Book An Introduction To Generative Grammar With Exercises
Book An Introduction To Generative Grammar With Exercises
1 Constituent structure 7
1.1 The grounds for constituent structure . . . . . . . . . . . . . . . . . . . . . . . . 7
1.1.1 Ambiguity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2 Making a hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3 Testing the hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3.1 Case 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3.2 Case 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3.3 Case 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4 Taking stock . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.5 Beyond syntactic ambiguity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.5.1 Lexical ambiguity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.5.2 Vagueness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2 X-bar Theory 17
2.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2 The bare bones of X-bar theory . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2.1 Complements vs. adjuncts: transparency to extraction phenomena . . . . 20
2.2.2 Complements vs. adjuncts: extraction from weak islands . . . . . . . . . 20
2.2.3 Complements vs. adjuncts: word order . . . . . . . . . . . . . . . . . . 21
2.2.4 Complements vs. adjuncts: ellipsis phenomena . . . . . . . . . . . . . . 22
2.3 Some refinements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.3.1 Binary branching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.3.2 Functional categories . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.3.3 Sentences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3 C-command 37
3.1 Beyond precedence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.2 Binding theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.2.1 Anaphors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.2.2 Pronouns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3
4 CONTENTS
4 Transformational rules 59
4.1 Classical transformational grammar . . . . . . . . . . . . . . . . . . . . . . . . 60
4.1.1 Phrase structure grammar . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.1.2 Deep Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.1.3 Transformational component . . . . . . . . . . . . . . . . . . . . . . . . 64
4.1.4 Surface Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.2 Revising the framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.2.1 Phrase structure grammar . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.2.2 Deep structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.2.3 Transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.2.4 S-structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.3 The minimalist program . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.3.1 Basic components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.3.2 Lexicon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.3.3 Operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.3.4 Copy theory of movement . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.3.5 Economy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5 Empty categories 89
5.1 Traces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
5.1.1 The internal subject hypothesis . . . . . . . . . . . . . . . . . . . . . . . 91
5.2 Null pronouns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
5.2.1 Subjects of nonfinite verbs . . . . . . . . . . . . . . . . . . . . . . . . . 94
5.2.2 Null subject languages . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
5.3 Ellipsis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
5.3.1 Identity problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
5.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
6 Sentence syntax 99
6.1 Intransitive verbs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
6.1.1 Unergatives and unaccusatives . . . . . . . . . . . . . . . . . . . . . . . 99
6.1.2 Formal proposal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
6.2 Transitives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
6.3 Class changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
6.3.1 Passive . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
6.3.2 Impersonals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
6.3.3 Causative alternation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
CONTENTS 5
Constituent structure
We can easily appreciate that this utterance has two different interpretations:
(2) a. Old men and old women will receive a money reimbursement.
b. Women and old men will receive a money reimbursement.
This is a typical case of syntactic ambiguity, which we can describe in informal terms saying that
the adjective old is related to the conjunct of men and women in the former case, and saying that
it is only related to men in the latter. Yet, even these informal terms presuppose a hierarchical
relation, for we need to relate old to another unit resulting from the combination of men and
women:
One can think of a simple rule for saying that an adjective modifies the first noun to its right,
which works fine for (3)-a , but then we need the rule to identify β as ‘the first noun to its right’,
which cannot be done on purely linear terms.
Moreover, if we allow such a rule to affect any noun to its right, we would get into trouble
with examples like the following:
(4) I am a 22 year old man and women my age do not like me.
1
On the differences between syntactic ambiguity, lexical ambiguity, and vagueness, see 1.5.
7
8 CHAPTER 1. CONSTITUENT STRUCTURE
Here, it is pretty obvious that old just combines with man, whereas women is a different sentence.
A surface linear analysis not taking constituent structre into account would predict the combination
of [ old [ man and women ] ] to be possible, against fact.
Let’s take again the case considered before, where we are making the hypothesis that a
different constituent structure is associated to each interpretation. In the first case, the adjective
forms a constituent α with the noun to its right, but not with the constituent β:
In the second case, the adjective forms a constituent a constituent α with a constituent β containing
both nouns:
These two proposal are reasonable for they explain why the adjective old only affects men in (5),
but men and women in (6). Now, we must test the hypothesis.
1.3.1 Case 1
Take the first meaning and its associated structure of the sentence introduced above, namely the
one where the adjective old only affects men:
We can predict that if α is a real constituent –say a noun phrase– it will be affected as a whole
by syntactic operations like movement. For instance, since the order of the coordinates by and
is indifferent (Mary and John came = John and Mary came), we can predict that the change of
order between old men and women will have no impact on the meaning of the sentence:
This is indeed the case: (8) retains the meaning where old just affects men.
We can also expect this structure to associate with certain continuations ((9)-a), but not with
others ((9)-b):
(9) [ α Old men ] and [ β women ] will receive a money reimbursement, so. . .
a. young men will be discriminated again.
b. #young women will be discriminated again.
Obviously, since old is not affecting women, the continuation in (9)-b is a blatant contradiction.
Since the predictions are correct, we could confirm that the hypothesis is on the right track,
namely that old men do form a constituent.
Consider now, the other meaning and its associated structure, namely the one where both men
and women are old:
Here, rearranging the order of men and women will have no impact, and the same meaning will
be obtained:
As predicted from the structure, since the adjective affects both nouns, we are excluding from
the reimbursement both young men and young women.
Moreover, note that the syntactic structure is a typical case of distributive relation:
(13) A × (B + C) = A × B + A × C
As a consequence, we can predict that the following two sentences will be synonymous:
Hence, since the predictions we made for this analysis are confirmed, we can confirm our
hypothesis that the constituent structure associated for this meaning is [α Old [β men and
women] ].
1.3.2 Case 2
Consider now another clear case:
Again, we have two interpretations –Mary hit the man who was holding a hammer or Mary used
a hammer to hit the man–, each associated to a different dependency of the prepositional phrase
[=PP] with a hammer, on the noun man or on the verb hit:2
Let test each proposal in turn. If (17)-a is correct, we expect it to be able to move (18) or
pronominalize (19) as a whole:
(18) a. [ DP The man [ PP with a hammer ] ] , Mary hit, not the one with a gun.
b. [ DP The man [ PP with a hammer ] ] was hit by Mary.
c. It was [ DP the man [ PP with a hammer ] ] that Mary hit.
(19) Mary hit him, namely [ DP the man [ PP with a hammer ] ] .
Hence, the predictions are fulfilled, so we can safely conclude that the man with a hammer forms
a constituent in the studied case.
Now, consider the other analysis, which takes the PP with a hammer to be a manner adverbial
depending on the verb, not on the noun:
We can predict that the DP and the PP will be affected by syntactic operations independently, as
for instance movement (topicalization, passivization and clefting, respectively):
(21) a. [ DP The man ] , Mary hit [ PP with a hammer ] , not the woman.
b. [ PP With a hammer ] , Mary hit [ DP the man ] , not with an umbrella.
(22) [ DP The man ] was hit [ PP with a hammer ] by Mary.
(23) a. It was [ PP with a hammer ] that Mary hit [ DP the man ] .
b. It was [ DP the man ] that Mary hit [ PP with a hammer ] .
2
DP stands for Determiner Phrase, namely a nominal constituent headed by a determiner. Determiners include
articles (the woman), possessives (her book), and demonstratives (that woman).
1.3. TESTING THE HYPOTHESIS 11
Obviously, since pronouns cannot be modified (*I talked to him tall, *She with a dress was
Italian), we can safely conclude that the PP must depend on the verb, just as in the case of a
proper name:
This sentence is not ambiguous: the PP can only depend on the verb.
In conclusion, if two units form a constituent, we expect them to be affected as a whole
by standard syntactic operations (movement, pronominalization, ellipsis). If the can be affected
independently, then they form two different constituents.
1.3.3 Case 3
Consider the following ambiguous sentence in Spanish:
As in the previous cases, the ambiguity stems from the dependency of the last unit, in this case the
adjective phrase (=AP) relucientes ‘shiny’. Let’s consider the first meaning, where the adjective
is a predicative related to the verb, and it is not dependent on the noun zapatos ‘shoes’:4
4
Note that Spanish allows changing the order of the two units under this reading:
5
Again, Catalan can pronominalize both constituents:
12 CHAPTER 1. CONSTITUENT STRUCTURE
units, they can be affected by syntactic operations separately. In turn, if they are forming a
constituent, they will be affected as a block.
Since the noun bank can have several meanings, the utterance in (34) is ambiguous, for instance
between a ‘finantial institution’ and a ‘sitting piece of furniture’. However, note that we are
not saying that the source of the ambiguity is the existence of several syntactic structures:
independently of the lexical meaning of bank, we will assign the very same structure to (34).
(35) TenseP
DP Tense’
D’ Tense VP
D NP + PAST /3 SG V
the N burned
bank
Note that this structure is identical to the one assigned to the non ambiguous sentence in (36):
DP Tense’
D’ Tense VP
D NP + PAST /3 SG V
the N opened
bank
14 CHAPTER 1. CONSTITUENT STRUCTURE
It is the semantic combination of open with its subject what forces us to select, or at least strongly
favors, the meaning of ‘finantial institution’.
1.5.2 Vagueness
A different source of difficulties comes from VAGUENESS, namely from lexical items with just
one meaning, but a very unspecified one. One typical case is polar adjectives like hot, tall,
big, and their opposites. The crucial problem here is that it is unclear what counts as ‘tall’
in general, for its meaning seems variable depending on the particular context of use, as the
following definition from the Cambridge Dictionary shows: “being higher than most other people
or things.”
Obviously, in Catalonia, we would agree that a 2 meters tall person is tall, whereas a 1.5
meters isn’t. But what about 1.8, 1.83 or 1.79 meters? Moreover, if we where a tribe of pigmies
or a basketball team, would these values apply equally to define a tall person? Therefore, the
adjective tall is semantically vague, and its standard value must be fixed by the context of use.
This is why we can say (37) without contradicting ourselves:
Namely, it is small for the standard of elephants, but it is still big for the standard of animals in
general.
This phenomenon of vagueness relaying on context-dependence is widespread. For example,
when we say It is raining, we implicitly restrict our assertion to the context of the utterance:
‘it is raining here now’. The same happens with the sentence Everybody was half sleep, where
the universal quantifier everybody is not taken literally as ‘all persons in the world’, but as ‘all
persons in the current relevant situation’. So then, if I utter this sentence in our Foundations’
class, everyone is intended as ‘everyone in the Foundations’ class.’
1.6 Exercises
E XERCISE 1 Are the following sentences ambiguous? Justify your answer applying standard
tests. Choose the language you prefer.
(38) Catalan
a. Hem parlat amb els estudiants de física.
b. Vam trobar els nens i les nenes molt bufones.
c. Quan et va explicar que havia fet aquella bestiesa?
d. Vam trobar les nenes i els nens molt bufons.
e. Quan et va oferir una explicació convincent del que havia fet?
(39) English
a. When did she tell you that she did such nonsense?
b. Old men and women are discriminated.
1.6. EXERCISES 15
c. When did she offer you an explanation that she did such nonsense?
d. Mary stabbed the man with the knife.
(40) Spanish
a. Hablamos con los alumnos de griego.
b. Necesitamos a los niños sanos.
c. Fueron a la reunión de espías.
d. Necesitamos mujeres y hombres valientes.
e. Puso los libros en la mesa.
(41) a. Where did she offer you the explanation that she had bought the car?
b. On et va donar l’explicació que havia comprat el cotxe?
c. ¿Dónde te dio la explicación de que había comprado el coche?
(42) a. She looked for good books and magazines.
b. Cercava bons llibres i revistes.
c. Buscaba buenos libros y revistas.
(43) a. Where shouldn’t you take a bath after having lunch?
b. On no t’has de banyar després de dinar?
c. ¿Dónde no te tienes que bañar después de comer?
16 CHAPTER 1. CONSTITUENT STRUCTURE
Chapter 2
X-bar Theory
2.1 Background
Once we have agreed on the necessity of constituent structure and the crucial role of the head
of the constituent, we must design a system for representing constituents relations that is both
theoretically and empirically motivated. Here are the relevant facts we must account for:1
• Differences between the relation of heads with complements and with the adjuncts.
• Homogeneity: as a null hypothesis, we expect the same representation for all the categories.
If this is not the case, we must offer compelling evidence against.
• Simplicity: we must think of a system that is reasonably implemented as part of the innate
linguistic component (UG), and, hence, suitable for any natural language.
Chomsky (1970) reconsidered some amassed evidence on the close similarities across categorial
domains, which had been analyzed as instances of syntactic rules. He focused on examples like
the following:
17
18 CHAPTER 2. X-BAR THEORY
These examples suggest that (some of) the properties of the verbal head are inherited by the
nominal or adjectival derivative. Yet, instead of proposing syntactic rules for explaining these
similarities, Chomsky argued that they might follow from a common structural template to which
categories should conform uniformly, namely X-bar theory.
X-bar theory builds phrases from heads in three levels: the head level (X-zero, X0 ) , the intermediate
level (X-bar, X’), under where we find the head and its complements, and the maximal or phrase
level (X-phrase, XP), under which we place material not required by the head, namely adjuncts
and specifiers:
(6) XP
specifier X’ adjunct
X0 complement
In more detailed proposals, a distinction is introduced between the specifier and adjuncts:2
2
Indeed, in generative syntax, the concept adjunct is a positional one: elements can move into a predefined
specifier position, or they can adjoin to a maximal projection. Hence, as a rule, phrases have just one specifier,
whereas they admit as many adjuncts as necessary.
2.2. THE BARE BONES OF X-BAR THEORY 19
(7) XP
XP adjunct
specifier X’
X0 complement
This schema allows us to express the parallelism between categories quite easily:
AdvP A’
very A PP
interested P’
P NP
in poker
NP N’
her N PP
interest P’
P NP
in poker
Besides assuming a uniform phrase structure for all lexical categories (and functional categories
in later developments, see below), the crucial innovation of the system is encoding in the structure
the traditional functional distinction between complements and adjuncts (Eng. adverbials, Catalan
circumstancials, Sp. circunstanciales). Hence, from its position in the structure, we can know
the function of any constituent:
b. VP
V’ PP
V DP P’
buried D’ P DP
D NP in D’
the N D NP
corpse the N
garden
The DP the corpse is under X’, so it is the complement of buried, whereas the PP in the garden
is under XP, so it is an adjunct.
(11) complements
a. What do you think [ S that Mary bought t]?
b. Where do you think [ S that Mary bought the book t]?
(12) adjuncts
a. *What do you buy a book [ S while Mary bought t]?
b. *Who do you buy a book [ S while t bought a dress]?
Since the adjunct blocks any kind of extraction, it is called a STRONG ISLAND. Schematically:
(13) complements
2.2. THE BARE BONES OF X-BAR THEORY 21
Since these sentences are islands for adjuncts only, but allow extraction of complements, we
call them WEAK islands.
Finally, remind the case of verbal ellipsis with do so. This pro-verb can represent different parts
of the VP, with the exclusion of the verbal head (17)-a:
(17) Mary will put the dishes in the dishwasher this afternoon and. . .
a. *I will do so the glasses in the sink tonight.
b. I will do so in the sink tonight.
c. I will do so tonight.
d. I will do so too.
We are making the hypothesis that the structure of both coordinate sentence is the following (on
the iteration of the V’ projection, see section 2.3.1):
(18) a. Mary will put the dishes in the dishwasher this afternoon and. . .
b. VP
V’ DP
this afternoon
V’ PP
V DP in the dishwasher
put the dishes
Let’s see how do so ellipsis affects this VP structure. Firstly, we can appreciate that it doesn’t
target heads:
(19) *Mary will put the dishes in the dishwasher this afternoon and I will do so the glasses
in the sink tonight.
2.2. THE BARE BONES OF X-BAR THEORY 23
Now, we move to the higher constituent, the lower V’, and we can see that ellipsis may target
this intermediate projection, so that the verb and the lower complement can delete, leaving the
rest of the constituents in the VP unaffected:
(20) Mary will put the dishes in the dishwasher this afternoon and I will do so in the sink
tonight.
Now, we target the higher V’, hence we delete the verb and both complements, leaving the
adjunct hanging from VP unaffected:
(21) Mary will put the dishes in the dishwasher this afternoon and I will do so tonight.
24 CHAPTER 2. X-BAR THEORY
Finally, we target the VP, and as a consequence all the material is deleted:
(22) Mary will put the dishes in the dishwasher this afternoon and I will do so too.
Crucially, in all the good examples, the elided constituent is a verbal projection higher than
the verbal head. When we try to delete a non constituent, the result is bad. For example, we
cannot delete the verbal head and the higher complement in the dishwasher without deleting as
well the lower direct object:
(23) Mary will put the dishes in the dishwasher this afternoon and I will do so the glasses
tonight.
2.2. THE BARE BONES OF X-BAR THEORY 25
This is nicely predicted by our system: if you want to delete the higher V’, you must also
delete the lower one, but not conversely (20).
The same happens when we target the verb and the adjunct:
(24) *Mary will put the dishes in the dishwasher this afternoon and I will do so the glasses
in the sink.
(25)
Again, this follows from our system: if you want to delete the VP, you must also delete the V’,
but not conversely (20)-(21). Hence, if ellipsis with do so targets the verb and an adjunct, then it
necessarily affects the complement.
26 CHAPTER 2. X-BAR THEORY
The first refinement was limiting the branching possibilities to two. At the time, this was a
simplicity and elegance matter, for instead of clumsy structures like (26), we would obtain
recursive representations like (27):
(26) VP
V’ DP
this afternoon
V DP PP
(27) VP
V’ DP
this afternoon
V’ PP
V DP in the dishwasher
put the dishes
The second structure fits much better with our current understanding of the way we build sentences,
namely bottom-up. We begin with the verbal head:
(28) V
put
(29) V’
V DP
(30) V’
V’ PP
V DP in the dishwasher
put the dishes
(31) VP
V’ DP
this afternoon
V’ PP
V DP in the dishwasher
put the dishes
(32) VP
VP DP
in the kitchen
VP DP
this afternoon
V’ PP
V DP by hand
Obviously, this refinement is crucial for operations sensitive to c-command conditions, for while
we clearly have symmetric c-command between the two complements in the non-binary structure
in (26), it is open to discussion whether this is the case in (27). See chapter 3.
Besides the classical lexical categories (Noun, Adjective, Verb, Preposition, and Adverb), syntax
has included a range of categories without lexical content, but encoding grammatical information:
FUNCTIONAL CATEGORIES ; see Bosque & Gutiérrez-Rexach (2009, 3.1.3). The first proposal to
incorporate these elements to the X-bar schema was developed in the mid eighties, and affected
NPs. The proposal was that the determiner (article or demonstrative) was not the specifier of the
NP (33), but rather a head selecting the NP (34):
(33) NP
Det N’
the N
book
2.3. SOME REFINEMENTS 29
(34) DP
D’
Det NP
the N’
book
This allowed to account for predeterminers more easily:
(35) DP
all D’
Det NP
the N’
book
A similar proposal was applied to degree words (Degree Phrase or DegP) or quantifiers (Quantifier
Phrase or QP)
Q’
Q NP
many DegP N’
Deg’ N
Deg AP books
very A’
nice
30 CHAPTER 2. X-BAR THEORY
2.3.3 Sentences
The next affected element was the sentence, which represented an anomaly in the system, for it
is a constituent without a head:
(38) S
DP VP
Mary V
sleeps
This was solved in the late nineties, when Jean-Yves Pollock Pollock (1989) proposed that
the agreement and tense properties of the sentence, which were commonly represented as a
node AUX(ILIARY) hanging from the S, were the head of the sentence and projected a phrase
structure under the guidelines of the X bar theory. Hence, the structure of the sentence was
redesigned from (39) to (40):3
(39) S
DP AUX VP
Mary present/3sg V
sleeps
(40) TenseP
DP Tense’
Mary Tense VP
present/3sg
sleeps
The same idea was applied to the left periphery of the sentence in subordinate clauses, which
became projections of the subordinator particle (COMPLEMENTIZER), which projected a Complementizer
3
Actually, Pollock’s proposal was a bit more complex, for he proposed two projections: Agree(ment)P and
TenseP, the former dominating the latter. Other authors represent the sentence as a projection of Inflection
(=agreement+tense), namely InflP or IP.
2.3. SOME REFINEMENTS 31
Phrase (CP):
(41) CP
C’
TenseP
DP Tense’
Mary Tense VP
present/3sg
sleeps
(42) TenseP
DP Tense’
I
Tensepresent/1sg VP
V’
V CP
think C’
C TenseP
that
DP Tense’
Mary Tense VP
present/3sg
sleeps
32 CHAPTER 2. X-BAR THEORY
(43) CP
DPi C’
what
V+T TenseP
did
DP Tense’
Mary Tense VP
past/3sg
V ti
buy
In the nineties, further developments of sentence structure have been proposed, leading to a
explosion of functional projections which were designed to encode several grammatical categories.
We will consider them briefly in chapter 7.
2.4 Conclusions
The main advantage of X-bar theory is proposing a very simple, yet powerful pattern for all
syntactic structures, so that we can build very complex sentences with very few structural positions
which repeat themselves as many times as necessary. From the point of view of computational
processing and cognitive effort this seems a welcome result.
Moreover, such recursive pattern seems widespread in the natural world, where it is studied
under the label of FRACTAL RECURSIVITY:
Just compare these cases with a recursive syntactic tree:
2.5 Exercises
E XERCISE 3 Given the restrictions on phrase structure following from Kayne (1994), what
would be the explanation for the impossibility of a tree like the following:
4
You can note that the head position of CP is typically occupied by a verbal head in interrogative sentences, be
it the first auxiliary present in the structure (Mary has bought a book: Has Mary bought the book? What has Mary
bought?) or by an inserted do auxiliary, as in the example.
2.5. EXERCISES 33
E XERCISE 4 Given the restrictions on phrase structure following from Kayne (1994), why
should multiple adjuncts (in red in the tree) not be allowed?
*VP
XP VP
YP VP
V DP
34 CHAPTER 2. X-BAR THEORY
C-command
As the subindex shows, we can understand that him=Peter in (1)-a, but not in (1)-b.1 On the
basis of this contrast, we can make the following generalization:
(i) Johni arrived late. I was slept, but I am pretty sure that hei was drunk.
37
38 CHAPTER 3. C-COMMAND
Here, Peter can either precede or follow the pronoun, but the relation is possible in both cases.
Again, something similar happens in Romance languages:
(7) Spanish
a. La mujer que rechazó a Pedroi loi invitó a un café.
the-F woman that rejected to Peter him invited to a coffee
b. La mujer que loi rechazó invitó a Pedroi a un café.
the-F woman that rejected him invited to Peter to a coffee
(8) Catalan
a. Les coses que va dir la Joanai no lai van perjudicar.
the-F. PL things that PAST.3 SG say the Joana not her PAST.3 SG penalize
b. Les coses que ∅i va dir no van perjudicar la Joanai .
the-F. PL things that PAST.3 SG say not PAST.3 SG penalize the Joana
To account for these cases, (Langacker, 1969, 167) introduced a new relation regulating the
antecedent-pronoun link: COMMAND.2 However his formulation wasn’t accurate enough, and it
become substitued by c-command, where c stands for constituent or category:
Definition 1 C-command =def α c-commands β iff neither α nor β dominates the other, and the
first branching node γ that dominates α dominates β.
C-command can help us explain the above data by means of the following generalization:
(9) S1
NP VP
Peteri V’
V DP
hated
D NP
the
NP S2
Here the pronoun does not c-command its antecedent, for the first branching node dominating he
is not dominating Peter. On the other hand, Peter c-commands the pronoun: the first branching
node immediately dominating Peter –namely S1– also dominates the pronoun.
Consider now, the situation when the position of the antecedent and the pronoun gets reversed:
(10) * S1
NP VP
Hei V’
V DP
hated
D NP
the
NP S2
N’
who Peteri rejected
woman
He commands Peter: the first branching node that immediately dominates he –namely S1– also
dominates Peter, hence the sentence violates Generalization 2.
Now consider the following case:
40 CHAPTER 3. C-COMMAND
(11) S1
DP VP
D NP is hated by himi
the
NP S2
N’
who rejected Peteri
woman
In this configuration, neither the pronoun nor Peter c-command each other: as a consequence,
Generalization 2 is satisfied.
Finally, take the same structure, but with the position of the pronoun and the antecedent
reversed:
(12) S1
DP VP
D NP is hated by Peteri
the
NP S2
• reflexives and reciprocals (‘anaphors’): himself, herself, itself, themselves, myself, yourself,
ourselves, yourselves; each other, one another
• non-reflexive pronouns (‘pronominals’): he, she, it, him, her, I, us, you, me, his, your, my,
our
• full NPs including names (‘r(eferential)-expressions’): the baroness, Peter, this, a disinherited
Russian countess,. . .
The idea is that only lexical NPs have inherent reference, whereas anaphors and pronominals
must obtain their reference from an antecedent. The syntactic conditions regulating this process
is what we call BINDING THEORY Büring (2005); Reuland (2006); Sportiche (2013); Truswell
(2014), where binding entails c-command and a sharing referential index.
Let’s consider the simplest case, anaphors.
3.2.1 Anaphors
Anaphors need an antecedent to obtain reference, and the antecedent must bind them:3
NP T’
Johni Tense VP
V’
V DP
likes himselfi
(15) a. *Himselfi likes Johni .
3
The same happens with reciprocals:
b. TP
NP T’
Himselfi Tense VP
V’
V DP
likes Johni
Here, c-command and precedence go hand in hand: the antecedent precedes and c-commands
the anaphor in (14) but not in (15). So we need one critical case where the antecedent precedes,
but does not c-command, the anaphor. Consider now the following case:
NP1 T’
likes himselfi
John precedes, but does not c-commands, himself : the first branching node dominating John is
NP1 , not TP, hence it is not c-commanding what TP dominates, but only what NP1 does. As a
consequence, our Binding principle A, based on c-command is doing a good job.
However, we must refine it a bit, for the antecedent of the anaphor cannot be too far:
So, the domain where the antecedent binds the anaphor is important.
Note that we have generally assumed that the relevant domain for binding principles to apply is
the sentence, but consider the following case:
If the domain for the application of principle A were the whole sentence, we would expect John to
be able to bind the anaphor himself within the object DP. However, the only possible antecedent
is the possessor Mary’s. This suggests that in this case, the relevant domain is the DP.
3.2.2 Pronouns
Let’s move now to pronouns. In general, pronouns are OK in the contexts where anaphors are
not, and are bad, where anaphors are fine.
As a consequence, we can formulate the principle regulating the referential dependence of pronouns
as follow (‘cannot be bound’ = ‘must be free’):
As mentioned in the case of anaphors (see 3.2.1), the relevant domain for binding may be a
DP:
(23) Johni likes very much Maryk ’s book about himi /*herk .
If the domain for the application of principle A were the whole sentence, we would expect John
to be unable to bind the pronoun him within the object DP. This suggests that in this case, the
relevant domain is the DP, and the pronoun him is free in this domain.
In all the cases in (24) the referential expression is bound an antecedent, yielding a bad result.
44 CHAPTER 3. C-COMMAND
We can summarize the pattern obtained from the application of binding principles in the
following table:
(26) VP
V’
V XP YP
complement 1 complement2
(27) VP
V’
V’ YP
V XP complement2
complement 1
Take one particular case: ditransitive constructions in English (similar results obtain in Catalan
or Spanish):
Since we are dealing with complements, we would assume structures like the following:
3.3. ADDING BINARY BRANCHING TO THE PICTURE 45
(30) V’
V’ PP
V DP to the boys
introduced Mary
(31) V’
V’ PP
V DP to the boys
Yet, if we take our original definition of c-command, repeated here as (32), we must conclude
that the rightmost complement (to the boys) c-commands the leftmost one (Mary/to the girls):
(32) C-command =def α c-commands β iff α and β do not dominate each other, and the first
branching node γ that dominates α dominates β.
V’ PP
V DP to herselfi
introduced Maryi
• Since Mary does not c-command the PP in (30), we predict (34) to be possible, for Mary
is not binding the pronoun (but it is not):
(34) V’
V’ PP
V DP to heri
introduced Maryi
46 CHAPTER 3. C-COMMAND
• Since the girls does not c-command the PP in (31), we predict (35) to be wrong, for the
girls cannot bind the reciprocal (but it is OK):
(35) V’
V’ PP
V DP to each otheri
• Since the girls does not c-command the PP in (31), we predict (36) to be possible, for Mary
is not binding the pronoun (but it is not):
(36) V’
V’ PP
V DP to themi
(37) V’
V’ PP
V DP to Maryi
introduced heri
(38) V’
V’ PP
V DP to the girlsi
introduced theyi
Henceforth, binding data suggest that the direct object c-commands the indirect object in these
structures (see Barss & Lasnik 1986; Larson 1988). One solution to this problem was modifying
3.4. EXERCISES 47
the definition of c-command, changing the ‘first branching node’ requirement by ‘the first maximal
projection’. This was labeled ‘m(aximal)-command’
(39) m-command =def α m-commands β iff α and β do not dominate each other, and the
first maximal node γ that dominates α dominates β.
Given this, in our case the direct object m-commands the indirect object (and any adjunct within
the VP), which would explain the data in (38)-(39), and (40):
The object pronoun m-commands the referential expression within the adjunct, resulting in a
principle C violation.
Other scholars suggest a different line of analysis that introduces more structure in these
cases, so that the first complement is higher in the structure that the second, accounting for the
binding facts without modifying the definition of c-command. See Kayne (1984); Larson (1988).
These authors suggest structures like the following (irrelevant details omitted):
In this configuration, the direct object c-commands the PP, which explains the binding phenomena
just described.
3.4 Exercises
E XERCISE 5 Consider the following tree and choose the correct option:4
4
The definition of c-command is the following:
C-command =def α c-commands β iff the first branching node γ that dominates α dominates β.
48 CHAPTER 3. C-COMMAND
(41) TP
NP T’
Mary
T VP
NP VP
Mary
V’ PP
V NP P’
eats vegetables P NP
at home
E XERCISE 6 Consider the following tree and choose the correct option:
3.4. EXERCISES 49
(42) TP
DP T’
A neighbor of Mary’s
T VP
NP VP
DP V’
A neighbor of Mary’s V NP
eats N’
N PP
vegetables P’
P NP
from home
E XERCISE 7 Consider the following tree and choose the correct option:
CP
DP C’
did
DP T’
D’ T VP
D NP V’
the AP N’ V DP
elder N read
which book about Lisa
N PP
cousin P’
P NP
of Barnie’s
• which book about Lisa c-commands the elder cousing of Barnie’s, and read: TRUE /
FALSE.
• about Lisa c-commands the elder cousing of Barnie’s, and read: TRUE / FALSE.
• elder c-commands cousin and Barnie’s: TRUE / FALSE.
• elder c-commands read and Lisa’s: TRUE / FALSE.
• which book about Lisa c-commands Barnie’s: TRUE / FALSE.
• which book about Lisa c-commands about Lisa: TRUE / FALSE.
3.4. EXERCISES 51
E XERCISE 8 Consider the following tree. Can either Jane or Mary bind herself? Is there any
other antecedent available? Justify your answer.
CP
DP C’
did
DP T’
D’
T VP
D NP
the AP N’ VP AdvP
elder N V’ yesterda
N PP V’ PP
cousin P’ V DP P’
P send t P NP
P NP to herself
of Mary
52 CHAPTER 3. C-COMMAND
E XERCISE 9 Consider the following tree. Can either Jane or Mary bind herself? Justify your
answer.
CP
DP C’
C TP
which book about herself
did NP T’
Jane
T VP
V’
V’ PP
V DP P’
send t P DP
to D’
D NP
the AP N’
elder N
N PP
cousin P’
P NP
of Mary
3.4. EXERCISES 53
E XERCISE 10 Consider the following tree. Can Jane and her have the same reference? What
about Mary? Justify your answer.
CP
DP C’
did NP T’
Jane
T VP
V’
V’ PP
V DP P’
send t P DP
to D’
D NP
the AP N’
elder N
N PP
cousin P’
P NP
of Mary
54 CHAPTER 3. C-COMMAND
E XERCISE 11 Consider the following tree. Can she refer to the fronted NP Mary? Justify your
answer.
CP ]
NP C’
Mary C TP
NP T’
she
T VP
VP PP
V’ P’
V’ DP P DP
E XERCISE 12 Explain whether the following sentences satisfy Binding Principles under the
intended reading:
E XERCISE 13 How do we deal with the following cases, according to the binding principles
just introduced?
3.4. EXERCISES 55
E XERCISE 14 What will happen in the following sentences if we change the reciprocal/reflexive
by a pronoun?
E XERCISE 15 Adapted from (Büring, 2005, 5). In the following sentences, Φ designates an NP
with the index given. For each sentence, determine by intuition what Φ can/must be (there may
be more than one option in some cases). Then give the local clause and the antecedent for Φ and
demonstrate that the Binding Conditions are met.
5. Hermann4 tried to be nice, and Gallia quite liked Φ4 . Now Φ4 and Gallia go out to see a
mud wrestling show.
E XERCISE 16 Consider the following tree and describe the c-commanding domain of the elements
given:
C-command =def α c-commands β iff the first branching node γ that dominates α dominates β.
56 CHAPTER 3. C-COMMAND
(56) TP
DP T’
D’
D NP T VP
The N PP VP
books by James
V’ PP
P’
V DP
P NP
were-written D’
by him
D NP
the N PP
books by James
E XERCISE 17 According to the results of the previous exercise, answer the following questions:
E XERCISE 18 Consider the following tree and answer the questions, justifying your response:
• If we replaced him by himself, would the sentence be possible? Would the antecedent of
himself be the same as that of him?
TP
DP T’
D’
D NP
T VP
The N PP
books by James
VP PP
V’ P’
P NP
V’ PP by Peter
P’
V DP
P NP
were-send D’
to him
D NP
the N PP
books by James
E XERCISE 19 Consider the following tree and answer the questions, justifying your response:
• If we replaced herself by her, would the sentence be possible? Would the antecedent of
her be the same as that of herself ?
58 CHAPTER 3. C-COMMAND
CP
NP C’
herself
C TP
NP T’
Mary
T VP
V’
V’ PP
V NP P’
invited herself P NP
to POSS NP
Mary’s party
Chapter 4
Transformational rules
In the 50’s, syntax had been developed to a quite successful and complex system in America,
mainly by the work of Leonard Bloomfield (Bloomfield 1933). For Bloomfield, syntax was
based on a strict analysis in terms of constituent structure, and linguists where supposed to
construct grammars by analyzing the distributional properties of constituents. Namely, they
were interested in kind of elements that could alternate, say, in subject position, or the positions
that a constituents, say, an NP could fill. Crucially, this method was strictly formal, without any
resort to the meaning of sentences, for meaning was considered a very messy stuff not amenable
to formal analysis (Bloomfield 1943). The result, called DISTRIBUTIONALISM, was proven very
successful at building grammar for native American languages, where the main information
was obtained from informants by means of inquiries about combinations and distributions of
constituents without any reference to meaning. In a sense, it was a taxonomic approach, for it
just listed the sentences of a language, without caring too much about the processes leading to
forming these sentences beyond purely constituent structure. For distributionalists, languages
were sets of sentences recorded from corpora or by oral interviews and grammar was a set of
generalizations deduced from these sentences.
Yet, such a system was static, and based on formal superficial differences, and quite different
from the European comparative grammar tradition, which focused on the evolution and derivation
of forms from more abstract underlying forms. Note, for example, the common assumption that
the forms sing, sings, sang and sung are different realization of a common morpheme to sing.
Or the different realizations of the consonant b in Spanish burro ‘donkey’ ([’buro]) and abuelo
‘grandfather’ ([a’Bwelo]), which were linked to a common abstract form (PHONEME) /b/.
These different approaches were already discussed by the American syntactician Zelig Harris
(Harris 1957, 2013), who was Noam Chomsky’s professor and mentor:
59
60 CHAPTER 4. TRANSFORMATIONAL RULES
Harris notes the different perspective we can take when studying language: a more static one,
concentrating on descriptions at different layers (phonetics, phonology, morphology, syntax),
or a more dynamic one, concentrating on the processes linking different steps of a derivation.
This second line of research was pursued by Noam Chomsky in his TRANSFORMATIONAL
GRAMMAR .
The main tenet of Transformational Grammar is considering languages as generative devices,
namely a system of rules capable of generating the sentences of a given language. Here, the
emphasis is placed on the mechanisms underlying sentences, not on the sentences themselves.
The reason for this change of perspective was twofold. On the one hand, Chomsky remarked that
languages could create an infinite number of sentences, and people could understand sentences
that they have not heard before. Namely, the language faculty is CREATIVE. Obviously, it makes
little sense describe language as a set of sentences if this set is infinite. It makes more sense
assuming that the crucial part is the mechanism allowing us to create these sentences.
On the other hand, even though we could create an infinite number of sentences, we can do
it with finite means. We have a limited number of categories, of sounds, of positions in a tree.
Language is, thus, GENERATIVE.
• a transformational component, which took the syntactic constituents created by the phrase
structure grammar as an input and created new syntactic constituents,
• a phonetic component that converted syntactic structures into oral utterances, and
In the seminal book Syntactic Structures, Chomsky proposed the following architecture,
where Deep Structure was the output of the phrase-structure grammar and the input to semantics,
and Surface Structure was the output of transformational component and the input to phonetics:
1. S → DP VP
2. DP → D NP
3. NP → A N
4. VP → V
5. VP → V DP
62 CHAPTER 4. TRANSFORMATIONAL RULES
Moreover, they also have lexical insertion rules that assign a lexical item to any head:
1. S
2. DP VP
3. D NP VP
4. D A N VP
5. D A N V
6. a A N V
7. a sad N V
8. a sad girl V
However, PSGs where soon considered problematic. First of all, they generate ungrammatical
sentences (i.e. they OVERGENERATE). For example, our toy grammar generates:
Namely, as it stands, the grammar cannot distinguish intransitive and transitive verbs. In order to
solve this problem, Chomsky was forced to include rules sensitive to the context, namely to the
presence of certain features (Chomsky, 1965, ch.2):
4.1. CLASSICAL TRANSFORMATIONAL GRAMMAR 63
These rules avoid the generation of the illformed sentences above, but are still unable to
express the following contrast:
(3) a. Mary asked what time it was/the time/*that it was ten o’clock.
b. Mary said what time it was/the time/that it was ten o’clock.
c. Mary changed *what time it was/the time/*that it was ten o’clock on her clock.
Namely, the restrictions that verbs impose on their complements go beyond categorial selection
(subcategorization), but call for a semantic restriction as well. Hence, PSG must be also enriched
to encode this information.
The second problem of our PSG is that it cannot express the infinite capacity of human
language. Therefore, we must incorporate RECURSIVITY to solve the problem:
1. S → DP AUX VP
2. S’ → COMP S
3. NP → A N
4. NP → N S
5. VP → V S
(4) This intelligent girl will believe that the man would like a tall woman.
However, even though PSG can be enriched to avoid overgeneration and express recursivity, they
still have a problem of UNDERGENERATION, namely they are unable to generate grammatical
sentences. One case is auxiliary inversion in interrogatives:
One could try to enrich the PSG with the following rule:
(6) S’ → AUX S
64 CHAPTER 4. TRANSFORMATIONAL RULES
The basic idea is that at D-structure both active and passive sentences are identical (Brutus killed
Caesar), and differences arise in the transformational component, where a transformation rule of
passive applies, which affects the transitive sentence in three different ways:
Formally:
In AUX we can have either an auxiliary verb or tense features. We have the following
derivation, where -ed stands for past tense, and -en for past participle:
As you can easily appreciate, transformations in early TG where a very powerful mechanism to
generate new syntactic structures from the output provided by phrase structure rules. Note that
the passive transformation involves reordering, deletion , and insertion of constituents. As it will
become obvious in class discussion, this was positive for overcoming the limitations of phrase
structure rules (DESCRIPTIVE ADEQUACY), but it had a negative downside: transformations
like the one just discussed were too powerful (and baroque) to be a real explanatory solution.
Indeed, after Chomsky’s initial work, almost any complex syntactic phenomenon was assigned
a transformation rule, leading to a very complex grammar, which could correctly generate the
sentences of the language, but at the cost of very ad hoc rules. Hence, after the explosion of
transformation rules in the sixties, it became clear that they should be constrained by general
principles if we wanted TG to offer a real explanation of human language faculty.
passive
D-structure [N P 1 Brutus] [AU X -ed] [V kill-] [N P 2 Caesar]
passive rule yes
S-Structure [N P 2 Caesar] [AU X -ed] be+-en [V kill-] by [N P 1 Brutus]
morphonological rules yes
output Caesar was killed by Brutus
The active sentence is ambiguous between the following two interpretations, where Alice, John,
Candice, Peter, and Andreas are all the students in the context:
• Alice and John speak Italian, Candice and Peter, French, Andreas, Catalan.
• There is a Romance language, say Catalan, that Alice, John, Candice, Peter, and Andreas
speak.
However, in the passive sentence, the ambiguity disappears and only one reading survives,
namely:
4.2. REVISING THE FRAMEWORK 67
• There is a Romance language, say Catalan, that Alice, John, Candice, Peter, and Andreas
speak.
Yet, this is not supposed to happen if transformations are meaning-neutral. The only way to
explain this contrast is assuming that part of the interpretation takes place after the transformational
component, namely at S-structure.
This leads us to a major shift in the architecture of the system:
4.2.3 Transformations
The first attempts to restrict the power of transformations where formal conditions on the formulation
of rules, but soon it was evident that a stronger approach was needed, and X-bar theory was
crucial. Even though it was conceived as a scheme restricting the output of the PSG component,
it could be applied at the output of transformations as well, so that all syntactic structures must
conform to the same X-bar pattern, regardless of the fact that they were base-generated or the
result of a transformation.
Hence, a transformation wasn’t allowed to move a head to the specifier of an phrase or to
adjoin a phrase to a head, for the output would not comply with the X-bar schema:
68 CHAPTER 4. TRANSFORMATIONAL RULES
(11) a. *XP
Y X’
X YP
Y’
Y
b. *XP
X’
X YP
ZP X Y’
Y ZP
As a consequence, the only allowed movement operations were the following:
(12) a. XP
X’
X YP
Y X Y’
Y
b. XP
ZP XP
X’
X YP
X Y’
Y ZP
4.2. REVISING THE FRAMEWORK 69
c. XP
ZP X’
X YP
X Y’
Y ZP
Let us see some real examples. A typical case of head-adjunction is auxiliary inversion in
English, where the verbal adjoins to the complementizer:1
1
Remember that there is a layer above TP where we find conjunctions, and wh-words (i.e. interrogatives,
exclamatives and relatives). The head of this projection is C(omplementizer), and, hence, we have a CP.
2
Even though both cases are formally identical, the nature of the landing site of the movement will prove
important. In the case of passive, the specifier of TP is a typical argumental position, in this case one associated
with subject properties (e.g. nominative case, agreement with the verb). This is why this kind of movement is
called Argument-movement (A-movement, for short). In the case of wh-movement, the specifier of CP is a typical
non-argumental position, totally unrelated to the properties of the arguments of the verb. This is why this kind of
movement is called non-Argument-movement (non-A-movement or A’-movement, for short).
70 CHAPTER 4. TRANSFORMATIONAL RULES
(13) passive
a. John sent Mary a letter.
b. Mary was sent a letter.
c. A letter was sent to Mary.
d. *Mary a letter was sent.
Notwithstanding, some languages do allow having two wh-words in the initial position. Note the
case of Russian:
This has raised several proposals. Some scholars suggest that in languages like Russian wh-
movement is XP-adjunction to TP or CP (see below), which is typicaly recursive. Others have
suggested that languages like Russian allow more than one specifier. The issue is subject to
empirical and theoretical controversy.
Finally, let us consider a case of XP adjunction to a YP: adverbial fronting.
(16) [ TP On the Idus of March, [ TP at the Senate, [ TP treacherously, [ TP Brutus killed Caesar
]]]] .
Deciding whether a movement is substitution into a specifier or adjunction is not always clear,
and some authors have even suggested that the distinction is nonexistent (e.g. Kayne 1994).
As you can appreciate, in this revised model, transformations where in fact reduced to a
simple movement rule, which was considered to apply freely. That is, it could move anything to
anywhere, but its output was filtered by several principles, like X-bar.
Islands Even though movement is ubiquitous in many languages, there are certain configurations
that block it completely. They are called STRONG ISLANDS. The classical research work on
islands is Ross (1967).
72 CHAPTER 4. TRANSFORMATIONAL RULES
(17) *What did you see [ .NP the man [ .CP who baked what]] ?
(18) *What did you meet [ .NP people [ .CP owning what]] ?
(19) What did [ S Bill cook] and what did [ S Peter wash] ?
(20) *What did [ S Bill cook what] and [ S Peter wash the dishes] ?
(21) *What did [ S Bill cook supper] and [ S Peter wash what] ?
(22) What did [ S Bill cook what] and [ S Peter wash what] ?
(23) *What was [ O that the police would arrest what] a certainty?
(24) *What was [ O to read what in peace] all he wanted?
4.2. REVISING THE FRAMEWORK 73
We can make the hypothesis that some languages have movement rules, others lack them or we
can make the alternative one: All languages have movement rules, but they apply at different
levels of representation.
Crucial evidence: are in situ wh-elements in Japanese or Chinese affected by the same
restrictions, say islands, as wh-elements in English or Catalan? Evidende is quite ellusive.
As Watanabe (1992) shows, unlike English, Chinese seems insensitive to the Complex NP
constraint:
Japanese seems to be insensitive to islands, but Watanabe shows that the form of the wh-word
makes a difference:
‘*What the hell did Mary meet the person who gave to John?’
(34) suggests that Japanese may indeed be sensitive to islands. Moreover, even though the wh-
word is not moving in these languages, they are interpreted as if the had moved to the correct CP,
as in the following cases:
4.2.4 S-structure
As far as transformations affected meaning, it became clear that s-structure was the central
syntactic module, for it served as input to both semantic and phonetic interpretation. One
consequence of placing semantic interpretation after S-structure was the introduction of TRACE
THEORY , namely the hypothesis that the elements moved by a transformation leave a trace at
the original position. Since, as we discussed at class, some elements are not interpreted in its
position at S-structure, we need a register of the original position.
(40) Mary told John that she came for some reason.
a. Why did Mary tell John [ CP that she came t ] ?
b. Why did Mary tell John [ CP that she came ] t ?
In future work, traces are redefined as deleted copies of the moved element:
(41) Mary told John that she came for some reason.
a. Why did Mary tell John [ CP that she came why ] ?
b. Why did Mary tell John [ CP that she came ] why ?
Figure 4.1: Transformational grammar architecture in the eighties, apud Hornstein et al. (2005)
This was a golden age for generative grammar, which extended its coverage to an impressive
amount of languages and problems (Newmeyer, 1997, 2014; Harris, 1995; Culicover, 2014).
However, the empirical coverage was not paired with simplicity, elegance and economy,
and several authors, leaded by Chomsky raised concerns about the conceptual foundations of
the framework. The motivation for the Minimalist Program (MP) was trying to reduce the
computational system (syntax) to the bare minimum required to do the job and satisfy the needs
of the interfaces (LF and PF). This is a kind of intellectual exercise to build a more elegant and
simpler system. In Howard Lasnik’s words (Lasnik, 2002):
The Minimalist program maintains that the derivations and representations constituting
linguistic competence conform to an ‘economy’ criterion demanding that they be
minimal in a sense determined by the language faculty (ultimately by general properties
of organic systems): that is, there are no extra steps in derivations, no extra symbols
in representations, and no representations beyond those that are conceptually necessary.’
The inspiration is similar to the studies about optimal design in the natural sciences. For
examples, the hexagon is a recurrent form in nature, as you can appreciate in beehives or basalt
rocks.
What these different contexts have in common is the necessity for optimal packing under
pressure conditions, and, as Joseph Louis Lagrange showed in 1773, the densest packing of
circles is an hexagonal lattice:
In prose, the emergence of hexagonal structures in the natural world is the optimal result of
physical forces imposed on the original circular form. We can think of syntax in similar terms:
76 CHAPTER 4. TRANSFORMATIONAL RULES
it is the optimal result of the conditions imposed by the two interfaces, the phonetic one and the
semantic one.
section 4.3.5.
4.3.2 Lexicon
The conception of the lexicon in the MP didn’t change from previous frameworks. In mainstream
generative grammar, the lexicon is simply a repository of material for syntax. The only important
contribution of the MP is the restriction of access: syntax can only access the lexicon once, before
any operation takes place. The idea is that when building a sentence, we pick out all the lexical
items we need from the lexicon, and we carry them to syntax in a single step. The set of lexical
items is called a NUMERATION, and its the only possible source of building blocks for deriving
our sentence: once we have our numeration, we have no further access to the lexicon (this will
be important later, when discussing economy conditions in section 4.3.5).
For example, when we build the sentence John sleeps, we build the following numeration:
From now on, syntax can only work with these pieces. The numbers indicates the times we pick
out a certain item. For instance, when building the sentence The girl kissed the other girl, the
78 CHAPTER 4. TRANSFORMATIONAL RULES
Figure 4.6: Subset relationship among derivations, apud Hornstein et al. (2005).
(43) Numeration: {hthe, 2i, hgirl, 2i, hT, 1i, hkissed, 1i, hother, 1i, }
4.3.3 Operations
Select The operation SELECT picks lexical items from the numeration to be combined by
Merge. Every time that selects picks up a lexical item from the numeration, it reduces its value.
(44) Numeration: {hGirl, 1i, hBoy, 1i, hLove, 1i, hM any, 1i, hAll, 1i}
Merge The operation åmerge combines two items and creates a new one with a label.
(45) γ
α β
γ = α or γ = β
For example, to construct the sentence Mary likes linguistics, we begin with the following
numeration:
(46) Num0 : {hM ary, 1i, hT, 1i, hlike, 1i, hlinguistics, 1i}
Now we select like and linguistics, and we merge them, forming a new object headed by the
verbal item (a V’ or a VP in the P&P model):
likes linguistics
4.3. THE MINIMALIST PROGRAM 79
b. Num1 :{hM ary, 1i, hT, 1i, hlike, 0i, hlinguistics, 0i}
T likes
likes linguistics
2
b. Num : {hM ary, 1i, hT, 0i, hlike, 0i, hlinguistics, 0i}
Mary likes
likes linguistics
3
b. Num : {hM ary, 0i, hT, 0i, hlike, 0i, hlinguistics, 0i}
In this derivation, all merging operations involved selecting an element from the numeration:
this is labeled EXTERNAL MERGE. However, we can also merging two elements already in the
structure (also called INTERNAL MERGE):
kissed Mary
In other words, movement is just a variant of the basic operation merge. However, if we stick to
minimalist concerns, we are into trouble: syntax cannot create new entities, but just modify what
you have in the numeration. Obviously, traces are created in syntax, after movement applies.
The solution is Copy Theory, which we discuss in section 4.3.4.
Conditions on Merge
Extension condition Applications of Merge can only target root syntactic objects.
Num0 : {hJohn, 1i, hkiss, 1i, hM ary, 1i}
kissed John
80 CHAPTER 4. TRANSFORMATIONAL RULES
kissed John
Extension condition Strict parallelism between syntactic and semantic composition:
kissed<e,t> → kissed<t>
kissed<e,<e,t>> John<e>
Mary<e> kissed<e,t>
kissed<e,<e,t>> John<e>
Minimality Good configuration:
(53) a. [ CP who [ TP who wondered [ CP how you fixed the car how]]]
b. [ CP who [ TP did you wonder [ CP how [ TP who fixed the car how]]]]
Bad configuration:
(54) [ CP how [ TP did you wonder [ CP who [ TP who fixed the car how]]]]
Minimality
(56) T
T was
was kissed
kissed Mary
The first step involves internal merge of a copy of the element in the final position (i.e. movement
of the former object to the subject position):
4.3. THE MINIMALIST PROGRAM 81
(57) T
Mary T
was
was kissed
kissed Mary
(58) T
Mary T
was
was kissed
kissed Mary
Consider now a case of wh-movement. We begin with external merge of came and who:4
Now we merge T:
4.3.5 Economy
We have seen that syntax is principally an application of merge (internal and external) for
satisfying the conditions of the interfaces, but we haven’t considered if there is more than way to
reach such goal. Let’s a turn to the set of possible syntactic objects:
Figure 4.7: Subset relationship among derivations, apud Hornstein et al. (2005).
Syntax can create sentences which are not legible at the interfaces (set D; uninterpretable for
semantics and/or unpronounceable for phonetics), but we want a system ensuring that its output
can be read at the interfaces (set C). But we can impose still a harder condition on our sentences:
they must be legible at the interfaces, but they must involve the most economical derivation
(set A). That is, in cases where more than one derivation is possible, we must introduce a set
of conditions for choosing the optimal one. Hence, we introduce ECONOMY principles over
derivations.
Consider for instance, the superiority effect:
Informally, when two wh-elements compete for moving to the front of the sentence (i.e. the
specifier of CP), only the highest can move. Since who c-commands what, but not conversely,
only who can move.
If we think of the movement of wh-elements in terms of attraction, namely that the interrogative
C (C[+wh] ) must attract a wh-element to its specifier, the superiority effect reduces to an economy
condition minimizing the distance of movements:
It is clear that the distance between C and who is shorter than the distance between C and what.
Hence, under the economy condition S HORTEST MOVE we predict that only who would be able
to move:
Note that this contrast is also at the roots of minimality contraints (Rizzi, 2001, 1990), which
forbid wh-elements to cross over other wh-elements.
Which candidates? The concept of economy conditions involves comparing different derivations,
so we must be certain that we compare the proper objects. For instance, we do not want to
compare the following two sentences:
It is clear that we cannot compare them, because they depart at the lexical level: they involve two
different numerations. Hence, we must make sure that only derivations from the same numeration
compete.
Let us consider a different case, were the same numeration is involved:
If these two derivations compete, should not (72) be most economical than (73) since it involves
less operations (it does not move Mary)? The answer is that only CONVERGENT [=well-formed]
derivations compete. In this case, the derivation in (72) is not convergent, for it violates a
grammatical principle of English, namely the requirement that the subject position be filled.
Hence, the conditions for two derivations to enter into economy evaluations are the following:
1. they must have the same numeration (lexical items picked up from the lexicon)
2. they must converge, namely they must be properly derived following all principles of
grammar, economy aside. That does not mean they are grammatical: convergent+most
economical=grammatical.
Consider one classical example (there are very few, indeed). With the numeration in (74)-a
we can build two convergent derivations, but we will show that only derivation A is optimal, for
it satisfies economy conditions:
We operate bottom-up, and the first steps are common to both derivations: we select someone
and here and merge them.
At this point, we have two possibilities for filling the subject position of the embedded sentence:
external merge of there or internal merge of someone.
(78) derivation A
a. Num4 = {T, seem}
b. [ to there [ to to [ be be [ here someone here ]]]]
(79) Derivation B
a. Num4 = {there, T, seem}
b. [ to someone [ to to [ be be [ here someone here ]]]]
• Derivation A
• Derivation B
It is clear that both derivations are well-formed (= they are convergent, in Chomsky’s terms), so
the difference must lie in the step where one derivation selects and merges of there while the
other resorts to internal merge of someone:
The economy condition is quite intuitive: since movement is merge+copy, it is a more complex
operation than simple merge. Hence, at this point it is more economical derivation A than
derivation B, hence it is the only grammatical derivation.
This economy condition is very problematic and debated on theoretical and empirical grounds
(see Collins 2008; Motut 2010). Just to note the major point of discussion, note that it is quite
a complex condition to apply. On the one hand, it involves global post-hoc evaluation of full
derivations: we must be sure that the derivations are convergent, for being comparable. On the
other hand, we evaluate economy locally: we consider a particular choice in one particular step
of the derivation. Note that if we just count the final number of uses of movement and merge,
both derivations fare equally.
Just to close the discussion on economy conditions, consider the following pair:
• H2: These two sentences do compete for economy evaluation, but they are equally economical.
7
Movement of someone would be an option, but then we would not exhaust the numeration, namely we would
leave there unusued, and, as we will see in a moment, selecting and merging there is more economical than moving
someone.
86 CHAPTER 4. TRANSFORMATIONAL RULES
For H1 to be correct, these sentence should involve different numerations, but if we consider
the lexical items involved, this seems very difficult to maintain. We are thus confronted to
H2: they involve equally economical derivations. Indeed, we can easily see that the operations
involved are identical, only differing in the initial choice of the object to merge with the verb:
love+Mary in the former, and love+John in the latter. From this initial step, all the operations
would follow in a strict parallel fashion until the merge of the subject, which will be the remaining
noun in the numeration: John in the former, and Mary in the latter. In sum, since we do not have
any contrast in terms of movement, economy considerations are irrelevant.
4.4 Conclusions
In this chapter we have seen that movement transformations are a powerful tool to express
relations between sentences which cannot be (easily) expressed by phrase-structure grammars.
Moreover, movement transformations cannot extract constituents from strong islands, and they
display cross-linguistic variation. Finally, we have stressed that the Minimalist Program is an
attempt to simplify syntax to its minimum, as an optimal derivational system to connect the
lexical material with the two interfaces (LF and PF). As a consequence, syntax is reduced to a
combination of lexical items by means of Merge, which are driven by the legibility requirements
of the interfaces, and constrained by economy considerations.
4.5 Exercises
E XERCISE 20 According to what you know about extracting elements from within different
positions in the sentence structure, indicate which sentences will be fine and which will be
ungrammatical, and explain the reason for the illformedness of ungrammatical ones. Choose
the language you prefer.
(89) a. What did you insist on buying?
b. What did you ask Mary not to insist on buying?
c. What did you make a proposal not to buy?
d. What did you make the proposal that nobody should buy?
e. What do you think that buying is useless?
f. What do you think that Mary left without buying?
g. What do you need that she be calm and buy?
(90) a. Què vas insistir a comprar?
‘What did you insist on buying?’
b. Què vas demanar a la Maria que no insistís a comprar?
‘What did you ask Mary not to insist on buying?’
c. Què vas fer una proposta de no comprar.
‘What did you make a proposal not to buy?’
d. Què vas fer la proposta que ningú comprés?
‘What did you make the proposal that nobody should buy?’
4.5. EXERCISES 87
E XERCISE 21 Indicate whether the following movements are affected by an island. Be very
specific about the exact configuration involved.
2. What did Jane insist that every child should learn at school?
3. Where did Jane had the belief that Peter hide the money?
4. Which knife does Mary knows that Jane stabbed his husband with?
5. What did Jane make the assumption that everybody wanted to know?
6. Whom did Mary confess to Bill her belief that aliens had abducted?
E XERCISE 23 Indicate which sentences will be fine and which will be ungrammatical, and
explain the reason for the illformedness of ungrammatical ones. Choose the language you prefer.
Empty categories
5.1 Traces
Crucially, if we interpret sentences after movement transformations, we need some device to
mark the original position for interpreting certain elements:
The solution was proposed by Robert Fiengo Fiengo (1977), who argued that movement
leaves a TRACE at their original position:
Traces are not just a notation device nor a theoretical entity, but they have syntactic reality. One
empirical argument for their existence is the problematic extraction of a subject of a sentence
introduced by that. Whereas object extraction is insensitive to the presence of that (3), subject
extraction is affected (4):1
Whereas the presence or absence of the complementizer that doesn’t affect the extraction of
what, the subject who cannot be extracted unless the complementizer is omitted.
This contrast was described as the that-trace effect Perlmutter (1971), and it is found with all
kinds of movements:
1
The notation (*that) means that the sentence is ungrammatical if we remove the parenthesis –with that–, but
grammatical otherwise –without that.
89
90 CHAPTER 5. EMPTY CATEGORIES
(4) a. This is the person who I thought (*that) met Sue. (relativization)
b. Mary we think (*that) t met Sue. (topicalization)
c. It is Mary that we think (*that) t met Sue. (cleft)
d. More people like Mahler than we think (*that) t like Bruckner. (comparative)
What’s the constituent structure of the coordination in (10) Burton & Grimshaw (1992); McNally
(1992)?
Since we interpreted the second coordinate in the future (“they will sing and will be acclaimed”),
we must assume that the coordination is under the level of the modal verb:
But, crucially in the second coordinate we have a passive sentence, which must have a trace/copy
of the subject in object position:
Yet, we know that (12) cannot be the correct structure, for it would involve a movement from
the object position of be acclaimed to the specifier of TP, violating the Coordinate Structure
Constraint, which forbids extraction from one part of a coordination. If both coordinates were
passives, we would have a standard case of across the board movement, that is movement from
both parts of the coordinations at the same time, which is unaffected by the Coordinate Structure
Constraint:
The problem is that in (12) we don’t have a movement of they in the first coordinate.
(14)
92 CHAPTER 5. EMPTY CATEGORIES
This is the Internal Subject Hypothesis Kitagawa (2018); Koopman & Sportiche (1991),
which generates subjects as adjuncts or specifiers of the VP:
(15) TP
T’
T VP
NP VP
Mary V’
sings
We have seen that this analysis was crucial for explaining the coordination case, but much more
evidence exists. Huang (1993) discussed the case of fronted elements which contain a reflexive.
In English, as we have discussed, the antecedent of the reflexive may vary in cases like the
following:
However, Huang observed that this wasn’t the case with VP fronting:
5.1. TRACES 93
If we apply the analysis for DP fronting, as before, we would expect both Bill and John to be
possible antecedents for the reflexive: Bill would bind the reflexive in the lowest copy (19) and
John would bind it in the intermediate copy (20):
(19) criticize himself, John thinks criticize himself Bill will not criticize himself.
(20) criticize himself, John thinks criticize himself Bill will not criticize himself.
However, the expected reading in (20) is not found. Why? The answer, again, has to do with the
internal subject: when we move the VP, we also move the copy of the internal subject, which is
the only possible antecedent in all the positions:
The fronted VP (criticize himself ) will always contain a copy of the internal subject Bill (i.e.
[V P Bill criticize himself]), so only this name will count as the antecedent for the reflexive.
Interestingly, Huang shows that the same contrast is found in Chinese. DP fronting shows
the same ambiguous pattern as English:
In contrast, the fronted VP (piping taziji) will contain a copy of the internal subject Lisi (i.e. [V P
Lisi piping taziji]), so only this name will count as the antecedent for the reflexive.
(23) a. Talking about yourself for an hour won’t convince the council that you are not
egocentric.
b. Peter insisted on buying himself a present for his birthday.
c. Haunted by scary thoughts, Jane remained awake the whole night.
This subject position, which is syntactically active for binding or agreement purposes is typically
represented by the symbol PRO(NOUN):
However, it is the basic case in most Romance languages, like in the following examples from
Catalan and Spanish:
These languages are not required to express their subject, which can easily be identified by means
of the agreement markers on the verbal head. Indeed, whereas English or French must fill the
subject position with a lexical noun phrase or a lexical pronoun, Catalan or Spanish tend to omit
the subject Rigau (1988); Mayol (2010):
‘You, eat!’
In this case, we are building a contrast between I and you, which requires a full pronoun.
A typical property of null-subject languages is free inversion of the subject. Namely, depending
on the context, we can have preverbal or postverbal subjects:
The position of the subject is related to the informational content: when preverbal, the subject
is typically a topic, encoding old information; in contrast, in postverbal position, the subject is
focus, namely new information. Hence, each structure is found in a different context:
On the light of the discussion about the internal subject hypothesis, we can analyze inverted
subjects as in situ subjects:
TP
T’
T VP
V T NP VP
canta María V’
canta
96 CHAPTER 5. EMPTY CATEGORIES
The difference between English/French and Catalan/Spanish would reduce to the requirement
of filling the higher subject position:
TP
NP T’
Mary T VP
V T NP VP
sings Mary V’
sings
TP
T’
T VP
V T DP VP
canta la Maria V’
canta
5.3 Ellipsis
Ellipsis is another major phenomenon suggesting that syntactic structure is not always phonetically
realized:
Interestingly, even though null, the site of VP-ellipsis is typically sensitive to islands. For
instance, in (36), cats is extracted from within a relative clause in the ellipsis, yielding a bad
result:
(36) *Abby knows five people who have dogs, but cats, she doesn’t know [NP five people [CP
5.4. EXERCISES 97
(37) *Abby wants to hire someone who speaks a Balkan language, but I don’t remember which
(Balkan language) Ben does want to hire [NP someone [CP who speaks which (Balkan
language) ]]
Similar effects are found with gapping (38) and contrastive sluicing (39):
(38) *Some wanted to hire the woman who worked on Greek, and others wanted to hire the
woman who worked on Albanian.
(39) She knows a guy who has five dogs, but I don’t know how many cats.
a. = the guy who has the five dogs has how many cats
b. 6= she knows a guy who has how many cats
All these data suggest that the elliptical part has syntactic structure and is affected by the very
same syntactic effects as full pronounced material. This gives support to analyses of ellipsis as
instances of postsyntactic deletion at PF.
(41) a. Mary crashed Mary’s car, and Jim too crashed Mary’s car.
b. Mary crashed the car of a salient third person, and Jim too crashed the car of such
a salient third person.
c. Mary crashed Mary’s car, and Jim too crashed Jim’s car.
Readings (41)-a and (41)-b are, so-called, strict readings: the elided part is an exact copy of
the main sentence. However in (41)-c we have a modified copy, a sloppy reading. Obviously,
since the different readings are typically based on possible antecedents and c-command, it seems
plausible to assume that the deleted part contains syntactic structure.
5.4 Exercises
E XERCISE 24 Consider the following sentences. Does the wh-phrase leave an intermediate
trace? Can we prove it somehow?
98 CHAPTER 5. EMPTY CATEGORIES
(42) *Which books about Maryi did shei said that John had read?
a. *“Quins llibres sobre la Mariai proi va dir que en Joan havia llegit?”
b. *“¿Qué libros sobre Maríai proi dijo que Juan había leído?”
(43) Which manuscripts by himselfi did Caesari claim that Cleopatra had destroyed?
a. “Quins manuscrits de si mateixi Cèsari va afirmar que Cleòpatra havia destruït?”
b. “¿Qué manuscritos de sí mismo i Césari afirmó que Cleopatra había destruido?”
E XERCISE 25 Consider the following sentences. Does the wh-phrase leave an intermediate
trace? Can we prove it somehow?
(44) Which books about himi did she said that Johni had read?
a. “Quins llibres sobre elli va dir (ella) que en Joani havia llegit?”
b. “¿Qué libros sobre éli dijo (ella) que Juani había leído?”
Chapter 6
Sentence syntax
• Predicates describing willed or volitional acts: fight, pray, whistle, smile, dance.
• Certain (normally involuntary) bodily processes: sneeze, cough, pee, burp, vomit.
The unaccusative class typically includes verbs with the following features:
• Intransitive predicates whose thematic role is a ‘patient’: burn, fall, drown, float.
• Intransitive inchoative verbs that involve a change of state: melt, die, perish, freeze,
evaporate.
99
100 CHAPTER 6. SENTENCE SYNTAX
• Non-voluntary emission of stimuli that impinge on the senses (light, noise, smell, etc.):
glitter, shine, smell.
There-sentences in English
(2) *There cried a beautiful girl at our house yesterday.
(3) There arrived a beautiful girl at our house yesterday.
However, note the case of there-sentences in Icelandic, which are not restricted:
(4)
Locative inversion
Locative inversion is possible with unaccusatives, but impossible with unergatives:
(7) unergative
6.1. INTRANSITIVE VERBS 101
Resultative constructions
Resultative constructions in English are possible with objects but not with subjects:
(10) unergative
a. *John cried sick.
b. *Dora shouted hoarse.
(11) unaccusative
a. The water froze solid.
b. The bottle broke open.
Absolute participle
Just as happened with past participle relatives, absolute participles are possible with unaccusatives,
but not with unergatives:
(12) unergatives
a. *Una vegada xisclats els nens, va començar la classe.
a time screamed the children PST.3 SG begin the class
‘Once the children ended screaming, the class began.’
b. *Amb els nens queixats, el dinar va ser un desastre.
with the children complained the lunch PST.3 SG be a disaster
‘With the children having complained, lunch was a disaster.’
(13) unaccusatives
102 CHAPTER 6. SENTENCE SYNTAX
Subject pronominalization
As a rule, subjects cannot pronominalize (unlike objects). However, unaccusatives can pronominalize
indefinite subjects by means of the partitive clitic in several Romance languages, but unergatives
cannot:
Impersonal with se
Unergatives can construct impersonal constructions with se, but unaccusatives cannot:
Cognate objects
Many unergative verbs may be constructed as transitives with a cognate object, namely an object
whose meaning is included in the lexical meaning of the verb:
(17) a. Malinda smiled her most enigmatic smile. Levin & Rappaport Hovav (1995)
b. Last night the candidate slept a restless sleep. Macfarland (1995)
(18) a. *The actress fainted a feigned faint. Levin & Rappaport Hovav (1995)
b. *She arrived a glamorous arrival. Levin & Rappaport Hovav (1995)
6.1. INTRANSITIVE VERBS 103
Auxiliary selection
• it.: ha pianto/protestato/urlato/sorriso
• It.: è andato/venuto/arrivato/caduto/nato
• Dutch: Ze is gegaan/gekomen/aangekomen/gevallen/geboren
Such a distinction has disappeared in English, Spanish and most Catalan dialects.
Dutch and German allow passives of intransitive verbs, however the verb must be unergative:
Many languages of the world allow the incorporation of the nominal object into the verbal root:
(21) TP
T’
T VP
NP VP
Mary V’
complains
(22) TP
T’
T VP
NP VP
Mary V’
V NP
arrives Mary
In other words, the subject of unaccusative verbs is an object in many respects, but a formal
subject (it agrees with the verb).
6.2 Transitives
From a syntactic point of view, a transitive verb must express a direct object.
However, such a view is too simplistic. We find many transitive verbs selecting an internal
argument (direct object), even though this argument is not always expressed in syntax.
Note that these intransitive uses are typically associated to activities where the object is implicitly
106 CHAPTER 6. SENTENCE SYNTAX
entailed:
(25) activities
a. Mary eats.
b. I am reading!
(26) states
a. *Mary does not have.
b. *Peter hates.
One way to understand this alternation is the concept of FRAMES developed by Fillmore & Atkins
(1992). The idea is that situations are associated to typical frames of participants, which can be
syntactically realized in different ways. This is the case of the ’commercial transaction’ frame:
Crucially, not all the verbs involved require the same participants nor they realize them always
as arguments. For instance, the money participant is realized as an optional adjunct with buy and
sell, but as an obligatory argument with charge, expend and cost (it is less clear with pay).
6.3.1 Passive
Passive is an intransitivizer operation: it takes a transitive verb and converts it into an intransitive
one:
In this case, the choice of the auxiliary may indicate a distinction between a dynamic and a stative
passive:
(31) Dutch
a. De deur wordt gesloten. dynamic
‘The door is closed.’ = The door is in the process of being closed.
b. De deur is gesloten. stative
‘The door is closed.’ = The door is in a closed state (result of the process).
(32) German
a. Die Reifen werden aufgepumpt. dynamic
the tires become inflated
b. Die Reifen sind aufgepumpt. stative
the tires are inflated
(33) Spanish
a. Los neumáticos són inflados. dynamic
the tires become inflated
b. Los neumáticos están inflados. stative
the tires are inflated
In Romance languages we can build syntactic passives by means of the clitic se (impersonal
passives or reflexive passives):
6.3.2 Impersonals
Just as passive, impersonal sentences are an argument reduction operation: it takes a verb with a
subject and returns a subjectless verb:
Impersonals cannot express the agent complement (unlike passives), but they maintain an agentive
reading:
In Romance languages, this alternation is typically marked by means of the clitic SE, as in the
following examples from Catalan:
(49) a. He left.
b. Mary made him leave.
(50) a. He opened the door.
b. Mary will make him open the door.
(51) a. Juan desayunó.
Juan ate.breakfast
b. María le hizo desayunar a Juan.
María to.him made ate.breakfast to Juan
‘María made Juan eat breakfast.’
(52) a. Juan leyó la carta.
Juan read the letter
b. María le hizo leer la carta a Juan.
María to.him made read the letter to Juan
‘María made Juan read the letter.’
6.4 Summary
b. TP
NP T’
Mary T VP
NP VP
Mary V’
V NP
eats vegetables
The subject is generated in an internal VP position and obligatorily moves to the preverbal
position in languages like English or French. In contrast, the subject can remain in the original
position in null subject languages, which combined with the movement of the verb to T yields
the typical subject-verb inversion pattern:
NP T’
María
T VP
V T
NP VP
vendió T
María V’
V DP
vendió el coche
b. TP
T’
T VP
V T
NP VP
vendió T
María V’
V DP
vendió el coche
6.5 Exercises
114 CHAPTER 6. SENTENCE SYNTAX
Chapter 7
7.1 Wh-movement
In many languages wh-elements must move to the periphery of sentence, usually forcing inversion:
We typically analyze these facts by means of a complementizer phrase CP, hosting the wh-word
and the verb:
CP
what C’
is TP
Mary T’
T VP
V VP
V’
V what
doing
The head C also host declarative complementizers, like that:
Obviously, since that associates with declarative sentences, it is typically incompatible with wh-
words:
115
116 CHAPTER 7. THE PERIPHERY OF SENTENCE
Combinations
(6) Italian
a. A G IANNI dovresti dare questo libro, non a Piero.
‘T O G IANNI you should give this book, not to Piero.’
b. *A G IANNI che cosa dovresti dare, non a Piero?
‘T O G IANNI what you should give, not to Piero?’
c. *A G IANNI IL LIBRO dovresti dare, non a Piero il disco.
‘T O G IANNI THE BOOK you should give, not to Piero the record.’
(7) Sardinian
a. *A kie SU JOCÁTULU as datu?
to whom the toy have.2SG given
b. *S U JOCÁTULU a kie as datu?
the toy to whom have.2SG given
‘To whom did you give the toy?’
This suggests that both wh- and focused phrases compete for the same position.
Compare with the case of topic constituents, which can precede wh-elements:
(13) With no job, Mary would be happy. = ‘If jobless, Mary would be happy.’
(14) With no job would Mary be happy. = ‘There is no job making Mary happy.’
Whereas PP fronting seems closer to topics, negative inversion seems closer to focus fronting
(inversion is required). As a consequence, we expect both constructions to be possible, but in a
rigid ordering:
(15) a. During my sabbatical, on no account will I read e-mail. PP fronting > neg-
inversion
b. *On no account, during my sabbatical will I read e-mail. neg-inversion > PP
fronting
On conclusion:
If we take topics to appear in a fixed position, one could conclude that two complementizer
positions may exist. Indeed, we find languages where two complementizers may coappear:
(21) Dywedais i mai fel arfer y dynion a fuasai’n gwerthu’r ci. Welsh:
said I that as usual the men that would-ASP sell-the dog
Rizzi & Cinque (2016)
‘I said that it’s as usual the men who would sell the dog.’
(22) María preguntó que el lunes si había periódicos. Spanish
María asked that the Monday whether there.were newspapers
‘Maria asked whether there were newspapers on Monday.’
(23) Taroo-wa Ziroo-ni [CP dare-ga kare-no ie-ni kuru ka to] tazuneta.
Taroo-TOP Ziroo-DAT who-NOM he-GEN house-to come ka to asked
Japanese: Saito (2012)
Hence, we can speculate that a second higher complementizer may appear above the topic
position:
One well-known proposal for the left periphery is due to Rizzi (1997):
(25) ForceP
Force TopP
Top FocP
Foc FinP
Fin TP
• The left periphery of sentence hosts elements with specific informational functions (topic,
focus).
7.3. EXERCISES 119
• Complementizers are sensitive both to the selectional requirements of the main clause, but
also to the tense and modality of the subordinate clause.
7.3 Exercises
120 CHAPTER 7. THE PERIPHERY OF SENTENCE
Key to the exercises
Constituent structure
K EY TO EXERCISE 1 Are the following sentences ambiguous? Justify your answer applying
standard tests. Choose the language you prefer.
C ATALAN
Hem parlat amb els estudiants de física. The sentence is ambiguous. On the one hand, the
PP de física may be a modifier of the noun estudiants, so that the bigger PP can be fronted: Amb
els estudiants de física, hi he parlat.
On the other hand, the PP can be a second complement of the verb, which allows us to predict it
will be independent of the PP amb els estudiants: De física, n’he parlat amb els estudiants. He
parlat de física amb els estudiants.
Vam trobar els nens i les nenes molt bufones. The sentence is not ambiguous, for the
feminine agreement on the adjective tells us that it is modifying les nenes only:
If the adjective had modified the coordination els nens i les nenes, it would be in masculine,
which is the unmarked gender: els nens i les nenes bufons.
Quan et va explicar que havia fet aquella bestiesa? The sentence is ambiguous. On the
one hand, the wh-word quan may be a modifier of the verb va explicar (e.g. M’ho va explicar
ahir):
On the other hand, the wh-word quan may be a modifier of the lower verb havia fet, so it moves
from the subordinate clause (e.g. Em va explicar que l’havia feta ahir, aquella bestiesa):
121
122 CHAPTER 7. THE PERIPHERY OF SENTENCE
Quan et va oferir una explicació convincent del que havia fet? The sentence is not ambiguous.
The wh-word quan may be a modifier of the verb va oferir, but it cannot modify the lower verb
havia fet, because it is within a nominal clause, which is a strong island forbidding extraction:
(31) Quan et va oferir [ DP una explicació convincent de [ CP el que havia fet]] quan?
(32) *Quan et va oferir [ DP una explicació convincent de [ CP el que havia fet quan]]?
E NGLISH
When did she tell you that she did such nonsense? The sentence is ambiguous. On the one
hand, the wh-word when may be a modifier of the verb tell (e.g. She told me that yesterday):
(33) When did she tell you [ CP that she did such nonsense] when?
On the other hand, the wh-word when may be a modifier of the lower verb did, so it moves from
the subordinate clause (e.g. She told me that yesterday she did such nonsense):
(34) When did she tell you [ CP that she did such nonsense when]?
Old men and women are discriminated. The sentence is ambiguous. On the one hand, the
adjective old may be a modifier of men only:
On the other hand, the adjective old may be a modifier of the coordination of both nouns:
When did she offer you an explanation that she did such nonsense? The sentence is not
ambiguous. The wh-word when may be a modifier of the verb offer, but it cannot modify
the lower verb did, because it is within a nominal clause, which is a strong island forbidding
extraction:
(39) When did she offer you [ DP an explanation [ CP that she did such nonsense]] when?
(40) *When did she offer you [ DP an explanation [ CP that she did such nonsense when]] ?
Mary stabbed the man with the knife. The sentence is ambiguous. On the one hand, the PP
with the knife can be interpreted as an instrumental of the verb stabbed, so it will be independent
from the direct object the man: With the knife, Mary stabbed the man. On the other hand, the the
PP with the knife can be interpreted as a modifier of the noun man, so that the direct object must
7.3. EXERCISES 123
On the other hand, the PP can be a second complement of the verb, which allows us to predict it
will be independent of the PP con los alumnos: De griego, hablamos con los alumnos.
Necesitamos a los niños sanos. The sentence is ambiguous. On the one hand, the AP sanos
may be a modifier of the noun niños, so that the bigger PP can be fronted: A los niños sanos, los
necesitamos (a los enfermos no).1
On the other hand, the AP can be a predicative of the direct object, which is independent of the
PP a los alumnos: A los niños, los necesitamos sanos. Es sanos como necesitamos a los niños.
The analysis of predicatives is not evident. Most scholars argue that it involves a predicative
phrase (PredP):
Fueron a la reunión de espías. The sentence is ambiguous. On the one hand, the PP de
espías may be a modifier of the noun reunión, so that the bigger PP can be clefted: Fue a la
reunión de espías a donde fueron.
On the other hand, the PP can be an adjunct of the verb (a manner adverbial: as spies), which
allows us to predict it will be independent of the locative PP a la reunión: Fue de espías, como
fueron a la reunión.
Necesitamos mujeres y hombres valientes. The sentence is ambiguous. On one reading, the
adjective valientes modifies hombres only:
On the other reading, the adjective will modify the result of coordinating both nouns:
1
Note that this a case of prepositional direct object, also known as Differential object marking, so the analysis as
a PP is probably inaccurate. However, this issue doesn’t affect the analysis in the text.
124 CHAPTER 7. THE PERIPHERY OF SENTENCE
Puso los libros en la mesa. The sentence is not ambiguous: the PP en la mesa can only depend
on the verb as a locative complement. As a result, it will be able to move independently of the
direct object: En la mesa puso los libros. Los libros los puso en la mesa.
(49) a. Where did she offer you the explanation that she had bought the car?
b. On et va donar l’explicació que havia comprat el cotxe?
c. ¿Dónde te dio la explicación de que había comprado el coche?
No. The interrogative word where/on/dónde can only be an adjunct of the VP headed by offer/va
donar/dio. Hence it moves from within the main clause to the specifier of CP.
(50) [ CP Where [ C’ did [ TP she [ VP offer you [ DP the explanation [ CP [ C’ that [ TP she had bought
the car]]]] where]]]]?
Where cannot be moving from the lower VP headed by bought, because that would imply
extraction from a strong island, in this case a Complex NP.
(51) [ CP Where [ C’ did [ TP she [ VP offer you [ DP the explanation [ CP [ C’ that [ TP she had bought
the car where]]]]]]]]?
Since this is forbidden, we don’t have this second interpretation, and the sentence is not ambiguous.
(52) a. She looked for good books and magazines.
b. Cercava bons llibres i revistes.
c. Buscaba buenos libros y revistas.
Yes. The adjective can modify either the NP books (53) or the coordinated NP books and
magazines (54):
(53) is synonymous of the unambiguous She looked for magazines and good books, and would
allow a modifier like bad to be inserted without contradiction:
Changing the ordering of the NPs in (54) would make no difference: She looked for [good
[magazines and books] ], and the insertion of bad would lead to a contradiction: “good books
and good bad magazines”.
The sentence is not ambiguous. We can only get the reading where the locative wh-word depends
on taking a bath:
(58) Where shouldn’t you take a bath where [ CP after having lunch]?
The reading where the locative wh-word depends on having lunch is not available for it would
entail extracting a wh-word from within an adjunct sentence, which we know is a strong island:
(59) *Where shouldn’t you take a bath [ CP after having lunch where ] ?
X-bar Theory
K EY TO EXERCISE 3 Given the restrictions on phrase structure following from Kayne (1994),
what would be the explanation for the impossibility of a tree like the following:
We didn’t get into the details of the proposal by Kayne, but the
basic idea is that if an XP asymmetrically c-commands another YP,
then the heads dominated by XP must precede the heads dominated by
YP. Namely if XP is the subject DP The boy from Barcelona and YP
the VP eats an apple, it entails that the set of heads the, boy,
from, Barcelona must precede the set of heads eats, an, apple (we
say nothing about the respective order of these heads at this point).
Obviously, this is in contradiction with crossing configurations,
which would allow, for instance, from Barcelona to be dominated by
the subject DP, but appear, say, after the VP: The boy eats an apple
from Barcelona (where from Barcelona depends on boy, not on apple).
K EY TO EXERCISE 4 Given the restrictions on phrase structure following from Kayne (1994),
why should multiple adjuncts (in red in the tree) not be allowed?
126 CHAPTER 7. THE PERIPHERY OF SENTENCE
We didn’t get into the details of the proposal by Kayne, but the
basic idea is that when we have multiple adjuncts they c-command
each other symmetrically. However, since we need asymmetric c-command
for obtaining a proper linearization of the heads (see the previous
exercise), this configuration could not be disambiguated for linearizatio
resulting in a bad derivation.
*VP
XP VP
YP VP
V DP
C-command
K EY TO EXERCISE 5 Consider the following tree and choose the correct option:2
(60) TP
NP T’
Mary
T VP
NP VP
Mary
V’ PP
V NP P’
eats vegetables P NP
at home
K EY TO EXERCISE 6 Consider the following tree and choose the correct option:
(61) TP
DP T’
A neighbor of Mary’s
T VP
NP VP
DP V’
A neighbor of Mary’s V NP
eats N’
N PP
vegetables P’
P NP
from home
K EY TO EXERCISE 7 Consider the following tree and choose the correct option:
CP
DP C’
did
DP T’
D’ T VP
D NP V’
the AP N’ V DP
elder N read
which book about Lisa
N PP
cousin P’
P NP
of Barnie’s
• which book about Lisa c-commands the elder cousing of Barnie’s, and read: TRUE.
• about Lisa c-commands the elder cousing of Barnie’s, and read: FALSE.
K EY TO EXERCISE 8 Consider the following tree. Can either Jane or Mary binding herself?
Is there any other antecedent available? Justify your answer.
CP
DP C’
did
DP T’
D’
T VP
D NP
the AP N’ VP AdvP
elder N V’ yesterda
N PP V’ PP
cousin P’ V DP P’
P send t P NP
P NP to herself
of Mary
Neither Jane nor Mary c-command the reflexive, hence they do not count as possible antecedents.
The only possibility is the DP the elder cousin of Mary.
7.3. EXERCISES 131
K EY TO EXERCISE 9 Consider the following tree. Can either Jane or Mary bind herself?
Justify your answer.
CP
DP C’
C TP
which book about herself
did NP T’
Jane
T VP
V’
V’ PP
V DP P’
send t P DP
to D’
D NP
the AP N’
elder N
N PP
cousin P’
P NP
of Mary
Mary does not c-command the reflexive, hence it does not count as a possible binder. However,
Jane does c-command the reflexive (in its original complement position), so it can bind the
reflexive.
132 CHAPTER 7. THE PERIPHERY OF SENTENCE
K EY TO EXERCISE 10 Consider the following tree. Can Jane and her have the same reference?
What about Mary? Justify your answer.
CP
DP C’
did NP T’
Jane
T VP
V’
V’ PP
V DP P’
send t P DP
to D’
D NP
the AP N’
elder N
N PP
cousin P’
P NP
of Mary
Since Mary does not c-command the pronoun, it may corefer. However, Jane cannot be a
proper antecedent, for it binds it in a local domain (violation of principle B).
7.3. EXERCISES 133
K EY TO EXERCISE 11 Consider the following tree. Can she refer to the fronted NP Mary?
Justify your answer.
CP ]
NP C’
Mary C TP
NP T’
she
T VP
VP PP
V’ P’
V’ DP P DP
K EY TO EXERCISE 12 Explain whether the following sentences satisfy Binding Principles under
the intended reading:
NO. He binds Bill in its the original position: Hei never had confidence in Billi . Hence, it is a
Principle C violation.
YES. Bill does not c-command him, not conversely. Hence, they can corefer without violating
any binding principle.
YES. Hillary binds her from outside the sentence, in accordance with Principle B.
YES. Hillary binds she from outside the sentence, in accordance with Principle B.
134 CHAPTER 7. THE PERIPHERY OF SENTENCE
YES. Neither her nor Hillary bind each other, so they can corefer, in accordance with binding
principles.
NO. Bill does not c-command himself in its local domain. Hence, we have a violation of Principle
A.
NO. Bill does not c-command himself in its local domain. Hence, we have a violation of Principle
A.
NO. Hillary does not c-command herself in its local domain. Hence, we have a violation of
Principle A.
YES. Hillary c-commands herself in its local domain (from its original object position).
Hence, we satisfy Principle A.
NO. She c-commands Hillary in its local domain (both from its original object position and
from the subject position). Hence, we have a violation of Principle C.
YES. Hillary c-commands herself in its local domain. Hence, we satisfy Principle A.
YES. Bill binds him from outside its local domain. Hence, we satisfy Principle B.
K EY TO EXERCISE 13 How do we deal with the following cases, according to the binding
principles just introduced?
1. *Sue1 said that Mary2 liked each other1+2 . This is a curious case. The antecedent
of the reciprocal must be in the same local domain (i.e. sentence),
but in this case, only Mary is: Sue is outside the local domain.
As a consequence, the sentence violates Principle A.
2. If Sue1 arrives late, John will be angry at her1 . Sue does not c-command her,
nor her, John. As a consequence, both Principle B (the pronoun
7.3. EXERCISES 135
3. After talking about herself1 for hours, Sue1 became silent. Sue cannot be the antecedent
of herself: it is in a different sentence, and it does not c-command
the reflexive. The only possibility is that the reflexive is
bound by the null subject of the gerund sentence: After PRO1
talking about herself1 for hours, Sue1 became silent. This null
subject corefers with Sue, resulting in a correct sentence: herself
gets bound in its local domain, and the pronoun is free in its
local domain.
4. His1 mother loves John1 . Since his is within the subject NP it cannot
c-command the object position, hence it doesn’t bind John (no
principle C violation). Obviously, John does not bind the pronoun
(no principle B violation). Then, the relation between John
and his is not of binding, but coreference: the pronoun picks
out an available antecedent from discourse, in this case John.
K EY TO EXERCISE 14 What will happen in the following sentences if we change the reciprocal/reflexive
by a pronoun?
1. *Sue1 said that Mary2 liked each other1+2 . The sentence is still bad: *Sue1
said that Mary2 liked them1+2 . However, it is not apparent what
the problem is. If the reciprocal was bad because it wasn’t
bound in its local domain, one would expect the pronoun to be
OK, for it would be free in its local domain. So, it seems that
both parts of the antecedent must be outside the local domain
(or within it for anaphors).
2. After talking about herself1 for hours, Sue1 became silent. The sentence will be
fine: After talking about her1 for hours, Sue1 became silent.
But crucially, the null subject of talking cannot be Sue: After
PRO2 talking about her1 for hours, Sue1 became silent. Since
136 CHAPTER 7. THE PERIPHERY OF SENTENCE
3. *John1 ’s mother loves himself1 . The sentence will be fine: John1 ’s mother
loves him1 . Since John does not c-command the pronoun, Principle
B is respected, and him and John may corefer.
K EY TO EXERCISE 15 Adapted from (Büring, 2005, 5). In the following sentences, Φ designates
an NP with the index given. For each sentence, determine by intuition what Φ can/must be (there
may be more than one option in some cases). Then give the local clause and the antecedent for
Φ and demonstrate that the Binding Conditions are met.
2. Masha2 believes that the swamp elks admire Φ2 . Φ2 =her. The pronoun is bound
by Masha from outside its local domain (the antecedent is in
the higher sentence), in accordance with Principle B.
3. Masha2 believes that [the swamp elks]3 admire Φ3 . Φ3 =themselves. The anaphor
is bound in a local domain (the same sentence) by the swamp elks,
in accordance with Principle A.
5. Hermann4 tried to be nice, and Gallia quite liked Φ4 . Now Φ4 and Gallia go out to see
a mud wrestling show. Φ4 =him/he(=Hermann). The pronoun is free in
its local domain (it has no binder), in accordance with Principle
B. Hence, it can freely corefer with a previous discourse antecedent
(Hermann).
for the pronoun would bind a proper name. Hence, the only option
is assuming that we have Cecilia takes care of Cecilia’s business
and that each instance of the proper name is referentially independen
resulting in a marked sentence with a marked interpretation:
we generate the inference that Cecilia only cares about herself.
Note that this inference is lacking in the unmarked Cecilia takes
care of her own business.
K EY TO EXERCISE 16 Consider the following tree and describe the c-commanding domain of
the elements given:
C-command =def α c-commands β iff the first branching node γ that dominates α dominates β.
(75) TP
DP T’
D’
D NP T VP
The N PP
books by James V’ PP
P’
V DP
P NP
were-written D’
by him
D NP
the N PP
books by James
• the PP by him c-commands were written and its complement the books by James.
K EY TO EXERCISE 17 According to the results of the previous exercise, answer the following
questions:
• Do James and him have the same reference? They may have the same reference.
• Does James bind the pronoun him? No, because James does not c-command him.
• Could we replace him by himself ? Why? No, because the reflexive would have no local
antecedent: neither James or its copy c-commanded him, hence they would not bind the
himself either. Therefore, the substitution would yield a Principle A violation.
7.3. EXERCISES 139
K EY TO EXERCISE 18 Consider the following tree and answer the questions, justifying your
response:
• Who is the antecedent of him? James. Since James does not bind him, they can corefer
without violating principle B (the pronoun is not bound). It cannot be Peter, because it
seems that c-commands the PP to him, as we will confirm in the next point.
• If we replaced him by himself, would the sentence be possible? Would the antecedent of
himself be the same as that of him? Since we saw that James is not c-commanding the PP
to him, we expect that James will not be able to bind himself in the very same position.
The only possibility will be Peter, even though we must somehow assume that it is the
whole PP who binds the reflexive (the NP cannot c-command outside its own PP).
TP
DP T’
D’
D NP
T VP
The N PP
books by James
VP PP
V’ P’
P NP
V’ PP by Peter
P’
V DP
P NP
were-send D’
to him
D NP
the N PP
books by James
K EY TO EXERCISE 19 Consider the following tree and answer the questions, justifying your
response:
140 CHAPTER 7. THE PERIPHERY OF SENTENCE
• Who is the antecedent of herself ? How can we prove it? It can only be Mary, who binds
the original position of the reflexive (herself). Mary is too low in the structure, so it cannot
c-command the object position. To see this, we can substitute John for Mary, and the
result should be wrong, for herself will have no antecedent at all. This is indeed the case:
*Herself, John invited to Mary’s party.
• If we replaced herself by her, would the sentence be possible? Would the antecedent of
her be the same as that of herself ? Once again, when we have a pronoun, things get more
complicated. First, Mary would bind the pronoun in the original position, but not in the
higher, so we could expect coreference to be possible in the second position (somewhat
strange). As for Mary, since no binding is involved, coreference should be possible (it is).
CP
NP C’
herself
C TP
NP T’
Mary
T VP
V’
V’ PP
V NP P’
invited herself P NP
to POSS NP
Mary’s party
Transformations
K EY TO EXERCISE 20 According to what you know about extracting elements from within
different positions in the sentence structure, indicate which sentences will be fine and which will
be ungrammatical, and explain the reason for the illformedness of ungrammatical ones. Choose
the language you prefer.
(76) a. What did you insist on buying? It is OK: what is extracted from a complement
sentence.
7.3. EXERCISES 141
b. What did you ask Mary not to insist on buying? It is OK: what is extracted from a
complement sentence.
c. What did you make a proposal not to buy? It is wrong: what is extracted from
within an island, a nominal sentence.
d. What did you make the proposal that nobody should buy? It is wrong: what is
extracted from within an island, a nominal sentence.
e. What do you think that buying is useless? It is wrong: what is extracted from
within an island, a subject sentence.
f. What do you think that Mary left without buying? It is wrong: what is extracted
from within an island, an adjunct sentence.
g. What do you need that she be calm and buy? It is wrong: what is extracted from
within an island, a coordinate structure.
(77) a. Què vas insistir a comprar?
‘What did you insist on buying?’ It is OK: Què is extracted from a complement
sentence.
b. Què vas demanar a la Maria que no insistís a comprar?
‘What did you ask Mary not to insist on buying?’ It is OK: Què is extracted from
a complement sentence.
c. Què vas fer una proposta de no comprar.
‘What did you make a proposal not to buy?’ It is wrong: what is extracted from
within an island, a nominal sentence.
d. Què vas fer la proposta que ningú comprés?
‘What did you make the proposal that nobody should buy?’ It is wrong: what is
extracted from within an island, a nominal sentence.
e. Què penses que comprar és inútil?
‘What do you think that buying is useless?’ It is wrong: what is extracted from
within an island, a subject sentence.
f. Què penses que va marxar sense comprar, la Maria?
‘What do you think that Mary left without buying?’ It is wrong: what is extracted
from within an island, an adjunct sentence.
g. Què necessites que estigui tranquil·la i compri?
‘What do you need that she be calm and buy?’ It is wrong: what is extracted from
within an island, a coordinate structure.
(78) a. ¿Qué insististe en comprar?
‘What did you insist on buying?’ It is OK: qué is extracted from a complement
sentence.
b. ¿Qué pediste a María que no insistiese en comprar?
‘What did you ask Mary not to insist on buying?’ It is OK: qué is extracted from a
complement sentence.
c. ¿Qué hiciste una propuesta de no comprar.
‘What did you make a proposal not to buy?’ It is wrong: what is extracted from
within an island, a nominal sentence.
142 CHAPTER 7. THE PERIPHERY OF SENTENCE
1. What was Mary convinced that eating wasn’t a good idea? It is wrong: what is extracted
from within an island, a subject sentence.
2. What did Jane insist that every child should learn at school? It is OK: what is extracted
from within a complement sentence.
3. Where did Jane had the belief that Peter hide the money? It is wrong: where is extracted
from within an island, a noun complement sentence.
4. Which knife does Mary knows that Jane stabbed his husband with? It is OK: which knife
is extracted from within a complement sentence.
5. What did Jane make the assumption that everybody wanted to know? It is wrong: what is
extracted from within an island, a noun complement sentence.
6. Whom did Mary confess to Bill her belief that aliens had abducted? It is wrong: whom is
extracted from within an island, a noun complement sentence.
7. What did Mary want John to buy in the store? It is OK: what is extracted from within a
complement sentence.
8. What did Mary asked who brought to the party? It is wrong: what is extracted from within
a wh-island.
The first case is unproblematic: Mary moves to the subject position (it is a passive sentence), and
the object moves to the initial specifier of CP by means of the intermediate position in the lower
CP:
(80) [ CP What [ C’ do [ TP you think [ CP t that [ TP Mary [ VP was given Mary t ]]]]]] ?
The second case is a prototypical that-trace effect: the object what moves to the subject position
(it is a passive sentence), and then it moves to the initial specifier of CP by means of the
intermediate position in the lower CP. However, this leaves a trace in the preverbal subject
position (in red), yielding the that-trace effect:
If the analysis is correct, we expect the sentence in b to be OK if that is dropped: What do you
think was given to Mary? This is indeed the case.
Since only one wh-element can move to the specifier of CP, we have two possibilities:
The difference has to do with the distance involved in each movement: if we move the subject
the distance is shorter than if we move the object. In other terms, a wh-word cannot cross over
another wh-word. This phenomenon was originally labelled as the Superiority Effect.
(85) a. What did Mary bring you from London and you accepted?
b. *What did Mary bring you a present from London and you accepted?
This is a prototypical strong island effect, particularly the Coordinate Structure Constraint by
Ross. Whenever we have a coordination, extraction must take place from both coordinates at the
same time (across the board movement). This is what happens in (85)-a:
(86) What did [ TP Mary bring you what from London] and [ TP you accepted what]?
However, in (85)-b we extract from the second coordinate only, violating the condition:
(87) *What did [ TP Mary bring you a present from London] and [ TP you accepted what]?
(88) [*Which book does she thinks that reading would help you?]
a. Quin llibre pensa que t’ajudaria llegir?
144 CHAPTER 7. THE PERIPHERY OF SENTENCE
This is a case similar to the last one in exercise 1. Again, the wh-word cannot be extracted from
a PREVERBAL subject sentence, which is a strong island. However, if the subject clause appears
in postverbal position, this effect disappears and extraction from this position becomes much
better. Obviously, this second option is generally unavailable in English, where the preverbal
subject position must be filled.
K EY TO EXERCISE 23 Indicate which sentences will be fine and which will be ungrammatical,
and explain the reason for the illformedness of ungrammatical ones. Choose the language you
prefer.
(89) a. Which book did they phone you while reading? Wrong: which book is extracted
from within the adjunct sentence while reading, which is a strong island.
b. Who are they saying that will be killed? Wrong: who leaves a trace in the subject
position of the passive sentence, creating a that-trace effect.
c. How many hours are you required to work after lunch? OK: How many hours is
extracted from the main sentence.
(90) a. Quin llibre et van trucar mentre llegies? Wrong: quin llibre is extracted from within
the adjunct sentence mentre llegies, which is a strong island.
b. Qui diuen que serà assassinat? OK: quin llibre is extracted from the postverbal
subject position (i.e. the VP internal position), which avoids the that-trace effect.
c. Quantes hores has de treballar després de dinar? OK: Quantes hores is extracted
from the main sentence.
(91) a. ¿Qué libro te llamaron mientras leías? Wrong: de qué libro is extracted from within
the adjunct sentence mientras leías, which is a strong island.
b. ¿Quién estan diciendo que será asesinado? OK: quin llibre is extracted from the
postverbal subject position (i.e. the VP internal position), which avoids the that-
trace effect.
c. ¿Cuántas horas has de trabajar después de comer? OK: Cuántas horas is extracted
from the main sentence.
Empty categories
K EY TO EXERCISE 24 Consider the following sentences. Does the wh-phrase leave an intermediate
trace? Can we prove it somehow?
(92) *Which books about Maryi did shei said that John had read?
a. *“Quins llibres sobre la Mariai proi va dir que en Joan havia llegit?”
b. *“¿Qué libros sobre Maríai proi dijo que Juan había leído?”
7.3. EXERCISES 145
We have no way to confirm that we have an intermediate trace, for she would c-command and
bind Mary both in the original position of the wh-phrase and in an intermediate position, yielding
a violation of Principle C in both cases. The evidence would come for a contrast between the
original position and the intermediate one.
(93) Which manuscripts by himselfi did Caesari claim that Cleopatra had destroyed?
a. “Quins manuscrits de si mateixi Cèsari va afirmar que Cleòpatra havia destruït?”
b. “¿Qué manuscritos de sí mismo i Césari afirmó que Cleopatra había destruido?”
Here we do have a contrast. In the original position of the wh-phrase the reflexive is not bound by
Caesar, yielding a violation of Principle A. In contrast, in an intermediate position, Caesar does
bind the reflexive, in accordance with Principle A. Hence, we have evidence that the wh-phrase
stayed in an intermediate position from where binding was possible.
K EY TO EXERCISE 25 Consider the following sentences. Does the wh-phrase leave an intermediate
trace? Can we prove it somehow?
(94) Which books about himi did she said that Johni had read?
a. “Quins llibres sobre elli va dir (ella) que en Joani havia llegit?”
b. “¿Qué libros sobre éli dijo (ella) que Juani había leído?”
As far as the binding of the pronoun is concerned, the only problematic position is the original
one, where the pronoun would be bound in its local domain, violating the Principle B:
(95) did she said that Johni had read which books about himi ?
Once we move the wh-word over the subject John, the problem disappears, for John will no
longer bind the pronoun, as required. Hence, we don’t have any cue whether the movement is in
one step (96) or in two (97):
(96) [ CP which books about himi [ C’ did [ TP she said [ CP [ C’ that [ TP Johni had read which
books about himi ]]]]]]?
(97) [ CP which books about himi [ C’ did [ TP she said [ CP which books about himi [ C’ that [ TP
Johni had read which books about himi ]]]]]]?
In order to see the effect of the intermediate position, we could change him by herself. Then
the reflexive will create a principle A violation in original position (she binds it, but not in its
local domain), and in the final position (she does not bind it) (98). The only way for herself to
be bound by she is landing in the intermediate position, from where she will bind it, as required
(99):
(98) [ CP which books about herselfi [ C’ did [ TP she said [ CP [ C’ that [ TP Johni had read which
books about herselfi ]]]]]]?
(99) [ CP which books about herselfi [ C’ did [ TP she said [ CP which books about herselfi [ C’
that [ TP Johni had read which books about herselfi ]]]]]]?
146 CHAPTER 7. THE PERIPHERY OF SENTENCE
References
147
148 CHAPTER 7. THE PERIPHERY OF SENTENCE
Bibliography
Barss, Andrew & Howard Lasnik. 1986. A note on anaphora and double objects. Linguistic
Inquiry 17. 347–354.
Bloomfield, Leonard. 1933. Language. New York: Holt, Rinehart and Winston.
Bloomfield, Leonard. 1943. Meaning. Monatshefte für deutschen Unterricht 35(3/4). 101–106.
Bosque, Ignacio & Javier Gutiérrez-Rexach. 2009. Fundamentos de sintaxis formal. Madrid ::
Akal. doi:10.1007/s13398-014-0173-7.2.
Burton, Strang & Jane Grimshaw. 1992. Coordination and vp-internal subjects. Linguistic
Inquiry 305–313.
Chomsky, Noam. 1965. Aspects of the theory of syntax. Cambridge, Massachusetts: M.I.T.
Press.
Chomsky, Noam. 1981. Lectures on government and binding. Dordrecht, The Netherlands: Foris
Publications.
Chomsky, Noam & Howard Lasnik. 1993. The Theory of Principles and Parameters. In
Syntax: Ein internationales handbuch zeitgenossischer forschung. an international handbook
of contemporary research, vol. 1, 506–569. de Gruyter. doi:10.1515/9783110095869.1.9.506.
Culicover, Peter W. 2014. The history of syntax. In The routledge handbook of syntax, 483–507.
Routledge.
149
150 BIBLIOGRAPHY
Fillmore, Charles J & Beryl T Atkins. 1992. Toward a frame-based lexicon: The semantics of
RISK and its neighbors. In Frames, fields, and contrasts: New essays in semantic and lexical
organization, vol. 103, 75–102. Hillsdale: Lawrence Erlbaum.
Harris, Randy Allen. 1995. The linguistics wars. Oxford University Press on Demand.
Harris, Zellig S. 1957. Co-occurrence and transformation in linguistic structure. Language 33(3).
283–340.
Hornstein, Norbert, Jairo Nunes & Kleanthes K Grohmann. 2005. Understanding minimalism.
Cambridge University Press.
Huang, C.-T. James. 1982. Logical Relations in Chinese and the Theory of Grammar: MIT
dissertation.
Huang, C-T James. 1993. Reconstruction and the structure of VP: Some theoretical
consequences. Linguistic Inquiry 103–138.
Kayne, Richard. 1994. The Antisymmetry of Syntax. Cambridge: Mass.: MIT Press.
Kayne, Richard S. 1984. Predicates and Arguments, Nouns and Verbs. In Glow newsletter, Paris.
Keenan, Edward & Matthew S Dryer. 2007. Passive in the world’s languages. In Timothy
Shopen (ed.), Clause structure, language typology and syntactic description, vol. 1, 325–361.
Cambridge University Press.
Kenstowicz, Michael. 1989. The null subject parameter in modern arabic dialects. In The null
subject parameter, 263–275. Springer.
Koopman, Hilda & Dominique Sportiche. 1991. The position of subjects. Lingua 85(2-3). 211–
258.
Langacker, Ronald W. 1969. On pronominalization and the chain of command. In Dan Reibel &
Sanford Schane (eds.), Modern studies in english, 160–186. Englewood Cliffs, NJ: Prentice-
Hall.
Larson, Richard K. 1988. On the double object construction. Linguistic Inquiry 19(3). 335–391.
BIBLIOGRAPHY 151
Lasnik, Howard. 2002. The minimalist program in syntax. Trends in Cognitive Sciences 6(10).
432–437. doi:10.1016/S1364-6613(02)01977-0.
Macfarland, Talke. 1995. Cognate objects and the argument/adjunct distinction in English:
Northwestern University dissertation.
Mayol, Laia. 2010. Contrastive pronouns in null-subject Romance languages. Lingua 120(10).
2497–2514.
McNally, Louise. 1992. Vp coordination and the vp-internal subject hypothesis. Linguistic
Inquiry 336–341.
Motut, Alexandra. 2010. Merge over Move and the empirical force of economy in Minimalism.
Toronto Working Papers in Linguistics 33. 1–54.
Perlmutter, David. 1971. Deep and surface constraints in generative grammar. Holt, Rinehart
and W inston, New York.
Pesetsky, David. 1982. Paths and categories: Massachusetts Institute of Technology dissertation.
Pollock, Jean-Yves. 1989. Verb movement, universal grammar, and the structure of IP. Linguistic
inquiry 20(3). 365–424.
Reuland, Eric. 2006. Binding theory: Terms and concepts. In Blackwell companion to syntax.
malden, ma: Blackwell, Wiley Online Library.
Rizzi, Luigi. 1997. The Fine-Structure of the Left Periphery. In Elements of grammar, 281–337.
Springer Netherlands. doi:10.1007/978-94-011-5420-8-7.
Rizzi, Luigi. 2001. Relativized minimality effects. The handbook of contemporary syntactic
theory 89–110.
Rizzi, Luigi & Guglielmo Cinque. 2016. Functional Categories and Syntactic Theory. Annual
Review of Linguistics 2(1). 139–163. doi:10.1146/annurev-linguistics-011415-040827.
152 BIBLIOGRAPHY
Saito, Mamoru. 2012. Sentence types and the japanese right periphery. In Günther Grewendorf
& Thomas Ede Zimmermann (eds.), Discourse and grammar: From sentence types to lexical
categories, 147–175. Walter de Gruyter Berlin.
Truswell, Robert. 2014. Binding theory. In The routledge handbook of syntax, vol. 1, 214–238.
Watanabe, Akira. 1992. Subjacency and S-structure movement of wh-in-situ. Journal of east
Asian linguistics 1(3). 255–291.