter is the more
difficult task (the simple rule S=Word* generates all grammatical sentences in any language), so knowledge of what is
ungrammatical is crucial.
A step-by-step introduction to GB
November 1998
4
Adjectives, nouns, and prepositions also subcategorize for their complements.
AP examples:
(11)
red / *red that Sylvia would win
(12)
afraid (of snakes) / *afraid to this issue
(13)
orthogonal (to this issue)
(14)
ambivalent ((to Joe) about her feelings)
(15)
certain (that Sylvia would win)
(16)
insistent (to Joe) (that we leave)
NP examples:
(17)
group (of scientists)
(18)
individual
(19)
book (about photography / *to Fred)
(20)
generosity (to Fred)
(21)
dislike of Fred
(22)
ambivalence ((to John) about my feelings)
(23)
rumor (that all is well)
(24)
message (to the Contras) (about the guns)
PP examples:
(25)
about [the talk]
(26)
before [we leave]
(27)
from [over the hill]
(28)
[looked] up
We can generalize that the lexical categories (V, N, A, P):
a. Subcategorize for their complements.
b. Precede their complements in the phrase.2
c. Co-occur with other constituents.
Heads and complements are not the only parts of phrases.  For example, NPs can be
preceded by words (or sometimes phrases) like:  the, no, some, every, John’s, my mother’s.
APs can be preceded by degree words such as:  very, extremely, rather, quite.  These items
differ from complements because they precede the lexical category and they are not
subcategorized for.  They are called specifiers.
                                               
2This ordering of lexical categories before their complements is obviously not a universal.  Section 6 will deal with
how to account for languages with different word orders.
A step-by-step introduction to GB
November 1998
5
4. X-bar Theory
GB seeks to capture the similarities between different categories of lexical phrases by
assigning the same structure to them (as shown in (30)).  Rather than having different phrase
structure rules for VPs, NPs, etc., just the two basic rules in (29) cover all the lexical
categories.
(29)  Phrase Structure Rules:
(for any lexical category X, X0=Head)
XP →  Specifier   X'
X'  →  X0    Complements (=YP*)
In the trees generated by these rules, the top node (corresponding to left side of the rule) is
known as the mother, with the two daughters introduced by the right side of the phrase
structure rule.  The daughter nodes at the same level are known as sisters.  In (30) one of the
daughters, X', is also a mother with daughters of her own, just as in normal family
relationships.
(30) Basic X-bar Structure
 XP--maximal projection
specifier
X'--intermediate projection
X0--head
complement(s)
Claims involved in this schemata:
1. All phrases are projected from lexical categories in the same way (i.e. the PSRs in (29).
a. For conjunction:  Xn  → Xn  Conj  Xn.
b. For adjunction:    Xn  →  Ym  Xn.3
2. A head (=X0) subcategorizes for all and only its sisters.
a. The subcategorized complements are always phrases.
b. Heads and their maximal projections share features, allowing heads to
subcategorize for the heads of their sisters (i.e. rely).
3. In general, specifiers are optional.  Evidently, specifiers may be words or phrases.
The following trees illustrate how X-bar theory works.  We apply the X-bar rules to
specific categories.  First find the head, which determines the type of phrase, then look for
specifiers, complements, adjuncts, and conjunctions.  In (31), interpretation is the head, the
                                               
3 n may be any bar level (0,1,2=X0, X’ or XP), m may only be 0,2 since only heads or maximal projections may move
or adjoin.  Also, the right side of the adjunction rule is unordered; adjectives adjoin on the left, but other NP adjuncts such
as relative clauses adjoin on the right and VP adjuncts such as adverbs may adjoin on either side.
A step-by-step introduction to GB
November 1998
6
musician's is the specifier, and of that sonata is the complement in the NP.  The specifier and
complement are each phrases themselves which are also diagrammed via the X-bar phrase
structure rules.4
(31)
     
     NP
NP[+poss]
    N'
D
N'[+poss]
    N0
       PP
 |
 |
     |
        |
the
       N0[+poss]      interpretation       P'
 |
  musician’s
      P0
NP
      |
    of              D
      N'
              |
       |
           that
      N0
          |
              sonata
In (32), afraid is the head, extremely is the specifier, and of snakes is the complement in
the AP.
(32)
AP
Deg
    A'
   |
    extremely
A0
PP
 |
  |
    afraid
 P'
    P0
 NP
      |
   |
    of
  N'
   |
  N0
   |
     snakes
                                               
4 In the trees throughout, a category X by itself (i.e. not X' or XP) is equivalent to X0.
A step-by-step introduction to GB
November 1998
7
(33)  shows a VP headed by talked with two PP complements.5
(33)
 VP
   |
  V'
   V0
    PP
  PP
    |
|
    |
talked
    P'
    P'
              P0
   NP       P0
NP
              |
 
    |
        |
             to
  Sue    about
 N'
  |
N0
  |
   politics
Tree (34) illustrates how conjunction and adjunction fit into the X-bar schemata.  The
conjunction rule is shown for black and white;6 huge, black and white; and extremely angry are
all adjuncts which are adjoined to N', showing how the adjunction rule is recursive; the is in the
specifier position and dog is the head of the whole NP.
(34)
NP
  D
   N'
   
the
AP
N'
  
 A'
    AP
 N'
  
 A0
 AP    Conj    AP
AP
   N'
   
  
       
      
    
      huge
 A'
    and
     A'
  Deg 
A'
   N0
  
      
     
 
    
 A0
     A0 extremely   
A0   dog
  
      
 
black
  white
     angry
                                               
    5 Proper names and pronouns are shown as NPs since in English they do not have specifiers or complements.  In general,
a triangle under a phrasal node means that further structure is not shown because it is irrelevant to the point being made.
6 In this case, the conjunction could have been shown at either the A' or A0 level instead.
A step-by-step introduction to GB
November 1998
8
At this point, even though we can draw trees for some complex phrases, we still cannot do
even a simple complete sentence such as John hit the ball.  The rule S  →  NP VP does not fit
the X-bar schemata.  We also cannot draw a tree diagram for a clausal complement to a verb,
such as the that-clause in Bill read that the economy is poor.  In order to make sentences and
clauses fit X-bar theory, we need to determine the head, specifier, and complement for each.
This will be the next topic addressed.
5. Extending X-bar theory to sentences and clauses
5.1 Clausal complements within X-bar theory
In the last section we saw that verbs choose or subcategorize for which complements can
follow them.  Consider the verb read.  One can read (the book) (to John) or read that the
economy is poor.  In the first instance, we can say that read subcategorizes for an optional NP
complement and an optional PP complement headed by the preposition  to.  The lexical entry
would be:7
read, V,
[ _ (NP) (PP[to]
 ) ]
But what can we do about the complement that the economy is poor?  What kind of phrase is
it and how does it fit into X-bar Theory?
Transformational grammar assumes that clauses are built up from sentences using the rule:
S' → COMP S.  (35) shows the traditional tree for the VP, ignoring for the moment the internal
structure of S.
(35)
VP
   |
  V'
V0
 S'
 |
               read
   COMP
S
      |
    that
the economy is poor
Under Transformational grammar, the head of the clause is the sentence and the
complementizer is a specifier.  The sentence cannot be the head of any phrase in X-bar theory,
since it is not a lexical item or word; it is most likely a complement.  Further, the X-bar
schemata allows more positions within the phrase than the S' rule does, so we need evidence to
determine whether the complementizer is a specifier or a head.
                                               
7 Refer back to section 3 for explanation of the subcategorization frames within the lexical entry.
A step-by-step introduction to GB
November 1998
9
Examine the following data:8
(36)
a.
Everyone insisted that the store would close on Thursdays.
b.
*Everyone insisted for the store to close on Thursdays.
c.
*Everyone insisted whether the store would close on Thursdays.
(37)
a.
They arranged for their children to be happy.
b.
*They arranged that their children would be happy.
c.
*They arranged whether their children would be happy.
(38)
a.
Sue wondered whether the smoke would clear before daylight.
b.
*Sue wondered that the smoke would clear before daylight.
c.
*Sue wondered for the smoke to clear before daylight.
What accounts for the distribution above?  Each of these verbs not only takes a clausal
complement but it chooses which complementizer the clause must have.  This is reminiscent of
a verb like rely which subcategorizes for a PP complement which must have on as the
preposition.  Therefore, the main verb’s subcategorization is what allows the grammatical
examples and rules out the ungrammatical ones.  By claims 2 and 2(b) in (30) we know that:
(a) only heads can subcategorize, (b) any sisters of the head must be subcategorized for (ruling
out the starred examples), and (c) if something is subcategorized for, it must be either a
complement or the head of a complement.
Since the specific complementizer is subcategorized for by the verb, the complementizer
must be the head of the complement clause.  (It cannot be the specifier because specifiers are
never subcategorized for.)  Further, if the complementizer is the head of the clausal
complement, then according to X-bar theory the clausal complement is a complementizer
phrase or CP.
The revised structure for (35) is shown in (39).
(39)
VP
  |
 V'
  V0
CP
   |
  |
read
C'
  C0
   S
   |
 that
  the economy is poor
                                               
8 Throughout, a * before an example indicates ungrammaticality.  As previously noted, it is crucial that the theory rule
out ungrammatical examples as well as correctly generate the grammatical data.
In (38), note that the sense of wonder used here is that of questioning.  The other sense, expressing amazement, would
subcategorize for a that-clause like insist, as in:
(I)
Sue wondered that the smoke had cleared before daylight.
A step-by-step introduction to GB
November 1998
10
C0 is a non-lexical or functional head.  We need to expand the range of categories to which
the basic PSRs apply.  This can be accomplished by deleting the phrase ‘from lexical
categories’ in claim 1 in (30) so that it reads simply, ‘All phrases are projected in the same
way’ and by removing the word ‘lexical’ within the parenthetical note in (29).
We can now write the following lexical entries for the main verbs in (36)-(38):9
insist,
V,
[ _ CP[that] ]
arrange,
V,
[ _ CP[for] ]
wonder,
V,
[ _ CP[whether] ]
At this point, complement clauses almost fit within X-bar theory.  One  problem remains:
if C0 is a head, its complement must be a phrase by claim 2(a) in (30).  Yet the complement of
C0 seems to be a sentence (S).  Is S a phrase in X-bar terms?
5.2 Reanalyzing sentences within X-bar theory
The traditional phrase structure rule for a sentence is:  S → NP VP.  In order to reanalyze
this rule in X-bar terms, its head, complement, and specifier must be determined.  Neither of
the constituents on the right side of this rule can be the head of a phrase because they are
phrases themselves, not lexical items or words.  To find out what the head of a sentence is, we
need to look again for evidence of subcategorization.
Consider this data:
(40)
a.
Everyone insisted that the store would close on Thursdays.
b.
Everyone insisted that the store was closed last Thursday.
c.
*Everyone insisted that the store to close on Thursdays.
(41)
a.
They arranged for their children to be happy.
b.
*They arranged for their children were happy.
c.
*They arranged for their children would be happy.
(42)
a.
Sue wondered whether the smoke would clear before daylight.
b.
Sue wondered whether the smoke cleared before daylight.
c.
*Sue wondered whether the smoke to clear before daylight.
When the complementizer is either that or whether, the sentence that follows is a regular finite
sentence, and to cannot be present as shown in (40c) and (42c).  In contrast, when the
complementizer is for (41), to must be present and a finite or tensed verb is not allowed in the
following sentence.  We can say that the complementizers that and whether subcategorize for a
finite complement, whereas for requires a nonfinite complement.  But the head of that
complement still needs to be determined.
We saw that to must be present when the complementizer is for.  We can conclude that to
is the marker for nonfinite clauses in English.  Thus,   for subcategorizes for a nonfinite
                                               
9 These subcategorization frames are not meant to say that these particular types of CPs are the only types of complements
that these verbs can select.  These are simply the ones relevant to the current discussion.
A step-by-step introduction to GB
November 1998
11
complement that must have to, so to must be the head.  Further evidence that to is a head can
be seen in (43)-(44).  Since to subcategorizes for the bare form of the verb following it, to
must be a head.
(43)
a.
They arranged for their children to be happy.
b.
*They arranged for their children to were happy.
c.
*They arranged for their children to are happy.
(44)
a.
We would hate for him to leave.
b.
*We would hate for him to left.
c.
*We would hate for him to leaves.
We still need a category for to and for its counterpart in finite sentences.  GB posits that
the tense and agreement features fill the same head position in finite sentences that to fills in
nonfinite sentences.  The category is therefore called Inflection, or Infl or I for short.  This
means a sentence is an Inflection Phrase (IP).
The lexical entries for the three complementizers and nonfinite to can be given as:10
that,
C,
[ _ IP[+fin] ]
for,
C,
[ _ IP[ -fin] ]
whether, C,
[ _ IP[+fin] ]
to,
I[-fin],
[ _ VP[+bare] ]
I[+fin] is never filled by a lexical word at D-structure in English, so it does not have a lexical
entry.  It always takes a VP as its complement just as nonfinite to does.  The subject NP is
assumed to fill the specifier position in the IP.  We can now draw trees for any of the sentences
discussed so far.
The tree for (44a) is exemplified in (45), where I assume that pronouns are NPs and
auxiliaries are verbs which take a VP whose head has a certain form as their complement.
                                               
10  The feature [+bare] means the infinitive or uninflected form.
A step-by-step introduction to GB
November 1998
12
(45)
IP
NP
I′
we
    I0  +past
VP
   1pers
  |
    -sg
 V′
V0[+aux]
   VP[+bare]
 |
|
  would
     V′
V0[+bare]
CP[for]
 |
 |
hate
C′
    C0[for]
IP[-fin]
     |
   for
NP
I′
him
  I0[-fin]
VP[+bare]
  |
  |
 to
 V′
  |
 V0[+bare]
  |
leave
English sentences and clauses, as well as lexical phrases, now conform to X-bar theory in
that they can be generated from the two basic phrase structure rules in (29), possibly coupled
with the conjunction and adjunction rules in claims 1(a-b) of (30).  Our next step is to see how
this theory of phrase structure can account for languages with other word orders.
6. X-bar theory applied to languages with other word
orders
We have seen that X-bar Theory can account for the phrase structure of lexical phrases,
sentences, and clauses in English, based upon the subcategorization of a head for its
complement(s), using only the two basic rules repeated in (46) plus the rules for conjunction
and adjunction.
(46)
XP → Specifier  X'
  X' → X0  Complements
A step-by-step introduction to GB
November 1998
13
We can refer to English and other SVO languages as head-initial and specifier-initial, since
the specifier comes before X' and the head comes before its complements.  This generalization
holds in all phrases in English.  For example, in sentences the subject is initial in the specifier
position and the VP complement follows the head containing nonfinite to or the inflection
features. Within the VP, the head V precedes the NP object complement.  The tree for a simple
sentence is given in (47).
(47)  
IP
        NP
  I'
       John I0 +past
   VP
   3pers
      |
   +sg             V'
 V0
  NP
 |
 hit
    the ball
In nominal phrases, either the possessor or the determiner is in the specifier position and
they again come before the head, whereas complements follow the head noun.
(48)
NP
   NP[+poss] 
 N'
Mary’s         N0
       PP
         |
     letter
to Bill
Also, in prepositional phrases, words like right or just may precede the head in the specifier
position and the preposition precedes its NP complement.
(49)
 PP
     Adv
     P'
       |
    right
  P0
  NP
   |
               before     the trial
But not all languages have these orders.  Can the X-bar rules work for them, too?  We will
look at each major word order in turn.
A step-by-step introduction to GB
November 1998
14
6.1 SOV Languages
Consider the following data from Abaza, a Northwest Caucasian language (O’Herin 1993).
(50) gives an example of a sentence with a transitive verb, (51) shows a possessed noun
phrase, and example (52) is a PP.11
(50)
H-pa
xsjº
yº¹ºn.
our-son milk
drank
‘Our son drank the milk.’
(51)
Ahmet
y-tdzº
Ahmet
3SM-house
‘Ahmet’s house’
(52)
awiy
a-m¸taxi
that
3SI-after
‘after that’
What is consistent about the phrases in this SOV language?  In the sentence, noun phrase,
and ‘prepositional’ phrase, the head is always final but the specifier is initial.  We can capture
these generalizations of head-final and specifier-initial by simply changing the order of the
head and complements in the X'  rule.12  Therefore, for regular SOV languages like Abaza,13
the basic phrase structure rules are:
(53)
XP → Specifier  X'
X' → Complements  X0
The trees generated by the rules in (53) for each of the Abaza examples will be given to
show how the SOV version of the trees look.  Sentence (50) is shown in (54), where the
subject is in the initial specifier position (as in SVO languages) but the object is also before the
verb in the complement position.  (Note that we always read the word order from left to right
beginning at the top left and going down and back up the tree.)
                                               
11Abbreviations:  3SM = third person singular masculine pronoun; 3SI = third person singular inanimate pronoun.
12This parameterization allowing us to change the order of the elements on the right side of the phrase structure rule
parallels the distinction between Immediate Dominance and Linear Precedence in Generalized Phrase Structure Grammar
(Gazdar, Klein, Pullum and Sag 1985).
13I use ‘regular’ here to mean a language in which all phrases in the language follow the same basic phrase structure rules.
In some languages, nominals have a slightly different order than the rest of the phrases do.  We will see an example of this
in the OVS language, Hixkaryana.  Other splits in ordering, or languages where some or all phrases lack a strict ordering,
are also possible.
A step-by-step introduction to GB
November 1998
15
(54)
    IP
  NP
   I'
h-pa
 VP
I0    +past
                       our son
   |
     3pers
  V'
         +sg
NP
V0
  |
xsjº
     yº¹ºn
milk
    drink
The possessor fills the specifier position in the noun phrase, which is initial:
(55)
NP
NP[+poss]
N'
 |
        Ahmet
N0
         Ahmet’s
 |
   y-tdzº
   house
The head of the PP comes after the complement, so it is a postposition:
(56)
PP
 |
 P'
NP
P0
 |
          awiy
a-m¸taxi
          that
  after
A step-by-step introduction to GB
November 1998
16
6.2 VOS Languages
Tzotzil, a Mayan language spoken in Mexico, provides data for a VOS language (Aissen
1987).  (57) shows a sentence with a transitive verb, (58) shows a possessed noun phrase, and
(59) gives a prepositional phrase.14
(57)
7i-s-pet 
lok'el 7antz
ti
t'ul-e.
CP-A3-carry away woman the rabbit-CL
‘The rabbit carried away the woman.’
(58)
s-tot
li Xun-e
A3-father the Xun-CL
‘Xun’s father’
(59)
ta
bala
with bullets
‘with bullets’
All the phrases here are head-initial (like English) but specifier-final.  The X-bar phrase
structure rules for VOS languages are:
(60)
XP → X'  Specifier
X' → X0  Complements
The tree for sentence (57) is given in (61), where the verb is first as the head of the VP,
followed by the object in the complement position, with the subject last in the (final) specifier
of IP position.  VOS languages are just like SVO languages except for the ordering of the
specifier.
(61)
IP
I'
   NP
 I0  +past
VP
ti t’ul-e
     3pers
  |
 
the rabbit
     
   +sg
  V'
     V0
 NP
      |
            7i-s-pet lok'el
    7antz
             carry away  
    woman
                                               
14The segment 7 is a glottal stop, whereas C' is a glottalized consonant.  Abbreviations:  CP=completive aspect; CL=clitic;
A3 = third person absolutive.
A step-by-step introduction to GB
November 1998
17
In the possessed noun phrase, the possessor fills the specifier position, so it is final, as
shown in (62):15
(62)
NP
N' 
     NP[+poss]
 |
N0
    li Xun-e
 |
       Xun's
         s-tot
        father
Finally, the PP is head-initial as expected:
(63)
PP
 |
P'
      P0
     NP
      |
    ta
  bala
   with
  bullet
6.3 OVS Languages
(64)-(66) give examples of a simple transitive sentence, a possessed noun phrase, and a PP
from one of the rare OVS languages, Hixkaryana, a Southern Guiana Carib language spoken in
Northern Brazil (Derbyshire 1985).
(64)
Kuraha yonyhoryeno b®iryekomo.
bow
he-made-it
boy
‘The boy made a bow.’
(65)
Masar® hokru
Masar® child-of
‘Masar®’s child’
(66)
mokro
yakoro
that-one with
‘with that fellow’
                                               
15Note that within the possessor NP a determiner is present.  In English, we have said that the determiner is also a specifier
of the NP which alternates with the possessor.  In Tzotzil and many other languages, possessors and determiners may co-
occur and have different positions with respect to the noun.  Since the determiner comes before the noun in Tzotzil, a head-
initial and specifier-final language, it seems more likely that the determiner is a head.  We will see how this works in
section 11.2.1 where the DP Hypothesis will be presented as one of the more recent additions to the theory.
A step-by-step introduction to GB
November 1998
18
What is consistent about the three phrases here?  The heads V, N, and P are all final.
Looking at the position of the subject in (64) we see that the specifier is also final.  The X-bar
rules that generate head-final and specifier-final trees are:
(67)
XP → X'  Specifier
  X' →
 Complements  X0
The tree for sentence (64) is shown in (68), where the object as the complement to the verb
is the leftmost element, followed by the head verb and then the subject in the specifier of IP
position.  The only difference between OVS and SOV languages is the order of the specifier.
(68)
IP
I'
   NP
   VP
 I0  +past
b®iryekomo
     |
     3pers 
boy
    V'
         
+sg
      NP
  V0
  |
        kuraha         yonyhoryeno
  bow
make
This simple change in ordering within the X-bar rules accounts for a good part of the
typological expectations based on the word order.  For example, head-initial languages, such as
SVO and VOS languages, have prepositions while head-final languages have postpositions.
(69) shows that Hixkaryana, as an OVS language, has postpositions as expected.
(69)
PP
 
 P'
            NP
      P0
 
         mokro
  yakoro
        that one
     with
A closer look at the possessed noun phrase in (65) shows that the possessor is initial in
Hixkaryana.  This is contrary to expectation, since the subject (which is also a specifier
A step-by-step introduction to GB
November 1998 
19
position) is final.  Hixkaryana can be analyzed as a head-final language that has the specifier in
final position in all phrases except nominal phrases,16 where the specifier is initial.  The tree for
(65) is shown in (70).
(70) 
        NP
NP[+poss]
   N'
   
          Masar® 
  N0
         Masar®’s
         
     hokru
child
The simple parameterization of the order of elements on the right side of the two basic
phrase structure rules has taken us a long way in accounting for the various underlying word
orders found in languages.  SVO, SOV, VOS, and OVS languages can all be analyzed in this
way, as summarized in the following chart:
SVO: XP → Specifier X' 
VOS: XP → X'  Specifier
X'  → X0  Complement(s) 
X'  → X0  Complement(s)
SOV: XP → Specifier X' 
OVS: XP → X'  Specifier
X'  → Complement(s) X0
X'  → Complement(s) X0
There are still two more word orders that do not fall out directly from a change of order
within the X-bar rules: VSO and OSV.
6.4 VSO and OSV Languages
There are only a few documented languages with OSV order, but a significant number of
languages from various language families have dominant VSO word order.  (71)-(73) give
examples from Quiegolani Zapotec, an Otomanguean language spoken in Mexico (Regnier
1989, Black 1994), that has VSO word order.17
(71) 
W-eey Benit mël.
C-take Benito fish
‘Benito took a fish.’
                                               
16 This division might be that all [-V] phrases, which would include PPs, have their specifiers initial.  More data would be
required to determine this.
17 Abbreviations:  C = completive aspect; 3RD = general third person pronoun; 1EX = first person exclusive pronoun.
A step-by-step introduction to GB
November 1998
20
(72)
W-nii men disa
lo
noo.
C-speak 3RD
language face 1EX
‘She spoke Zapotec to me.’
(73)
xnaa
noo
mother 1EX
‘my mother’
Urubú, of the Tupí family in Brazil (Derbyshire and Pullum 1981 from Kakumasu 1976),
provides data for an OSV language (74)-(75).
(74)
Pako
xuã u’u.
banana
João he-ate
‘John ate bananas.’
(75)
Koƒ
sepetu-pe
jurukã
Nexƒ mái
muji-ta.
tomorrow spit-on
ribs
Nexƒ mother she-will-roast
‘Nexƒ’s mother will roast the ribs on the spit tomorrow.’
In all of the phrases in Quiegolani Zapotec, the head is initial:  the verb is first in (71) and
(72); the preposition lo ‘face’ (used for ‘to’) comes before the pronoun noo ‘me’ in the
prepositional phrase in (72); and the noun comes before the possessor in (73).  In contrast, all
of the phrases in Urubú have the head in final position: the verb is last in both (74) and (75);
sepetu-pe ‘spit-on’ in (75) shows that the language has postpositions; and in the possessed
noun phrase, Nexƒ mái ‘Nexƒ’s mother’ (75), the noun is final.  We can generalize that VSO
languages are head-initial and OSV languages are head-final.
The problem with these two word orders for X-bar Theory is that the subject intervenes
between the verb and the object, something a specifier should not do.  For many years, it was
assumed that these languages were different from the others and had a flat structure rather than
a configurational one.  For example, (76), which follows, shows the flat VP structure
(otherwise following GB) that Woolford (1991) posits for the VSO language, Jacaltec:18
(76)
 
    IP
     |
     I'
             I0  tense
VP
    pers
   |
    num
 V'
V0
 NP
  NP
 |
        verb      subject       object
                                               
18 The OSV structure would be a mirror image of  (76).
A step-by-step introduction to GB
November 1998
21
However, research on VSO languages indicates that they really are similar to languages
with other word orders in a number of ways (such as subcategorization for complements,
subject-object asymmetries, and topicalization of a constituent allowed for a verb plus its
object but not a verb plus its subject), leading to the proposal that they start out with the same
configurational structure as other languages (Anderson and Chung 1977, Chung 1990,
McCloskey 1991, etc.).  From a theoretical point of view it is desirable to claim that all
languages have the same basic D-structure, where the subject is in a specifier position and the
sisters of the verb are all and only its complements.  Two main proposals have been made to
derive VSO word order from an underlying head-initial (SVO or VOS) D-structure.  We need
to understand more of the theory before addressing these proposals, however, so they will be
presented in section 11.1.
We have now seen how to account for the basic word order of a language using the X-bar
Theory of phrase structure.  In languages having SVO, SOV, VOS, or OVS word order, the
trees are simply generated in the proper order at D-structure from the phrase structure rules
which have been parameterized for the language.  In VSO and OSV languages, we must either
use a flatter structure where the subject is a sister of the verb or posit some as yet
undetermined movement.  Of course, not all sentences in any language have the basic word
order.  In the next section, we begin to look at the constraints on movement as we consider
question formation.
7. Constraints on movement: Question Formation
Government and Binding theory considers word order an important part of the syntax and
therefore seeks to account for how and why different word orders come about within a
particular language.19  Since it is a derivational theory, the various word order changes will be
assumed to arise from movement of one or more constituents, as was done in the predecessor
to GB, Transformational Grammar.  However, Transformational Grammar was strongly
criticized as being too powerful, therefore GB puts constraints both on the movement allowed
and on the structure resulting from the movement.  We will see what these constraints are as
we develop the analysis of Yes/No questions and content questions, one step at a time.
7.1 Yes/No Questions
  To form a Yes/No question in English, an auxiliary is moved in front of the subject, as in
the change from (77a) to (77b).  The distribution in (77c-g) shows that we need to be careful
in how we formulate this movement rule to be sure that only the grammatical Yes/No
questions are generated.  For example, only one auxiliary can move (an auxiliary only, not a
main verb), it has to be the first auxiliary, and that first auxiliary agrees with the subject and
determines the form of the following verb.
                                               
19 Even in languages which allow (fairly) free word order, usually due to a rich case and/or agreement system, there is an
unmarked, neutral, or more frequently occurring word order.  The other allowed word orders are more marked in that they
involve foregrounding or backgrounding of one or more constituents and/or would only be used in specific contexts.  The
unmarked word order would be the D-structure in GB, with the other orders being derived by movement.
A step-by-step introduction to GB
November 1998
22
(77)
a.
Sally has declined the job.
b.
Has Sally declined the job?
c.
*Has Sally declines the job?
d.
*Has Sally might decline the job?
e.
*Might have Sally declined the job?
f.
Might Sally have declined the job?
g.
*Declined Sally the job?
Sentences like:
(78)
Jill could have been playing the piano.
show that auxiliaries come in a certain order and that they subcategorize for the form of the
verb that comes after them.  This means that auxiliaries must be heads themselves, each
subcategorizing for a VP of a certain type.  For example, could, might, shall, etc. are modals
which may not be followed by another modal and require the verb that comes after them to be
in the bare form.  The lexical entry for could is given in (79).  Similar entries could be given for
the other modals and for the non-modal auxiliaries.
(79)
could
V[+aux /+modal ]
[ _VP[-modal /+bare ]]
The structure for (78) has four VPs, stacked one upon the other, as shown in (80).
(80)
IP
NP
I'
Jill
I[+fin]
VP
  |
 V'
V[+aux/+bare]
VP[+bare]
  |
  |
could
 V'
V[+aux/+bare]
VP[+en]
  |
        |
       have
       V'
V[+aux/+en]
VP[+ing]
  |
       |
       been
      V'
     V[+ing]
NP
      |
playing
 the piano
A step-by-step introduction to GB
November 1998
23
In Yes/No questions, the order and form restrictions still hold.  With a movement analysis,
we can account for these restrictions via subcategorization at D-structure.  Then the highest
auxiliary can move in front of the subject to obtain the surface word order.  We need to look at
more data to determine which position the auxiliary moves to.
Two other phenomena show that the highest auxiliary is important also:  VP deletion and
the placement of negation.  Consider the example in (81).
(81)
Jill couldn't have been playing the piano but
Bill could (have [been (playing [it])]).
We see in the first clause of (81) that negation follows the first auxiliary.  The options for the
second clause show that  VP deletion constructions carry the restriction that at least one
auxiliary (the highest one) must remain behind.  A coherent account of these facts yields that
the highest auxiliary must move to I0.
The D-structure tree for (77a and b) would be as shown in (82), with the movement of the
highest auxiliary has up to I[+fin] indicated by the arrow:
(82)
IP
NP
    I'
             Sally
I[+fin]
 VP
   |
  V'
V[+aux]
VP[+en]
  |
  |
has
 V'
     V[+en]
 NP
 |
     declined
the job
This movement of V[+aux] to I
0 is called HEAD MOVEMENT, since a head is moving to another
head position.  This movement does not give us the word order of a Yes/No question, so
something more is needed.  We know that the auxiliary moves in front of the subject, but we
do not know where it attaches.  Before we can propose what that movement is, we need to see
how movement is constrained in the theory.
Two basic principles are used to constrain movement.  First is the Principle of No Loss of
Information, which simply says that nothing can move to a position that is already phonetically
filled.  The second principle says that movement must be either STRUCTURE-PRESERVING or
ADJUNCTION.  Structure-preserving movement means that the moved constituent must fit into
A step-by-step introduction to GB
November 1998
24
the tree structure that is already there.  Further, it must fit the requirements of X-bar theory so
that it could have been generated there at D-structure.  This means that you cannot put a head
into a complement position or a phrase into a head position, and you cannot add a complement
that was not subcategorized for (and therefore filled) at D-structure.  In general, a head can
move to an empty head position, as we saw when the highest auxiliary moved to I0, or a
maximal projection can move to an empty specifier position.
Looking back at the tree in (82), we see that there is not another empty head position for
the auxiliary to move to.  We do have the option of adjoining to the IP, since movement by
adjunction is allowed, but the adjunction rule is recursive, which would incorrectly predict that
more than one auxiliary may be fronted ((77e) is ungrammatical).
We can find a better solution by looking at embedded Yes/No questions, as in (83):
(83)
I wonder whether Sally has declined the job.
We know from section 5.1 that the embedded clause following wonder is a CP.  Semantically,
both main clause Yes/No questions and embedded Yes/No questions have the same
interpretation; one only requires a more direct answer than the other.  Drawing on this parallel
between main clauses and embedded clauses, we can posit that main clauses are also CPs, and
that the CP and its head C for both main clause and embedded clause questions have the
feature [+q].  This means the D-structure for (77b) is as shown in (84).  Now, as the arrows
indicate, after the V[+aux] moves to I
0 it can move on to C[+q] to obtain the word order of
Yes/No questions.
(84)
CP[+q]
   |
  C'
    C[+q]
 IP
                NP
      I'
              Sally    I[+fin]
 VP
   |
 
  V'
V[+aux]
VP[+en]
  |
  |
has
  V'
   V[+en]
NP
    |
   declined
   the job
A step-by-step introduction to GB
November 1998
25
While we want to assume that the highest V[+aux] moves to I
0 in all cases, the movement of
I0 to C[+q] only occurs in questions, since declaratives and other non-questions will have a
C[-q].
20
Main clause and embedded Yes/No questions now have parallel structure, yet there is a
significant difference in their surface word order.  How do we account for the fact that no
Subject/Aux inversion occurs in embedded Yes/No questions? The tree in (85) shows that the
C[+q] position is filled in an embedded clause with whether.  The Principle of No Loss of
Information blocks the movement of has to the C[+q] position, without any additional
stipulations.
(85)
CP[-q]
   |
 C'
C[-q]
     IP
NP
    I'
  I
I[+fin]
VP
  |
 V'
 V
C[+q]
  |
  |
   wonder
C'
C[+q/ -root] 
IP
       |
 whether
      NP
I'
     Sally
   I[+fin] 
  VP
   |
  V'
V[+aux] 
VP
 
 |
  |
has
V'
  V
NP
   |
  declined     the job
                                               
20 In practice, the CP[-q] projection is usually omitted, so main clause declaratives simply have IP as the topmost node.
A step-by-step introduction to GB
November 1998
26
The examples in (86)-(88) from Black English (Jim McCloskey, p.c.) show that
Subject/Aux inversion is not a root-clause-only phenomenon universally.
(86)
Ask your father DOES HE want his dinner?
(87)
Do you remember DID THEY live in Rosemont?
(88)
I never found out WOULD HE really have come with me.
The only change we need to make to account for this data is to note that Black English has a Ø
C [+q/ -root] instead of whether.  Therefore, I
0 moves to C[+q] in embedded clauses just as in root
clauses in Black English, since no information is lost.
7.2 Content Questions
Content questions have the same Subject/Aux inversion as Yes/No questions, so the same
head movement of the highest auxiliary to I0 and on to C[+q] applies.  This movement takes
place automatically because both types of questions have a C[+q]. The difference between
Yes/No questions and content questions is that an additional movement takes place in content
questions:  a [+wh] phrase moves to the front. For example:
(89)
Which job has Sally declined?
Since there is an open specifier position in the CP[+q], the wh-phrase which job can move to that
position, as shown in the tree in (90).  This movement is called ¼-movement (where the bar in
this case is set complement notation), which is movement to a non-argument position.21
                                               
21 An argument is a subject or complement position.  Therefore, a non-argument position is a non-subject specifier
position or an adjoined position.
A step-by-step introduction to GB
November 1998
27
(90)
CP[+q]
 C'
C[+q]
   IP
         NP
I'
       Sally
 
  I[+fin]
VP
  |
 V'
 V[+aux] 
VP[+en]
  |
  |
has
 
 V'
     V[+en]
   NP[+wh]
 |
declined
  D[+wh] 
N'
   |
 |
which
N
 |
job
The movements required for questions can be stated simply:
(91)
a. The highest V[+aux] must move to I
0.
b.
I0 must move to C[+q].
c. An XP[+wh] must move to the specifier of a C[+q].
No conditions are needed on the rules, because the general principles rule out the
ungrammatical examples.22  For example, English allows multiple wh-phrases in a single
question, such as in (92), yet only one of these phrases is fronted.  Since there is only a single
specifier position for CP[+q], once one wh-phrase is fronted, the Principle of No Loss of
Information blocks further movement.
(92)
a.
Who(m) did John give what to?
b.
*Who(m) what did John give to?
c.
What did John give to whom?
d.
*What to whom did John give?
                                               
22 Other principles are needed to handle all the cases.
A step-by-step introduction to GB
November 1998
28
Of course, not all languages have movement of wh-phrases in questions while others allow
multiple fronting, so parameterization is needed to account for the variation seen (Black
1994:Appendix).  The motivation for ¼-movement seems to be the scope of the semantic
operator.  Other instances of ¼-movement involve focus, negation, and quantifier
constructions, all  of which are semantic operators.  Some languages, such as Zapotec and
Tzotzil require scope to be readable from the S-structure tree, whereas other languages allow
further covert movement to take place to establish scope for semantic interpretation.
7.3 Traces of Movement
Instead of drawing arrows on the D-structure tree to indicate movement, as we have done
so far, a separate S-structure tree is drawn with a coindexed trace  (indicated by the
subscripted t) left in place to show where the moved constituent came from.  The S-
structure tree for (89) corresponding to tree (90) is given in (93).
(93)
CP
NP[+wh] 
          C'
      which jobk   
C[+q]
IP
 |
    hasi      NP
    I'
   Sally
   I[+fin]
VP
   |
  |
   ti 
 V'
V[+aux] 
VP[+en]
 |
  |
 ti
  V'
   V
NP[+wh]
   |
   |
  declined
   tk
The traces also serve to insure that movement cannot move an item into a previously filled
position.
We have seen examples of both head movement and ¼-movement here.  The other type of
movement is A-movement:  movement to an argument position.  This type turns out to be all
movement to subject position, such as passive, unaccusative, and raising constructions.
Nothing is allowed to move to a complement position, since complement positions are
determined by subcategorization and filled at D-structure.  The full analysis of these
A-movement constructions requires more theoretical ‘machinery’ involving semantic roles and
Case theory.  These will be presented in the following two sections.
A step-by-step introduction to GB
November 1998
29
8. A-movement and Semantic roles
In the last section, we began looking at the constraints on movement.  We saw there
instances of head movement, such as movement of the auxiliary verb in front of the subject in a
question, and ¼-movement, where a phrase moves to a non-argument position as in the
fronting of a wh-phrase in content questions.  This section looks at the third type of movement,
A-movement or movement to the subject position, which is an argument position.  Cases of
A-movement to be considered here include passive, unaccusative, and raising constructions.
Passives are the most well-known constructions involving movement to the subject position.
The old Transformational Grammar analysis begins with a transitive deep structure, then
creates a passive surface structure by moving the subject to the by-phrase (or omitting it
completely); moving the object to the subject position; adding the passive be; and changing the
verb form appropriately.  This movement analysis of passives captured the generalizations that:
(94)
a. Most transitive verbs have passive alternates.
b. No intransitive verbs have passive alternates.
c. The ‘subject’ of a passive verb corresponds to the object of its transitive
alternate.
Government and Binding Theory does not account for passives in exactly the same way as
Transformational Grammar did, since movement of the object to a previously occupied subject
position would violate the Principle of No Loss of Information.  Besides accounting for the
generalizations in (94), GB also seeks to explain the synonymy between the active and passive
sentences, as seen in (95).
(95)
a.
The kids invited Sue to the party.
b.
Sue was invited to the party by the kids.
This is done by expanding the lexical entries to include semantic roles.
8.1 Semantic Roles in the Lexicon: Passive Constructions
In both (95a and b) the kids gave the invitation, Sue was the one invited, and the party is
what Sue was invited to.  General semantic roles, such as AGENT, THEME, RECIPIENT, GOAL,
LOCATIVE, etc., can be linked to arguments within the lexical entries to capture this synonymy.
For example, the lexical entry for invite would be:23
invite, V,
[ __     NP   (PP[to]) ]
      |
|
invite′
<AGT, THEME, GOAL>
                                               
23The line containing the semantic roles is introduced by the logical semantic predicate which is indicated by the syntactic
verb form followed by a prime.  The lexical entry thus specifies both syntactic and semantic subcategorization information.
A step-by-step introduction to GB
November 1998
30
where each syntactic complement (called internal arguments) must be linked one-to-one with a
semantic role, and one additional role (underlined) may be linked to the subject (called the
external argument).  Lexical entries apply at D-structure, so not all verbs assign a semantic role
to the subject position.24  As seen in sections 3 and 4, the internal arguments are sisters to the
head, as shown in the tree for the declarative sentence (95a).
(96)
     IP
NP
 I'
             the kids
  I[+past]
VP
   |
  V'
   |
V
  NP
 PP[to]
 |
   invite       Sue       to the party
Further assumptions about lexical entries in GB include the desire that the related forms of
a word share a single subcategorization frame and that there should not be any cross linking of
syntactic arguments and semantic roles.  For example—directly relevant to the analysis of
Passive—the THEME should not be assigned to the object in one case and to the subject in a
related entry.  This assumption is formalized under the Uniformity of Theta Assignment
Hypothesis (Baker 1988:46), which makes the broader claim that the THEME role is always
assigned to the direct object when it is present, since that is its position in normal transitive
verbs; the RECIPIENT role is assigned to the indirect object, etc.
Though the semantic roles are assigned at D-structure through the lexical entries, another
assumption forbids any movement from changing the linking between syntactic arguments and
semantic roles.  Therefore, movement of something out of a position is allowed, as we have
seen with wh-question formation, but nothing else can move into that position.  The semantic
role stays with the original position rather than moving with the phrase; the semantic role is not
part of the tree but part of the lexical subcategorization that goes with the D-structure
position.  Movement into a position linked to a semantic role is not allowed, since it would
cause the moved element to take on that semantic role, and thus alter the original linking.  The
coindexed trace left after movement provides the link between the moved element and the
position it occupied at D-structure. Therefore, both semantic roles and subcategorization
requirements are still recoverable at S-structure.
                                               
24For example, auxiliaries do not.  So in John may have hit a home run, there is only one external argument = John.  Which
verb assigns a semantic role to John?  It is the AGENT of the verb hit; a home run is the THEME of hit.  Auxiliaries do not
assign external arguments:  there is nothing anyone did or experienced to may.  Instead they simply select a VP complement
that is a type of EVENT or STATE.
A step-by-step introduction to GB
November 1998
31
The assumptions about the form of the lexical entries are made to capture generalizations
about language.  Given these assumptions, the GB account of passive must be partly done in
the lexicon and partly by movement.  A lexical rule,25 such as that given in (97), is used to
capture generalizations (94a-b) that passive verbs are related to transitive verbs.  No
intransitive verbs will have a passive counterpart generated by this lexical rule.  In order to
account for alignment of the semantic roles between entries as much as possible and to follow
the Uniformity of Theta Assignment Hypothesis, the object NP remains in position at D-
structure.
(97) V,  [ _  NP X ] → V+en[+pass], [ _ NP X (PP[by]) ]
The full lexical entry for the passive form invited, including semantic roles, is:
invited
V[+pass]
[ _ NP (PP[to] ) (PP[by] ) ]
|
|
|
invited′
<THEME, GOAL, AGENT>
Note that no external argument is assigned by the passive verb.
The D-structure for the passive sentence (95b) is as shown in (98), where the passive
auxiliary verb is inserted from the lexicon as the only head which subcategorizes for a VP[+pass]:
(98)
D-structure for passive
IP
 |
 I'
  I[+past]
VP
  |
 V'
    V
VP[+pass]
 
     |
  |
  was
 V'
V[+pass]         NP
       PP[to]
 PP[by]
 |
 invited
      Sue
  to the party
  by the kids
                                               
25A lexical rule takes a lexical entry that matches its left side and generates an additional lexical entry in the form of the
right side.  The original lexical entry is unchanged and still part of the lexicon.  A few exceptional passives will have to be
separately added to the lexicon, such as verbs which take clausal complements that can have a passive alternate.
A step-by-step introduction to GB
November 1998
32
Movement of the object to subject position is still needed.  This movement can take place
since there is no semantic role linked to the subject position at D-structure and the position
was not lexically filled so the Principle of No Loss of Information is not violated.  A coindexed
trace is left behind to maintain the linking of the object to its semantic role.  So the S-structure
tree looks like (99):
(99)        S-structure for passive
IP
NP
   I'
                Suek
I
VP
 |
  |
    wasi
 V'
V
VP[+pass]
 |
  |
ti
 V'
V[+pass]       NP
 PP[to]
     PP[by]
 |
|
  invited
tk
 to the party
by the kids
8.2 Unaccusative Constructions
The assumptions made about semantic roles in the lexical entries also require us to
distinguish between types of intransitive verbs, following the Unaccusative Hypothesis
(Perlmutter 1978).  The key distinction is whether the NP (which ends up) in subject position
performed the action or was acted upon.  Consider the examples in (100):
(100)
a. Bill sleeps well.
b. Bill died.
c.
The glass broke.
d. Bill broke the glass.
In (100a), sleep is assumed to be a regular intransitive verb with an AGENT subject.  In
contrast, Bill did not do anything to make himself die, so in (100b) the subject is assumed to
have the THEME role.  Similarly, in both (100c) and (100d) the glass is what the breaking
happened to so it fills the THEME role in both the unaccusative construction in (100c) and the
transitive construction in (100d).
The assumptions about semantic roles require that while verbs like sleep have an AGENT in
subject position at D-structure, verbs like die have an empty subject position at D-structure
with a THEME object, as shown in the lexical entries below.
A step-by-step introduction to GB
November 1998
33
Further, a verb like break has an optional AGENT.  Movement of the THEME to subject
position for the unaccusative verb die is exactly parallel to the passive movement, as shown in
the simple example in (101).
sleep,  V     [ _ ]
die,  V   [ _ NP ]
break,   V     [ _ NP ]
|
|
sleep'
<AGT>
die'        <THEME>
break'    <(AGT), THEME>
(101) 
   D-structure
S-structure
IP
IP
 |
 I'
⇒
 NP
I'
I[+past]
        VP
 Billi
  I[+past]
VP
          |
  |
         V'
 V'
V                NP
V
NP
 |
 |
  |
          die              Bill
    die
 ti
8.3 Raising Constructions
The third type of A-movement (movement to an argument position) involves raising
predicates such as seem and likely.  These predicates take either a finite or a nonfinite clause
complement and do not assign a semantic role to their own subject position.  The lack of a
semantic role assigned to the subject position can be seen by the presence of the dummy it
when there is a finite clause complement (102a).  In the case of a nonfinite clause complement,
the subject of the lower clause must raise to the main clause subject position (102b-c).
(102)
a.
It is likely that Sue will come.
b.
Sue is likely to come.
c.
*It is likely Sue to come.
The trees for (102a) are given in (103).  The dummy it is inserted in the main clause subject
position to obtain the S-structure, fulfilling the English-specific requirement that (at least) main
clauses have phonetically filled subjects.  Movement of the auxiliaries to I0 in both clauses is
also shown.
A step-by-step introduction to GB
November 1998
34
(103)
D-structure
S-structure
     IP
IP
      |
⇒
     I'
NP
I'
      I[+fin]
VP
it
I[+fin]
 VP
  |
  
|
   |
 V'
 
           isi
        V'
V[+aux]
      AP
V[+aux]
   AP
  |
        |
 |
|
 is
      A'  
 ti
     A'
       A
   CP
  A
CP
        |
    |
   |
  |
    likely
  C'
likely
 C'
 C
 IP
C
 IP
  |
 |
          that
 NP
  I'
    that  NP
  I'
Sue   I[+fin] VP
Sue   I[+fin]   VP
  |
   |
    |
 V'
willk
   V'
V[+aux]
  VP
     V[+aux]
    VP
 |
   |
      |
 |
will
  V'
tk
V'
   |
 |
  V
V
   |
 |
come
   come
Contrast the trees in (103) for sentence (102a) with the trees shown in (104) for sentence
(102b), where the subject of the nonfinite clause must raise to the main clause subject position.
A step-by-step introduction to GB
November 1998
35
(104)
D-structure
S-structure
      IP
IP
       |
      I'
⇒
   NP
I'
     I[+fin]
VP
  Suek
    I[+fin]
VP
 |
|
  |
V'
isi
 V'
V[+aux]
     AP
   V[+aux]
  AP
 |
       |
     |
    |
is
     A'
    ti
   A'
       A            IP[-fin]
A
   IP[-fin]
        |
 |
    likely      NP
    I'
   likely    NP
   I'
|
    Sue
 I[-fin]
   VP
tk
  I[-fin]
  VP
  |
    |
   |
    |
  to         V'
  to
   V'
    |
    |
   V
   V
    |
    |
come
come
Clearly, more than just a requirement that the main clause subject position be phonetically
filled is at work here.  Case Theory provides the motivation for A-movement of the particular
NP which raises in passives, unaccusatives, and raising constructions.
9. Case Theory
The English pronoun system gives us a glimpse of the positions that are assigned case and
which morphological case they receive.  Consider the data in (105)-(111).
(105)
She/*her went to the store.
(106)
John invited her/*she to the party.
(107)
John bought it for her/*she.
(108)
John would have liked for her/*she to come.
(109)
John was glad that she/*her came.
(110)
John wondered whether she/*her would come.
(111)
She/*her brought them/*they  their/*they/*them suitcases.
A step-by-step introduction to GB
November 1998
36
We can generalize from this data that:
a. she, they, I, etc. are used in subject position of main and embedded clauses, except
for embedded clauses headed by C[for].
b. her, them, me, etc. are used in object position and as the object of a preposition and
in subject position after C[for].
c.
their, his, my, etc. are used in possessor position.
As usual, GB rephrases the generalizations in terms of phrase structure.  Also, even though
only pronouns show overt morphological case in English, it is assumed that all NPs have Case
(called abstract case) that matches the morphological case that shows up on pronouns.  Appeal
is made to other languages with much richer case systems than English to back up this claim.
In phrase structure terms:
a. Nominative Case is assigned to the NP specifier of I[+fin].
IP
NP
I'
I[+fin]
 VP
 …
b. Accusative case is assigned to the NP sister of V or P.  The C[for] which is
homophonous with the preposition for acts like P for Case assignment.  Note
that the subject of a nonfinite clause could not receive Case from I[-fin] since only
I[+fin] assigns Nominative Case.
VP
or
PP
or
CP
  
|
 |
  |
 V'
 P'
 C'
V        NP
 P
       NP
   C
IP
     
  |
    
for
NP
I'
I[-fin]
 VP
 
|
to
  …
A step-by-step introduction to GB
November 1998
37
c. Genitive Case is assigned to the specifier of N.
NP
NP
N'
 
 |
N
What is the same about these positions that receive Case and the positions that assign
Case?  Chomsky observed that every maximal projection (=XP) that dominates the NP that
receives Case also dominates the head that assigns it (if we do not count the IP that intervenes
between the C[for] and the NP).
Our first definition of government 26 comes from this observation.
^  GOVERNS  _ iff
a.
^ is a head [±N,±V] or I[+fin] or C[for], and
b.
every XP that dominates ^ also dominates _, and
c.
every XP (other than IP) that dominates _ also dominates ^.
In this definition, ^ and _ stand for particular categories.  Clause (a) requires that ^ be one of
the heads N, V, A, P, I[+fin]  or C[for] .  Almost always, _ is an NP, since NPs need Case, which is
assigned under the government relation.  Clause (b) determines how high up the tree a head
may govern:  if every maximal projection above the head must also dominate the NP in
question, then the NP must be below the maximal projection of the head (e.g. VP for V, IP for
I[+fin]).  Clause (c) provides the lower limit of government by not allowing the head to govern
down into another maximal projection other than IP.  Together, clauses (b) and (c) establish
locality constraints on the government relation for each head.
These locality constraints are illustrated in the tree in (20) by the dashed lines.  (Indexes
have been added to some of the categories to aid discussion.)
                                               
26This definition will be changed slightly in section 10 on Binding Theory.  At that point, the reader should be able to
understand the full name of the theory.
A step-by-step introduction to GB
November 1998
38
(112)
IP-1
    NP-1
I'
    John
   I[+fin]
VP-1
   |
  V'
   V
CP
    |
  |
  likes
 C'
    C
IP-2
     |
   for
NP-2
I'
 her
  I[-fin]
VP-2
   |
   |
  to
  V'
   |
  V
   |
come
There are four heads that can govern in this tree:  I[+fin], V likes, C for, and V come (recall that
I[-fin] to cannot govern by clause (a) of the definition).  I[+fin], governs up to its maximal
projection IP-1, so it governs its specifier NP-1 John, and I[+fin] also governs its complement
VP-1 (but not anything inside VP-1).  V likes governs up to its maximal projection VP-1 and
its complement CP, but it does not govern anything within the CP.  C for governs up to its CP
maximal projection and down all the way through IP-2 (which is invisible for government) to
and including the next maximal projection VP-2, crucially governing the NP-2 her.  Finally, V
come governs everything else within its own maximal projection, which in this case is nothing.
Since we will be using government for Case assignment to NPs, we can think of the
definition in simpler terms as:
A head (N, V, A, P, I[+fin], C[for]), GOVERNS its NP specifier and its NP complement and the
NP specifier of an IP[-fin] complement.
Note that government is stronger than subcategorization because a head governs its
specifier (and the specifier of its complement for IP complements) as well as its complements.
The Case assignment rules in terms of government are simply:
a.
I[+fin] assigns nominative case to the NP specifier that it governs.
b.
N assigns genitive case to the NP specifier that it governs.
c.
V, P, C[for] assign accusative case to the NP that they govern.
A step-by-step introduction to GB
November 1998
39
GB requires that all NPs must have Case at S-structure by the Case Filter in (113).
(113)
Case Filter:  *NP if it does not have Case at S-structure.
With one further assumption, we will have the motivation for A-movement in passive,
unaccusative, and raising constructions.  Burzio’s Generalization (Burzio 1986) states that
predicates which do not assign a semantic role to their external argument cannot assign Case to
their complement(s).  This provides the answer to why the passive object must move.  Passive
verbs do not assign a semantic role to their external argument position, so they have lost the
ability to assign Case to their complement.  Therefore, the NP object cannot remain in place at
S-structure and must move to a position where it can get Case:  the specifier of I[+fin] where
nominative case is assigned.  The same reasoning accounts for the unaccusative and raising
constructions:  the NP which cannot receive Case in its D-structure position is the one which
must move; and it may only move to a position which does assign Case, the specifier of I[+fin].
Note that A-movement is motivated by the need for Case, so the moved NP will be in a
Case-assigning position at S-structure but its trace will not.  ¼-movement is just the opposite:
the moved NP is not in a Case-assigning position at S-structure but its trace is.  We must
therefore allow the ¼-moved phrase to pass the Case Filter via its coindexed trace.  In general,
either the moved NP or its trace must be assigned Case at S-structure, but not both.  Another
way of stating this is to say that the moved element and its coindexed trace form a chain, and
only one Case is assigned to a chain.27
We have now completed the basic restrictions on the three types of movement allowed in
GB.  We also learned what government is in phrase structure terms.  The next section deals
with the second part of the name of the theory:  binding constructions.
10. Binding Theory
In this section, we finally learn what Binding Theory is and why it is so important that it is
part of the name of the overall Government and Binding framework.  In its narrowest
conception, binding involves reflexive constructions, such as (114).  Equi constructions, such
as (115), also fall under Binding Theory, however, as do the various types of movement
constructions considered in the last three sections, exemplified in (116)-(118).  Further, we will
see that pro-drop constructions (which English does not allow), shown for Spanish in (119),
are also accounted for by Binding Theory.
(114)
Sue likes herself.
(115)
Bill tried to win the race.
(116)
What did Jill give to you?
                                               
27The only construction which seems to allow two Case assignments to a chain is tough movement, where the object NP of
an embedded nonfinite clause may move to the main clause subject position:
i a. The bricksi are hard for Lisa to cut ti.
b. The bricksi are hard to cut ti
The bricks would receive nominative Case in its S-structure subject position, but its trace would also receive accusative
Case from cut.  This construction remains a problem for the theory.
A step-by-step introduction to GB
November 1998
40
(117)
The homerun was hit by Joey.
(118)
Kim is likely to win the prize.
(119)
Hablo español .      ‘I speak Spanish’
It is most likely unclear at this point what the examples in (114)-(119) have in common.
Let’s start with reflexive constructions and build the Binding Theory step by step in section
10.1, and then seek to unite all the constructions above in section 10.2.  We will also work
through the analysis of the equi constructions (115) in section 10.3.
10.1 Binding and Command Relations in Reflexive Constructions
We saw in section 9 that Case Theory determines whether a nominative pronoun, such as
she or he, is used instead of an accusative pronoun, her or him, or a genitive pronoun like his.
It is Binding Theory’s job to determine when a reflexive anaphor, for example, herself, is used
instead of one of the pronouns, she or her.
Consider the following data, where ‘her/*herself’ means her is grammatical but herself is
not (similarly for ‘she/*herself’ and other combinations).
(120)
She/*herself shuddered.
(121)
Sally enjoyed herself at the party.
(122)
Sally left a note for herself.
(123)
Sally thought that Max disliked her/*herself.
(124)
Sally talked to John about himself.
(125)
Sally talked to John about herself.
(126)
Sally believed that she/*herself would succeed.
(127)
Sally herself/*she/*her couldn’t read what she/*herself had written.
 (128)
That Sally might succeed amazed her/*herself.
(129)
That we had seen Sally in the street amazed her/*herself.
(130)
That Sally enjoyed herself/*her surprised her/*herself.
What basic generalizations can be gleaned from this data?
a.
Reflexive pronouns must corefer with some NP before them in the sentence.
b.
There is a locality condition for this coreference relationship.  Examples (126)-
(130) suggest that the antecedent, which is the NP that the reflexive corefers
with, must be within the same minimal clause (=CP) as the reflexive.
The S-structure tree for (130) is given in (131) to make the same/different clause
distinction clearer.  The NP following enjoyed (marked as NP-2) can be the reflexive herself
because it is coreferent with the NP-1 Sally in the same CP.  NP-3 following surprised may
also refer back to NP-1 Sally, but the reflexive herself is ungrammatical in that position
because the locality condition is not met.
A step-by-step introduction to GB
November 1998
41
(131)
IP-1
CP
I′
  |
 C′
   I[+fin]
VP
  |
  C
IP-2
 V′
   |
that
    NP-1
I′
V
  NP-3
 |
   Sally
 I[+fin]
   VP
surprised
her/*herself
     |
    V′
   V
NP-2
    |
enjoyed
   herself/*her
We need more definitions before we can formulate the binding conditions more precisely.
Just like the government relation, the following command and binding relations are based on
the phrase structure.
The first definition is that of  C(onstituent)-COMMAND (Reinhart 1976), which formally
expresses the notion of ‘higher in the tree than’.
^  C-COMMANDS _ iff
a.
^  does not dominate _, and
b.
the first branching node that dominates ^ also dominates _.
In this definition again (and in others to follow) ^ and _ stand for particular categories.  For
example, in tree (131) we can let ^ be NP-1 Sally and see if it  C-commands NP-2 herself
(=_).  Clause (a) of the definition requires that NP-1 does not dominate NP-2.  This is true
because NP-1 is not directly above NP-2 in the same branch of the tree.  Clause (b) requires
that the first branching node that dominates NP-1, which is IP-2, also dominates NP-2.  IP-2
does dominate NP-2, so NP-1  C-commands NP-2.  (Note that NP-1 also  C-commands
everything else under IP-2 on the right branch.)
Now let’s check whether NP-1 Sally  C-commands NP-3 her according to the definition.
This time we set ^ = NP-1 and _ = NP-3.  Clause (a) is met because NP-1 does not dominate
NP-3.  Clause (b) fails, however, because the first branching node that dominates NP-1 is still
IP-2 and NP-3 is not under (dominated by) IP-2.
A step-by-step introduction to GB
November 1998
42
A simple way to think of  C-command is to start with your ^ category, go up the tree one
level to where it branches, then ^  C-commands everything down in the other branch.  So, if the
category you are concerned about (_) is in that other branch, ^  C-commands _.
As you might have guessed,  C-command is one of the conditions on binding.  Before we
go on with the specifics of binding, though, we are ready to understand a similar command
relation, called M(aximal)-COMMAND, that is used for government.28
^  M-COMMANDS _ iff
a.
^  does not dominate _, and
b.
the first maximal projection that dominates ^ also dominates _.
We can see how  C-command and  M-command differ using the simple X-bar tree in (132).
If we choose the NP specifier to be ^, we know from above that it  C-commands everything in
the right branch below the maximal projection XP.  M-command will give exactly the same
results for this choice of ^ ; in simple terms, M-command says to go up the tree from ^ until
you reach a maximal projection, then ^  M-commands everything in the other branches below
that maximal projection.  The reason C-command and M-command give the same results in this
case is that the first branching node above the specifier is also the first maximal projection
above it.
(132)
XP
NP
X′
      specifier
X0
YP
head
complement
The difference between these two types of command relations shows up when ^ is a head X0.
Now the first branching node above X0 is X′, so the head only C-commands its complements,
as in the subcategorization relationship.  M-command, however, reaches up to the maximal
projection and then goes down the other branches, so both the complements and the specifier
are included.
This second relation is exactly what is needed for government.  In fact, M-command
provides the same upper limit as clause (b) of the definition of government given in section 9,
so the revised definition is:
^  GOVERNS _ iff
a.
^  is a head [±N,±V] or I[+fin] or C[for], and
b.
^  M-commands _, and
c.
every XP (other than IP) that dominates _ also dominates ^.
                                               
28Review section 9 if necessary for explanation of the notion of government and see the revised definition below.  Also,
there is actually a whole family of command relations.  See Barker and Pullum (1990) for formal discussion.
A step-by-step introduction to GB
November 1998
43
Returning now to the binding conditions, the official definition of binding simply adds
coindexing to the C-command relation.  Coindexing is marked in the tree via subscripts and
indicates that the two NPs refer to the same entity.
^  BINDS _ iff
a.
^  C-commands _, and
b.
^  and _ are coindexed.
We can further distinguish between A-binding and ¼-binding just as we did with movement.
A-binding is binding by an antecedent in argument (=subject or complement) position, and ¼-
binding is when the antecedent is in a non-argument position.
^  A-BINDS _ iff
a.
^  is in an argument position, and
b.
^  binds _.
A simpler way to think of A-binding that works most of the time is:
An NP is A-BOUND if it is coindexed with a higher NP in either a subject or object position.
We can illustrate A-binding using the tree in (133).  NP-1 Sally and NP-2 Max are both in
argument position and they both C-command the object of disliked, NP-3, and are coindexed
with it.  Therefore, NP-3 is A-bound.
(133)
IP-1
NP-1
I′
       Sallyi
    I[+fin]
 VP
  |
 V′
         V
CP
          |
   |
    thought
  C′
C
IP-2
 |
    that
   NP-2
        I′
Maxk
   I[+fin]
VP
  |
 V′
   V
NP-3
    |
     disliked      heri/*herselfi/himselfk/*himk
A step-by-step introduction to GB
November 1998
44
But A-binding alone does not explain all the options for filling NP-3.  Why must we use her
and not herself to refer back to Sally, while just the opposite is true with respect to Max?  We
still need conditions to rule out the ungrammatical cases.
The Principles of Binding Theory determine whether a pronoun or an reflexive anaphor is
correct in a particular position.
Principles of Binding Theory
A. Anaphors (e.g. reflexives and reciprocals) must be A-bound in their governing
category.
B.
Pronouns must not be A-bound in their governing category.
C.
Full NPs (also called denoting expressions or R(eferential)-expressions) must
not be A-bound.
Principle A says that an anaphor can only be used when the position that A-binds it is local
enough:  In tree (133), NP-2 Max is close enough to NP-3 so that the anaphor himself is
correct; NP-1 Sally is too far away to use herself in NP-3.
Principle B says that a pronoun can only be used if it is not A-bound at all, or if its A-binder
is far enough away. This is why him cannot be used in NP-3 to refer back to NP-2 Max but her
may refer back to NP-1 Sally in (133).
Finally, Principle C says that nonpronominals may not be A-bound at all.   This is to rule
out repetition of full nominals.
(134)
*Johni hit Johni.
(135)
*Sallyi thought that Maxk disliked Sallyi/Maxk.
Defining this local domain that requires an anaphor and cannot have a coreferent pronoun
has been problematic.  We saw above that a basic generalization is that the antecedent and the
anaphor must be in the same clause.  This works for most cases, but there are a few exceptions
(e.g., Sally is eager for herself to succeed) where the anaphor and antecedent are not in the
same clause.  The local domain is therefore defined in terms of government and subjects,29
since most anaphors have antecedents that are subjects.
The GOVERNING CATEGORY is a local domain which denotes the minimal category which
contains both a subject and the governor of the element in question.  This minimal category
is usually a finite IP or an NP containing a possessor (which qualifies as the subject).
In tree (133) above, the governing category for NP-3 is IP-2, since it contains both a subject
(=NP-2 Max) and the governor for NP-3 (=V disliked).
                                               
29 Actually, the formal definition requires that there be a SUBJECT, which includes an NP subject, and NP possessor for
binding within an NP, or agreement features in I[+fin].
A step-by-step introduction to GB
November 1998
45
Working with the binding principles can get confusing since the definitions are nested one
within another, but, in about 99 percent of the cases, these simplified principles will work:
Simplified Principles of Binding Theory
A. Reflexives and traces of A-movement must be coindexed with the closest subject 
above them in the tree.
B. Pronouns cannot be coindexed with the closest subject above them in the tree.
C. Full NPs and traces of ¼-movement must not be coindexed with any subject or 
object above them in the tree.
10.2 Extending Binding Theory Beyond Reflexives
As seen in the last section, the Principles of Binding Theory recognize that the class of
nominal phrases is partitioned into three different types:  anaphors, pronouns, and full NPs.
These partitions are characterized by the two features [±anaphoric] and [±nominal], where
reflexives and reciprocals are [+ana,-pro], pronouns are [-ana,+pro], and full nominal phrases
are neither pronominal nor anaphoric so they are [-ana,-pro].
The chart in (136) shows these featural distinctions and which Principle of Binding Theory
applies to each.  Empty categories are also included, since both Extended Standard Theory
(which came out of Transformational Grammar) and GB claim that the chain coindexing
established by movement is equivalent to the coindexing in binding relationships between overt
nominals.  Four types of empty categories are recognized, corresponding to the four possible
feature specifications, as explained further below.
(136)
Featural Distinctions for Overt NPs and Empty Categories
Features
Binding
Principle
Overt
Nominals
Empty
Categories
[-pro/+ana]
A
e.g. himself
trace of
A-movement
[+pro/-ana]
B
e.g. him
pro
[-pro/-ana]
C
e.g. John
trace of
¼-movement
[+pro/+ana]
A and B
PRO
We talked about the trace of A-movement in passive, unaccusative, and raising
constructions in section 8.  By saying that the trace of A-movement is anaphoric and thus
subject to Principle A, we restrict the movement to only local domains.  We also saw the trace
of ¼-movement in the formation of content questions in section 7.2.  This trace is subject to
Principle C, which requires that it cannot be bound by an element in an argument position.  But
¼-movement is movement to a nonargument position by definition, so this requirement is
clearly met.
A step-by-step introduction to GB
November 1998
46
The two new empty categories are not traces, but empty elements in the lexicon.  The first
of these, pro, is the empty pronoun allowed in pro-drop languages, usually because of
agreement morphology on the verb to specify the person and number of the subject,  e.g. pro
No hablo español.  (See (119).)  This empty pronoun shows up in all the same places that an
overt pronoun does and is therefore subject to Principle B.
Finally, PRO is the empty subject in non-finite clauses, sometimes called controlled PRO.
Since it is both anaphoric and pronominal, PRO is subject to both Principles A and B of
Binding Theory.  From this is derived the fact that PRO must be ungoverned:  the only way it
could be A-bound inside its governing category and not be A-bound in its governing category is
if it doesn’t have a governing category because it doesn’t have a governor.  Further, if PRO
doesn’t have a governor, it cannot receive Case, since Case is assigned by the governor.  Overt
NPs of all types are required to have Case at S-structure by the Case Filter (see section 9).
This explains why there cannot be an overt counterpart to PRO with the [+pro/+ana] feature
specifications.
The requirement that PRO must not be governed at S-structure means that it can only be in
the configuration shown in (137).  This position is ungoverned because the V cannot govern
past the CP maximal projection, and neither the null C nor the I[-fin] to may govern.  Note also
that since this position is ungoverned and therefore cannot receive Case either, only PRO can
fill it.
(137)
V′
V
CP[-fin]
  |
   |
verb
  C′
           C[-fin]
IP[-fin]
             |
       for
   NP
I′
 PRO
     I[-fin]
   VP
|
     to
    …
Other NPs and traces filling the specifier of IP[-fin] cannot have a CP headed by a null
C[-fin] above them as in (137); either the complementizer for must be present, or the main verb
simply subcategorizes directly for an IP[-fin].
Both pro and PRO have intended reference, even though they are null.  Since they are
referential, they bear a semantic role, unlike the dummy it seen in constructions like It is likely
that Sue will come.  (This construction was analyzed in section 8.3.)  We’ll see examples of
how and where PRO is used in the next section.
A step-by-step introduction to GB
November 1998
47
10.3 Equi Constructions
How can we analyze (138)-(140)?
(138)
I would like to leave by noon.
(139)
I hate to swim in that pond.
(140)
I would like to have finished the homework by midnight.
Non-derivational theories such as Head-Driven Phrase Structure Grammar (HPSG) (Pollard
and Sag 1994) assume that, in examples like (138), like simply subcategorizes for a VP[-fin],
which in English must be marked by to.  Consider what happens if we try such an analysis
within the assumptions of GB.  Possible subcategorization frames for the two main verbs in
(138) are shown, followed by the tree structure in (141).30
like V
[ __ VP[-fin/ to] ]
leave
V
[ __ ]
  |
like′
    <EXP, EVENT> 
leave′
<AGT>
(141)
IP
NP
I′
I
I[+fin]
VP
  |
 V′
V
VP[+bare]
 |
  |
   would
 V′
  V[+bare]
  VP[-fin]
     |
   |
 like
  V′
    V[-fin]
VP[+bare]
     |
    to
VP[+bare]
PP
  |
 V′
  by noon
  |
 V[+bare]
  |
     leave
                                               
30 EXP stands for the Experiencer semantic role.
A step-by-step introduction to GB
November 1998
48
There are two problems for this analysis within the GB framework.  First, we have said earlier
that to is I[-fin].  Allowing to to be either I[-fin] or V[-fin] is possible, but less constrained.  The
second and greater problem is that there is no place for the semantic role for the external
argument of leave to be assigned.
Considering more data will aid us in finding a better proposal:
(142)
I would hate to be chosen by the committee.
(143)
I would like to be appreciated by someone.
We need to take into account the meaning of these sentences, our assumptions about the
linking of semantic roles to syntactic positions (see section 8.1), and the fact that (142)-(143)
are synonymous with (144)-(145).
(144)
I would hate for the committee to choose me.
(145)
I would like for someone to appreciate me.
Based upon examples (144)-(145), we can propose that the subcategorization frame for
like is instead:
like
V
[__ CP[-fin] ]
         |
like′
   <EXP, PROP>
This allows the presence of a full embedded clause so that passive can take place within the
lower clause, as in (142)-(143), and subjects may be expressed, as in (144)-(145).
A step-by-step introduction to GB
November 1998
49
The D-structure for (144) is straightforward, as shown in (146).
(146)    D-structure
IP
NP
I′
 I 
I[+fin]
VP
  |
 V′
V[+aux] VP[+bare]
 |
  |
     would
 V′
V[+bare] CP[-fin]
 |
   |
      hate
  C′
C[-fin]
IP[-fin]
 |
for       NP
I′
  the committee
I[-fin]
VP[+bare]
 |
  |
 to
 V′
   V[+bare]
NP
    |
  choose
me
But this subcategorization is still not the full story for (142)-(143), nor does it answer the
question of how we generate (138)?  Compare:
(147)
*I would like for me/myself to leave by noon.
The standard Transformational Grammar account of (138) is called Equi Deletion, where
(in GB terms) the D-structure for (138) has a full CP[-fin] as the complement to like, as in (147),
and then the complementizer for and the subject NP delete when the NP is coindexed with the
subject of the main clause.
GB chooses another way, because it does not like deletions.  This option is to assume the
positions were never filled since we have the possibility of using an empty category for the
coindexed subject NP:  PRO.  The C[-fin] is always null when PRO is the subject, but it is
A step-by-step introduction to GB
November 1998
50
always for when there is an overt subject (e.g., *I would hate the committee to choose me).31
The null C[-fin] is not a governor, so the specifier of IP[-fin] would not be governed, nor would it
receive Case.  This exactly meets the requirements for PRO (as seen in section 10.2) but is
disallowed for overt NPs.
The D-structure for (138) is shown in (148).  The S-structure would be identical, except
that the auxiliary would in the main clause would have moved to the I[+fin] position (as
discussed in section 7).
(148)
    D-structure
IP
NP
I′
 Ii
 I[+fin]
    VP
|
     V′
    V[+aux]
VP[+bare]
     |
  |
would
 V′
  V[+bare]
  CP[-fin]
     |
     |
 like
   C′
    C[-fin]
      IP[-fin]
     |
     0
      NP
       I′
    PROi
  I[-fin]
VP[+bare]
   |
  to
VP[+bare]
    PP
   |
  V′
      by noon
   |
  V[+bare]
   |
     leave
                                               
31 Some speakers prefer I would like someone to appreciate me over the version with the complementizer for present, as in
(145).  In this case, like would have to subcategorize for an IP[-fin]  as in the tree (156) for expect, as well as the CP[-fin/0]
shown in (148).
A step-by-step introduction to GB
November 1998
51
To review how Passive movement interacts with these equi constructions, the derivation
for (142) is shown in (149).
(149)
D-structure
S-structure
IP
IP
   NP
I′
   NP
 I′
   ⇒
    Ii
   I[+fin]     VP
    Ii
  I[+fin]
 VP
|
   |
   |
     V′
    wouldk
  V′
V[+aux]
 VP[+bare]
V[+aux]
  VP[+bare]
  |
   |
  |
    |
would   V′
 tk
   V′
V[+bare]     CP[-fin]
   V[+bare] CP[-fin]
  |
 |
|
  |
hate
C′ 
    hate        C′
   C[-fin]     IP[-fin]
C[-fin]
   IP[-fin]
    |
     |
 |
    0
     I′
 0
  NP
   I′
    I[-fin]       VP[+bare]
PROi
I[-fin]     VP[+bare]
|
  |
  |
  |
     to
 V′
 to
 V′
     V[+bare]  VP[+pass] 
    V[+bare]
 VP[+pass]
 |
   |
 |
   |
be
  V′
be
  V′
V[+pass]   NP
 PP
   V[+pass]   NP      PP
 |
    |
      |
chosen 
PROi by the
   chosen
     ti 
by the
     committee
  committee
Compare the D-structure in (149) with that given for the synonymous sentence (144) in (146).
PRO is filling the same semantic role in (142) as the overt pronoun me is in (144), so they both
begin as the object of choose.  But, even though the passive verb chosen cannot assign Case to
PRO, it can still govern it, forcing PRO to move to meet the Binding Principles by S-structure.
A step-by-step introduction to GB
November 1998
52
GB claims that the various types of predicates which take embedded clause complements
can be accounted for with the correct subcategorization frame.  In the case of nonfinite clause
complements, either PRO or A-movement (raising) will be involved when there is no overt NP
in the embedded subject position in the surface string.  The key distinction between the
A-movement constructions (known earlier as Subject-to-Subject Raising and analyzed in
section 8.3) and the constructions involving PRO (also known as Subject-Subject Equi) is
whether or not the main predicate assigns a semantic role to its subject position.32  Idiom
chunks and the possibility of the dummies it and there provide ‘tests’ to help determine this.
Consider:
(150)
a.
John is likely to win the race.
b.
The roof is likely to cave in.
c.
It is likely that John will win the race.
d.
There is likely to be no solution to her dilemma.
(151)
a.
John tried to win the race.
b.
?The roof tried to cave in.
c.
*It tried that John will win the race.
d.
*There tried to be no solution to her dilemma.
The predicate is likely is a raising predicate that does not assign a semantic role to its subject
position; either of the dummy NPs can fill the position (150c-d), or an idiom chunk can raise to
it and still maintain the idiomatic reading (150b), or a regular NP can raise to the subject
position (150a).  In contrast, try does assign a semantic role to its subject position and it also
requires that the subject of its embedded nonfinite clause complement be coreferent with its
own subject.  Thus, try subcategorizes only for a CP[-fin] that is headed by the null C[-fin] (as in
tree (148)), so the coindexed PRO will always be the subject of the embedded clause.33
Two other main types of predicates should be mentioned.  These were known in
Transformational Grammar as Object-Subject Equi and Subject-to-Object Raising.  Here it is
important to distinguish whether or not the main predicate assigns a semantic role to the NP
following it.  The dummy there and idiom chunk tests can be applied again, as well as checking
whether passive in the embedded clause yields a synonymous result.
(152)
a.
Sue persuaded Bill to fix the sink.
                                               
32These distinctions seem to apply cross-linguistically in that predicates meaning seem or likely will not assign a
semantic role to their subject position in any language.  Whether they raise the lower clause subject to the main clause
depends on the syntax of the particular language, however.  Similarly, predicates with the same meaning as try will assign
a semantic role to their subject position and require that the subject of the embedded clause be coreferent with the main
clause subject.  Depending upon the binding conditions of the language, however, the coreferent subject may or may not be
PRO.  For example, the VSO language, Quiegolani Zapotec, simply allows the main clause subject to be missing in
‘raising’ constructions and overtly repeats the coreferent subject in ‘equi’ constructions (Black 1994:Ch. 4-5).
33Note that in both raising and equi constructions, apparent long-distance binding of reflexives is allowed:
(i) Bill seems to like himself.
(ii) Bill tries to humble himself.
This is because either the coindexed trace or the coindexed PRO acts as the antecedent for the reflexive within the lower
IP, which is the governing category.
A step-by-step introduction to GB
November 1998
53
b.
?Sue persuaded the sink to be fixed by Bill.
c.
*Sue persuaded there to be no solution to her dilemma.
d.
?Sue persuaded the roof to cave in.
(153)
a.
Sue expected Bill to fix the sink.
b.
Sue expected the sink to be fixed by Bill.
c.
Sue expected there to be no solution to her dilemma.
d.
Sue expected the roof to cave in.
It should be clear from (152) that persuade does assign a semantic role to the NP following
it, so it subcategorizes for both an NP and a CP[-fin] headed by the null C[-fin], which will have a
PRO subject that is coindexed with the object of the main clause. The tree for (152a) is given
in (154).
(154)
IP
      NP
 I
     Sue
I[+fin]
VP
  |
 V′
V
      NP     CP[-fin]
 |
|
      |
         persuaded     Billi      C′
C
  IP[-fin]
 |
 0
   NP
  I′
 PROi
   I[-fin]
VP
    |
  |
   to
 V′
V
  NP
 |
fix       the sink
In contrast, the examples in (153) show that expect does not assign a semantic role to the
NP following it, which would lead us to believe that expect simply subcategorizes for an IP[-fin]
complement.  Data such as (155) led to the Transformational Grammar proposal that the
subject of the lower clause subsequently moves to the object position in the main clause.
(155)
Sue expects herself to give a perfect performance.
A step-by-step introduction to GB
November 1998
54
Movement to a complement position is not allowed in GB, since the complement position
can only exist if it is subcategorized for, and therefore filled, at D-structure.  Instead, Principle
A of the Binding Theory and the Case Filter can be met without movement, as shown in the D-
structure tree for (155) given in (156).
(156)
IP-1
       NP-1
I′
       Suei
     I[+fin]
VP
  |
 V′
    V
     IP-2[-fin]
     |
      expects    NP-2
   I′
 herselfi
  I[-fin]
VP
   |
  |
  to
 V′
V
NP
  |
     give
  a perfect
performance
In this tree structure, the NP-2 herself is governed by the verb expects (since IP-2 is excluded
from blocking government by clause (c) of the definition).  Further, expects is not a passive
form and it does assign a semantic role to its external argument (subject), so it can assign
accusative case to NP-2 (e.g., Sue expects me to give a perfect performance), allowing it to
pass the Case Filter.  Finally, NP-2 can act as if it is part of the main IP-1 for the Principles of
Binding Theory:  since its governor is above IP-2, the governing category in which the
coindexed antecedent for NP-2 must be found is IP-1, legalizing the reflexive in NP-2.
This concludes our introduction to most of GB theory as developed through 1986.34  The
next section deals with more recent developments in the theory, many of which were necessary
due to consideration of other languages, especially non-Indo-European languages.
                                               
34In Barriers (Chomsky 1986), Chomsky reformulates the definition of government in terms of barriers.  Discussion of
barriers, subjacency, bounding theory, and the Empty Category Principle has been omitted from this introductory series.
The reader is referred to the original sources or to textbooks (such as Haegeman 1994) for more information.
A step-by-step introduction to GB
November 1998
55
11. More recent additions to the theory
After the theory seemed fairly adequate for English, attention shifted to how it would
account for other languages.  Many other linguists joined in the task of analyzing phenomena
not seen in English to determine the coverage of the theory and to propose needed
modifications or extensions.
This section covers some of the additions to the basic GB theory that were proposed in the
late 1980s and early 1990s.  First, we return to the issue of how VSO or OSV word order can
be obtained from a configurational phrase structure.  Section 11.2 then introduces the
additional functional projections (besides IP and CP) which have been proposed.  Finally,
section 11.3 discusses several ways of dealing with the interaction between morphology and
syntax.
11.1 Phrase structure for languages with VSO/OSV word order
In section 6 we saw that the phrase structure of all the basic word orders except VSO or
OSV can be generated by simply changing the order of the elements on the right side of the
two basic X-bar phrase structure rules:
XP  →   Specifier  X'
X'   →   X0  Complements (=YP*)
We are now able to understand the proposals for VSO or OSV word order which allow
these languages to have an underlying configurational structure (rather than a flat structure)
like the others.  The example data is repeated here for reference.
One of the many languages exhibiting VSO word order, Quiegolani Zapotec, an
Otomanguean language spoken in Mexico (Regnier 1989, Black 1994), is exemplified in (157)-
(159).35
(157)
W-eey Benit
mël.
C-take Benito
fish
‘Benito took a fish.’
(158)
W-nii men disa
lo
noo.
c-speak 3RD
language face 1EX
‘She spoke Zapotec to me.’
(159)
xnaa noo
mother 1EX
‘my mother’
                                               
35 Abbreviations:  C=completive aspect; 3RD=general third person pronoun; 1EX=first person exclusive pronoun.
A step-by-step introduction to GB
November 1998
56
Urubú, of the Tupí family in Brazil (Derbyshire and Pullum 1981 from Kakumasu 1976),
provides data from one of the very rare OSV languages, shown in (160)-(161).
(160)
Pako xuã u’u.
banana John he-ate
‘John ate bananas.’
(161)
Koƒ
sepetu-pe
jurukã Nexƒ mái
muji-ta.
tomorrow spit-on
ribs
Nexƒ 
mother
she-will-roast
‘Nexƒ’s mother will roast the ribs on the spit tomorrow.’
Two main proposals have been made to account for VSO surface order36 from an
underlying configurational structure where the verb and its complements form a constituent
distinct from the subject, which is in a specifier position.  Each proposal works well for
particular VSO languages but makes incorrect predictions for others.37
11.1.1 Subject Adjunction
The Subject Adjunction proposal was developed for Chamorro in Chung (1990) and was
originally proposed by Choe (1986) for Berber.  An underlying SOV structure is assumed.
The surface order is obtained by movement of the subject down to adjoin to the right of the
verb, leaving behind a coindexed null expletive (i.e. a null form of the dummy it).38  The D- and
S-structures under this proposal are shown in (162).39
(162)
D-structure
S-structure
IP
 IP
⇒
I′
      NP
I′
        NP
 I0  +tense
    VP    subject
   
I0   +tense
  VP
 expli
      +agr
       |
   
     +agr
    |
      V′
      
   V′
   
 
           V0
NP
      
  V0
     NP
            |
         verb
 object
V0
NPi 
   object
 
 |
    
    verb     subject
                                               
36OSV order can be obtained by taking the mirror image of either proposal.
37 For full argumentation, refer to the works cited with each proposal and/or Black  (1994:Sections 6.2, 9.3, 11.1).
38 The use of this questionable null element rather than a trace is necessary because downward movement is ruled out by
the Empty Category Principle (ECP), which basically requires that, in addition to being governed by a proper head governor
(including lexical heads and those functional heads allowed by the specific language), a trace must be governed by the
moved element that it is coindexed with.  Since clause (b) of the definition of government (in section 10.1) requires that the
governor M-command the category in question, movement downward into another maximal projection is ruled out.