Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Relativization, Complementation, And Deep Structure Conjunction In The Grammar Of English
(USC Thesis Other)
Relativization, Complementation, And Deep Structure Conjunction In The Grammar Of English
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
72-6090
NICHOLAS, James Karl, 1939-
RELATIVIZATION, COMPLEMENTATION, AND DEEP
STRUCTURE CONJUNCTION IN THE GRAMMAR OF
ENGLISH.
University of Southern California, Ph.D.,
1971
Language and Literature, linguistic
; University Microfilms, A X E R O X Company, Ann Arbor, Michigan
© COPYRIGHT BY
James Karl Nicholas
1971
THIS DISSERTATION HAS BEEN MICROFILMED EXACTLY AS RECEIVED
RELATIVIZATION, COMPLEMENTATION,
AND DEEP STRUCTURE CONJUNCTION
IN THE GRAMMAR OF ENGLISH
by
James Karl Nicholas
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
Doctor of Philosophy
(English)
September 1971
UNIVERSITY O F SOUTHERN CALIFORNIA
THE GRADUATE SCHOOL
UNIVERSITY PARK
LOS ANGELES, CALIFORNIA S 0 0 0 7
This dissertation, written by
J am es_ _ Karl. Nicholas........................................
under the direction of h D i s s e r t a t i o n Com
mittee, and approved by all its members, has
been presented to and accepted by The Gradu
ate School, in partial fulfillment of require
ments of the degree of
D O C T O R OF P H I L O S O P H Y
t
Dean
Date..§.
DISSERTATION COMMITTEE
Chairman
PLEASE NOTE:
Some Pages have i n d i s t i n c t
p r i n t . Filmed as received.
UNIVERSITY MICROFILMS
ACKNOWLEDGMENTS
Strangely enough, this dissertation started out to be an inves
tigation of the historical development of the English adjective. The
study of the adjective necessarily led to the examination of relative
clauses; relative clauses led to the conjoined structure theory; and
in trying to substantiate that theory, I found nyself struggling with
complementation and presupposition. Thus the prolegomena to the in
tended dissertation became the dissertation itself. Needless to say,
I have had a very understanding dissertation committee. Professors
Finegan, Downing, and Brown have been, at turns, patient, encouraging,
consoling, and demanding— all in the proper amounts. They have not
always agreed with me; they have always kept me scampering.
Particular thanks must go to Ed Finegan, who three years ago
rescued me from Victorian literature by piquing ny curiosity about
language. That is a debt that will be a long time in the repaying,
and I hope that this work marks a beginning.
Special thanks must also go to Sandra Annear Thompson, from whom
I first learned about the possible conjunction source for relatives
and complements and whose ideas I have borrowed freely.
And finally, the unsung heroine of this entire project, disser
tation and Ph. D. program, has been my wife, Veronica, who with the
greatest equanimity has put up with me, reminding me of the ultimate
good, which I, more than occasionally, lost sight of, and who post
poned her cwn ambitions for higher education to earn money in support
of mine. Thanks, Ron, it’s your turn now.
TABLE OF CONTENTS
Page
ACKNOWLEDGEMENTS............................................ii
CHAPTER O N E .................... . . . . . ................ 1
1.0 Introduction.................................... . 1
1.1 Outline........................................ 4
1.2 The Conceptual Framework ......................... 5
CHAPTER T W O .............................................. 8
2.0 Revising the Model........................... 8
2.1 The Standard Doctrine........................ 8
2.1.1 The Restrictive Clause................ 8
2.1.1.1 The Restrictive Relative Analysis 8
2.1.1. 2 Counterexamples.....................12
2.1.2 The Non-restrictive Relative Clause . . . 21
2.1.2.1 Selectional Restriction Analysis 22
2.1.2.2 Conjoined Structure Analysis . . 24
2.1.2. 3 Counterexamples.....................27
2.2 Relatives as Conjoined Structures............28
2.2.1 Straight-forward Conjunction.......... 29
2.2.1.1 Restrictive Relative Clauses . . 30
2.2.1.2 Non-restrictive Relative Clauses 32
2. 2.1.3 Ordering of Conjuncts...........3 3
2. 2.1.4 Postal's Argument.................. 34
Pa5e
2.2.2 Counterexamples.............................. 36
2.2.2.1 Order of Conjuncts .............. 36
2. 2. 2. 2 Numbers.............................. 37
2.2.2.3 The Non-restrictive Analysis . . 38
2.2.3 Basic Elemental Structure ................ 39
2.3 Revision of Straight-Forward Conn unction
Hypothesis............................................41
2.3.1 Order in Deep Structure.....................42
2.3.2 The Handling of Numbers.....................46
2.3.3 Non-restrictive Analysis.....................56
2.4 Further Conclusions...................................70
Footnotes...................................................74
CHAPTER THREE ........................................... 77
3.0 Extension of the Analysis to Complement
Structures.........................................7 7
3.1 Demonstration..........................................77
3.1.1 Complement Formation and Extraposition. . 77
3.1.2 Extraposition from NP Collapsed with
Extraposition .............................. 88
3.1.3 Countering Ross's Argument..................103
3.1.4 Analysis Expanded to Include
Non-restrictives........................... 126
3.1.5 Additional Derivations, Relative Clause
Reduction.................................. 133
3.1.6 Complexity in Derivations ................ 152
3.2 Ordering of Transformations......................15 2
Footnotes................................................ 159
CHAPTER FOUR............................................... 164
4.0 Conjunction Attachment, Prior Conjunct Deletion,
and Presupposition................................164
4.1 Conjunctions and Prior Conjuncts ............... 16 6
v
Page
4.1.1 The Attachment of a n d ......................16 6
4.1.2 Prior Conjuncts.............................167
4.1.2.1 Prior Conjunct Deletion...........167
4.1.2.2 Non-recoverable Deletions. . . .17 3
4.2 Presupposition......................................179
4.3 New Directions......................................187
4.3.1 Comparison with the Prague Circle
Analysis......................................188
4.3.2 Two Avenues for Future Inquiry.............195
4.4 Conclusion.......................................... 197
Footnotes................................................. 198
REFERENCES................................................. 200
APPENDIX....................................................204
CHAPTER ONE
1.0 Introduction
My purpose in this dissertation is to examine
sentences containing relative clauses and noun phrase com
plements— sentences of the following type:
(1.1) a. The man that ate ragweed died of
indigestion.
b. (The claim) That I was a good boy
astonished my mother.
Sentences such as these have presented transformational
grammarians with a variety of problems, not the least of
which is the determination of a suitable deep structure for
them. The relative clause structure and the noun phrase
complement structure both appear to be instances of a
sentence embedded into a noun phrase, a syntactic config
uration something like this:
(1.2)
NP S
1
A phrase structure rule that rewrites NP as NP (S) might be
assumed to generate both (1.1a) and (1.1b).
But if we are to maintain the hypothesis that is
fundamental to all transformational analyses, that sen
tences (surface structures) which have ambiguous readings
have differing deep structures, then this must be the case
for sentences like (1.3).
(1.3) The fact that I know is a secret.
This sentence has one reading if that I know is construed
as a relative clause and another if the same locution is
interpreted as a noun phrase complement. In the first
reading that has as its antecedent the fact, but in the
second that has no antecedent; instead, the fact is the
antecedent of the entire clause that I know.
It would seem clear that the same phrase structure
rule may not be used to generate the structures underlying
relative clauses and noun phrase complements. Various
linguists have solved the problem in varying ways; perhaps
the most popular device being that stated by Jacobs and
Rosenbaum (1968), one which uses the phrase structure rule
NP— ^ NP (S) to generate relative clauses and a slightly
different rule, NP >(Det) N (S), to generate noun phrase
complements.
It is my contention, however, that both relative
clause sentences and noun phrase complement sentences are
3
derived from underlying structures that do not involve
embedding; they are derived instead from structures which
are conjoined S's in the deep structure. This line of
investigation sidesteps the issue posed in the preceding
paragraphs concerning the proper phrase structure rule for
the expansion of NP. Crucial to this analysis is my belief
that the sentences of (1.1) have the paraphrases shown in
(1.4) and that these conjoined structures are more primi
tive, more like the deep structures, than their paraphrases
that involve embedding.
(1.4) a. A man ate ragweed, and the man died
of indigestion,
b. I was a good boy and that astonished
my mother.
I believe that the conjoined structures are more primitive
than their paraphrases which contain embedding because,
although every relative clause sentence and every noun
phrase sentence may have a compound sentence as its para
phrase, the reverse is not true, i.e., the compound sen
tence (1.5) has no relative clause or noun phrase comple
ment sentence for a paraphrase.
(1.5) Mike stole a motorcycle, and Pat washed
the dishes.
T ’
1.1 Outline
The analysis begins in Chapter Two with a reexamina
tion of the standard transformational treatment of relative
clauses, both restrictive and non-restrictive, and several
shortcomings of that treatment are pointed out. Next the
conjoined structure analysis for relative clauses is intro
duced, and several transformations are formulated which
serve to map the conjoined deep structures onto surface
structures containing relative clauses. The transforma
tions which are formulated include Definitization, Restric
tive Relative Clause Formation, Post-definitization
Deletion and Non-restrictive Relative Clause Formation.
The third chapter expands the analysis by demon
strating that the rules of Definitization and Post-
def initization Deletion, transformations formulated in the
preceding chapter, are rules of greater generality, i.e.,
that they are independently motivated because they are
necessary also in the derivation of noun phrase complement
sentences. Such an analysis represents a significant
simplification of the grammar. Additionally, it is shown
that within this analysis the transformational rules of
Extraposition and Extraposition from NP may be collapsed
into one general rule of Extraposition, another measure
that produces greater economy in the grammar.
Chapter Four relates the notion of presupposition
to the conjoined sentences of the deep structure. That is,
for the sentences of (1.1) it must be assumed that the
speaker who utters them, and the hearer who interprets
them, suppose the truth of the underlying initial conjuncts
in the deep structure, namely:
(1.6) a. A man ate ragweed,
b . I was a good boy.
Finally, it is suggested that the notion of presupposition
is in some ways parallel with the Prague School concept of
theme and rheme, new and old information.
1.2 The Conceptual Framework
The conceptual framework of this dissertation is
firmly fixed in the conservative generative-transforma
tional school. I have accepted the hypothesis that the
most adequate grammar contains a finite number of context
free phrase structure rules, which will generate infinitely
many structure trees. I assume that after the phrase
structure rules have generated a syntactic structure,
lexical items are inserted into terminal nodes by lexical
insertion transformations, and that the resulting deep
structures contain all the information that is necessary
for their semantic interpretation.
Interpretive rules of the semantic component of the
grammar interpret these deep structures. The shape of such
rules is unclear to me at the present time, although I
shall tentatively propose one in section 4.3.1.
I assume that the transformational component, which
maps the deep structures onto surface structures, consists
of ordered transformational rules, each having a) a struc
tural index, i.e., a syntactic structure to which it is
sensitive and which may optionally or obligatorily trigger
its b) structural change, a set of orders which may delete,
permute, or add features to nodes of the structural index.
These ordered transformations apply cyclically, running
once through for each S node in the structure tree, begin
ning at the lowest S node. And these transformations are
meaning preserving; they do not alter the meanings of the
deep structures upon which they operate.
I further assume that the transformational component
acts as a filter to block the derivation of ungrammatical
deep structures generated by the phrase structure rules.
That is, it is the transformational component, acting upon
the output of the phrase structure rules , that determines
grammaticality.
The works that best characterize the conceptual
framework for this study are Noam Chomsky's Aspects of a
Theory of Syntax (1965) and John R. Ross's Constraints on
Variables in Syntax (1967b). The reader is expected to
have a thorough familiarity with the former work and at
least a nodding acquaintance with the latter. Exotic terms
such as filters , Chomsky-adjunction, constraints, etc.,
will be used freely in the following pages. Such terms are
fully explained in the works just cited, and I have not
unduly extended the length of this work by pausing to
define those terms which have been so well defined there.
The bibliography which follows the final chapter
represents my readings in transformational syntax which
have had a direct bearing on this work.
CHAPTER TWO
2.0 Revising the Model
In this chapter I shall propose a revised transforma
tional model for the generation of relative clause struc
tures, both restrictive and non-restrictive. I shall begin
by examining the standard transformational analyses for
these two structures, together with the counterexamples
that call their validity into question. I shall then exam
ine a more recently proposed theory which views relative
clause sentences as being derived from underlying conjoined
structures. Here too counterexamples will be discussed.
Finally, I shall propose a new model which bears a great
deal of resemblance to the second type, but without its
shortcomings.
2.1 The Standard Doctrine
2.1.1 The Restrictive Clause
2.1.1.1 The Restrictive Relative Analysis
A restrictive relative clause is characterized as a
sentence embedded as a modifier of a NP. It has in its
structure a NP which is coreferential with the NP it
modifies. The NP in the embedded sentence is replaced by
a relative pronoun or is deleted. The overall effect of
the embedding is that the scope or meaning of the higher
NP is restricted by the embedded modifier sentence. The
deep structure most often ascribed to this construction is
that shown in (2.1). This configuration is produced by a
(2.1) NP
NPi
•NPi
branching rule of the base, which allows NP to be rewritten
as NP (S). The subscripts in (2.1) indicate coreferential-
ity.
Deep structures of this sort underlie the sentences
of (2.2).
(2.2) a. The boy who ate the goldfish received
a reprimand.
b . I reprimanded the boy who ate the
goldfish.
c. I reprimanded the boy whose dog ate
the goldfish.
d. I gave a licking to the boy who ate
the goldfish.
10
e. The boy coughed up the goldfish which
he ate.
More explicitly, the deep structure associated with (2.2a)
would be (2 . 3).
The Relative Clause Formation Transformation applies oblig
atorily to this structure, doing two things: a) it
attaches the features +PR0 and +WH to the lowest NP under
identity and b) it moves that NP to the initial position in
the constituent S. (In this case, the second step of
Relative Clause Formation may be thought of as applying
vacuously since the boy already occupies the initial
position in the constituent S. Notice that in (2.2e)
above, the second step of the transformation would apply
to move the goldfish, which is the direct object in the
constituent S, to the initial position.)
(2.3) S
NP VP
received
V NP
V NP
ate the goldfish
11
The Relative Clause Formation Transformation might
be formalized in the following way:
(2.4) Relative Clause Formation Transformation
X - [ NP[Y - NP - W] ] - Z
NP S S NP
SI: 1
SC: 1
2 3
2 i+
+PR0
+WH
4
3
• (obligatory)
condition: 2=4
After application of relative clause formation the struc
ture tree would have the shape of (2.5).
(2.
N. VP
NP
VP received a reprimand the boy NP.
/'s
the boy
+PRO V
+WH I
N,
ate the goldfish
Thus far it has been shown how restrictive relative
clauses are thought to arise from deep structure embedded
configurations provided for by a branching rule of the base
component which allows a NP to be rewritten NP (S). Such
an embedding accounts for the way in which restrictive
relative clauses are said to restrict the meaning of the
12
NP; that is, the projection rules of the semantic component
will interpret the constituent sentence as an expansion of
the NP which immediately dominates it, thus syntactically
limiting the choice of available lexical distinguishers
associated with the lexical noun which is inserted into the
deep structure. (But cf. p. 33 below.) Thus the ball
which is made of rubber limits the lexical meaning of ball
to "spherical object" by ruling out the possible meaning
"gala affair." (See also Katz S Postal (1964).)
Furthermore, the transformation that maps these deep
structure embeddings onto grammatical surface structures
has been illustrated and formalized.
2.1.1.2 Counterexamples
2.1.1.2.1 The analysis that I have described seems to be
quite satisfactory, but it manages to skirt several rather
sticky issues. For example, take the identity condition on
the Relative Clause Formation Transformation. Just what
constitutes identity? If the deep structure represented in
(2.3) were altered to that shown in (2.6) by substituting
as indefinite article in place of a definite one, would the
identity condition still be fulfilled?
( 2. 6)
HP
ate the goldfish
13
Again the subscripts indicate coreferentiality, but is
coreferentiality sufficient to establish identity, or must
the identity be syntactic as well? If the latter, the
Relative Clause Formation Transformation will not apply,
and the ungrammatical (2.7) will result.
(2.7) * The boy a boy ate the goldfish
received a reprimand.
Perhaps a model that would rule out (2.6) as a suit
able deep structure is just what is desired. This seems to
be what Chomsky is claiming in the following passage (1965:
138-40).
Putting aside questions of formalization,
we can see that not all generalized Phrase-
markers generated by the base will underlie
actual sentences and thus qualify as deep struc
tures. What, then, is the test that determines
whether a generalized Phrase-marker is the deep
structure of some sentence? The answer is very
simple. The transformational rules provide
exactly such a test, and there is, in general,
no simpler test. A generalized Phrase-marker
is the deep structure underlying the sentence
S, with the surface structure M , just in case
the transformational rules generate Ms from M^.
The surface structure Ms of S is well formed
just in case S contains no symbols indicating
the blocking of obligatory transformations. A
deep structure is a generalized Phrase-marker
underlying some well-formed surface structure.
Thus the basic notion defined by a transforma
tional grammar is: deep structure underlies
well-formed surface structure Mg. The notion
"deep structure" itself is derivative from this.
The transformational rules act as a "filter"
that permits only certain generalized Phrase-
markers to qualify as deep structures.
14
Under this interpretation (2.6), although it might be
generated by the base component, will be eliminated from
consideration as a possible deep structure because it will
be filtered out by the transformational rules, specifically
the obligatory Relative Clause Formation Transformation.
But now consider the sentences of (2.8).
(2.8) a. A dog I own bit me.
b. A girl Elmo married got pregnant.
c. A professor Bill admired was fired.
These sentences seem to me to be ambiguous in much the same
way. On one reading I have one dog, Elmo has one wife, and
Bill is rather stingy with his admiration. But on another
reading I have several dogs, Elmo is a bigamist at least,
and Bill might have admired more than one professor. Work
ing under the assumption that surface structures that are
ambiguous have multiple deep structures, one for each pos
sible semantic interpretation, we must assume a set of at
least two deep structures for each of the sentences in
(2.8). (2.9) is my proposal of what these deep structures
must be for (2.8a).
(2.9) a . ____________ S____________
NP VP
nF I s V FTP
a dog NP ^ bit me
I V NP.
1 /H
own a dog
own dogs
The question that must be answered is how does (2.9b)
become the surface structure (2.8a)? It should have been
filtered out by the transformational rules in exactly the
same way that (2.6) was. One alternative is to say that
determiners and number are not crucial to the identity con
dition for the Relative Clause Formation Transformation.
But such a concession would result in monstrosities such as
(2.10).
(2.10) a. *Elmo married some girls who had a
baby.
b. *1 own the geese that laid a golden
egg-
c. *Roscoe ate a catfish that weighed
two pounds apiece.
Chomsky, it seems, anticipated such difficulties. He
suggests (Chomsky, 1965:181) that
features introduced transformationally into
lexical formatives are not to be considered in
determining when deletion is permitted; a form
ative in other words, is to be regarded as a
pair of sets of features, one member of the
pair consisting of features that are inherent
16
to the lexical entry or the position of lexical
insertion, the second member of the pair con
sisting of features added by transformation.
Only the first set is considered in determining
the legitimacy of deletion. Second, what is
involved in determining legitimacy of deletion
is not identity but rather non-distinctness in
the sense of distinctive feature theory. Thus
consider once again the case of "I know several
lawyers"— "Bill is a lawyer." [The sentence
under examination had been "I know several more
successful lawyers than Bill." in which a por
tion of the underlying structure, namely, . . .
is a lawyer of Bill is a lawyer has been trans
formationally deleted m apparent conflict with
the rule of identity.] The predicate nominal
of the latter is not singular, in the base struc
ture; rather it is unspecified with respect to
number exactly as a nasal is unspecified with
respect to point of articulation in the lexical
representations of the formatives king, find,
lamp, etc. Hence, it is not identical with the
corresponding nominal element of "I know several
lawyers"; it is, rather, non-distinct from it,
and the example suggests that it is sufficient
to permit deletion.
And the footnote (p. 2 34) is also pertinent.
I
Similar considerations may account for another
apparent violation of the general condition on
recoverability of deletions. As has frequently
been observed, the identity condition for rela-
tivization involves not only the Noun, and not
the Determiner of the deleted Noun Phrase. Thus
from "I have a [#the friend is from England#]
friend." we can form "I have a friend (who is)
from England." The deleted Noun Phrase is "the
friend", and the problem is the deletion of the
Article, which differs from the Article that is
used to erase it by the relative transformation.
The embedded sentence could not be "a friend is
from England," in which case the problem would
not arise, since definiteness of the Article is
automatic in this position. But the fact that
definiteness is obligatory suggests that in the
underlying Phrase-marker the Article be left
unspecified for definiteness, this being added
by a "redundancy rule" (in this case an oblig
atory transformation). If this is the correct
................... '..17
analysis, then by the principle just established,
deletion of the Article will be permissible,
since in its underlying form it is non-distinct
from the Article of the Noun Phrase of the Matrix
Sentence.
This analysis seems reasonable enough, but it falls
short of solving all the problems I have raised. The
introduction of definiteness by transformational rule could
account for the actual deep structure identity of (2.3) and
(2.6), but in (2.8) the plurality of dogs, girls, and pro
fessors does not arise, so far as I can see, from any sort
of agreement transformation, and must be accounted for in
some other way. And any machinery based on non-distinct
ness that might be introduced to allow the generation of
the sentences of (2.8) would have to be constrained in some
way to prevent the generation of the sentences of (2.10), .
a problem I can see no way out of, given the present formu
lation of Relative Clause Formation. Yet the notion of non
distinctness and its corollary, definiteness by transfor
mation, are interesting, and I shall return to them later.
2.1.1.2.2 The identity condition for relative clause for
mation under the NP (S) analysis runs into further diffi
culty in sentences like the following.
(2.11) a. Fourteen diagrams that Downing drew
were incorrect,
b. All policemen who beat up prisoners
are pigs.
18
Notice that these sentences do not necessarily entail the
the sentences of (2.12).
(2.12) a. Downing drew fourteen diagrams.
b. All policemen beat up prisoners.
c. All policemen are pigs.
While it is possible that Downing did draw only fourteen
diagrams, all of which were wrong, and while it is possible
that all policemen beat prisoners and are, accordingly,
pigs, the speaker, in uttering the sentences of (2.11), is
by no means committed to those readings. He could just as
easily, and perhaps more likely, mean that Downing drew a
number of diagrams, fourteen of which were wrong, and that
a certain number of policemen beat prisoners and that all
of those who do so might be labeled pigs. But only the
first of these readings is possible, i.e., the readings
which entail the sentences of (2.12), if the identity con
dition for Relative Clause Formation is interpreted
strictly, as it is shown in figure (2.4), in which case the
deep structure for (2.11a) would be (2.13).
(2.13) S
Downing V
drew
19
In much the same way, and in violation of the identity
condition, it seems to me that the sentences of (2.14) do
not exactly^ entail those of (2.15). And stranger still,
to me at least, the sentences of (2.16) seem to entail
those of (2.15) rather than those of (2.17).
(2.14) a. A boy who left early missed the fun.
b. Ten girls who left early missed the
fun.
c. Forty-seven mean, miserable zombies
with freckles on their knees who
left early missed the fun.
(2.15) a. A boy left early.
b. Ten girls left early.
c. Forty-seven mean, miserable zombies
with freckles on their knees left
early.
(2.16) a. The boy who left early missed the
fun.
b. The ten girls who left early missed
the fun.
c. The forty-seven mean, miserable
zombies with freckles on their knees
who left early missed the fun.
20
(2.17) a. The boy left early.
b. The ten girls left early.
c. The forty-seven mean, miserable
zombies with freckles on their knees
left early.
To say that the sentences of (2.14) entail exactly and only
those of (2.15), which must be the case if the identity
condition for relativization is to be satisfied, is to rule
out the possibility, and in fact the implication, that some
boys, some girls, and some zombies managed to have some fun
despite an early departure. To say that the sentences of
(2.16) entail those of (2.17) is to overlook the impli
cations , which are quite apparent, that some boys, girls,
and zombies remained and had a very good time indeed. The
sentences of (2.16) entail those of (2.17) only when the
relative clauses are intended non-restrictively, in which
case it is the speaker's intent to be non-committal about
those who remained and to express an opinion only about
those who left. But the sentences of (2.16) do entail
exactly those of (2.15) inasmuch as they imply that only
one boy, only ten girls and only forty-seven zombies left
early.
To capture this difference we would need to claim
that a deep structure like (2.18) underlies (2.14b) while a
deep structure like (2.19) underlies (2.16b).
NP
ten^gxrls
s missed the
fun
NP VP
girls left
early
(2.19) S
NP VP
missed the
fun
NP
A
S
NP
ten girls left early ifi^girl
2.1.1.2.3 Paul M. Postal (196 7) has presented another
argument that weakens the case for the derivation of
restrictive relative clauses from embedded structures.
Because this argument is an appropriate counterexample to
the NP (S) analysis, I mention it here, but because it is
useful also to buttress the case for the conjoined struc
ture theory which I shall elaborate in section 2.2 below,
I have decided to delay until then the presentation of this
analysis. See section 2.2.1.4 (p. 34) below.
2.1.2 The Non-restrictive Relative Clause
Non-restrictive relatives differ semantically from
restrictive relatives in that they do not limit the.NP with
which they share a coreferent; they instead relate
22
additional information about the NP. As such, they are
generally thought to be the result (a) of a selectional
restriction between the determiner of the NP and the fol
lowing relative clause or (b) of a conjoined structure.
2.1.2.1 Selectional Restriction Analysis
In her analysis of relative clauses and determiners,
Carlotta Smith (1969) states that there are three classes
of determiners: unique, specified, and unspecified, and
that there exist selectional restrictions between these
determiners and relative clauses. These restrictions may
be observed in the following paradigm, (2.20).
(2.20) a. Bruce, who is a good bartender, mixed
a Harvey Wallbanger. (Non-restrictive)
b. *Bruce who is a good bartender mixed
a Harvey Wallbanger. (Restrictive)
c. Ed picked up a book, which was about
the sex life of wombats. (Non-restrictive)
d. Ed picked up a book which was about
the sex life of wombats. (Restrictive)
e. The movie, which is about the middle
ages, will interest Bill. (Non-
restrictive)
f. The movie which is about the middle
ages will interest Bill. (Restrictive)
23
g. *Any heavyweight, who attacks Tom,
is a friend of mine. (Non-restrictive)
h. Any heavyweight who attacks Tom is a
friend of mine. (Restrictive)
The unique determiner is the 0 determiner associated
with the Proper Noun, and it allows only non-restrictive
clauses, as the ungrammaticality of (b) demonstrates. The
ungrammaticaliiy of (g) illustrates the selectional
restriction that prevents the unspecified determiners any,
all, each, every, some, etc. from taking non-restrictive
relatives. On the other hand, the specified determiners
the and a/an permit either kind of modification for their
NP. The speaker's intuition then is called upon to deter
mine the uniqueness condition signalled by these deter
miners ; that is, whether the NP is clearly definite and
possesses a known referent (like a Proper Noun) or clearly
indefinite and without an established referent (like any
heavyweight). Once the uniqueness condition of the spec
ified determiner is decided upon, it is a simple matter to
determine the nature of the relative clause that must be
associated with it.
While this analysis certainly captures an undeniable
truth about the uniqueness condition of NP's, it fails to
be a complete analysis. Calling into play the speaker's
intuition concerning the definiteness of specified
m
determiners has the unfortunate effect of shunting off to
the semantic component a decision that perhpas might be
made on syntactic grounds— as we shall see.
2.1.2.2 Conjoined Structure Analysis
John R. Ross (1967b) and George Lakoff (1966) pro
posed that non-restrictive relative clauses be viewed as
conjoined structures. Their analysis hinges on the fact
that the following sentences seem to be paraphrases.
(2.21) a. Enrico, who is the smartest of us
all, got the answer in seven seconds,
b. Enrico, and he is the smartest of us
all, got the answer in seven seconds.
Ross further asserts that non-restrictive clauses are
formed by inserting the right-hand or second conjunct into
the first one, as in (2.22).
(2.22) Non-restrictive Clause
Transformation
—
Optional
and
Enrico is the Enrico got the
answer in seven smartest of us
seconds all
25
got the answer
in seven seconds
TP
Enrico is the
smartest of us
all
Lakoff, additionally, introduces evidence that the
Non-restrictive Clause Transformation is a last cycle rule
because all non-restrictive clauses are conjuncts of the
highest S node. His proof is highly involved, but his most
crucial argument hinges on sentences of the following sort.
(2.23) John denied that Harry, who robbed the
bank, shot Bill.
He begins by positing a plausible deep structure for this
sentence, one that could be generated by any adequate
grammar of English.
(2.24)
NP VP
John
NP VP NP VP
Harry Harry
shot Bill robbed bank
~ 26
He first argues that the non-restrictive clause
transformation rule must not be cyclic because nothing
could prevent it from functioning at the SQ cycle to yield
"Harry, who robbed the bank, shot Bill." which would be
tantamount to (2.25) since, given (2.24), John is denying
both conjuncts.
(2.25) John denied that Harry shot Bill and
robbed the bank.
Since (2.23) and (2.25) are not synonymous, (2.24) may not
be the deep structure of (2.23). The deep structure of
(2.23) must instead be (2.26).
(2.26) S
NP VP NP VP
John V
bank
Harry V NP
I
shot Bill
Thus, not only is the non-restrictive transformation rule
non-cyclic, it must also be associated with the highest S
node. The non-restrictive relative may not, therefore,
occur in embedded position in the deep structure.
A corollary to this analysis is the claim that non-
restrictives, because they are associated with the highest
S node, always embody an assertion by the speaker: in
(2.26) that Harry robbed the bank and in (2.21) that Enrico
is the smartest of us all.
2.1.2.3 Counterexamples
Perhaps the most damaging claims brought against the
Conjoined Structure analysis on non-restrictives are those
mentioned by Ross (1967b:239-241). While it is well known
that a declarative and an interrogative sentence may not be
conjoined, such a conjunction does seem to be possible via
the non-restrictive clause transformation, because such a
conjunction seems to underlie (2.27).
(2.27) a. *Is even Clarence a swinger, and he
wears mauve socks.
b. Is even Clarence, who wears mauve
socks, a swinger?
Other constructions may, of course, be added to the list.
(2.28) a. *Arrest Clarence, and he wears mauve
socks.
b. Arrest Clarence, who wears mauve socks.
c. *Would that Clarence were dead, and
he wears mauve socks.
■ ‘ ~ "" ' 28
d. Would that Clarence, who wears mauve
socks, were dead.
Ross theorizes, without much enthusiasm, that the under
lying structure for sentences like (2.27) might be two
contiguous but non-conjoined sentences in discourse. I
find this not implausible in view of claims which I shall
make later.
It also seems that Ross is incorrect in his claim
that the non-restrictive clause transformation moves the
right hand conjunct into the left. For (2.29) seems as
good as (2.21), and the sentences are in all respects
synonymous.
(2.29) a. Enrico, who got the answer in seven
seconds, is the smartest of us all.
b. Enrico, and he got the answer in
seven seconds, is the smartest of
us all.
But this claim is a bit too strong, as we shall see.
2.2 Relatives as Conjoined Structures
Recently several linguists1 * have proposed a strik- i
ingly different analysis for relative clauses, one that
posits underlying conjoined structures for both restric
tive and non-restrictive relatives. In this section I
shall recount briefly the arguments in favor of such an
29
analysis, cite the counterexamples that have been proposed,
and relate how Thompson (1970) has adjusted the original
theory, and in some ways impaired it.
2.2.1 Straight-forward Conjunction
This is the name that Sandra Annear Thompson gave to
her earlier treatment (1967) of relative clauses. Postal
(1968) referred to a similar process as 'repetition
bindingThe theory is based upon the anaphoric occur
rence of the definite article.
(2.30) a. I saw a dog this morning. This
afternoon I saw the dog again.
b. I saw a dog this morning, and this
afternoon I saw the dog again.
c. I saw a dog this morning. This
afternoon I saw iit again.
d. I saw a dog this morning, and this
afternoon I saw it again.
The preceding paradigm demonstrates two principles: a)
that definitization is quite similar to pronominalization, !
and b) that definitization, like pronominalization, may
cross sentence boundaries. The first of these principles
is of immediate interest here.
30
i2.2.1.1 Restrictive Relative Clauses
Thompson’s straight-forward conjunction source for
restrictive relative clauses claims that sentences like
(2.31) will have deep structures like (2.32), two conjoined
sentences, each containing at least one coreferential NP.
(2.31) Kim typed the paper which he had just
written.
2.3
Kxm had just
written a paper
Kim typed a paper
The derivation of (2.31) proceeds as follows: a) S-^ is
optionally embedded into S2 by copying to yield (2.33).
(2.33)
NP NP-
a paper a paper
■NP
a paper
b) As a result of anaphora, the determiner a in S2 is
optionally changed to the.
31 :
(2.34)
^— NP—^
a paper
NP
the paper
a paper
c) And is optionally deleted.
(2.35) S
NP- NP
Z\ L--NP—^
the paper a paper
a paper
Thus the deep structure (2.32) will yield not only the sur
face structure (2.31) but three other possible surface
structures, if we assume that a restrictive relative clause
forming transformation has applied.
(2.36) a. Kim typed a paper which he had just
written.
b. Kim had just written a paper and he
typed the paper which he had just
written.
c. Kim had just written a paper, and he
typed a paper which he had just written.
And if the optional transformation does not apply to create
the restrictive relative, then the conjoined or compound
sentence will surface in one of the following variations.
(2.37) a. Kim had just written a paper, and
2.2.1.2 Non-restrictive Relative Clauses
The obvious advantage of such an analysis, and one of
the strongest arguments for it, is that the reverse of the
Restrictive Relative Clause Formation Transformation
applied to the same deep structure, i.e., to embed Sg into
S-p yields a semantically equivalent non-restrictive clause
sentence, (2.38),
(2.38) Kim had just written a paper, which he
which is derived in the following way, discussed above in
2.1.2.2.
he typed the paper,
b. Kim had just written a paper, and
he typed it.
typed.
written a paper a paper Kim typed
a paper
I........" '..' . 33'"
2.2.1.3 Ordering of Conjuncts
The concept of linear ordering of conjuncts in deep
structure puts the whole discussion of relative clauses
into a new perspective. It has always been the contention
of grammarians, of whatever bias, that restrictive clauses
limited the scope or meaning of a NP while non-restrictive
clauses enlarged or added to that scope or meaning. But
until now the terms limiting and adding had vague conno
tations at best. My earlier reference (pp. 11-12) to
semantic projection rules which interpreted the NP in terms
of its embedded modifier was as puny a straw-man as was
ever devised, because non-restrictive relatives place no
less stringent interpretive restrictions on the NP's they
modify. Now the terms limiting and adding may be under
stood in syntactic terms. Restrictive relatives limit the
NP because they embody information originally contained in
a left-hand conjunct, and hence occurring within the sen
tence prior5 to the information in the right-hand conjunct,
which is eventually to become the surface structure matrix.
Conversely, non-restrictive relatives add information about
the NP because their original position as right-hand con
juncts demands that they follow the initial conjunct in
linear or temporal order.
i2.2.1.4 Postal’s Argument
Yet another argument for preferring the conjoined
analysis over the embedding analysis comes from Postal
(1967).6
This sentence has two possible readings, one in which
Charley is mistaken, and the other in which Charley thinks
a contradiction. One possible deep structure that a stand
ard grammar of English might generate for this sentence
would be the following:
(2.40) Charley assumed that the book which was
burned was not burned.
(2.41) S
NP VP
Charley V NP
assumed
NP VP
A
was not
burned
NP VP
book
was
burned
Now the question is which of the readings does this deep
structure represent? It could not be the reading in which
Charley is mistaken because that reading involves a dual
Iclaim on the part of the speaker; that is, a) that Charley
assumed something, and b) that the book was burned. As we
have seen (see pp. 25-27 above) this type of claim is most
logically represented by a conjoined structure at the top
most S, such as (2.42).
NP VP NP
Charley V book was burned
assumed it
NP VP
book
was not
burned
This must mean that (2.4-1) is the deep structure for the
reading in which Charley is contradicting himself. But if
this is the case, the grammar is contradicting itself as
well. We have just seen that the preferred method for
representing a dual assertion on the part of the speaker is :
a conjoined structure embodying each of the assertions as a
conjunct. What is good for the speaker should be good for
Charley, who in his confused state is making two assertions,
namely, that the book was burned and that it was not burned.
The true deep structure for this reading would be not
(2.42) but (2.43).
36 !
(2.43) S
NP VP
NP Charley V
assumed i
S S
NP VP NP VP
book book
was not
burned
was
burned
(2.41) would then be the intermediate structure that would
result from the application of the relative clause trans
formation, which has been illustrated in (2.32) - (2.35).
2.2.2 Counterexamples
Straightforward conjunction analysis finds its
severest critic in one of its original authors, Sandra
Thompson.
2.2.2.1 Order of Conjuncts
She does not find her original argument concerning
the order of conjuncts altogether convincing.
. . . there does not seem to be any linguistically
motivated reason why the difference between two
sentences like
(10) The driver who was speeding caused
the accident.
(11) The driver, who was speeding, caused
the accident.
should be related to the difference between the
following constructions:
37
(12) S
a driver was a driver caused
speeding the accident
(13) S
a driver caused a driver was
the accident speeding
In other words, if there is any significance to
the order in which conjuncts appear in conjunction,
it is far from clear that this has anything to do
with the distinction between restrictive and non-
restrictive clause sentences. (Thompson, (1970:
44-1+5))
2.2 . 2.2 Numbers
She further claims that the "inability of the
straightforward conjunction analysis to provide an under
lying representation which correctly accounts for the mean
ing of numbers in relative clause sentences is its crucial
detracting feature" Thompson (19 70:46). A sentence such as
(2.44) is by no means equivalent in meaning to the sen
tences of (2.45).
(2.44) Two boys who came to the party have
beards.
(2.45) a. Two boys came to the party. Two
boys have beards.
b. Two boys came to the party, and they
have beards.
Referential indices are not enough, for although the under
scored NP's in (2.45) are coreferential, the sentences
imply that only two boys came to the party, an implication
not included in (2.44).
2. 2. 2. 3 The Non-restrictive Analysis
The third counterexample for the straightforward con
junction analysis concerns non-restrictive clauses and is
similar to the one brought against Ross's analysis cited
earlier in 2.1.2.2. It appears that the claim that only
the second conjunct may be used in non-restrictive relative
clause formation is too strong. Thompson adds the clincher
with the following argument (1970:6 8-69):
(2.46) a. George noticed that Marjie refused
the candy, and he didn't take any
either.
b. George, who didn't take any either,
noticed that Marjie refused the
|
candy.
c. George, who noticed that Marjie |
|
refused the candy, didn't take any
either. '
1 ' ".... ............ ...... ... .. 39]
| d. *George didn't take any either, and
he noticed that Marjie refused the
; candy.
The argument makes use of the stressed either, which seems
to apply only to second conjuncts. The second conjunct can
be used successfully to derive the non-restrictive relative
sentence (b) from (a), but notice- that (c) cannot be so
derived. To maintain the second conjunct hypothesis would
necessitate the positing of (d) as the underlying conjoined
structure, a construction that is unacceptable because it
violates the either rule. Apparently, what must be done is
to derive both (b) and (c) from (a), allowing the second
conjunct to become the non-restrictive clause in the former
case and the first conjunct to become the non-restrictive
clause in the latter.
2.2.3 Basic Elemental Structure
As a result of such shortcomings, Thompson has
adopted a new theory which still maintains relative clauses
as derived from underlying conjoined structures, but
regards the ordering of the conjuncts as not semantically
justified; she regards numbers as predicates, and chooses
to disregard determiners. To do this she posits highly
abstract underlying structures, which she calls Basic
Elemental Structures (BES), having the following properties:
1 “..................... 40
1) No lexical nouns are represented.
2) They represent the basic relations between ele
ments of a sentence.
3) They represent the basic relationships among the
simplex components of a complex sentence.
She adopts a notation similar to Fillmore’s (196 8) to
capture these basic semantic relationships.
An example of this notation and the preferred anal
ysis may be seen in (2.47), her BES for sentence (2.44),
which I repeat here for ease of reference.
(2.44) Two boys who came to the party have
beards.
(2.47) BES: (came (boys)D (party)^) .
(have (beards)Q (boys)D) .
(two (boysset)0)
The BES consists of three essentially unordered but con
joined propositions, each with a predicate (came, have, and
two) and a varying number of arguments. The arguments are
not lexical nouns but predicates of a more abstract struc
ture (although this is not readily apparent from the
notation). The subscripts represent case relationships
(following Fillmore (1968)). The BES may be interpreted in
the following way: There is a set of boys such that they
are two in number and such that they came to the party and
: “ ..' ........ ' "" 41
isuch that they have beards. Subsequent transformations
will map (2.47) onto (2.44).
Having established the BES, Thompson goes on to
examine a wide variety of data concerning relative clause
structures and notes that they all yield quite nicely to
this revised hypothesis. She says, "By assuming a level at
which two sentences which differ only in that one contains
a definite determiner where the other contains an indefi
nite one have identical underlying representations, it
becomes possible to highlight these structural properties
which relative clause sentences exhibit independently of
whether the head noun is definite or indefinite."
2.3 Revision of Straightforward Conjunction Hypothesis
The BES analysis just described strikes me as a case
of throwing out the baby with the bath water. Thompson has
abandoned the most insightful part of her theory for a more
powerful version, one so powerful that she can do anything
she wants with it. Unfortunately, this new analysis does
not allow her any longer to distinguish between restrictive
and non-restrictive relatives when the coreferents of the
latter are other than a proper name. If the propositions
of (2.47) are unordered with respect to each other, then
there can be no appeal to the "prior information" contained
in a left-hand conjunct which restricts the information
contained in a following or right-hand conjunct. (Cf. the
42
discussion of the advantages arising from ordering of
conjuncts in deep structure given on p. 33). By scrapping
the theory of ordering, she foregoes as well the treatment
the definite articles in relative clause sentences as cases
of anaphora, another insightful analysis.
What I propose to do in this section is to renovate
the straightforward conjunction hypothesis so that it will
overcome the complaints brought against it, or to show, in
some cases, that the counterexamples are merely apparent.
The result will be a weaker hypothesis than the BES, but
one which will give a more adequate explanation of the
data, including the transformational placement of definite
determiners, and an underlying differentiation between
restrictive and non-restrictive relatives.
2.3.1 Order in Deep Structure
We shall begin by examining more closely the straight
forward conjunction analysis proposed by Thompson (et al.).
It will be recalled that restrictive relative clauses were
thought to be derived from sentence conjuncts by left to
right embedding and non-restrictive relatives by right to
left embedding. The operations are illustrated in (2.48)
and (2.49).
(2.48) Restrictive Clauses
a. Kim had just written a paper, and
Kim typed a paper. >
43
b. Kim had just written a paper and Kim
typed the paper which he had just
written. (or)
c. Kim typed the paper which he had just
written.
(2.49) Non-restrictive Clauses
a. Kim had just written a paper and Kim
typed a paper. = = = >
b. Kim had just written a paper, which
he typed.
Thus each restrictive relative clause has a non-restrictive
clause sentence as its paraphrase. And the restrictive
clause lives up to its name because it embodies information
which in deep structure was a conjunct ordered before the
surface structure matrix, and hence it is this notion of
prior knowledge that limits the coreferent which the super
ordinate sentence contains. The non-restrictive nature of
the non-restrictive clause is nicely captured also because
the non-restrictive clause embodies information that in
deep structure was a conjunct ordered after the surface
structure matrix, and hence it contains only additional
information about the coreferent contained in the super
ordinate sentence.
Yet to account for the relationships of relative
clauses on the basis of order is a mistake, according to
44
Thompson. She reasons that the difference between the
sentences of (2.50) must have some stronger motivation than
the order of underlying conjuncts, as may be seen in (2.51).
(2.50) a. The driver who was speeding caused
the accident,
b. The driver, who was speeding, caused
the accident.
(2.51) a. A driver was speeding and a driver
caused an accident,
b. A driver caused an accident and a
driver was speeding.
I think that this is only partly true. There are sentences
in which the ordering of conjuncts in deep structure is of
negligible bearing, while in others it makes considerable
difference. The sentences of (2.52) are like those of
(2.50) and (2.51) with respect to the unimportance of the
order of conjuncts, but those of (2.53) are quite different.
(2.52) Surface Structures
a. The Laurel and Hardy movie which we
saw was hilarious.
b. The Laurel and Hardy movie, which
we saw, was hilarious.
Deep Structures
c. We saw a Laurel and Hardy movie, and
it was hilarious.
d. A Laurel and Hardy movie was hilar
ious, and we saw it.
(2.53) Surface Structures
a. The boy who became ill took the
medicine.
b. The boy, who became ill, took the
medicine.
Deep Structures
c. The boy became ill, and the boy took
the medicine.
d. The boy took the medicine, and the
boy became ill.
In (2.53) sentences (a) and (c) imply that the medicine is
an expected cure, while sentences (b) and (d) imply that
the medicine is the cause of the affliction.
Also I would suggest that in analyzing the sentences
of (2.50) and (2.51) Thompson did not pay strict attention
to the process of anaphoric definitization. Anaphora may
not work backwards in conjunct sentences, so Non-restrictive:
Relative Clause Transformation will not yield sentence
(2.50b) but (2.54) .
46
(2.5 4) A driver, who was speeding, caused the
accident.
The difference in meaning, slight though it is, is attrib
utable to the order of conjuncts.
2.3.2 The Handling of Numbers
Thompson further says that the inability of straight
forward conjunction to handle the meaning of numbers in
relative clause sentences is its crucial detracting feature.
And as evidence she presents a relative clause sentence
like (2.55a), which would, according to her earlier
analysis, have as its deep structure the underlying con
juncts of (2.55b).
(2.55) a. Two boys who came to the party have
beards.
b. Two boys came to the party, and they
have beards.
She notes that sentence (a) does not entail the conjuncts
of (b), which, according to straight-forward conjunction
analysis, it should.
If this is the case— if (2.55a) does not have a
synonymous conjoined structure like (2.55b)--then what can
be said about (2.55b)? Is it ungrammatical? Hardly.
Given the straight-forward conjunction analysis, though,
we might expect that a structure such as (2.55b) might
; 47
still be developed into a semantically equivalent relative
clause structure via the Restrictive Clause Formation
Transformation, which we saw earlier on pp. 30-31. I think
that we can demonstrate this possibility quite easily.
Compare (2.56) with (2.48), which I repeat here.
(2.48) a. Kim had just written a papery and
Kim typed a paper.; . -— ■>
b. Kim had just written a paper and
Kim typed the paper which he had
just written. (or)
c. Kim typed the paper which he had
just written.
(2.56) a. Two boys^ came to the party and
two boys| have beards. >
b . Two boys came to the party and the
two boys who came to the party have
beards. (or)
c. The two boys who came to the party
have beards.
The conjoined structure shown in (2.55b) and (2.56a) does
indeed have a semantically equivalent relative clause sen
tence paraphrase--(2.5 6 c). Now, if the relative clause
sentence paraphrase of (2.55b) is (2.56c) and not (2.55a),
may we assume that (2.55a) has a conjoined structure
48
paraphrase? I think so, and I think that a clue to what
that structure is may be found by returning to examine the
sentences of (2.14-2.16). Recall that the sentences of
(2.16) seemed to entail exactly those of (2.15); e.g.,
(2.16b) entailed precisely (2.15b), while (2.14b) did not;
rather it implied that perhaps more than ten girls may have
left early.
(2.14) b. Ten girls who left early missed
the fun.
(2.15) b. Ten girls left early.
(2.16) b. The ten girls who left early missed
the fun.
These data seem to show that relative clause sentences like
(2.16), whose relative clauses follow a noun preceded by
the definite article, have in their underlying structure
conjoined sentences, each containing the head noun, and
that those head nouns are identical— with respect to
referential index and number. That is, the conjoined
structure underlying (2.16b) would be (2.57), a structure
quite similar to those shown in (2.48a) and (2.56a).
(2.57) Ten girls^ left early and ten girls^
missed the fun.
49
On the other hand, the relative clause sentence (2.14b),
which does not imply that the same number of girls as left
early missed the fun, must have a different conjoined
structure underlying it. The sentence does state that ten
girls missed the fun, but it remains noncommittal as
regards the number who left early. Therefore, we would
expect the underlying conjuncts to be something like (2.58)
(2.58) Girls left early and ten girls missed
the fun.
Similarly, in the sentence (2.55a), the statement is
fairly clear that two boys have beards, but not specific
about how many boys came to the party. We might, therefore
on the analogy of (2.58), posit (2.59) as the conjoined
structure underlying (2.55a).
(2.59) Boys came to the party and two boys
have beards.
If the conjoined structures of (2.58) and (2.59) are indeed
the underlying forms of (2.14b) and (2.55a), respectively,
then the Restrictive Clause Formation transformation must
be able to relate these structures to their relative clause
sentence paraphrases, just as it related the sentences of
(2.48) and those of (2.56). The way in which it related
the sentences of (2.48) and of (2.56) might be formalized
as follows:
(2.60) Restrictive Relative Clause Formation
[X NP Y] ; [W NP
S S S
Z]
S
SI: 1 2 3 6 (optional)
SC: (1 2 3) 4 5 [ 2 1 3] 6
+def +def
+WH
+PR0 condition: 2=5
Optional Instruction:
Delete the first conjunct
1 2 3
This formalism may be interpreted as follows: if two
conjoined S’s share a coreferential NP, then the first S
may be recopied and Chomsky-adjoined to the coreferential
NP of the second S; the coreferential NP of the copied S
will be moved to the initial position in that S, and the
features +definite, +WH, and +PR0 will be affixed to it,
while the feature +definite will be affixed to the corefer
ential NP of the second S. Optionally, the first conjunct
may be deleted.
While this transformation will operate satisfactorily
to explain the derivation of the relative clause sentences
of (2.48bSc) and (2.56bSc), it will not account for the
relative clause sentences of (2.14b) and (2.55a), which I
have just supposed were derivable from (2.58) and (2.59).
The reason is immediately clear. The NP's in each of the
latter sentences are not coreferential, a condition which
51
is crucial to the structure index of the transformation as
it is stated in (2.60).
What adjustment must be made in order to make this
analysis work? Or must the analysis be abandoned, as
Thompson has suggested? I think that we may save the
analysis and gain a certain explanatory advantage as well,
if we keep, at least temporarily, the transformation shown
in (2.60) and examine the sentences of (2.58) and (2.59)
more closely. We have just noted that these sentences fail
to meet the structure index for the Restrictive Relative
Clause Formation transformation, (2.60), because their con
juncts do not share coreferential NP's. But this need not
be the case. There are two ways of interpreting the sen
tences of (2.58) and (2.59). The second conjunct of each
may be interpreted as a complete non-sequitur, in which
case the judgment passed in the preceding paragraph--that
the conjuncts do not share a coreferential NP--is quite
surely justified. Yet I do not feel that the sentences
will be interpreted in this way very often. More likely
than not they will be interpreted in the following way.
(2.58) will be interpreted to mean that a certain unspe
cified number of girls left early and that of that unspe
cified number who left early, ten missed the fun. And
(2.59) will be interpreted to mean that an indeterminate
number of boys came to the party and of that number, two
have beards.
What I am claiming is that the most likely inter
pretation of (2.58) or (2.59) is one in which a corefer
ential NP may be thought to exist in each of the conjuncts.
In order to demonstrate this claim I rely on a recent
analysis of the phenomenon of plurality in which D. T.
Langendoen (1969:138), (1970:117) has proposed that the
plurality of NP's may be thought of as a conjunction of
nouns, each being identical except for referential indices.
We might thus think of the deep structure of (2.56)
as a structure tree having the following shape, (2.61), in
which the subscripts below the lexical items represent
referential indices.
(2.61)
NP vp VP
came to
the party have beards
boy
367
boy
368
boy
368
boy
367
Because the NP in dominates two conjoined N's, each
having a referential index which is matched by those of the
two conjoined N's dominated by the NP in S2 , we may think
of the two conjunct S's as sharing a coreferential NP and
as satisfying, therefore, the structure index of the
Restrictive Relative Clause Formation transformation.
I now claim that (2.5 8 ) and (2.59) also display a
similar coreferentiality with respect to shared NP's.
(2.62), I propose, is the structure tree that represents
(2.59).
(2.62)
NP VP VP
have
beards the party |
boy. . -boyn boy. . .boy3 6 ?,boy368. . *boyn
The bracketed N's represent a conjunction of N's, making up
a set within an unspecified number of members, n. At the
point in the derivation of the sentence when lexical inser
tion occurs, the lexical item boy is inserted n times to
complete the sets dominated by NP^ and NP2 , respectively.
And if each of the N's of the set dominated by NP^ has an
equivalent N, with respect to referential index, in the
set dominated by NP2 , then the sets may be thought to be
coreferential. At the same time, and in some way that is
yet unclear to me, two lexical items inserted in the set
dominated by NP2 must receive a special specification or
marking. The result is that the sets dominated by NP-^ and
NP2 are identical and coreferential, and thus SQ answers
the structure index for Restrictive Relative Clause Forma
tion, with the following result:
NP VP VP
[N .
|n the party NP have
beards boy. . .boy
boy. . .boy3 6 7 boy3 6 8 boy.
+def
'Pi ' VP
came to
the party
boy. . .boy
i _ _ _ _ _ _ _ Hi
+def
+WH
+PRO
NP^’ and NPj are assigned the definite marker, and the
former the additional markers, +WH and +PRO. At this point
the two specially marked N's are abstracted from the set
dominated by NP2 and moved to the left of it, and an appro
priate feature segment is transformationally attached to
NP25 which will eventually surface as the preposition of.
The result, if S1 is optionally deleted, is the familiar
partitive construction (2.64), which, at least to me, is
synonymous with (2.55a),
(2.64) Two of the boys who came to the party
have beards.
(2.55) a. Two boys who came to the party have
beards.
just as (2.65) is a paraphrase of (2.14b).
(2.65) Ten of the girls who left early missed
the fun.
(2.14) b. Ten girls who left early missed the
fun. 7
This analysis not only reveals that the straight
forward conjunction analysis can indeed handle numerals,
but provides at least a partial statement of the manner
which partitives are derived.
This analysis also explains sentences (2.11) and
(2 .8 ), which, as we have seen, the standard transforma
tional doctrine handles unsatisfactorily. The sentences
of (2.11) have as their paraphrases those of (2.67),
(2.11) a. Fourteen diagrams that Downing drew
were incorrect,
b. All policemen who beat up prisoners
are pigs.
(2.67) a. Fourteen of the diagrams that Downing
drew were incorrect,
b. All of the policemen who beat up
prisoners are pigs.
and the sentences of (2 .8 ) have as their paraphrases (para
phrases of at.least one set of ambiguous readings) those of
(2 .6 8 ).
(2.8) a. A dog I own bit me.
b. A girl Elmo married got pregnant.
c. A professor Bill admired was fired.
(2.68) a. One of the dogs I own bit me.
b. One of the girls Elmo married got
pregnant.
c. One of the professors that Bill
admired was fired.®
2.3.3 Non-restrictive Analysis
Now that the matter of numerals has been clarified,
the only remaining impediment to the straight-forward con
junction analysis is that concerning the order of conjuncts
in non-restrictive clauses. It was demonstrated (on p. 28
and again on p. 38) that the transformation which creates
non-restrictive relative clauses does not always move the
right-hand conjunct into the left-hand conjunct of the
underlying structure. However, a closer look at the
examples cited, (2.29) and (2.H6), reveals that the option
of left to right non-restrictive embedding is possible only
when the shared coreferents are uniquely specified (Proper
Names) or otherwise marked as definite.
; .... ~.... ' ... 5 7 !
I contend that the conjoined sentences shown in (2.69)
do not yield to the left to right non-restrictive embedding.
(2.69) Tom gave Bill a cookie, and the cookie
tasted awful.
According to straight-forward conjunction analysis, left to
right, or restrictive embedding, yields (2.70), while right
to left, or non-restrictive embedding, yields (2.71).
(2.70) The cookie which Tom gave Bill tasted
awful.
(2.71) Tom gave Bill a cookie, which tasted
awful.
To say that (2.70) might derive from (2.69) as a result of
non-restrictive embedding, even if appropriate pauses were
inserted, is to have the transformation operate counter to
the way in which anaphoric processes are generally thought
to behave.
At this point I must digress, but I shall return
momentarily (p. 6 *+) to the matter at hand. It was men
tioned earlier (pp. 29-30) that the straight-forward con
junction analysis made crucial use of the process of
anaphora. This was so because it was thought that the same
process which attached the definite article to dog in
(2.30aSb) allowed that NP to be pronominalized in (2.30cSd).
.... '.... 58
It was also felt that the same process was involved in
Restrictive Relative Clause Formation. In (2.32) it was
assumed that the coreferential NP's were unmarked for
definiteness in the underlying conjoined structure, and
that as an optional part of the Restrictive Relative Clause
Formation transformation the definite article was attached
to the coreferent NP in the second conjunct.
I disagree with this part of Thompson's analysis.
Anaphora must obligatorily definitize the NP dog in (2.30a).
Otherwise the resultant sentence (2.72) could only be
interpreted to mean that the dogs were not identical.
(2.72) I saw a dog this morning, and this
afternoon I saw a dog again.
If we are correct in assuming that the status of NP's in
deep structure is that they are unmarked for definiteness,
the operation of anaphoric definitization must be oblig
atory and not optional.
Assuming this premise, (2.36a) and (2.36c) are not
possible derivations from the deep structure shown in
(2.32), and, indeed, I do not find them to be acceptable
paraphrases of (2.31) or (2.36b).
The occurrence of definitization as one of the
several processes involved in the Restrictive Relative
Clause Formation transformation is a result of the con
current operation of anaphora. As the transformation is
now conceived, i.e., (2.60), the initial conjunct is first
Chomsky-adjoined to the coreferent NP of the following con
junct. Then anaphoric definitization takes place. The
coreferent NP in the second conjunct, together with the
coreferent NP in the attached copy of the first conjunct
receive the definite marker. Because the latter NP
receives the definite marker, it may be subsequently
replaced by the relative pronoun. (See p. 64 below for
further discussion of this point.)
At this point a problem arises. While the Restric
tive Relative Clause Formation transformation satisfacto
rily accounts for the restrictive relativization of the
conjunct sentences we have seen so far, it must be admitted
that anaphoric definitization is a process far more general
than that of relativization. Consequently, it might be
wise to establish Definitization as a separate transforma
tion and Restrictive Relative Clause Formation as a subse
quent transformation, one that applied to the output of the
former. It is clear that such a transformation is needed
in the grammar anyway to account for the occurrence of the
definite article in such sentences as (2.30a).
Definitization might be stated in the following way:
(2.73) Def initization
[X NP Y] ; [W NP
S S S
Z]
S
SI: 1 2 3 4 5 6 1 -> (obligatory)
SC: 1 2 3 4 5 [1 2 3] 6
+def +def
Condition: 2=5
This is a much simpler transformation than Restric
tive Relative Clause Formation was. Here if two conjoined
sentences of the deep structure share a coreferent, then a
copy of the first conjunct is Chomsky-adjoined to the
coreferential NP of the second conjunct, and that NP,
together with the coreferent in the copy, is assigned the
definite marker. As we have noted above, this transforma
tion must be obligatory.
The output of Definitization would provide the struc
ture index for an optional Restrictive Relative Clause
Formation transformation. That transformation might be
stated as follows:
(2.74) Restrictive Relative Clause Formation
SC: (1 2 3) 4 5 7 6 8 9
+WH
+PRO
Condition: 5=7
Optional Instruction:
Delete 1 2 3.
[X NP Y] ; [W NP [ X NP Y] Z]
S S S +def S +def S S
SI: 1 2 3 4 5 6 7 8 9=»(optional)
Here the definitized coreferent in the copied S is moved
to the front of that S, and the features, +WH and +PRO, are
assigned. Optionally, the initial conjunct may be deleted.
As an illustrative example of the way these two
transformations complement each other, we may begin with
the deep structure of (2.32) again.
Definitization moves into S2, Chomsky-adjoining it to
NP2. NP2 and NP^' receive the definite marker.
(2.32) S
had just
written
i
paper±
NP VP NP VP
Kim V NP
had just
written
papery
+def NP VP
NP1» Kim V
had just paper,
written +def
62
Restrictive Relative Clause Formation will now move NP-^!
to the initial position in S-^' , adding the features +WH and
+PR0, and optionally deleting S^. If is deleted, the
highest node, Sq , which no longer branches, would be
deleted by the general convention of pruning, described by
Ross (1969a). The resulting structure would be (2.76) and
its surface structure manifestation, (2.77).
(2.76)
typed NP
paper
+def NP VP
Kim paper
+def
+WH had just
+PRO written
(2.77) Kim typed the paper which he had just
written.
Should Restrictive Relative Clause Formation not
apply, then a subsequent transformation is needed to oblig
atorily delete the copy of S-^ embedded by Definitization.
This transformation might be called Post-definitization
Deletion and could be formalized this way:
(2.78) Post-definitization Deletion
[X NP Y] * , [W NP [X NP Y] Z]
S S S +def S +def S S
SI: 1 2 3 4 5 6 7 8 9^( obligatory)
SC: 1 2 3 4 5 0 0 0 9
Condition: 5=7
This transformation would apply to the structure tree shown
in (2.75) in the event that Restrictive Relative Clause
formation did not apply, and the resulting sentence that
would surface would be (2.79).
(2.79) Kim had just written a paper and Kim
typed the paper.
Notice that the only overt signal that remains after
Definitization and Post-definitization Deletion have
applied is the definite article, which marks the coreferent
in the second S as having occurred in an earlier sentence.
We have posited a very general transformation which
accounts for the presence of definite articles in conjoined
sentences that share a coreferent NP, and which accounts
equally well for the appearance of the definite article in
restrictive relative clause sentences. The rule of
Definitization is followed in the transformational cycle in
the optional rule of Restrictive Relative Clause Formation,
which is in turn followed by Post-definitization Deletion,
which is obligatory. Furthermore, it is only by virtue of
: .....'..."" " 64
the operation of Definitization that Restrictive Relative
Clause Formation or Pronominalization may occur to change
(2.75) into (2.77) or (2.79) into (2.80).
(2.80) Kim had just written a paper and Kim
typed it.
Postal (196 9) and Kuroda (1969) have both argued convinc
ingly that the presence of the definite marker is a pre
requisite for pronominalization. If their claims are valid,
and I believe that they are, then this is another reason
for establishing Definitization as a transformation ordered
cyclically prior to Restrictive Relative Clause Formation,
for were there no mechanism to definitize the copied
coreferent NP embedded in the succeeding conjunct, then
there would be no motivation for attaching the feature +PR0
to it, an operation that must occur if it is to become a
relative pronoun.
Having proposed this much of an alteration of
Thompson's straight-forward conjunction analysis, we may
return to the matter of non-restrictive clauses.
I have contended that the sentence (2.69) is not
subject to left to right non-restrictive embedding,
although examples of such embedding did occur in figures
(2.29) and (2.46).
(2.69) Tom gave Bill a cookie, and the cookie
tasted awful.
; ' 65 '
Actually, (2.69) is the sentence as it might have surfaced
after undergoing Post-Definitization Deletion. Let us
return to the deep structure of this sentence, which I
assume must be something like (2.81),
(2.81) Tom gave Bill a cookie^ and a cookie^
tasted awful.
where the underscored NP's are coreferents.
Definitization will apply to yield (2.82).
(2.82) Tom gave Bill a cookie and the cookie
[Tom gave Bill the cookie] tasted awful.
And if Restrictive Relative Clause Formation applied
optionally next, (2.83) would result.
(2.83) The cookie which Tom gave Bill tasted
awful.
Should Restrictive Relative Clause Formation not apply,
then Post-definitization Deletion would apply obligatorily
to yield (2.69).
(2.69) Tom gave Bill a cookie and the cookie
tasted awful.
At this point the last cyclic rule of Non-restrictive
Relative Clause Formation may apply to add the features +WH
; ' ' " 66
and +PRO to the coreferent NP in the second conjunct, which
has already been definitized. The result will be (2.84).
(2.84) Tom gave Bill a cookie, which tasted
awful.
And it is only because the coreferent in the second con
junct had received the definite marker as a result of the
earlier application of the Definitization transformation,
that it is now capable of being assigned the additional
features +WH and +PR0, which will cause it to surface as a
relative pronoun. And it is precisely because the
coreferent NP in the first conjunct has not received the
definite marker as a result of the Definitization trans
formation, or any other transformation, that it is incapa
ble of participating in any pronominalization operation,
including left to right non-restrictive embedding.
We might thus state the Non-restrictive Relative
Clause Formation transformation in the following way:
(2.85) Non-restrictive Relative Clause Formation
[X NP Y] ; [ W NP Z]
S S S +def S
SI: 1 2 3 4 5 6 1 ~~> (optional)
SC: 1 2 3 5 4 6
+WH
+PR0
Condition: last cycle
2 = 5
i ” " .........." 67 '
making the definiteness of the coreferent NP in the second
conjunct an essential factor in the structure index of the
transformation. A situation to the contrary would run
counter to the anaphoric process; i.e., a structure of the
type (2.86) or (2.87) would be impossible, given the cor
rectness of the definitization process— which I have just
outlined— and whose operation is based on the general pro
cess of anaphora.
(2.86) [X NP Y] ; [W NP Z]
S +def S S -def S
1 2 3 4 5 6 where 2=5
(2.87) [X NP Y] ; [W NP Z]
S -def S S -def S
1 2 3 4 5 6 where 2=5
But (2.85) is not an entirely satisfactory formulation
because it overlooks a possible structure index, i.e.,
(2.88),
(2.88) [X NP Y] ; [W NP Z]
S +def S S +def S
1 2 3 4 5 6 where 2=5
and this is precisely the one that is involved in sentences
like (2.29) and (2.46), in which the coreferent NP’s are
Proper Names, or NP1s previously marked as definite, such
as might appear in a sentence like (2.89).
68
(2.89) The cow, which I sold for some beans,
has a crumpled horn.
In (2.29) and (2.4-6), the referents of the coreferent NP’s
are uniquely specified. We might assume that the proper
names receive the definite marker by redundancy rule. In
(2.89), the cow, we must assume, has been definitized
earlier by some prior application of the Definitization
transformation, so that immediately preceding the operation
of the Non-restrictive Relative Clause Formation trans
formation there must have been an intermediate structure of
the following form.
(2.90) The cow has a crumpled horn and I sold
the cow for some beans.
But this sentence answers to the structure index (2.88) as
do the sentences of (2.29) and (2.47), and like them may
participate in left to right non-restrictive embedding, to
yield (2.91),
(2.91) I sold the cow, which has a crumpled
horn, for some beans.
which, I must admit, is fully synonymous with (2.89).
We must therefore adjust the Non-restrictive Relative
Clause Formation transformation to read as follows:
(2.92) Non-restrictive Relative Clause Formation
[X NP Y] 5 C W NP Z]
S Idef S S +def S
SI: 1 2 3 5 6 (optional)
SC: 1 2 3 5 4 6
+WH
+PRO
Condition: last
cycle 2=5
(or)
SC: 0 0 0 5 6 + 2 1 3
+WH
+PRO
Condition on second
SC: 2 = +def
Notice that as it stands (2.92) contains no mechanism for
embedding the relativized conjunct into the matrix. I do
not think that such an operation should be included as part
of the Non-restrictive Relative Clause Formation trans
formation, for remember the synonymy of the sentences of
(2.21) .
(2.21) a. Enrico, who is the smartest of us
all, got the answer in seven seconds,
b. Enrico, and he is the smartest of us
all, got the answer in seven seconds.
I think that a preferable strategy would be to leave (2.92)
as it stands and to formulate a subsequent transformation
that would embed not only the output of Non-restrictive
70
Relative Clause Formation but also a second conjunct, one
which has not undergone Non-restrictive Relative Clause
Formation.^
2.4- Further Conclusions
At this point all the impediments offered against the
straight-forward conjunction analysis have, I think, been
successfully countered. The system has admittedly been
altered slightly but not beyond recognition. The system
does offer a satisfactory explanation for the derivation of
restrictive and non-restrictive relative clauses, and it
does not neglect the handling of determiners, as does
Thompson's analysis.
Because the analysis I have offered assumes the
underlying structure of restrictive and non-restrictive
clauses alike to be conjoined sentences, I am not as dis
concerted by the sentences of (2.27) or (2.28) as Ross is.
Indeed all that needs to be stated is a constraint to the
effect that if one of the underlying conjuncts contains a
question, an imperative or an exclamation marker, then the
sentence must undergo either restrictive or non-restrictive
relativization— or some other such transformation11— which
would prevent it from surfacing as a compound sentence.
Additionally, this analysis serves to explain the
ambiguity of such sentences as (2.8). We have seen that
the reading which gives me several dogs, Elmo several
71
wives, and Bill several professors, has underlying it con
junct S’s containing coreferential sets of NP's which may
optionally emerge as a partitive construction after the
operation of Restrictive Relative Clause Formation.
(2.93) a. One of the dogs I own bit me.
b. One of the girls Elmo married got
pregnant.
c. One of the professors Bill admired
was fired.
On the other hand, the readings which give me one dog, Elmo
one wife, and Bill one professor, are the result of the
application of Non-restrictive Relative Clause Formation.
The path for this reading of (2.8a) is traced from its deep
structure in the paradigm which follows.
(2.94) a. A dog^ bit me and I own a dog^,
Definitization, followed by Post
Definitization Deletion
b. A dog bit me and I own the dog.
Non-restrictive Relative Clause
Formation
c. A dog bit me, which I own. — /
Conjunct Incorporation
d. A dog, which I own, bit me. "y
Relative Pronoun Deletion
e. A dog I own bit me.
I am claiming that the only restrictive interpretation
available for a relative clause following a head noun pre
ceded by the indefinite article is one in which a multi
plicity of referents is assumed, from which one is singled
out, as the paraphrases, shown above in (2.93), of one set
of readings of (2.8),— the restrictive interpretation—
clearly demonstrate. If a multiplicity of referents is not
assumed, the relative clause is in no way restrictive, but
offers only commenting, additional, or non-restrictive
information. In other words, the first set of readings for
the sentences of (2.8) contains information concerning the
number of dogs, wives, and professors in their initial con-
juncts, information to the effect that the number of each
is indeterminate. The nature of this information is sub
stantially altered by the second conjunct, which limits the
number of dogs, wives, and professors to one. No such
limiting or restricting takes place in the second set of
readings. A predication is made concerning a dog, a wife,
and a professor in the initial conjuncts, and further
information concerning them is added in the second.
Thompson (1970:52) has also remarked this peculiarity
in relative clauses when she pointed out ’in an indefinite
sentence like
(M-3) I saw a child who was carrying a
huge dog get out of a Volkswagen.
restrictiveness plays no part at all.’ The only way in
which such a clause could be construed restrictively would
be in the event that several children were carrying a huge
dog, a possibility, certainly, but one so remote as to be
ruled out by most listeners.
The possibility that the embedding of a non-restric
tive clause into a noun head preceded by an indefinite
article may create a structure index that would permit the
transformational erasure of pauses and the subsequent
substitution of that for the relative pronoun, or even the
deletion of the relative pronoun is highly interesting,
even a little alarming, and certainly requires further
study.
Finally, the revised straight-forward conjunction
analysis I have proposed renders the anomalous sentences of
(2 .1 0 ) acceptable, or provides a suitable surface structure
for their underlying structures.
(2.95) a. Elmo married girls, one of whom had
a baby.
b. I own the geese, one of which laid
a golden egg.
c. Roscoe ate one of the catfish that
weighed two pounds apiece.
CHAPTER TWO
Footnotes
1. The Relative Clause Formation Transformation, in
addition to attaching the features +WH and +PR0 to the
coreferent NP in the embedded S and permuting that NP
to the initial position in that S, also erases the sen
tence boundary markers (#) that Chomsky in his analysis
conventionally places on either side of the embedded S.
If Relative Clause Formation does not apply, as it may
not to (2 .6 ), then the sentence boundaries will remain,
and the resulting sentence will be filtered out by a
general convention which recognizes structures contain
ing such boundary markers as not well formed (Chomsky
1965:138).
2. By exact entailment I mean that the antecedents of the
relative pronouns in the sentences of (2.14) must be
interpreted to be exactly one boy, exactly ten girls,
and exactly forty-seven zombies. This eventuality,
which is necessary if the identity condition for rela-
tivization is to be met, gravely limits the interpre
tation of the sentences, however. Now the sentences
can mean just the following: that only one boy, and
only ten girls, and only forty-seven zombies left early.
And they are not open to the interpretation that more
than one boy, or more than one girl, or more than
forty-seven zombies left early— an interpretation that
I feel is reasonable for the sentences of (2.14).
3. The sentences of (2.29) do not constitute sufficient
counterevidence to support such a claim, i.e., that
Ross is incorrect in assuming that non-restrictive rela
tive clauses always derive from a right-hand conjunct,
(as long as the conjuncts are reversible, there is no
strong argument for claiming intrinsic rightness or
leftness of conjuncts in deep structure), but see the
sentences of (2.46) below (pp. 38-39).
74
75
4. Postal (1967), (1968); Thompson (1967), (1970);
Thompson (1969); Langendoen (1969), (1970).
5. The notion of 'prior conjunct' rests on my belief that
whatever we may feel about the process involved in
thinking, and hence speaking, we must admit that the
events involved in each process occur consecutively
rather than simultaneously. More specifically, if we
assume the basic unit of communication to be the sen
tence, then it is my additional hypothesis that in
strings which consist of more than one sentence, these
sentences must be linearly ordered. These sentences in
deep structure are arranged consecutively as conjuncts
and are subsequently acted upon by transformations,
which either preserve the coordinate relationship or
produce embedding. Because the phrase structure rules
of the base generate sentences in a random way, without
regard for order, we may not construe the order of con
juncts in deep structure to be a function of syntax.
Thus in the following pages when I mention 'prior con
juncts' or 'earlier generation' I am making use of a
convenient fiction, a distinction that appears to be
syntactic because of the way in which structure trees
are conventionally drawn, but which is in reality a
matter of interpretation, and hence semantics. That is,
the order of conjuncts in deep structure, although
established randomly by the base rules , is of semantic
significance to the speaker who interprets the deep
structure and assigns readings which are crucially
dependent on order. More will be said concerning the
peculiar status of 'prior conjuncts,' those 'generated
earlier,' in Chapter Four.
6 . I hope that in fabricating this argument I am doing
justice to Postal. Try as I might, I have not been
able to lay hands on his article, which contains the
sentence about Charley and the accompanying argument
for conjoined structure. I have come across his exam
ple, however, several times in my reading, and each
time the allusion has been made that his sentence is a
strong argument for conjoined structure and against
deep structure embedding. The argument is never fully
recapitulated though; so what follows is Postal's argu
ment only if I have construed the facts as he did. If
I have deviated, the concoction is my own.
7. Actually the structure after the operation of Restric
tive Relative Clause Formation and the operation that
abstracts the marked lexical items is more nearly some
thing like (2 .6 6 ).
76'
(2.66) a. Two boys of the boys who came to the
party have beards.
b. Ten girls of the girls who left early
missed the fun.
These structures may then be reduced transformationally
to either (2.64) or (2.55a), or, (2.65) or (2.14b),
respectively.
8 . The fact that sentences of (2.8) behave similarly to
those containing numerals, i.e., (2.11), (2.5 5a), and
(2.14b), together with the fact that the sentences of
(2 .6 8 ), which contain the numeral one are synonymous
with those of (2.8), seems to bear out Perlmutter's
(196 8 ) contention that the numeral one and the indefi
nite article are realizations of the same lexical item.
9. For further speculation concerning the manner in which
this might be accomplished, see Chapter Four, p. 19 6 .
10. For the formulation of this transformation, see Chapter
Three, p. 143.
11. Examples of other such transformations will be treated
in the following chapter.
CHAPTER THREE
3.0 Extension of the Analysis to Complement Structures
In the preceding chapter I have formulated a principl
of Definitization, a process by which semantic information
from an earlier conjunct in deep structure is transforma
tionally inserted into a subsequent conjunct. I have
demonstrated that such an analysis is helpful in accounting
for the occurrence of restrictive and non-restrictive
clauses in English. In this chapter I intend to show that
the rule of Definitization is a rule of even greater gener
ality than has thus far been demonstrated, that it is
independently motivated by other requirements of the
grammar.
3.1 Demonstration
3.1.1 Complement Formation and Extraposition
Consider the following sentences:
(3.1) a. That Mike has neat tattoos is well
known.
B. Gill proposed that classes should be
dismissed.
77
78
c. Bruce agreed that he would read the
paper.
d. The claim that Terry slugged Gloria
is false.
e. Ed thought that Bill knew that Tom
believed that Max advocated that
Estelle should buy a wombat.
Each of these sentences is an example of what standard
transformational theory would treat as sentences containing
noun phrase complements. The deep structures for these
sentences would be represented by trees of the following
types.
(3.2) a S
NP VP
it S
is well
known NP VP
Mike V NP
has neat
tattoos
b S
Gill V
I
proposed it
NP
should^e
S
NP
classes
dismissed
Bruce would reac
the paper
I
slugged Gloria
e. S
Ed V NP
./X
thought it S
Bill V NP
1
knew it S
Tom V
I
believed
advocated it
Estelle V
should a
buy wombat
It has generally been assumed that noun phrase com
plement constructions are generated by a base rule which
allows NP to be rewritten as NP (S) or (Det) N (S), at any
rate, a rule similar to that which provides for the occur
rence of relative clauses. It has long been recognized,
however, that noun phrase complements and relative clauses
are quite different constructions inasmuch as in the latter
the S must contain a NP coreferential with the NP in the
expansion, whereas in the former there is no such require
ment. Numerous attempts have been made to capture the simi
larity of these constructions while preserving the differ
ence or distinction, the most often used being the assign
ment of the base rewrite rule NP— ► NP (S) as the generator
of relative clauses and the rule NP — ► (Det) N (S) as the
generator of noun phrase complements. But these methods
seem arbitrary at best.
Worse still, is the treatment of certain lexical
items which appear in the structure trees of noun phrase
complement sentences. I refer to the it, which is gener
ally called an 'abstract pronoun1 , and the ''complementizer'
that, which is transformationally introduced by the rule of
Complementizer Placement. Again these designations seem to
me thoroughly arbitrary and do not in any way account for
lexical items, that is, the 'abstract pronoun' is equiva
lent in phonological shape with the personal pronoun it,
and the 'complementizer' that is equivalent to the definite
81
article that. That one should be treated as an abstract
entity or dummy symbol and that the other should be treated
as a sort of pseudo-conjunction strikes me as counter
intuitive and, worse, lacking in explanatory power.
I believe that my analysis will correct this arbi
trary treatment of lit and that and supply the explanatory
power that the current theory is lacking. I propose that
the rule of Definitization applies to deep structures under
lying noun phrase complement sentences in such a way as to
account for the occurrence of it as a personal pronoun and
for the occurrence of that as a definite article. The deep
structure for the sentences of (3.1), however, must neces
sarily be construed as more abstract than those represented
in (3.2).
I claim that the deep structure underlying (3.1a)
would be approximated by something like (3.3),^-
(3.3) S
S
2
NP VP NP VP
Hike
is well
known
NP VP
Mike V NP
has
neat
tattoos
82
and that this structure meets the structure index for the
Definitization Transformation. The Definitization Trans
formation, it will be recalled has the following formalism:
If the S, Mike has neat tattoos, is in deep structure
a prior conjunct of a second S, Mike has neat tattoos is
well known, then Definitization will recognize the equiv
alence or identity of the first conjunct S and the S domi
nated by NP in the second conjunct S. The only adjustment
that needs to be made to the transformation as it now
stands is that the scope of the identity condition must be
2
expanded to read: 2=5 or 1 2 3=5.
If this is the case, the Definitization Transforma-
tion will render the intermediate structure (3.5)
(3.4-) [X NP Y] ; [W NP
S S S
SI: 1 2 3 4 5
Z] Definitization
g ----------------
6 =>(obligatory)
SC: 1 2 3 4 5 [1 2 3] 6
+def +def
Condition: 2=5
n:
I
Mi:
/ \ is w e l l
N P V P k n o w n
Mike,?
Mike V
+def
neat
tattoos
+def
83
and mark each of the subtrees, Sg and , dominated by NP
in the second conjunct as +definite in the same way that it
marked coreferential NP’s in the examples in the preceding
chapter. Compare the above operation with that which
occurs in the derivation of the relative clause sentence
(3.6).
(3 .6 ) a. The alligator which ate a cannibal
got sick.
NP NP VP
alligator NP alligator
got sick
ate cannibal
(obligatory) Definitization
VP NP
alligator V NP NP
got sick
'VP cannibal NP ate
alligator
+def NP
alligator I
+def ate cannibal
To the structure (3.5), formed by Definitization,
Post-definitization Deletion would apply to yield (3.7).
84
NP VP
Mike V S NP
is well
known has
/ X NP
neat I
tattoos Mike
VP
has / X.
neat
I __________________tattoos |
+def
A subsequent transformation would optionally delete the
first conjunct and another would copy the feature +definite
to the left of the sentential subject in the second con
junct, sister adjoining it to S3, which segment would be
realized in the surface structure as the article that, the
article which must accompany NP's that are S's and which
have been definitized by the process just described. The
surface structure manifestation would then be the desired
one, (3.1a).
(3.1) a. That Mike has neat tattoos is well
known.
Looking again at the Post-definitization Deletion
Transformation, we recall that its formalism, as stated in
the previous chapter, was that which is repeated here as
(3.8).
85
(3.8) [X NP Y]; [W NP [X NP Y] Z] Post-
S S S +def S +def S S definitization
Deletion
SI: 1 2 3 4 5 6 7 8 9 U p (obligatory)
SC: 1 2 3 4 5 0 0 0 9
Condition: 5=7
This formalism serves equally well for the process just
described, provided that the identity condition be expanded
to the statement: 5=7 or 5=6 7 8 . This is the same trans
formation that converts (3.7b) into (3.10),
(
VP NP NP
I
alligator alligator
+def
NP
got sick
ate cannibal
which will have the surface structure manifestation (3.11).
(3.10) An alligator ate a cannibal, and the
alligator got sick.
Notice that this general process, which I have called
Post-definitization Deletion, has obviated the necessity
for the transformation of It Deletion, which, according to
standard theory, was required to erase the so-called
'abstract pronoun.' This analysis also gives a reasonable
explanation for the occurrence of it. It is not an
86
'abstract pronoun' but an actual personal pronoun derived
from a definitized NP, the NP which was an S. It is
derived simply (by Postal's rules (Postal, 1969)) by the
introduction of the feature +PRO into its article segment,
with subsequent deletion of the NP.
It will appear in the surface structure when Extra
position applies to the structure tree which is the product
of the Definitization Transformation. Going back to figure
(3.5), we can see how the Extraposition Transformation
operates.
(3.11) a
(3.5)
is well
known Mike
Mike
Mike
neat
tattoos,
neat
tattoos
J
+def +def
Extrapos ition (optional)
V
87
b* Sn
Mike V
is well
known
Mike V
Mike V NP
has / tattoos
neat
tattoos| | _____ ____________
+def +def
In the second conjunct the first occurrence of the defini-
tized sentence NP will be reduced by pronominalization to
it, and the second will bear the definite marker still,
which will ultimately produce the article that. The first
conjunct may be transformationally deleted to effect a
structure which will manifest itself at the surface level
as (3.12), a sentence synonymous with (3.1a).
(3.12) It is well known that Mike has neat
tattoos.
That the 'abstract pronoun' it and the 'complemen
tizer' that should, in this analysis, turn out to be a
personal pronoun and an article, respectively, is hardly
surprising. They have the same phonological shape as the
personal pronoun and the article, and this treatment which
88
assumes that they are nothing more or less than they seem
to be, strikes me as being very naturally motivated, while
the standard theory is uneconomical in its multiplicity of
lexical items.
3.1.2 Extraposition from NP Collapsed with Extraposition
A further argument for my analysis is that the Extra
position Transformation I have just described, which may be
formalized as (3.13), may be used to account for the trans
formational movement of relative clauses as well as noun
phrase complements.
(3.13) Extraposition
X [NP S] Y
NP NP
SI: 1 2 3 4 "~S> (optional)
SC: 1 2 0 4+3
Compare the operation just illustrated in (3.11) with the
following one, which may be performed on the structure
shown in (3.6b).
Extraposition (optional)
>/ +PR0
1
alligator V NP alligator
1 +def got sick NP VP
ate cannibal
alligator V NP
+def | |
+WH ate cannibal
+PRO
With the deletion of the first conjunct, the surface struc
ture manifestation will be (3.15).
(3.15) The alligator got sick which ate a
cannibal.
This analysis again has the effect of reducing the
number of transformations required by the grammar by allow
ing the Extraposition Transformation shown in (3.13) to do
the work of two transformations of the standard theory:
Extraposition and Extraposition from NP.
A reasonable question at this point might be the
following: If the two processes of extraposition can be
successfully collapsed into one transformation in this way,
why was it not done long ago?
One argument (Burt, 19 71) against such an analysis
seems to be the belief that Extraposition must precede
Restrictive Relative Clause Formation and, as such, could
90
not operate on the output of the latter transformation.
The argument for the ordering of Extrapos ition prior to
Restrictive Relative Clause Formation runs as follows. To
generate the surface structure (3.16) the deep structure
(3.17) must be assumed (at least according to standard
theory).
(3.16) The cow which it is obvious that Roscoe
bought has a crumpled horn.
(3.17)
has a crumpled
horn
S is obvious
2
Roscoe V
bought cow
It has been argued that if Restrictive Relative Clause
Formation were to apply before Extraposition to such a
structure tree, the following configuration would result.
91
(3.18)
NP has a crumpled
horn
cow
NP NP
which
it is obvious
NP VP
Roscoe bought
Extraposition would then move S2 to the end of S-^ to yield
(3.19).
(3.19)
has a crumpled
horn cow
NP NP
is obvious which it
Roscoe bought
And this does, in fact, produce the desired surface struc
ture. But there is a problem. According to standard
theory, Restrictive Relative Clause Formation is an oblig
atory transformation while Extraposition is only optional.
92
Should Extraposition not apply, It Deletion would obliga
torily apply to (3.18) to yield (3.20),
(3.20)
VP NP
NP
has a crumpled ____ has
V? horn
is obvious
NP cow
which NP
Roscoe bought
which would have the ungrammatical surface structure (3.21).
(3.21) *The cow which that Roscoe bought is
obvious has a crumpled horn.
Because of this difficulty it is necessary either to
make the Extraposition Transformation obligatory under
certain circumstances or to remedy the situation by
reordering the transformations. By adopting the latter
course, we get the following results. Applying Extra
position first to (3.17), we derive (3.22).
(3.22)
NP
NP
has a crumpled
horn VP cow
VP NP
is obvious
NP
I
cow
Roscoe V
bought
93
Restrictive Relative Clause Formation may now apply to move
the lowest identical NP cow to the initial position in Si
and to attach +WH and +Pro to it to yield (3.23).
(3.23) SQ
NP " " " " " " ” ' " VP
NP has a crumpled
horn
NP NP VP cow
which is obvious NP VP
Roscoe bought
But strangely enough the same difficulty seems to persist
and has gone undetected in Burt’s analysis. If we go back
to the deep structure (3.17) and choose the option not to
extrapose S2, It Deletion will obligatorily apply and
Restrictive Relative Clause Formation, being obligatory,
must now take place to produce (3.20), the very structure
which the reordering of transformations was designed to
exclude.
Thus, it appears that the relative ordering of
Restrictive Relative Clause Formation and Extraposition
is still an open question since only the option of applying
Extraposition will yield the desired results, and, as I
have shown, it is irrelevant whether that option is taken
before or after the application of Restrictive Relative
Clause Formation. In either case the structure index for
Restrictive Relative Clause Formation is met, and because
it is obligatory, it must apply. It may not act as a
filter to eliminate an ungrammatical reading.
At this point it might be proposed that the second
alternative be adopted, that of constraining the Extra
position Transformation in such a way that it must apply
obligatorily in certain environments, an adjustment that
has already been proposed for sentences whose matrix VP is
one of a set of verbs including seem. J. R. Ross has set
up a constraint which seems to explain the present diffi
culty, but he does not make the constraint a condition on
the Extraposition Transformation. He proposes a general
constraint for the grammar of English, which he calls the
Sentential Subject Constraint: 'No element dominated by
S may be moved out of that S if the node S is dominated by
an NP which itself is immediately dominated by S.' (Ross,
1967b: 134) In the deep structure of (3.17) the NP cow in
S2 is dominated by an NP which is itself dominated by S^.
The constraint effectively blocks Restrictive Relative
Clause Formation until after Extraposition has moved S2
from under the dominance of the NP dominated by S^. Then
Restrictive Relative Clause Formation may apply. Is this
sufficient evidence for assuming that Restrictive Relative
Clause Formation follows Extraposition? Apparently so.
95
But let us now assume that the deep structure under
lying (3.16) is not (3.17), as was originally proposed
consistent with the standard theory, but is instead (3.24).
(3.24)
Roscoe V
IP VP cow has a crumpled
~ horn
NP Sj. is obvious
bought cow
Roscoe V
bought cow
This configuration of simplexes shows the relative order of
their generation, that is, left to right. The initial
predication about Roscoe is that he bought a cow, the next
that he bought the cow was obvious, and finally that the
cow had a crumpled horn. We may think of the transforma
tional cycle as applying in the normal way to these con
joined S's, that is, beginning at the lowest S node. No
transformations of any significance will apply at Sg, S3 ,
or S^. At S-j_ Definitization must obligatorily apply to
copy S3, which is identical with Sg , to the right of the NP
dominated by S^.
(3.25)
cow
NP NP
is obvious
Roscoe V
I
bought cow
Roscoe V
NP Roscoe V
bought cow | | ______ bought cow ^
has a crum
pled horn
+def +def
Extraposition may now optionally apply to move Sg' around
the VP is obvious.
(3.26)
VP
i / N
Roscoe V NP
1 I S is obvious NP
bought cow ^
cow has a crum
pled horn
Roscoe V
| bought cow|
Roscoe V
I
| bought cow|
+def
97
And pronominalization will reduce Sg to i/t.
This exhausts the transformations which are relevant
to the cycle. No transformations of any significance
occur at S2 . At SQ Definitization must again apply to move
into Sg under the dominance of the subject coreferential
NP. This will produce the structure shown in (3.27).
has a crum
pled horn cow
+def NP
is /
obvious NP
bought cow it is
obvious
NP VP I |
I bought cow
Roscoe V NP | ___________ +def|
I I
| bought cow | +def
+ def
And Restrictive Relative Clause Formation may now apply to
move the coreferential NP cow in Sg" to the initial
position in 1, attaching the features +WH and +PR0 to
yield (3.28).
V NP it is
obvious
/ \ + P R O
V + S ^ i s
obvious |
NP VP
bought cow
NP VP
' A
|Roscoe bought|
+def
Roscoe V NP
bought cowj
+def
If we assume the presence in the cycle of an optional
transformation which would delete the preceding conjunct
after Definitization has applied to copy that conjunct into
a subsequent S, then assuming the application of such a
transformation, all that would remain of the structure
shown in (3.28) would be S^, which would surface as the
desired sentence. Otherwise the prior conjuncts would
surface also to yield the synonymous (3.29).
(3.29) Roscoe bought a cow, and it is obvious
i
that Roscoe bought a cow, and the cow
which it is obvious that Roscoe bought
has a crumpled horn. j
99
Notice also that at the point in the derivation
illustrated by (3.26), it was a rather arbitrary choice
that was made in selecting as the conjunct to be moved
by Definitization. That choice was, of course, motivated
by the desire to derive the intended surface structure
(3.16). Actually Definitization might just as easily apply
on this cycle to Sg, S^’, or S^, each of which contains the
coreferential NP cow. The correct analysis would be that
the instruction to the structural change of the Definiti
zation Transformation is to apply to the highest S node
consistent with the structure index, in this case,
Then assuming no transformational deletion of prior S nodes
occurs, the derived surface structure would be (3.30).
(3.30) Roscoe bought a cow, and it is obvious
that Roscoe bought a cow, and the cow
which Roscoe bought and which it is
obvious that Roscoe bought has a crum
pled horn.
If, however, Sg were deleted on the cycle,
would be pruned, making Sl ^ the highest node consistent with
the structure index of Definitization at the SQ cycle. The
derived surface structure would then be (3.31), assuming
that is not deleted on the Sq cycle.
100
(3.31) It is obvious that Roscoe bought a cow,
and the cow which it is obvious that
Roscoe bought has a crumpled horn.
If S1 + is, in fact, deleted on the Sg cycle, (3.16) will
surface.
(3.16) The cow which it is obvious that Roscoe
bought has a crumpled horn.
In this analysis, there is no need to assume that
Extraposition must precede Restrictive Relative Clause
Formation, because the conjunct analysis assures that there
will be no circumstance under which the two rules will come
into conflict. If Extraposition does not apply on the S^
cycle at the point in the derivation shown in figure (3.25),
Post-definitization Deletion will apply to delete Sg*.
Definitization will be blocked at the Sg cycle by Ross's
Sentential Subject Constraint, and the derived surface
structure will be (3.32), not exactly a good sentence
because cow in the
(3.32) ?Roscoe bought a cow, and it is obvious
that Roscoe bought a cow, and a cow has
a crumpled horn.
final conjunct has not been definitized. But notice that
while the Sentential Subject Constraint prevents the
movement of by Definitization, Sg may still be moved.
This being the case, cow in the final conjunct may be
definitized and Restrictive Relative Clause Formation or
Post-definitization Deletion may apply to relativize or to
delete the copy of Sg. Should Sg have been deleted, Defi
nitization would nonetheless still be able to apply to move
9 *
3 •
Since in this analysis the ordering of Extraposition
and Restrictive Relative Clause Formation does not appear
to be crucial, the path still lies open to collapse Extra
position and Extraposition from NP. Thus at the S^ cycle
in (3.28) the collapsed Extraposition Transformation could
conceivably apply after Restrictive Relative Clause Forma
tion to move the relative clause to the right. Assuming
that prior conjuncts have been deleted, this would yield
(3.33).
(3.33) The cow has a crumpled horn which it
is obvious that Roscoe bought.
This sentence is perhaps acceptable in most dialects, but
it must be admitted that the sentence is ambiguous because
it could be construed as meaning that it was obvious that
Roscoe bought only the horn and not the cow. I think that
the grounds for arguing the unacceptability of (3.33) are
not the occurrence of Extraposition at this point in the
cycle and its optional applicability to the structure tree,
102
but is rather an output condition on the resultant tree
which would block, or at least mark, some derivations pro
ducing ambiguous sentences. Such an output condition is
already necessary in the grammar to rule on the possible
ambiguous readings that might be yielded by the extra
position of noun phrase complements, as (3.34) demonstrates.
(3.34) a. The fact that John knew was a secret.
b. The fact was a secret that John knew.
c. That John knew was a secret.
d. It was a secret that John knew.
This problem merely points up the distinction which
must be made in any adequate grammar between performance
and competence. I do not think that we should wish to
exclude from the grammar any transformational machinery
that would prevent the derivation of ambiguous sentences,
such as (3.33) or (3.34bSd), because such sentences do
commonly occur in casual speech and are rendered intelli
gible by the context in which they appear. They are
excluded from careful speech and writing by some form of
output condition which analyzes the string as grammat
ically well formed but ambiguous. We would no more wish to
block the derivation of such sentences than we would wish
to prevent the derivation of the sentences in (3.35),
simply because they are ambiguous..
103
(3.35) a. Roscoe wants to marry a rich girl.
b. The fierce dogs got the bones.
c. A dog I own bit me.
3.1.3 Countering Ross’s Argument
Ross (196 7b) has claimed that the rule of Extra
position and the rule of Extraposition from NP may not be
collapsed because the Question Transformation must inter
vene between them. He orders the rules that are crucial to
his analysis in the following way: (p. 155)
1. Particle Movement
2. Extrapos ition
3. Question
4-. Extraposition from NP
The necessity for this ordering can be
seen in conncetion with (5.24), which derives
from the intermediate structure (5.25) . . . .
(5.24) Which packages is it possible
that Sam didn't pick up which
are to be mailed tomorrow?
(5.25) S
Q
NP
1
VP
that NP'
is possible
Sair didn’t pick NP
which packages
which are to be
mailed tomorrow
[_
104
It should be obvious that Extraposition
must precede Question, for if S 2 has not been
moved out of NP]_ to the end of S]_, the ques
tioned element, NP3 will be contained in a sen
tential subject, NP^, and will be prohibited
from moving out of it by the constraint stated
in (4.254) Sentential Subject Constraint. But
it is not so obvious that Extraposition from NP
most follow Question.
. . . it might be argued that a collapsed rule
of extraposition could operate to move both
embedded sentences to the ends of the appropri
ate higher sentences, yielding a structure like
But notice that if the questioned NP, NP 2 , is
now moved to the front of S-j_ by the rule of
Question, and the subject and copula are in
verted, the resulting structure is (5.27), not
the intuitively correct (5.28).
(5.26)
is possible that NP VP
didn't
pick £
which
package
it
mailed tomorrow
(5.27) S
which packages is it V that NP VP
possible
didn* t
pick up
S
which are toi
be mailed_ tomorrow]
105
(5.28)
NP NP VP
which packages is
possible that Sam
didn't
which are to be
mailed tomorrow
pick up
The structure shown in (5.27) makes the
incorrect claim that the string didn't pick up
which are to be mailed tomorrow is a constit-
uent break after the particle up. (pp* 155-157)
Unfortunately, Ross does not illustrate the process
whereby the correct derived constituent structure is to be
effected.
I would presume that if we take his intermediate
structure (5.25) to be the derived constituent structure
which obtains after the application of Particle Movement,
then with the application of Extraposition, (3.36) would be
derived.
didn't
pick
which
packages ^HIcH" are to j
be mailed tomorrow j
106
Next, following Ross’s ordering of transformations, Ques
tion would permute NPg to the initial position to Chomsky-
adjoin it to S^, and subject and copula would be inverted.
(3.37)
which
possible
didn't
pick
packages is it
which are to be
mailed tomorrow
Now Extraposition from NP is supposed to apply. But notice
that the Question Transformation, in moving NPg to the
initial position in the structure tree has blocked Extra
position from NP, for the structure index for that trans
formation is no longer met. How has Ross managed to move
Sg U P t O Sg?
Clearly, what is involved in his analysis is his con
vention of Pied Piping, which allows him to move not only
the questioned NP, NPg, to the front of the sentence, but
to move as well the NP dominating it, NP2, together with
Sg. Then Extraposition from NP would apply to move Sg to
the end of SQ. But such a maneuver would demand the
obligatory operation of Pied Piping, and as the convention
is currently formulated, it is optional. Now if Ross is to
argue that surface structure which derives from (3.37), in
which Pied Piping has not applied, is merely a less accept
able variant of the sentence (5.28), in which Pied Piping
has applied, he will do so at the expense of admitting that
(3.37), which must be grammatical according to his analysis,
suffers from the same faultiness of derived constituent
structure which he attributes to (5.2 7). This point Ross
chooses to ignore.
Let us now examine the way in which the grammar I am
proposing would derive the sentence under discussion. We
must begin with a different deep structure, one consistent
with the conjunct analysis.
(3.38) S
Q
S
2
S
3
VP
NP
be mailed tomorrow Sam V NP S
is
possible
didn't
pick up
packages are \
to be mailed S
tomorr
108
One benefit which the conjunct analysis provides
which is immediately obvious is that the Particle Movement
Transformation may be simplified. In the standard theory
this transformation is usually stated in such a way that
its operation blocks if the NP around which the particle is
to move is a complex NP, that is, if it dominates an S.
The deep structures that will be generated in the conjunct
analysis will not contain complex NP ’ s, so the need for
this constraint is obviated. The Particle Movement Trans
formation may now be stated more simple as (3.39).
(3.39) Particle Movement
Part NP Y
SI: 1 2 3 4 (optional)
SC: 1 0 3+24
Particle Movement will apply optionally to Sg and Sg
in figure (3.38), and Definitization will apply on the S2
and Sg cycles to copy S^^ and Sy into Sg and Sg, respec
tively, and Restrictive Relative Clause Formation will
apply afterwards to produce (3.40) , p• 109.
At this point the structure index for the collapsed
version of Extraposition is met and may apply at the S2 and
Sg levels, to move S^' and SyT around the particles. Addi
tionally, on this cycle S^ and Sy may be transformationally
deleted with the result that S2 and Sg will be pruned. The;
resultant structure will be (3.41), p. 110. I
(3.HO)
packages are
to be mailed Sam V
tomorrow _____
didn’t
pick
is possible
packages
+def
packages are NP
to be mailed I
tomorrow '
Sam
packages are to be
+def mailed
didn’t NP
pick I
+WH
+PRO
tomorrow packages
+def NP VP
packages are to
+def be mailed
+WH tomorrow
+PR0 o
C O
110
(3.41)
NP VP NP VP
Sam NP up
is
possible
which are
to be mailed
tomorrow ^
NP
didn't
pick
packages
+def
VP
/^up ^
l V NP
didn't packages
pick
Sam
S 1
+def
which are
to be
mailed
tomorrow
Now that Extraposition and the deletion of the prior con-
juncts have applied at the S2 and Sg levels, the identity
condition is met so that Definitization may apply at S^.
Its obligatory application will yield (3.42).
L .
(3.42)
Sam didn't pick
the packages up
which are to be
mailed tomorrow is possible
Sam didn't pick
the packages up
which are to be
Sam didn't pick
the packages up
which are to be fmailed tomorrow|
imailed tomorrowi
+def
+def
Ill
Extraposition may now apply to move Sg» to the end of Sg, and
Sg may be transformationally deleted. Sg will pronominal-
ize, and the resulting structure will be that shown in
(3.43).
(3.43)
it is possible Sam V NP
‘ Q— A
which are to
be mailed
I
didn1t packages tomorrow
pick +def
+def
The Question Transformation may now permute the questioned
NP in Sg' to the front.
(3.44) S
which V NP VP
packages I 1 1
is it possible NP
Sam didn't
pick up
I _____________
which are to
be mailed
tomorrow_____
+def
Here there is no conflict between the two rules of
Extraposition because they apply on separate cycles, and
because I can fashion no sentence which would occasion the
use of two such rules within a single cycle, I must
J
112
conclude that the two rules of Extraposition are an added
expense that the grammar can do without. Also the rule of
Particle Movement has been greatly simplified by this
analysis.
(3.4-4) is not, however, the derived constituent
structure which is compatible with Ross's intuitions, that
is, (3.44) bears a strong resemblance to his (5.2 7) and is
unlike his (5.28). Along with (5.27), (3.44) seems to be
making the claim that Sam didn't pick up which are to be
mailed tomorrow is a constituent. Ross (pp. 149-150) cites
an argument of Chomsky's which would allow for the raising
of such nodes as S^" in (3.44) by surface structure adjust
ment rules, rules which are thought to be phonologically
motivated.
Ross argues against such rules on the grounds that
structures such as (3.45), his (5.30), in which Extra
position from NP may apply to yield (3.46), behave
(3.45) S
0
until it had stopped
raining
NP VP
Sam didn’t
pick
NP
a
s
a
which are to those which are to be
packages mailed tomorrow
113
peculiarly when they are embedded into a higher structure
and questioned.
(3.1+6) Sam didn't pick those packages up which
are to be mailed tomorrow until it had
stopped raining.
Such a structure is the one shown in (3.47).
(3.47)
is possible
it that
VP until it had
up stopped raining
Sam didn’t NP
pick
which
packages
which are to be
mailed tomorrow
Two variants of the resulting structure . . .
are possible: (5.35a), in which the relative
clause S3 has not been extraposed away from its
head NP, which packages, and (5.35b), in which
it has.
(5.35) a. Which packages which are to be
mailed tomorrow is it possible
that Sam didn't pick up until
it had stopped raining?
114
b. ?Which packages is it possible
that Sam didn't pick up until
it had stopped raining which
are to be mailed tomorrow?
While it is clear that (5.35a) is the more com
fortable version of the two, by far, I think
(5.35b) should be treated as being grammatical
but of low acceptability. For notice that the
acceptability of (5.35b) can be improved by
lengthening the extraposed relative clause, as
in (5.36).
(5.36) Which packages is it possible that
Sam didn't pick up until it had
stopped raining which he had arranged
with his agents in Calcutta to send
him here in Poplar Bluffs because
of his fear that someone in Saint
Louis might recognize him?
Note that in (5.35b) the extraposed clause follows
the until-clause, which the ungrammaticality of
(5.33) shows not to be possible when the structure
underlying (5.29) is not embedded [(5.33) is ?*Sam
didn't pick those packages up until it had stopped
raining which are to be mailed tomorrow, and the
structure underlying (5.29) is that represented by
his (5.30), my (3.46).] But more important is the
fact that the preferred order in the non-embedded
case, i.e., with the relative clause preceding the
until-clause, as in (5.29), [Sam didn't pick those
packages up which are to be mailed tomorrow until
it had stopped raining.] is absolutely impossible
in the embedded case, as the ungrammaticality of
(5.37) shows.
(5.37) *Which packages is it possible
that Sam didn't pick up which
are to be mailed tomorrow until
it had stopped raining?
In fact, if a relative clause has been extra
posed away from its head NP, that NP cannot be
questioned. So compare (5.29), which contains
such a head NP, with the ungrammatical (5.38) in
which this head NP has been questioned.
115
(5.38) *Which packages didn’t Sam pick up
which are to be mailed tomorrow
until it had stopped raining?
(pp. 158-160)
Thus, according to Ross, the following facts, a) that
the preferred order in the non-embedded case does not
appear to be grammatical if the sentence is embedded, and
b) that the head noun of an extraposed relative clause may
not be questioned, banish the possibility of surface struc
ture adjustment rules, because such rules could not convert
(5.37) into the more acceptable (5.35b) and because in the
event of a collapsed rule of extraposition, the head NP of
the relative clause could not be questioned because of a
general constraint of the grammar.
Let us examine these facts. The second is clearly
wrong. The head NP of an extraposed relative clause can
indeed be questioned, as the following set of sentences
demonstrates.
(3.48) a. The girl came in who wore a coat.
b. Which girl came in who wore a coat?
What effect does this have on (5.38)? I would rather mark ;
this sentence as ungraceful rather than ungrammatical.
As for Ross’s other claim, it is true that no adjust-i
ment rule would be able, on phonological grounds, or by \
1
analogy with phonological rules, to convert (5.37) into '
116
(5.35b), but I am not so sure that this is to be desired.
In my dialect, (5.37), while it is even more ungraceful or
inelegant than (5.38), is nonetheless preferrable to
(5.35b), which Ross treats as grammatical but unacceptable.
It must also be remembered that Ross's analysis makes
crucial use of the optional convention of Pied Piping. And
were this convention not employed (as it has every right
not to be) then the following sentence would result from
(3.47).
(3.49) *Which package is it possible that Sam
didn't pick which are to be mailed
tomorrow up until it had stopped rain
ing?
This sentence would result from the application of Particle
Movement, Extraposition and Question, but not Extraposition
from NP, because Question has destroyed the structure index
for that transformation. And even if it is claimed that
such a sentence as (3.49) is grammatical but unacceptable
owing to an output condition of the grammar which estab
lishes a preferred order for elements following the VP
(Ross, 1967b: 36, figure (3.41)), as may well be the case,
its derived constituent structure will make the claim that
Sam didn't pick which are to be mailed tomorrow up until it
had stopped raining is a constituent, the sort of claim
that Ross is attempting to avoid. And even if some means
117
could be devised for permuting the relative clause around
the particle, the derived constituent structure that would
result would still be inconsistent with Ross's claim that
the extraposed relative clause must be attached to the
highest S.
Now let us examine the same sentence using the model
I have proposed. (3.50) might be an equivalent deep struc
ture for (3.47).
(3.50)
"S until it had
K. stopped raining
NP VP NP VP
i A I
pack- are Sam
ages to be V NP
mailed /\ |
tomor-/^ packages NP
row
package
didivt
pick up
is possible
i
packages are to Sam V^
be mailed
until it had
n stopped raining
f \ .
NP VP
NP
tomorrow didn't packages
pick up
Here Particle Movement may apply at the S^ and S ^ levels.
Next Definitization and Restrictive Relative Clause Forma
tion may apply at S^ and Sg. At this point the derived
constituent structure will be (3.51).
^ stopped raining
until it had
packages are NP VE^up
to be mailed
tomorrow Sam V NP
didn't
pick
packages NP
+def
packages
+def
+WH
+PRO
packages are
to be mailed
tomorrow
~VP
are to be
mailed
tomorrow
until it had
stopped raining
10
pick
packages /
+def NP VP
packages
+def
+WH
+ PR0
are to be
mailed
tomorrow
118
119
The collapsed rule of Extrapos ition may also apply at the
and Sg levels to move Sg' and S-^q' around the particles,
and and may be transformationally deleted, necessi
tating the pruning of S^ and Sg. Then at Sg Definitization
will move S-^ into the NP dominating Sg to produce (3.52),
p. 120.
Extraposition will move S^' around the VP is possible
and Sg will pronominalize. With the deletion of S^ the
structure will be that shown in (3.53).
(3.53) SQ
NP VP
it
is possible
until it had
stopped rain
ing
NP
NP Sam
didn't Q packages S
pick + def
which are to
be mailed
j _______________________tomorrow_____ |
+ def ;
:When Question applies to this structure, the questioned NP j
of Sy' will move to the front of Sg to replace Q, producing;
|(3.54), whose derived constituent structure is (3.55). j
Sam V
didn’t
pick
until it had
,up ' — stopped raining
NF
Q packages
+def
which are
to be
mailed
tomorrow
until it had
up\stopped raining Sam
Sam V
didn't Q packages s t
pick +def 10
VP\ until it
up had
\ stopped
NP \raining
didn’t packages SI"
pick
which are to L
be mailed
tomorrow I
which are to be
mailed tomorrow.
+ def,
+def
120
121
(3,54) Which packages is it possible that
Sam didn't pick up which are to be
mailed tomorrow until it had stopped
raining?
(3.55) S
which V NP VP
packages I I I
is it possible
until it had
stopped rain
ing
NP VP
A
Sam V up S "
didn't which are to be
| pick mailed tomorrow
+def
If, at the point in the derivation shown in (3.51),
the option is not taken to extrapose the relative clauses
Sg ' and S-j^q ' , Def initization, Extraposition, and Pronomi-
nalization will occur at the Sg level, and Question will
permute the questioned NP to the initial position to yield
(3.5G). In (3.56a) Pied Piping has applied; in (3.56b)
it has not.
(3.56) a. Which packages which are to be mailed
tomorrow is it possible that Sam didn't
pick up until it had stopped raining.
122
b. *Which packages is it possible that Sam
didn’t pick which are to be mailed
tomorrow up until it had stopped rain
ing?
At this point we may assess the differences in the
derivations produced by my model and Ross’s. Ross's model
will yield the following sentences.
(3.57) a. Which packages which are to be mailed
tomorrow is it possible that Sam didn't
pick up until it had stopped raining?
b. *Which packages is it possible that Sam
didn't pick which are to be mailed
tomorrow up until it had stopped rain
ing?
c. ?Which packages is it possible that Sam
didn't pick up until it had stopped
raining which are to be mailed tomor
row?
And mine will yield the following.
(3.5 8) a. Which packages which are to be mailed
tomorrow is it possible that Sam didn't
pick up until it had stopped raining?
(= 3.57a)
123
b. *Which packages is it possible that Sam
didn’t pick which are to be mailed
tomorrow up until it had stopped rain
ing? (= 3.57b)
c. ?Which packages is it possible that Sam
didn’t pick up which are to be mailed
tomorrow until it had stopped raining?
It is readily seen that the only differing results
are the (c) sentences, and here the difference seems to me
more a matter of dialectal preference, if such a preference
may be said to exist between sentences so abominable. The
sentences of (3.59), which follow the pattern of (3.58c)
seem to me to be preferable to those of (3.60), which
follow Ross's pattern (3.5 7c).
(3.59) Iiy Analysis
a. Which gorilla is it necessary for the
girl to tie up which escaped from the
Bronx Zoo when she finishes icing the
cupcakes for Aunt Martha?
b. Which wheelbarrow is it obvious that
Claude didn't strip down which belonged
to Elmo after the spring plowing was
finished?
j
i
124
c. Who is it true that Roscoe called up
who ate the catfish now that the
Saturday night fish-fry is over?
d. ?Which gun is it obvious that Bonnie
didn’t pick up which Clyde stole until
after the gangland massacre had occur
red?
(3.60) Ross's Analysis
a. ?Which gorilla is it necessary for the
girl to tie up when she finishes icing
the cupcakes for Aunt Martha which
escaped from the Bronx Zoo?
b. ?Which wheelbarrow is it obvious that
Claude didn't strip down after the
spring plowing was finished which
belonged to Elmo?
c. ?Who is it true that Roscoe called up
now that the Saturday night fish-fry
is over who ate the catfish?
d. ?Which gun is it obvious that Bonnie
didn't pick up until after the gang
land massacre had occurred which Clyde
stole?
As for Ross's claim concerning clause length, (5.35), I
only counter with (3.61).
j
can I
(3.61) *Which packages is it possible that Sam
didn’t pick up until it had stopped rain
ing because he had left his galoshes and
slicker back in his room at his Uncle
Hugo’s in Buffalo the weekend before and
had not reclaimed them which are to be
mailed tomorrow?
The results having been tallied, there seems to be
pretty much of a stand-off. Our grammars each produce
three derivations from the deep structure. Two of them are
identical: the (a) sentences, which are fully grammatical,
and the (b) sentences, which are unacceptable owing to an
output condition which establishes a preferred order of
constituents following the VP. The (c) sentences are
different, and both seem to be of questionable accept
ability. As for Ross’s sentence, he is welcome to it, and
I am sure he would feel the same way about mine. Yet if
such matters are to be decided empirically, it must be
admitted that the model I have constructed with its simpli
fied rule of Particle Movement, its collapsed rule of
Extraposition, its inclusion of the rule of It Deletion
under the more general rule of Post-definitization Deletion|
is preferrable for reasons of economy.
3.1.4 Analysis Expanded to Include Non-restrictives j
As further evidence for the merit of the analysis I
have been proposing, let us look again at the sentence i
126
concerning Sam and the packages. Let us suppose that no
element of the sentence has been questioned. From the deep
structure (3.62) the sentences of (3.63) may be derived.
(3.62)
until it had
stopped raining
is
possible
pack- are to be Sam V NP pack- /L—
until it had
stopped raining
ages mailed
tomorrow
ages
didn't pack-
pick up ages
be
mailed
tomorrow
Sam V
didn't pack-
pick up ages
(3.6 3) a. That Sam didn't pick up the packages
which are to be mailed tomorrow until
it had stopped raining is possible.
127
b. That Sam didn't pick the packages up
which are to be mailed tomorrow until
it had stopped raining is possible.
c. It is possible that Sam didn't pick
up the packages which are to be mailed
tomorrow until it had stopped raining.
d. It is possible that Sam didn't pick
the packages up which are to be mailed
tomorrow until it had stopped raining.
e. Sam didn't pick up the packages which
are to be mailed tomorrow until it had
stopped raining, which is possible.
f. Sam didn't pick the packages up which
are to be mailed tomorrow until it had
stopped raining, which is possible.
It is the (e) and (f) sentences which concern us here. To
me it seems that these sentences are synonymous with sen
tences (a) through (d). If so, they must be transforma
tionally related and derived from the same deep structure.
I know of no transformational machinery in the standard
theory of transformational grammar which will derive sen
tences (3.6 3e&f) from the deep structure underlying sen
tences (3.63a through d). But under the analysis I am
proposing such derivations are possible. ;
128
For reasons of simplicity, I shall derive only
(3.6 3e) here; it is simpler because it does not involve
Particle Movement and Extraposition«
The first transformation of interest here is Defini-
tization, which applies obligatorily at and Sg. It is
followed by Restrictive Relative Clause Formation, thus
providing the intermediate structure (3.64), p. 129. At
this point Sg and S^Q may be optionally deleted and S^ and
Sg, because they no longer branch, may be pruned, (3.65),
p. 130. Now on the Sg cycle Definitization must apply
because of the identity of and Sg, (3.66), p. 131. At
this point Extraposition might occur, followed by Pronomi-
nalization. The first conjunct, S-^, might then be option
ally deleted, and the result would be (3.6 3c). Or if the
option of Extraposition is not elected, Post-definitization
Deletion will delete S^', and with subsequent deletion of
S-p (3.63a) will surface. Let us assume, however, that
Extraposition does not apply and that Post-definitization
Deletion deletes S-^'. Let us assume, furthermore, that the
first conjunct, S^, is not transformationally deleted. The
surface structure will be (3.67).
(3.67) Sam didn't pick up the packages which are
to be mailed tomorrow until it had stopped
raining, and that Sam didn't pick up the ;
(3.64)
NP
; pack-
I ages
is possible
S„ until it had
stopped raining
are to be Sam V
mailed
tomorrow
NP
I
S_' pack-
6 ages are to he S’ am
mailed
packages NP VP tomorrow
+def
until it had
stopped raining
didn’t NP
pick up I
packages are to be
+def mailed
+WH tomorrow
+ PRO
didn’t NP
pick up
packages NP
+def I
I are to be
packages mailed
+der tomorrow
+WH
+PRO
129
(3.65)
until it had NP VP
Sam V
didn't
pick up
stopped raining
NP
6
VP NP
+def
packages are to be
+def mailed
+WH
+PR0
tomorrow
is possible
11
until it had
stopped raining
NP VP
NP V
didn't NP
pick up I
10
NP VP packages
+def
packages are to be
+def mailed
+WH tomorrow
+PR0
C O
until it had
stopped raining
Sam V
didn't NP
pick up |
packages which are
+def to be
mailed
tomorrow
possible
Sam
Sam V
until it
had stopped / \
Np raining didn't NP
pick up |
didn't NP . ' S ’ >
until it had
stopped rain
ing
pick up |
packages which are I
+def to be
mailed
__________ tomorrow |
packages which are to
+def be mailed
_____________tomorrow____ |
+def
+def
132
packages which are to be mailed tomorrow
until it had stopped raining is possible.
This is a suitable surface structure even though it is a
bit unwieldy. But notice that the structure tree under
lying it has the definite NP, which is S3, in the second
conjunct, S2, and that Sg is identical with S^.
(3.68)
Sam V
VP until it had
stopped
fining
NP
is possible
didn't NP
pick up
packages which are
+def to be
mailed
tomorrow
Sam V
until it had
stopped
NP raining
didn't N
pick up
packages which are to
+def be mailed
______________tomorrow |
+def
It will be recalled that the Non-Restrictive Relative
Clause Formation Transformation was formalized in the
following way.
133
(3.69) Non-Restrictive Relative Clause Formation
SI:
SC:
[X NP Y]
S S
1 2 3
1 2 3
C W
S
5
+WH
+PR0
NP Z]
+def S
5 6 :
4 6
S (optional)
Condition: 2=5
(3.68) will meet the structure index for this trans
formation if we adjust the condition to read: 2=5 or 1 2 3=
5. The features +WH and +PRO will be added to Sg, the
relative pronoun will replace it, and the result will be
the desired surface structure (3.63e).
3.1.5 Additional Derivations, Relative Clause Reduction
Returning now to the sentences of (3.1), we may show
how the remainder of the sentences containing noun phrase
complements fall into line when the present analysis is
used. (3.1b) will have the following deep structure.
(3.70)
VP NP VP NP
Gill
NP should be classes
proposed
should be
dismissed
134
Definitization, applying at Sq , will move into S2 ,
Chomsky-adjoining it to the NP dominating S3 .
(3.71)
NP VP VP
Gill classes should be V NP
proposed NP
NP VP
NP VP classes should be
dismissed!
classes should be +def
| _________ dismissed |
+def
Post-definitization Deletion will delete S^1, and with the
transformational deletion of S^, the desired surface struc
ture will be derived. The structure of (3.71) is subject
to the Passive Transformation, and should this transforma
tion apply, the following surface structures might result.
(3.72) a. That classes should be dismissed was
proposed by Gill,
b. It was proposed by Gill that classes
, be dismissed. (With Extraposition)
135
The important fact to be recorded from this observation is
that Passive must precede Extraposition and Post-definiti
zation Deletion, or else Extraposition would not have been
able to move S-^ to the end of S2 in order to yield (3.72b).
The derivation of (3.1c) is straight-forward. The
only new twist is that the NP dominating the identical S is
not the direct object but an oblique object of some sort.
(3.73)
NP'
I
Bruce would rea
agreed to
the paper
Definitization
\
Bruce would read
the paper
(obligatory)
V
Bruce would read
the paper
Bruce V
I
agreed to
Bruce would
I read the paperl Ipaper
+def -
jruce would
read the
+def
J
136
If Post-definitization Deletion applies to the output of
Definitization, it seems that the preposition may be trans
formationally deleted. I am not at all sure of the status
of to as a preposition, for notice that if Passive applies
to this structure, to must remain behind. This fact
perhaps argues for the interpretation of to as a particle.
(3.74) a. That he would read the paper was
agreed to by Bruce,
b. It was agreed to by Bruce that he
would read the paper.
(3.Id) presents a certain amount of difficulty in its
analysis. Suppose we assume a deep structure like (3.75).
VP NP
claim false VP NP NP VP
Terry V NP claim V
I I I
slugged Gloria is
Terry V
I I
slugged Gloria
Definitization, applying at S-p will move Sg into the NP
dominating Sg, producing (3.76).
(3.76)
NP VP
claim Terry slugged NP
Gloria I
false VP
claim NP V
NP
Terry slugged
|Gloria________
S
Terry slugged
(Gloria |
+def
Post-definitization Deletion will delete Sg1, and Sg may be
optionally deleted, thus allowing for the pruning of S-p
(3.77)
VP NP
claim
VP NP
claim false
Terry slugged
Gloria_______
+def
138
Definitization will now apply at Sg, moving S^ into the
subject NP of Sg. Restrictive Relative Clause Formation
will apply next to add +WH and +PR0 to the subject NP of
(3.78)
NP VP NP VP
claim is
Terry slugged
Gloria________|
+def
claim NP
+def |
claim
+def
+WH
+PRO
is
false
Terry slugged
[Gloria________j
+def
S^ may be optionally deleted and the Relative Clause
Reduction rule may now apply to delete the subject NP of
Salong with the copula. This produces the desired
surface structure, (3.Id).
(3.Id) The claim that Terry slugged Gloria
is false.
139
There are, however, some residual difficulties. First, if
Relative Clause Reduction does not occur, we are left with
the surface structure (3.79).
(3.79) The claim which is that Terry slugged
Gloria is false.
This is not an altogether bad sentence, but it is one that
is subject to the general rule that restrictive relative
clauses may optionally substitute that for which. Should
such a substitution occur, the result, (3.80), does not
seem quite so desirable.
(3.80) ?The claim that is that Terry slugged
Gloria is false.
The questionableness of (3.80) suggests that we are perhaps
barking up the wrong tree in assuming that a restrictive
relative clause is involved in this derivation.
Secondly, the analysis does not provide a means of
deriving the synonymous sentence, (3.81).
(3.81) The claim is false that Terry slugged
Gloria.
If Extraposition applies to move S^' to the right of the VP,
in S2 in figure (3.78), then Relative Clause Reduction,
which is usually formalized in the following way, j
(3.82)
SI:
SC:
will be prevented from operating, because its structure
index will no longer be met, and (3.83) will surface.
(3.83) The claim is false which is that Terry
slugged Gloria.
If Extraposition does not apply and Relative Clause Reduc
tion is allowed to delete the relative and the copula, then
Extraposition may not apply to the resulting tree, because
its place in the transformational cycle has already been
passed.
Perhaps it could be argued that Relative Clause
Reduction is an 'anywhere rule' like Gapping and Conjunc
tion Reduction. Ross (1967a) and Lakoff (1966) have argued
effectively for the presence of such rules in the grammar,
rules that may apply at any point at which their structure
indices are met. Yet because of the ad hoc nature of such
a remedy in this case, and because of the additional prob
lem of the questionableness of the restrictive relative
clause substructure in this case, I hesitate to propose
such a solution.
140
Relative Clause Reduction
X E NP - C NP be - VP] ] Y
NP S +WH S NP
+PR0
< y - ----'
1 2 3 4 5 (optional)
1 2 0 4 5
Instead, let us examine the sentence again, assuming
this time that a non-restrictive relative clause is
involved in the substructure of (3.Id) instead of a restric
tive relative. The deep structure underlying (3.Id) would
be something like (3.84).
(3.84)
claim
claim
NP VP
I
Terry V NP
I
slugged Gloria
NP
Terry .V NP
slugged Gloria
Definitization will apply at Sg to copy Sg into S^
(3.85)
claim is false S
142
Post-definitization Deletion will delete Sg', and Sg can be
optionally deleted to produce (3.86), pruning S2 .
Definitization must now apply at Sq to definitize the sub
ject NP of S^, and Post-definitization Deletion may then
delete the copy of which was embedded in by Defini
tization. At this point, the structure index for Non-
restrictive Clause Formation Transformation, a last cyclic
rule as we saw in the previous chapter, has been met, and
it may apply to yield (3.37).
(3.8 6 ) S
claim is I ---
false claim V NP
Terry slugged
Gloria
+def
claim false claim V NP
+def
+WH
+PR0
is
Terry slugged
| Gloria 1
+def
143
This structure would have the surface manifestation (3.88).
(3.88) A claim is false, which is that Terry
slugged Gloria.
It must be remembered, however, that in the discussion of
Non-restrictive Relative Clause Formation in Chapter Two, a
transformation was proposed which would move the conjunct
containing the identical definite NP into the other con
junct. If we stated this transformation as follows,
(3.89) Conjunct Incorporation
[X NP Y ] [W NP Z]
S S S +def S
SI: 1 2 3 4 5 6 (optional)
SC: 1 2 [4 5 6 ] 3
Condition: a. 2=5
b. last
cyclic
we might have it apply after Non-restrictive Relative
Clause Formation to introduce the non-restrictive clause
into the other conjunct. The result would be (3.90a). In
the event that Non-restrictive Relative Clause Formation
did not apply, the structural index for Conjunct Incorpo
ration would still be met, and should it apply, the result
would be (3.90b).
i»m
(3.90) a. A claim, which was that Terry slugged
Gloria, is false,
b. A claim, and the claim was that Terry
slugged Gloria, is false.
Should both Non-restrictive Relative Clause Formation and
Conjunct Incorporation apply to (3.87), the derived con
stituent structure would be (3.91).
(3.91)
NP false
claim VP P
NP V
+def
+WH
+PR0 is
Terry slugged
I Gloria________|
+def
To this structure Relative Clause Reduction may apply to
:delete the relative and the copula, leaving after pruning
(3.92)
NP false
claim Terry slugged
Gloria
+def
There are still two difficulties with this analysis.
First, the output of (3.92) is (3.93).
(3.93) A claim that Terry slugged Gloria is
false.
This is not exactly the desired sentences (3.Id).
(3.Id) The claim that Terry slugged Gloria is
false.
In Chapter Two I mentioned that my analysis would not
account for such sentences as (2.89),
(2.89) The cow, which I sold for some beans,
has a crumpled horn.
where the head noun of the first conjunct is preceded by
the definite article. Such is the case here. For the
moment all that I can suggest is that the definite article
must be derived from a prior mention of claim as a result
146
of the Definitization Transformation. Such a mention might
have been in a conjunct S of the sort (3.94),
(3.94) There is a claim.
which was ultimately deleted by the transformation which
may optionally delete prior conjuncts. (This transforma
tion will be treated at greater length in the following
chapter.)
The second difficulty is the same one we encountered
earlier. Even if we are able to prove that a definite
article may be derived so that the proper surface structure
may be arrived at, my analysis has no mechanism at present
to derive the synonymous sentence (3.81).
(3.81) The claim is false that Terry slugged
Gloria.
As in the previous discussion in which the restrictive
relative clause substructure was used, I am prevented from
applying Extraposition to the tree in (3.92) because in the
ordering of transformations Non-restrictive Relative Clause
Formation must necessarily follow Post-definitization
Deletion, which has already applied at the Sq cycle, and it
has been shown that Post-definitization Deletion follows
Extraposition.
Nevertheless, I believe that, given the intuition ;
that the non-restrictive analysis of (3.Id) is preferrable j
147
to the restrictive analysis, the counterexample posed by
(3.81) is only an apparent one.
I propose that (3.81) is not the result of Extra
position, but rather the result of a more powerful Relative
Clause Reduction Rule. The usual formulation of this rule
has been stated in (3.82). In its present form it is used
to convert sentences of the type (3.95) into those of the
type (3.96).
(3.95) a. He is a man who is good at gardening.
b. He is someone who is difficult to handle.
c. I saw a man who was running down the hill.
d. My brother, who is John, sells used cars.
(3.96) a. He is a man good at gardening.
b. He is someone difficult to handle.
c. I saw a man running down the hill.
d. My brother, John, sells used cars.
The same transformation works in conjunction with Modifier
Shift to transform sentences like (3.97) into (3.98).
(3.97) a. I hit a ball which was red.
b. The man who was old sat down.
c. I gave it to the nurse who was ugly.
(3.9 8) a. I hit a red ball.
b. The old man sat down.
c. I gave it to the ugly nurse.
148
But notice that for reasons that are completely inde
pendent of the case at hand, the structure index for Rela
tive Clause Reduction may be made less constrained.
(3.99) Relative Clause Reduction
X [ NP be VP] Z
S +def S
+WH
+PRO
SI: 1 2 3 4 ---^ (optional)
SC: 1 0 3 4
If we order the transformation after Non-restrictive
Relative Clause Formation, it will apply not only to non-
restrictive clauses, but to restrictive clauses as well.
It may be used to convert the sentences of (3.95) into
(3.96), and, together with Modifier Shift, to convert the
sentences of (3.97) into (3.98). It may also be used to
transform the sentences of (3.100) into those of (3.101), a
type of conversion unavailable in the usual analysis of
English, with its highly constrained rule of Relative
Clause Reduction.
(3.100) a. ?John came home from the exam, who
was defeated,
b, Joe was a man of great dignity, who
was proud, who was aloof, who was
not given to casual friendships.
149
- c. The townspeople admired Bill, who
was a man of means, who was wealthy
beyond their wildest dreams.
d. It was a pond which was unassuming,
which was shallow, which was no more
than a few yards in diameter.
e. . The claim was utterly false, which
was unworthy of further discussion.
(3.101) a. John came home from the exam defeated.
b. Joe was a man of great dignity, proud,
aloof, not given to casual friend
ships .
c. The townspeople admired Bill, a man
of means, wealthy beyond their wildest
dreams.
d. It was an unassuming pond, shallow,
no more than a few yards in diameter.
e. The claim was utterly false, unworthy
of further discussion.
Given the more powerful statement of the Relative
Clause Reduction Transformation, it would be possible to
derive the desired surface structure, (3.Id), by appli
cation of this transformation at the point in the deriva
tion illustrated by (3.87), i.e., before Conjunct Incorpo- j
; i
iration has applied. It is also possible that such a j
transformation could apply to an extraposed restrictive
relative clause of the same type to produce an identical
surface structure. In such a case, the analysis works both
ways, and it is the matter of whether the substructure is
to include a restrictive or non-restrictive relative clause
that is to be settled, in the final analysis, by the
intuition of the speaker, i.e. , whether he chooses to con
strue the underlying conjuncts to have the order shown in
(3.75) or (3.84).
There is yet one more problem that results from the
strengthening of the Relative Clause Reduction rule. Such
a rule will produce the wrong results for sentences of the
type (3.6 3eSf). If Relative Clause Reduction is allowed to
delete a NP bearing the features +WH and +PRO, together
with the copula, in any environment, it will convert
(3.6 3eSf) into the sentences of (3.102), clearly ungram
matical sentences.
(3.63) e. Sam didn’t pick up the packages
which are to be mailed tomorrow
until it had stopped raining, which
is possible,
f. Sam didn't pick the packages up which
are to be mailed tomorrow until it had
stopped raining, which is possible.
151
(3.102) a. *Sam didn’t pick up the packages
to be mailed tomorrow until it had
stopped raining, possible,
b. *Sam didn’t pick the packages up
to be mailed tomorrow until it had
stopped raining, possible.
While the revised rule of Relative Clause Reduction will
work satisfactorily to delete the relative pronoun and the
copula in the base clause, it will also cause the deletion
of these same items in the non-restrictive clause, and such
a deletion is clearly undesirable.
We might pause to ask if there is any significant
difference between the sentences of (3.6 3eSf) and the
others that we have discussed. The answer, of course, is
that there is: in sentences (3.6 3eSf) the which derives
from the relative pronominalization of a NP which was a S.
In the other sentences , the which derives from the relative
pronominalization of a lexical NP. With this difference in
mind, it seems better to propose a constraint, or condition,
on the strengthened Relative Clause Reduction rule, (3.99),
than to abandon it in favor of the old rule (3.82). Such
a condition might be stated as follows:
(3.103) Relative Clause Reduction may apply to all
S's containing the string [ NP + be], j
+ def
+WH
+PR0
152
provided that the NP to be deleted does
not dominate an S.
With this condition added to the transformation, Relative
Clause Reduction will yield the correct results for the
sentences (3.Id), (3.96), (3.98), (3.100) and will avoid
the undesirable readings of (3.102).
3.1.6 Complexity in Derivations
The last sentence of (3.1) is not so difficult to
derive as it is cumbersome. Its deep structure may be
stated in my analysis by (3.104), p. 153. Definitization
will occur at each of the successive levels, Sg, Sg, S^,
and Sq to copy the preceding conjunct into the object NP of
the following one. Post-definitization Deletion will apply
at each cycle to delete the identical S, once Definitiza-
tion has occurred. And the prior conjunct may be deleted
after each of these cyclic operations has taken place. The
final result will be (3.1e).
(3.1e) Ed thought that Bill knew that Tom
believed that Max advocated that
Estelle should buy a wombat.
3.2 Ordering of Transformations '
The end result of this analysis is the following
!
group of ordered transformations.
I
I
j
(3.104)
VP
/ X
Bill V NP
I
NP knew S
l
believed S NP
Ed V
thought
Tom V
Bill V
Tom V
knew
Max V
VP believed
believed
Sr j advocated S
should Max V
buy a
:Estelle Estelle wombat
'should buy should
ja wombat buy a
wombat
advo- S Estelle
cated^ should buy Max
a wombat V
Estelle
should buy
a wombat
Estelle
advocated JS should buy Max V
a wombat |
advocated
Estelle
should buy
a wombat
Estelle
should buy
a wombat
153
154
tion (optional)
Post-definitization Deletion (obligatory)
Prior Conjunct Deletion (optional)
Non-restrictive Relative Clause
Formation (optional)
Conjunct Incorporation (optional)
Relative Clause Reduction (optional)
Modifier Shift (obligatory)
Definitization is the first of the transformations
and precedes Restrictive Relative Clause Formation, because
it provides the source for the embedded structure that is
necessary for the structure index of the latter transforma
tion.
Restrictive Relative Clause Formation must precede
Passive in order to derive such sentences as (3.10 6b) from
(3.106a).
(3.106) a. The dog bit the man who ate the
donuts.
b. The man who ate the donuts was bitten
by the dog.
155
Passive precedes Extraposition and Post-definitization
Deletion in order that sentences of the type (3.107b) may
be derived from those of the type (3.107a).
(3.107) a. Mary observed that Jack stole the
tarts.
b. It was observed by Mary that Jack
stole the tarts.
Restrictive Relative Clause Formation comes before
Extraposition because it provides the relative clause
structure which may be subsequently extraposed.
(3.108) a. The alligator which ate the cannibal
got sick.
b. The alligator got sick which ate the
cannibal.
Extraposition follows Definitization because it is
only S’s which have been affected by Definitization which
may be extraposed, i.e., restrictive relative clauses or
i f
that clauses.
That Post-definitization Deletion follows Definiti
zation is apparent by its name. It must also follow Extra
position to delete the redundant definitized S, in the
event that Extraposition does not apply. j
156
(3.109) a. That Bill is a good boy [that Bill
S
is a good boy] is true. ----
b. That Bill is a good boy is true.
Prior Conjunct Deletion, the name I have temporarily
assigned to the transformation that deletes the prior con-
juncts, follows Definitization, Restrictive Relative Clause
Formation, and Extraposition. It optionally deletes the
prior conjuncts which have been transformationally copied
into following S’s. It will be recalled that such a
deletion was originally posited as an optional mechanism
included in Restrictive Relative Clause Formation. But
since such deletion is also necessary, at least optionally,
in the substructures of sentences containing that clauses,
it seems desirable to propose a separate transformation
which will assume these two roles. More will be said about
this transformation in the following chapter.
Prior Conjunct Deletion must precede Non-restrictive
Relative Clause Formation and Conjunct Incorporation,
because its application blocks the structure index for
these transformations. Should Prior Conjunct Deletion not
apply, then these transformations may occur to relativize
the definitized coreferential NP or to move the clause con
taining the definitized coreferential NP, or both.
Non-restrictive Relative Clause Formation must follow
Post-definitization Deletion, because the latter trans
formation produces the appropriate structure index for the
former, as is illustrated in figures (3.110) and (3.111).
(3.110) a. [I hit boy] [ Boy [I hit boy ]
S S S +def S +def S
cried] S Post-definitization
S
Deletion
b. [I hit boy] [ Boy cried] mr^> Non-
S S S +def S
restrictive Relative Clause Forma
tion
c. I hit a boy, who cried.
(3.111) a. [John is a louse] [ [John is a louse]
S +def S S S +3e? S
[John is a louse] is obvious] 'S*
S +def S S
Post-definitization Deletion
b. [John is a louse] [ [John is a louse]
S S S S +def S
is obvious] — ' " V Non-restrictive
S
Relative Clause Formation
c. John is a louse, which is obvious.
Conjunct Incorporation will apply to the output of
Post-definitization Deletion in the event that
158
Non-restrictive Relative Clause Formation has not applied,
and, as such, must necessarily follow that transformation.
(3.112) a. [A boy is smart] [ [I know boy
S S S S +def
C boy is smart] ] Post-
S +def S S '
definitization Deletion
b. [A boy is smart] [I know the boy]3l^z^>.
S S S S
Conjunct Incorporation
c. A boy, and I know him, is smart.
Relative Clause Reduction must necessarily follow
both Restrictive Relative Clause Formation and Non-
restrictive Relative Clause Formation, because it is upon
the output of these transformations that it operates.
Modifier Shift, in turn, operates upon the output of
Relative Clause Reduction.
159
CHAPTER THREE
Footnotes
1. Such deep structures as this one, and others which I
shall propose on the following pages for sentences con
taining noun phrase complements, all rely on one con
stant assumption— that the S which is immediately domi
nated by NP, S3 here, has occurred as a prior conjunct
at some earlier point in the derivation, earlier in the
sense of linear ordering. Such a guarantee must be
incorporated into the grammar to prevent deep struc
tures like (1 ) from surfacing into ungrammatical
strings like (2 ).
Cl)
Mike has
neat tattoos
VP NP
well
known
Mike has
neat chickens
(2) *Mike has neat tattoos, and Mike has neat
chickens is well known.
The reason for the ungrammaticality of (2) is the
absence of the complementizer, or as I have supposed,
definite article, before the sentential subject of the
second conjunct. And this is because Definitization,
owing to the non-identity of S]_ and S3 , could not apply
to introduce the definite marker, which would even
tually be realized as that.
160
Notice that similar difficulties do not arise in
case the non-identity occurs at the NP level. A deep
structure, (3), containing conjoined S’s which share a
coreferential NP may surface as either a complex or
compound sentence, (4-) or (5), respectively.
dog chased
cat-;
cat^ ate
mouse
The cat which a dog chased ate a mouse.
(5) A dog chased a cat, and the cat ate a mouse.
Should S, and S2 not share a coreferential NP, i.e., if
we were to substitute policeman for cat in S2, Defini
tization could not apply, but the resulting surface
structure, (6 ), unlike (2 ), would still be grammatical.
(6 ) A dog chased a cat, and a policeman ate a
mouse.
What must be done in order to block the derivation
of sentences like (2 ) while allowing the derivation of
sentences like (6 )?
We might introduce a formal device such as Chomsky's
# convention (see Chomsky (1965:138)). This device
would cause the rewrite rule which allows NP to be
written as S to be stated as follows: NP > #S#. One
of the functions, then, of Definitization would be to
delete these boundary markers, in addition to assigning
definite markers. A general convention then would
analyze any structure emerging from the transformational
component and still containing #'s as unacceptable,
thus filtering out such sentences as (2 ), which would
still contain s, but allowing sentences like (6 ) to
pass, for the underlying structure of those sentences
would have contained no #'s in the first place.
An alternate method of blocking sentences like (2)
would be to state the base rule expanding NP as follows:
NP- ^ S^, where is a dummy symbol which serves two
functions: 1 ) at the point in the derivation during
which lexical insertion occurs, this symbol would
demand the insertion of a previously generated S,
161
complete with lexical items, in other words, a copy of
an earlier S, and 2) its presence would demand the
lexical insertion of a V with a strict subcategoriza
tion feature containing an S. With the rewrite rule
stated this way, deep structures like (1 ) could not be
generated, for the dummy symbol would insure that after
lexical insertion S3 would always be a copy of S^.
2. The condition 2=5 or 1 2 3=5 is inelegant and perhaps
even misleading. 1 2 3 is, after all, a S, and 5 is a
NP. 1 2 3=5 must be construed to mean that S dominates
a S equivalent to 1 2 3.
An alternate means of stating the transformation
might be (1 ).
(1) CX Y] ; [W IgPl Z]
x ) x x I P 3 x
SI: 1 2 3 4 5 6 (obligatory)
SC: 1 2 3 4 5 [1 2 3] 6
Condition: 2=5
Perhaps this notation more accurately captures the
essential similarity so often remarked concerning the
syntactic behavior of NP's and S's; see for example
Jackendoff (1968) and Chomsky (1970).
3. The convention I have adopted here of placing the
definite marker at the bottom of the structure trees
has the advantage of familiarity for it is similar to
other uses of the device, i.e., for marking terminal N
nodes. Because of this familiarity I have maintained
the convention even though it is clumsy. A more handy
convention might be to assign the feature to S node.
In either case the feature will eventually trigger a
segment transformation which will affix a feature seg
ment to the left of S, just as a segment transformation
is currently supposed to create the article segment to
the left of NP's (Postal 1969, Jacobs and Rosenbaum
1968).
4. A difficulty arises with this analysis when we con
sider that an intermediate structure like (1 ), which
has undergone Definitization, but which, optionally,
has not undergone Restrictive Relative Clause Formation,
meets the structural index for Extraposition as that
transformation is now stated (see (3.13)).
162
(1) S
NP VP NP VP
alligator NP
ate cannibal _ got
VP sick alligator NP
alligator
+ def ate
cannibal
If Extraposition were to apply, the result would be
(2 ), an ungrammatical surface structure.
(2) *An alligator ate a cannibal, and the alli
gator got sick, the alligator ate a cannibal.
To block the application of Extraposition in such
structures as (1 ), the structure index for that trans
formation must be limited in the following way:
(3) X [ NP S ] Y
NP +def NP
This states that only S’s bearing the definite
marker may be extraposed. This is already the case for
noun phrase complements and must also be the case for
relative clauses in order to prevent the occurrence of
sentences such as (2 ).
Thus an additional operation must be included in
the structural change of the Restrictive Relative
Clause Formation Transformation, one which affixes the
definite marker to the embedded S.
If this adjustment was prompted merely to block the
derivation of sentences like (2 ), it would certainly be
ad hoc and uneconomical, but the adjusted rule of
Restrictive Relative Clause Formation is independently
motivated for the following reason. If it is the case
that only definite NP's may be pronominalized, as I
have previously argued, then it must also be the case,
that only S's that have been definitized may be pro
nominalized. And it is upon this assumption that I
have based my claim for the personal pronoun nature of
163
it in sentences containing extraposed noun phrase com
plements. Now, if Postal's (19 69) claim that such is a
pro-relative is true, as I believe it is, it seems only
natural that relative clauses should also be definite
prior to pronominalization— and the adjustments I have
just introduced would effect such a provision.
CHAPTER FOUR
4.0 Conjunction Attachment, Prior Conjunct Deletion,and Presupposition
In the preceding chapters I have described the process of Defini
tization and the role it plays in the derivation of restrictive and
non-restrictive relative clause sentenoes, and also of sentences con
taining noun phrase complements. In arguing for such a process as
Definitization, as I have envisioned it, I have been obliged to assume
deep structures significantly different from those posited by more
standard transformational models; more precisely, I have been obliged
to assume that deep structures consisting of conjoined S's underlie
surface structures containing relative clauses and noun phrase comple
ments. That this must be the case is suggested by the fact that for
each relative clause sentence and for each noun phrase complement sen
tence in English, there is a paraphrase which involves conjoining,
while the reverse is not true. That is, complex sentenoes containing
relative clauses or noun phrase complements always have .compound sen
tence paraphrases, but compound sentences do not always have complex
sentence paraphrases. In the following examples, the conjoined struc
tures, i. e., the (b) sentences of (4.1) and (4.2), are paraphrases of
the corresponding (a) sentenoes, a relative clause sentence and a noun
164 . .j
165
phrase complement sentence, respectively; but (4.3), a conjoined sen
tence, does not have any relative clause or noun phrase complement
paraphrase.
(4.1) a. The horse that won the derby was raised on our farm,
b. A horse won the derby, and the horse was raised on
our farm.
(4.2) a. That Jeff caught the measles was a crying shame,
b. Jeff caught the measles, aind that was a crying
shame.
(4.3) Sally collects tropical fish, and Feed plays snooker.
The processes of Definitization, Restrictive and Non-restrictive
Relative Clause Formation, and Post-definitization Deletion, together
with the other transformations discussed in the preceding chapters,
have done much to account for the derivational histories of conjoined
deep structures which surfaoe as structures containing embedded S's.
Still there remain two processes about which I have remained deliber
ately vagua: the prooess by which a conjunct of the deep structure is
deleted (a maneuver which I have tentatively labeled Prior Conjunct
Deletion), and the placement of the coordinating conjunction in the
event that it is the conjoined paraphrase that is selected to surface.
It is to these two processes that I now turn ny attention.
166
4.1 Conjunctions and Prior Conjuncts
4.1.1 The Attachment of and
The placement of the coordinating conjunction appears to be a
trivial issue. I think that we may regard the coordinate conjunction
and, or its marker at least, as implicit in any structure having the
shape (4.4).
(4.4) X
X X
Should this branching arrangement of S's, which I assume to be
present in the deep structures underlying such surface manifesta
tions as (4.1a) and (4.2a), maintain its integrity, having under
gone the cyclic transformations, then we may suppose that a low
level rule of lexical insertion will select the coordinate conjunc
tion and from the second lexicon and position it appropriately,
producing surface structures of the type (4.1b) and (4.2b).
On the other hand, should some transformation such as Prior
Conjunct Deletion or Non-restrictive Relative Clause Formation, in
the case of conjoined S's, or Conjunct Movement, in the case of
noun phrases (see Lakoff and Peters, 1969), apply to delete or
to detach one of the branches, then, of course, the structure index
of this lexical insertion transformation will no longer be met,
and it will not apply.
167
4.1.2 Prior Conj uncts
4.1.2.1 Prior Conjunct Deletion
The matter of Prior Conjunct Deletion is not, however, a
trivial one. First of all, this rule clearly must occur in the
transformational cycle. In section 3.1.2 above, it became cibvious
that such a rule was needed in the cycle so that sentenoes of the
type (3.29), (3.3.0), and (3.31) would not surface, as a general
rule, but would simply be more cumbersome paraphrases of the more
acceptable (3.16).
(3.16) The cow which it is obvious that Roscoe bought
has a crumpled horn.
(3.29) Roscoe bought a cow, and it is obvious that Roscoe
bought a cow, and the cow which it is obvious that
Roscoe bought has a crumpled horn.
(3.30) Roscoe bought a cow, and it is obvious that Roscoe
bought a cow, and the cow which Roscoe bought and
which it is obvious that Roscoe bought has a crum
pled horn.
(3.31) It is obvious that Roscoe bought a ccw, and the
cow which it is obvious that Roscoe bought has a
crumpled horn.
Secondly, the transformation is an optional one, for although
(3.29), (3.30), and (3.31) are less acceptable than (3.16), they are
168
not ungrammatical, and henoe their derivation should not be blocked,
a possibility which would result were the rule an obligatory one.
Also, if the rule of conjunct deletion were made-obligatory, sentences
such as (4.1b) and (4.2b), which.are perfectly grammatical and
acceptable, would be prevented from surfacing.
In section 3.2 I plaoed this transformation in the cycle prior
to Non-restrictive Relative Clause Formation and immediately following
Post-definitization Deletion. Such a positioning is not very well
motivated, hcwever. I reasoned that because Prior Conjunct Deletion
could prevent the occurrence of the structure index for Non-restric
tive Relative Clause Formation by applying to delete the left-hand
conjunct, it should be ordered prior to Non-restrictive Relative
Clause Formation so that it could optionally perform this function.
But because Non-restrictive Relative Clause Formation, like Prior
Conjunct Deletion, must be an optional transformation, the ordering
of the two transformations does not appear to be crucial. Consider
the sentenoes of (4.5) and (4.6)
(4.5) a. I know a wombat, and the wombat tells obscene
stories.
b. I know a wombat, who tells obscene stories.
c. The wombat tells obsoene stories.^-
(4.6) a. Bill collects frogs, and that Bill collects
frogs is dumb.
b. Bill collects frogs, which is dumb.
c. That Bill collects frogs is dumb.
169
Now if we assume an ordering like that in (4.7),
(4.7) a. Non-restrictive Relative Clause Formation (optional)
b. Prior Conjunct Deletion (optional)
should Non-restrictive Relative Clause Formation apply, Prior Conjunct
Deletion would be blocked, and the (b) sentenoes of (4.5) and (4.6)
would result. Should Non-restrictive Relative Clause Formation not
apply, then depending on whether or not Prior Conjunct Deletion
applied, either the (c) or (a) sentenoes would result.
On the other hand, if we assume the order shown in (4.8),
(4.8) a. Prior Conjunct Deletion (optional)
b. Non-restrictive Relative Clause Formation (optional)
then should Prior Conjunct Deletion apply, Non-restrictive Relative
Clause Formation would block, and the (c) sentences would result.
Should Prior Conjunct Deletion not apply, then depending on whether
or not Non-restrictive Relative Clause Formation applied, either the
(b) or (a) sentences would result. Thus, because either ordering of
transformations yields the same results, I conclude that the ordering
2
of these two transformations is not crucial.
The reason for ordering Prior Conjunct Deletion after Post-
definitization Deletion was that the latter transformation ,as now
stated (I repeat figure (3.8) here),
(3.8) Post-definitization Deletion
[X NP Y] ; [W NP [X NP Y] Z]
S S S +def S +def S S
SI: 1 2 3 4 5 6 7 8 9 > (obligatory)
SC: 1 2 3 4 5 0 0 0 9 Condition: 5=7 or
5=6 7 8
contains the initial conjunct as part of its structure index, as does
the Restrictive Relative Clause Formation transformation which pre
cedes it. Should Prior Conjunct Deletion have applied earlier in the
cycle to eliminate the left-hand conjunct prior to Hie application of
either of the transformations, their operation would have been blocked.
This appears to be a more serious motivation for the ordering of
Prior Conjunct Deletion in the cycle. Examine now (4.9), an inter
mediate structure, which we may take to be the output of the Definiti-
zation transformation.
+def
If at this point in the derivation S-^ were to be deleted, then neither
Restrictive Relative Clause Formation nor Post-definitization Deletion,
las worms
,1 '
know elephant
whose structure indioes are shown in (4.10), would apply because the
171
first three elements of the structure index no longer exist.
(if.10) [X NP Y] ; [W NP [X NP Y] Z]
S S S +def S +def S S
£
1 2 3 4 5 6 7 8 9 where 5=7
If Prior Conjunct Deletion does not apply, of course, Restrictive
Relative Clause Formation and Post-definitization Deletion might
operate, but should this be the case, there would then be no way to
eliminate S - j _ after their operation,so that while (4.11a) and (4.12a)
could be derived, there would be no way to derive (4.11b) and (4.12b).
(4.11) a. I know an elephant, and the elephant which I know
has worms,
b. The elephant which I know has worms.
(4.12) a. I knew an elephant, and the elephant has worms,
b. The elephant has worms.^
This, then, was the reasoning upon which the tentative ordering
of transformations in section 3.2 was based. And this grammar seems
to yield satisfactory results.
Yet it seems to me that there is no real necessity to maintain
that the structure index for Restrictive Relative Clause Formation
and for Post-definitization Deletion must be that shown in (4.10).
There is no absolute requirement, so far as I can tell, that would
demand the presence of the left-hand conjunct in the structure indices
of either of these transformations. The only argument for the mainte
nance of the left-hand conjunct, at least until after the operation
of Post-definitization Deletion, might rest on the belief that after
the operation of Definitization at SQ (as shown in figure (4.9)), the
deletion of would reduce the structure to just that dominated by
S2 , and since S2 cycle has already been passed, no more transforma
tions might apply. Such reasoning would, however, be fallacious,
because the structure that would remain after the deletion of
would not, in fact, be S2 but a non-branching SQ, which would dominate
S2. The transformational cycle would continue at Sq level until all
cyclic transformations had been allowed to operate, at which point
Pruning, which is perhaps the last transformation in the cycle, would
apply to erase Sq.
Thus if the structure index for Restrictive Relative Clause
Formation and Post-definitization Deletion can be reduced from that
shown in (4.10) to (4.13),
(4.13) W NP [X NP Y] Z
+def S +def S
1 2 3 4 5 6 where 2=4
then we may argue that Prior Conjunct Deletion may occur as early
in the cycle as immediately following Definitization. This being the
case, it might be further argued that this optional deletion of the
left-hand conjunct is actually part of the Definitization process;
that is, that Definitization obligatorily copies the left-hand con
junct into an NP of the right-hand conjunct just in case that NP is
coreferential with one contained in the left-hand conjunct, or if
that NP is coreferential with the totality of the left-hand conjunct,
173
and it may optionally delete the first conjunct after the copy has
been embedded in the second.
What effect does this alteration have on the grammar? None that
I can discover, except that the structure indices of two rules are
made simpler, and the total number of transformations has been re
duced by one. We now consider Prior Conjunct Deletion to be an
optional suboperation of Definitization (just as Thompson (1970:42-43)
had earlier believed such a conjunct deletion maneuver to be an
optional suboperation of relativization).
Even though we have apparently settled the problem of the
optimum position of the conjunct deletion operation within the trans
formational cycle, it has not gone undetected that the combination of
the applications of conjunct deletion and Post-definitization Deletion
(regardless of the order in which they occur) apparently constitutes,
at least in some cases, a non-recoverable deletion. As it has long
been one of the axioms of transformational theory that all deletions
must be recoverable, this seeming difficulty must be explored.
4.1.2.2 Non-recoverable Deletions
First of all, we notioe that the problem of a non-recoverable
deletion does not occur in sentences of the noun phrase complement
type, i. e., sentenoes like those in (4.6). In such sentenoes the
coreferent shared by the conjuncrts in deep structure is an S, and
neither the conjunct deletion operation associated with Definitization
nor Post-definitization Deletion ever removes that coreferent from
the structure tree. But in sentenoes of the type (4.5) or (4.12),
in which the shared coreferent of the deep structure conjuncrts is
merely a simple NP, that coreferent NP remains in the tree as a
definitized NP, but the combined operation of conjunct deletion and
Post-definitization Deletion will completely erase the initial con
junct of the deep structure, an apparently non-recoverable deletion.
We can illustrate the operation this way. Assume that the
deep structure of (4.5) is (4.14).
onscene
stories
Definitization will obligatorily copy into NP2, Chomsky-adjoining
it, and optionally deleting the original S]_. Assuming that the
optional portion of the Definitization transformation applies, we
obtain the resulting structure tree, (4.15).
(4.14) S,
0
wombat
(4.15)
S
2
wombat
+def
New should Restrictive Relative Clause Formation apply to move NPg'
to the front of S-^', adding +WH and +PRO, the result would be (4.16),
which contains all the information found in the deep structure (4.14).
(4.16) The wombat who I know tells obscene stories.
But if Restrictive Relative Clause Formation does not apply (remember •
that it is optional), Post-definitization Deletion, being obligatory,
must apply. In this case the resulting structure tree after pruning
would be (4.17), which would surface as (4.18).
(4.17) S2
NP
Zb,
obsoene
stories
(4.18) The wombat tells obscene stories.
What has happened is that the left-hand conjunct of the deep
structure has disappeared completely; thus the model I have been pro
posing seems to be making the claim that the deep structure of (4.18)
is (4.14), which contains a conjunct S that does not surfaoe at all.
I contend, however, that the initial conjunct has not completely
disappeared. There is a trace which remains behind and which argues
convincingly for the earlier presence of the deleted conjunct. That
traoe is the definite article. The presence of the definite article
wombat
+def
176
implies that the wombat under discussion has been mentioned at some
time in the discourse prior to its mention here, making the occurrence
of the definite article here an instance of anaphora. This anaphoric
definite article actually triggers the recollection of the wombat.
What I am proposing then is that the speaker who utters (4.18)
is employing the following linguistic strategy4 in his effort to make
himself understood. He knows that at some earlier point in the dis
course— several sentences, perhaps even several days or weeks earlier
— he has communicated to the hearer a sentence to the effect I know a
wombat. New in uttering (4.18), he is depending on the hearer's
memory to make the associative leap over sentences, days, or weeks to
that earlier utterance and to relate this wanibat, marked +definite, to
the one mentioned earlier. More precisely, the speaker generates the
deep structure (4.14) underlying (4.18) and obligatorily applies
Definitization. At this point he may either trust or distrust his
hearer's memory. Should he distrust it, he has the option of allowing
the initial conjunct to remain, either in its original position or
in the copy posited by Definitization. In this case the sentence may
surfaoe as (4.5a), (4.5b) or (4.16), depending upon what subsequent
transformations are selected to apply.
(4.5) a. I know a wombat, and the wombat tells obscene
stories.
b. I know a wombat, who tells obscene stories.
(4.16) The wonbat who I know tells obscene stories.
Should the speaker trust the memory of the hearer, however, he
may elect to delete the initial conjunct, together with the copy
embedded in the second conjunct by Definitization, since both are
merely repetitions of an earlier predication which he expects his
hearer to remember. The speaker's success in conmunicating with his
hearer in this second situation is wholly dependent upon the ability
of the hearer to make the associative leap— to relate this wombat
to the one mentioned earlier, by intuiting the earlier predication
as a part of the deep structure of the sentence he is now hearing.
If the speaker is successful in his strategy, he is likely to
elicit the response from his hearer, He's a cad, in which the hearer
has a) demonstrated his understanding by associating the just-mentioned
wombat with the earlier-mentioned one, and b) demonstrated as well
his cwn linguistic expertise by duplicating the strategy employed by
the speaker. More exactly, he has taken up the statement (4.18) and
conjoined it with a statement of his own, the deep structure of which
would look something like (4.19).
(4.19)
NP-L VP
I *
^jp wombat
NP2 VP
I
wombat
y
NP
tells
is cad
obscene
stories
178
The second speaker will now definitize NP2 by copying Shinto
it, and he will not feel at all reluctant to rid himself of both
(via the conjunct deletion option associated with Definitization) and
its copy resulting from Definitization (via Post-definitization De
letion) . Left only with the structure (4.20), he pronominalizes NP2
to produce the surface structure he desires.
(4.20)
wonibat
+def
Going back to our first speaker, we may, on the other hand,
witness his consternation, if instead of He's a cad, or some such
similar response, he receives the response What waiibat? In this
case, his linguistic strategy has backfired, and he must backtrack
with The wombat I know, all the while wishing that he had not chosen
a strategy that plaoed such an imposition on the memory of his hearer.
There seems to be some connection between the initial conjunct
and prior utterances in the discourse— and if this is true, then the
recoverability of the deletion effected by the combined application
of the conjunct deletion option of Definitization and Post-definiti
zation Deletion is strictly dependent on the hearer's memory, i. e.,
upon his ability to intuit the content of the deleted S. To the
extent that the burden plaoed on the hearer's memory is not too
179
great, the transformational deletions employed by the speaker are
indeed recoverable. When the burden upon the hearer’s memory is such
that he can no longer make the association between what he is hearing
and what he has heard earlier, or what he is expected to have heard
earlier, then communication has broken down— the deletions made by the
speaker are not recoverable, at least for the hearer.
4.2 Presupposition
What this extended examination of a very ordinary linguistic
exchange has accomplished, I hope, is to demonstrate, first, that
certain deletions, which are sanctioned by this model and which
appear to be non-recoverable, are indeed recoverable when a larger
sample of discourse is examined, and second, that we are a little
closer to pinpointing the true function of the initial conjuncts in
the deep structures I have been proposing. The initial conjunct
embodies the part of the linguistic message that the speaker expects
the hearer to be aware of— the presupposition.
As we have seen, it is not always the case that the speaker will
depend on the hearer to recall a statement frcm an earlier stage of
the discourse. He may elect not to delete either the initial conjunct
or the copy embedded by Definitization. Retaining one or the other,
or both, he employs a strategy that is surer, one that reminds his
hearer of the message that has occurred earlier in the discourse.
Or the statement may be self-contained. (4.21), whose deep
structure we will assume is (4.22) may very well be an example of
such a self-contained statement.
180
(4.21) The man who sells used tooth brushes has a
black eye.
(4.22) SQ
man
used tooth
brushes eye
i In this case the speaker does not presuppose that the hearer is aware
j either of the man who sells used tooth brushes or about his black eye.
; Both conjuncts may be construed as containing new information for the
! hearer, so the speaker will not choose the options available to him
; which would result in the deletion.of the first conjunct. Instead,
ihe will adopt a strategy that will preserve all the new information in |
| the surface structure. Here he must obligatorily apply Definitization. ;
; He elects to delete S^, but he preserves the copy embedded by Defini-
i tization into NP2 by applying the optional Restrictive Relative Clause i
!Formation transformation, in turn blocking the operation of Post-
definitization Deletion, which otherwise would have deleted the copy.
j - “ I T T . . . .
! Notice now the deep structure (4.24), which according to my
I
i
janalysis underlies (4.23).
(4.23) That Veronica spilled the Kool Aid is false.
j
i
1
181
(4.24) SQ
Kool Aid
In this case it is the notion, Veronica spilled the Kool Aid,
that is being denied, but, even so, it is precisely this predication
that the speaker presupposes his hearer to believe and about which
he makes the further claim, is false. The strategy, as before, calls
for the definitization of the subject NP of the second conjunct,
after which S-^ and its copy in S2 may be deleted, leaving the de
sired structure (4.23).
Even if the accusation about Veronica does not figure in an
earlier discourse, if, for example, I should calmly utter the con
tents of (4.23) to a total stranger (a stranger both to me and to
Veronica), he should nonetheless manage to understand the utterance
fairly well. But I think that he would do so mainly because his
intuition would be that there existed an accusation, namely, Veronica
spilled the Kool Aid, and that I am informing him that that particular
accusation, that Veronica spilled the Kool Aid, is false.
The fact is that noun phrase complement sentences are more
182
easily interpretable with respect to presupposition than are sen
tences like (4.5), whose deep structure might develop variously into
a compound sentence, a relative clause sentence, or a simple sentenoe,
depending on the strategy employed by the speaker. Noun phrase
complement sentenoes are more easily interpretable because, regardless
of the strategy adopted by the speaker, he will not obliterate the
information contained in the initial conjunct, i. e., the presupposi
tion, because that predication is mirrored as a NP of the second
conjunct. (Various strategies that reinforoe the presupposition in
noun phrase complement sentences are available. Recall, for example,
sentenoes (3.29), (3.30),and (3.31).)
In the final analysis, then, the conjoined deep structures that
I have been proposing, and which I claim underlie both relative clause
sentences and noun phrase complement sentences, seem to have their
most convincing confirmation in this notion of presupposition.
Consider again sentences (4.1) and (4.2).
(4.1) a. The horse that won the derby was raised on
our farm.
r
b. A horse won the derby, and the horse was
/
raised on our farm.
(
(4.2) a. That Jeff caught the measles was a crying shame,
b. Jeff caught the measles, and that was a crying
shame.
Up till new I have been contending that the deep structures under-
183
lying such sentences as these consisted of conjoined S's. This
claim was based on the obvious fact that the sentenoes of (4.1) are
paraphrases of one another— as are those of (4.2). And because all
relative clause sentences and noun phrase complement sentences have
compound sentence paraphrases, while not all compound sentenoes have
relative clause or noun phrase complement sentence paraphrases, I
considered the conjoined structure to be primary.
Furthermore, I have treated the initial conjunct of these con
joined structures as being ’generated earlier* or as 'occurring
earlier in the discourse' so that through the process which I have
labeled Definitization, definite markers are attached to NP's and S's
! in succeeding conjuncrts, just in case coreferents for these NP's or
S's appear in preceding conjuncrts. The process of Definitization
; then is simply an instance of anaphora which marks NP's or S's as
having been mentioned earlier in the discourse, in this particular -
I i
case in a preceding conjunct S. ;
Notioe new that the notions 'generated earlier'1 and 'occurring
earlier in the discourse'— notions difficult to express or to incor- j
: porate into linguistic theory— find a satisfactory and altogether
inatural resolution in the notion of presupposition. I
| |
It is evident that in (4.1a) there is the underlying presupposi- ;
I !
!tion (4.25).
j (4.25) A horse won the derby. j
I
I And in (4.2a) there is the underlying presupposition (4.26).
t
i 1
(4.26) Jeff caught the measles. j
184
And in my analysis it is just these presuppositions which appear as
the initial conjuncts in the deep structure. It is just these pre
suppositions or initial conjuncts which manifest themselves as re
strictive relative clauses or noun phrase complements in surface
structure. And it is succeeding conjuncts, i. e., those that do not
originate in deep structure as initial conjuncts or presuppositions,
that manifest themselves in the surface structure as non-restrictive
clauses.
The conjoined structure analysis thus seems to provide a mecha
nism that the grammar of English needs for dealing with the notion of
presupposition. Using this mechanism we are at last able to handle
in a more satisfactory way one of the problems that gave initial
iirpetus to this study, the problem embodied in the sentences of
(4.27).
(4.27) a. A dog I own bit me.
b. The dog I own bit me.
And the problem is simply this: why is the (a) sentence ambiguous,
while the (b) sentence is not?
i
The explanation lies in the fact that the (b) sentence has
only one possible presupposition. It may come from an earlier dis- j
course and be repeated as a reminder to the hearer, or it may con- j
i
stitute new information delivered concurrently with that of the j
second conjunct. However the initial conjunct of the deep structure I
l
of the (b) sentence is to be interpreted by speaker or hearer, that
is, as information from an earlier discourse or new information, it
can only be (4.28).
185
(4.28). I own a dog.
In the (a) sentence, however, there is an ambiguity resulting
: from a decision that is thrust upon the hearer. He must decide
whether the speaker is the owner of a single dog or of a nuirber of
dogs. If the speaker and hearer are acquaintances of such standing
: that the hearer knows the number of dogs that the speaker owns, then
j
he will correctly interpret the utterance. If he does not knew this
; information, then he is forced to choose or to guess between two con-
I flicting presuppositions, one presupposing a multiplicity of dogs,
| the other presupposing only a single dog.
To complete this example and to draw this section to a close,
; the deep structure of the sentences of (4.27) are shown below in
(4.29). (4.29a) is the structure underlying the unambiguous (4.27b).
| (4.29b8c) represent the dual meanings associated with (4.27a).
(4.29) a S
0
NP VP NP VP
I
V NP
II II
own dog bit me
b S
NP VP NP VP
own dogs bit me
The deep structure shewn in (4.29a) presupposes a single dog
in its initial conjunct. Definitization occurs, followed by the
' optional deletion of S-^. Restrictive Relative Clause Formation next
; converts the copy of enbedded into S2 by Definitization into a
relative clause thus blocking its deletion by Post-definitization
Deletion. Tne sentence will now surface as (4.27b).
The deep structure shown in (4.29b) presupposes a nunber of
dogs in its initial conjunct. Again Definitization occurs, in the
i manner peculiar to partitives discussed in section 2.3.2 above, and
j is followed by Restrictive Relative Clause Formation. The result
; would be an intermediate structure which might conceivably surface
; as something like (4.30), but which would be transformationally
! reducible to (4.31) or (4.27a) by appropriate partitive reduction
; trans formations.
(4.30) A dog of the dogs which I cwn bit me.
(4.31) One of the dogs which I own bit me.
The deep structure shewn in (4.29c) presupposes a single dog
in its initial conjunct. Definitization occurs, but the option
187
which deletes S - j _ is not elected, nor does Restrictive Relative Clause
Formation apply. Post-definitization Deletion erases the copy of Sq
embedded into S2 by Definitization. Next, the last cyclic transfor
mation, Non-restrictive Relative Clause Formation applies to attach
the markers +WH and +PR0 to the definitized coreferent NP in the
second conjunct.(It could only be the coreferent NP of the second
conjunct since the ooreferent NP of the first conjunct remains unde-
finitized.) Conjunct Incorporation next moves the non-restrictive
clause into the first conjunct. A subsequent rule,which permits the
elimination of the relative pronoun, provided that the non-restrictive
clause is dominated by an undefinitized NP, will yield the desired
surface structure (4.27a).
4.3 New Directions
The notion of presupposition introduced in the preceding sec
tions is hardly a new concept. It has been the subject of granmatical
inquiry for a number of years. Reoently it has received enthusiastic
attention in transformational circles, which has occasioned a rash of
discussion concerning the subject. The most recent connentaries
with which I am familiar are Chomsky (1969), Morgan (1969), and
Muraki (1970).
The same concept has been studied by the Prague Circle and has
there received a different nomenclature. Presupposition to these
scholars has involved the relationship of the theme (the previously
known segment of discourse) with the rheme (the newly introduced
segment of discourse).^
188
More recently Wallace Chafe (1970) has adopted an analysis
similar to that of the Prague Circle’s but with a more transparent
terminology. He refers to old information and new information.
The conments that I make in this section are highly tentative
and speculative, the main body of iry thesis having already been
concluded with the preceding section. Here I shall speculate briefly
on certain interesting parallels that seem to me to obtain between
the Prague School analysis and iry own. I shall adopt Chafe’s termi
nology because the terms old and new information seem to me more ad
vantageous, for reasons of familiarity,than theme and rheme.®
Following this discussion I shall indicate two possible avenues
for continued research along the lines drawn in the preceding pages.
4.3.1 Comparison with the Prague Circle Analysis
The Prague Circle limited its concept of old and new information
to the attitude of the speaker toward— his semantic interpretation of
— the lexical items making up a sentence. It was assumed that such a
notion was necessary in ongoing discourse so that coherence could be
maintained, that new ideas could be introduced yet still be anchored
to the prior disoourse. Sentenoes thus consisted of old information
and new information, and perhaps transitional information. For
example, (4.32) offers an example of old and new, while (4.33)
furnishes an example of all three types of information.
(4.32) The hunter shot a wolf yesterday.
(4.33) He has turned out to be a good cock.
189
In (4.32) the hunter is the old information and shot a wolf
! yesterday is new information. The speaker presupposes that his
hearer knows about die hunter, most likely because the hunter has
been introduced earlier in the discourse or because the speaker and
hearer share a mutual understanding about the referent of the hunter.
The speaker, however, does not presuppose that the hearer knows about
the shooting of the wolf, nor about the time of its occurrence. These
items constitute the new information and carry a greater degree of
’ ccmmunicative dynamism,' a quality evidenced by a number of markers:
. higjier pitch, stronger amplitude, and occurrence in the predicate, to
mention a few. Old information carries less communicative dynamism:
: lower pitch, lesser amplitude, and a frequent occurrence as subject.
In (4.33) the subject, he, is the old information, and a good
cook is new information, while has turned out to be is transitional
: information, information characterized by a lesser degree of oommuni-
; cative dynamism than that which is labeled new.
Notice also that certain lexical items are obligatorily repre- i
i |
| sentative of old information— personal pronouns like he in (4.33)—
i j
; while others are obligatorily representative of new information— nouns j
; indicative of time, such as yesterday in (4.32). The articles also j
' j
| enter into this system, the definite articles most often indicating j
i i
j old information, the indefinite cnes most often indicating new infor- j
! !
| nation.
j
With this much background, perhaps certain similarities between
the Prague School system and the one I have been developing may become
apparent. If we assume the functional strategy of a sentence to be j
19 0
the speaker's attempt to convey new information while maintaining the
continuity and coherence of the discourse by references to previously
mentioned, and hence old, information, we might well develop a rewrite
rule to capture this notion. We might rewrite S as 0 + N, where 0
• 8
equals old information and N equals new information.
In the Prague School analysis the terms equivalent to 0 and N
night roughly equate with NP and VP. In the more abstract system I
have been developing, 0 and N might best be viewed as S's conjoined
at the highest level.
Locking again at the sentences of (4.1) and (4.2) is helpful.
(4.1) a. The horse that won the derby was raised on our
farm.
b. A horse won the derby, and the horse was raised
on our farm.
(4.2) a. That Jeff caught the measles was a crying shame,
b. Jeff caught the measles and that was a crying
shame.
If we consider the (b) sentenoes more nearly a manifestation of the
deep structure than the (a) sentences, then the information which
is contained in the first conjunct S in each, and which we have
described as embodying the presupposition implicit in the (a) sen
tences, surely answers to the specification old information. And
the second oonjunct S in the (b) sentences, which ends up as the
matrix S in the (a) sentences, surely corresponds to new information.
191:
I think that it would be unwise, however, to maintain that the
rule S — > 0 + N is a base rule. To do so would mean that constituent;
status must necessarily be assigned to 0 and N, and that they must be
rewritten again as S’s but with the stipulation that those S's might
not be again expanded as 0 + N. That is, that S— > 0 + N is strictly
a rewrite rule for the highest S in a structure tree. Such a maneuver
would constitute an unwarranted complication of the grammar.
Actually the concept of old and new information is not really
| a syntactic one at all, and as such should not figure in the base
rules. It is a semantic concept, one used to interpret the struc
tures generated by the base. As was noted in Chapter Itoo, the notion '
; of prior or first conjunct has very little significance in a strictly
syntactic sense, for these conjuncts are generated in random order;
: it is the semantic or interpretive component that assigns significance
; to the order of conjuncts in the deep structure. Here we seem to have;
| a fairly clear distinction between syntax and semantics, between
! generation and interpretation. What the rule S— ► 0 + N must be is
! a semantic or interpretive rule.
| Ihe operation of this rule might be likened to a template. If :
! 1his rule were manifested in the form of a structure tree, it would |
S have the following shape: j
I
i
<4.3»0 S !
j Now if this figure were superimposed, like a template, over a syn-
192
tactic structure, it would serve as a powerful device for interpreting
that structure.
For example, in the deep structure underlying (4.1), which we
will assume is that shown in (4.35), the interpretive rule will mark
S-^ as Old information and S2 as New information.
(4..35) S0
horsey was raised
on our farm
; Ihe superimposing of (4.34) onto (4.35) would look roughly like
| (4.36).
horse^ won derby horse^ was raised
on our farm.
But in a sentence like (4.37), whose deep structure is represented by
(4.38), the rule would interpret the syntactic structure differently.
!
t
i
! (4.37) A cat ate a rat.
193
(4.38)
It would superimpose the semantic structure (4.34) onto the syntactic
structure in the following way:
C4.39) S
0 N
S0
NP VP
I / X
cat V NP
I I
ate rat
In (4.39) the interpretive rule has operated to stipulate that all
the information contained in Sq is information new to the discourse.
The node dominated by 0 is thus null or empty.
Such also might be the interpretation assigned by the rule to
structure (4.35). If the structure is to be interpreted as a 'self-
g
contained sentence' in the sense of (4.22), the interpretive rule
will analyze it as follows:
(4.40) S
O' *N
Sn
194
In this case all the information in the sentences dominated by Sq is
new information, and the 0 node is again empty.
Additionally, we may expect to find structures like (4.41),
which are indicative of rote memorization.
(4.41) S
0 N
pledge allegiance . . .
j But a very unsettling prospect new arises. If we are to claim
j that (4.40), as a self-contained sentence containing all new informa- ;
; tion, is somehcw different semantically from (4.36), in which the
! . ;
; initial conjunct or presupposition is taken to be old information,
1 then we have a different situation than the one we observed earlier
i
; !
| in sentence (4.28). j
i
I I
! (4.28) A dog I own bit me.
! The ambiguity in this sentence rested upon the determination con- j
i cerning- the initial conjunct or presupposition— whether it referred I
| to a single dog or a set of dogs. This seems to be a clearly syn-
|
j tactic matter— a matter of what lexical item gets inserted into the
i
: structure tree.
I
I i
| The ambiguity here is a result of the application of a semantic !
I
interpretation rule. To claim that the syntactic structure (4.35) may J
195
have two separate meanings is to violate the conceptual framework
established in Chapter One, which assumes that ambiguous surface
structures have distinct deep syntactic structures.
To remedy this inconsistency it seems necessary to retreat to
the position described earlier (p.190 ), which views the old and new
information as syntactic material and which establishes S > 0 + N
as a base rule, a r*’ le accompanied by the added complexity mentioned
on p. 191.
An alternative solution might be a base rule which rewrites
S as Sold + S“, where Sodd, the initial S in the expansion, is
; specified or marked as the carrier of old information-1 - 0 and where
S* is a rule schema representing an indefinite nunber of ccnjoined
S's. Again it would be necessary to stipulate that this rule would
be the first of the rewrite rules, to be used only once in the gener
ation of a sentence. Necessary also would be a rule which would allow:
S to be rewritten as 0, the situation obtaining in (4.39), (4.40),
and (4.41).
At this point I am not altogether satisfied with either of the
alternatives available: a) complicating the base rules with the
matter of old and new information, or b) altering the conceptual
: framework of the grammatical model I have been describing. For the
i moment, I can only state the problem, recognize its gravity, and
\ I
! stake it out as an area for further investigation.
i i
i i
! 4.3.2 Itoo Avenues for Future Inquiry ;
; |
Regardless of the way in which the above problem is solved, the j
| approaches to further investigation are fairly clear in at least two j
directions.
First, I think that we may expect to find constraints on the
transformations which may apply to S's containing new information
alcne which will not apply to S’s containing old and new information.
For example, the conbined operation of the optional conjunct deletion
associated with De finitization and the Fost-definitization Deletion
transformation have to be blocked in structures like (4.40) although
both might apply to structures like (4.36). It might even be ex
pected that a universal constraint cn all deletion transformations
might is some way obtain in structures containing strictly new
information.
Second, we might hope to produce a more comprehensive grammar
of the definite article. It has been shown in this study how the
process of definitization may account for the presence of the definite'
article in sentences such as (4.22) where the speaker is presenting
new information.
(4.22)
NPj_ VP NP2 VP
I I /■"-—
man y ^p man y
l
has
black eye
tooth brushes
j We called this type of sentence a 'self-contained' sentence. And I
|
j have hinted that other occurrences of the definite article may be
accounted for by deleted presuppositions that have their actual mani- J
197
festations in prior discourse. By further close study along the lines
down which I have been moving, we may finally be able to capture the
behavior of the definite article in a series of formal statements.
We may find that some structure like (4.43) always underlies the
heretofore mysterious sentences like (4.42), mysterious at least in
the origin of the definite article.
(4.42) The cat scratched the dog.
(4.43) S
NP +old
NP cat
| | scratched dog
cat dog
Without the deletion of the old information, (4.43) migjrt have
| conceivably surfaced as (4.44).
(4.44) There was a cat, and there was a dog, and the
' cat scratched the dog.
! 4.4 Conclusion
I The material dealt with in section 4.3, in all its roughness,
| sketchiness and vagueness, constitutes a beginning rather than a oon-
; elusion for this work. In contemplating the work on this subject yet
| to be done, I feel not unlike Alice, who, having travailed bravely in
!
j Wonderland in search of the delightful garden, found it at last,
I
| filled with maniacs playing an anomalous game.
198
CHAPTER FOUR
Footnotes
1. The derivation of this surface structure, of course,
appears to involve a non-recoverable deletion. I con
tend that this problem is merely apparent. I shall
treat the problem fully in section 4.1.2.2.
2. Even if the ordering were crucial, this type of infor
mation would not prove to be extremely valuable as
regards the placement of Prior Conjunct Deletion in the
in the transformational cycle, for, as we have seen in
section 2.1.2.2, Non-restrictive Relative Clause Forma-
is a last cycle transformation. We should have sue-
ceeded only in positioning Prior Conjunct Deletion with
respect to a transformation that occurs only once, in
the last cycle.
3. Again, this sentence is an example of what seems to be
a non-recoverable deletion. See note 1.
4. I use the term ’strategy* to mean the decision on the
part of the speaker to use or not to use optional
transformations in deriving a surface structure.
5. See Firbas (19 66a£b) and Wagner (1969).
6. Muraki’s suggestion (1970:16) that presupposition is a
two place predicate with two sentences as arguments,
such that one is presupposed for the other, seems to
come fairly close to the analysis I am proposing. His
analysis certainly deserves more attention than the
casual mention I give it here. But for the moment I
must limit myself to this brief comparison with the
Prague School Analysis.
7. This term is borrowed from Firbas (1966a).
199
8. Wagner (1969) has in fact done something like this in
establishing the following rules:
(1 ) s * - Theme + Nucleus
(2) Nucleus — - » Rheme + Residue
Here theme is old information, rheme is new information,
and residue is something akin to transitional informa
tion.
9. See the discussion above concerning self-contained
sentences, pp. 179-182.
10. The convention for assigning features to non-terminal
notes has already been introduced in Chapter Three,
note 3. Presumably +old would be attached to the
appropriate S node as a result of the just-stated
rewrite rule, and this feature would serve as a crucial
part of the structure index in transformations involv
ing deletion. See the discussion concerning deletions
in section 4-. 3. 2.
200
REFERENCES
Annear, Sandra. (See Thompson, Sandra Annear).
Bach, Emmon. (1968). Nouns and noun phrases. In Bach and Harms, eds.,
pp. 90-122.
Bach, Emmon, and Robert Harms, eds. (1968). Universals in Linguistic
Theory. New York: Holt, Rinehart £ Winston.
' Baker, Leroy. (1970). Notes on the description of English questions:
the role of an abstract question morpheme. Foundations of
Language. 6.197-219.
Burt, Marina K. (1971). From Deep to Surface Structure: An Introduc
tion to Transformational Syntax. New York: Harper 6 Row.
‘ Chafe, Wallace. (1970). Meaning and the Structure of Language.
Chicago: University of Chicago Press.
Chomsky, Noam. (1957). Syntactic Structures. The Hague: Mouton.
: Chomsky, Noam. (1965). Aspects of a Theory of Syntax. Canbridge:
MIT Press.
1
I Chomsky, Noam. (1969). Deep structure, surface structure, and
semantic interpretation. Unpublished paper, reproduced by
the Indiana University Linguistics Club.
| Chomsky, Noam. (1970). Remarks on nominalization. In Jacobs and
! Rosenbaum, eds., pp. 184-221.
| Fillmore, Charles J. (1963). The position of embedding transforma-
I tions in a grammar. Word. 19.208-231.
| Fillmore, Charles J. (1968). The case for case. In Bach and Harms,
j eds., pp. 1-88.
Firbas, Jan. (1966a). On defining the theme in functional sentence
analysis. Travaux De Linguistiques De Prague. 1.267-280.
201
Firbas, Jan. (1966b). Non-thematic subjects in contemporary English.
Travaux De Linguistiques De Prague. 2.239-56.
Jackendoff, Ray S. (1968). Speculations on presentences and deter
miners . Unpublished paper, reproduced by the Indiana
University Linguistics Club.
Jacobs, Roderick A., and Peter S. Rosenbaum. (1968). English Trans
formational Grammar. Waltham: Ginn.
Jacobs, Roderick A., and Peter S. Rosenbaum, eds. (1970). Readings
in English Transformational Grammar. Waltham: Ginn.
Kartunnen, Lauri. (1968). What do referential indioes refer to?
Rand Corporation Publication No. P-3854, Santa Monica.
Kartunnen, Lauri. (1969). Problems of Reference in Syntax. Unpub
lished monograph.
Katz, Jerrold J., and Paul M. Postal. (1964). An Integrated Theory of
Linguistic Descriptions. Cambridge: MIT Ftess.
KLima, Edward S. (1969). Relatedness between gramuatical systems.
In Reibel and Schane, eds., pp. 227-246.
Kuroda, S-Y. (1969). English relativization and certain related
problems. In Reibel and Schane, eds., pp. 264-287.
Lakoff, George. (1966). Deep and Surface Grammar. Unpublished mono
graph, Harvard University.
Lakoff, George. (1968a). Pronouns and reference. Unpublished paper,
Harvard University.
Lakoff, George. (1968b). Counterparts, or the problem of reference
in transformational grammar. Unpublished paper, reproduced
by the Indiana University Linguistics Club.
Lakoff, George. (1969). Generative Semantics. Unpublished monograph,
reproduced by the Indiana University Linguistics Club.
Lakoff, George. (1970). Irregularity in Syntax. New York: Holt,
Rinehart 8 Winston.
Lakoff, George, and Stanley Peters. (1969). Fhrasal conjunction and
symmetric predicates. In Reibel and Schane, eds., pp. 113-142.
Langendoen, Terence D. (1969). The Study of Syntax. New York: Holt,
Rinehart g Winston.
202
Langendoen, Terence D. (1970). Essentials of English Grammar. New
York: Holt, Rinehart 8 Winston.
Langendoen, Terence D., and Thomas G. Bever. (1971). The Interaction
of Speech Perception and Graiunatical Structure in the
Evolution of English. Unpublished monograph.
McCawley, James D. (1968). The role of semantics in grammar. In Bach
and Harms, eds., pp. 243-269.
McCawley, James D. (1970). Where do noun phrases come from? In Jaccfos
and Rosenbaum, eds., pp. 166-183.
Morgan, Jerry L. (1969). On the treatment of presupposition in trans
formational grammar. Papers from the Fifth Regional Meeting of
the Chicago Linguistic Society. Department of Linguistics,
Chicago University.
Muraki, Masatake. (1970). Presupposition, Pseudo-clefting, and The-
matization. Unpublished Ph. D. dissertation, University of
Texas.
Perlmutter, David M. (1968). On the article in English. Unpublished
paper, Brandeis University.
Paul M. (1967). Crazy notes on restrictive relative clauses
and other natters. Unpublished paper, IBM.
Paul M. (1968). Notes on repetition binding. Unpublished
paper.
Paul M. (1969). On so-called 'pronouns' in English. In Reibel :
and Schane, eds., pp. 201-224.
Paul M. (1971). Cross-over Phenomena. New York: Holt, Rine- ■
hart 8 Winston.
David A., and Sanford A. Schane, eds. (1969). Modern Studies ‘
in English: Readings in Transformational Granroaru Englewood :
Cliffs: Printioe Mall.
I
Rosenbaum, Peter S. (1967). The Grammar of English Predicate Comple- !
ment Constructions. Cambridge: MIT Press. |
i
Ross, John R. (1967a). Gapping and the order of constituents. Un- !
published paper, Massachusetts Institute of Technology. j
j
Ross, John R. (1967b). Constraints on Variables in Syntax. Unpub- j
lished Ph. D. dissertation, Massachusetts Institute of Tech- i
nology, reproduced by the Indiana University Linguistics Club, i
Postal,
Postal,
Postal,
Postal,
Reibel,
203
Ross, John R. (1969a). A proposed rule of tree pruning. In Reibel
and Schane, eds., pp. 288-300.
Ross, John R. (1969b). Adjectives as noun phrases. In Reibel and
Schane, eds., pp. 352-360.
Smith, Carlotta. (1969). Determiners and relative clauses in a
generative grammar of English. In Reibel and Schane, eds.,
pp. 247-263.
Stockwell, Robert P., Paul Schachter, and Barbara H. Partee. (1968).
UCLA Syntax Project. AF project -2801.
Thompson, Sandra Annear. (1967), Relative clauses and conjunctions.
Unpublished paper, Ohio State University.
Thompson, Sandra Annear. (1970). On Relative Clause Structures in
Relation to the Nature of Sentence Complexity. Unpublished
Ph. D. dissertation, Ohio State University.
Vendler, Zeno. (1967). Linguistics in Philosophy. Ithaca: Cornell
University Press.
Vendler, Zeno. (1968). Adjectives and Nominalizations. The Hague:
MDuton.
f
Wagner, Karl H. (1969). Generative Grammatical Studies in the Old
English Language.Heidelberg: Julius Groos Verlag.
204
APPENDIX
1. Definitization
[X NP Y] ; [W NP Z]
S S S S
SI: 1 2 3 4 5 6
SC: (1 2 3) 4 5 [1 2 3] 6
^ (obligatory)
+def
Condition: 2=5
or 1 2 3 = 5
Optional instruction:
Delete original 1 2 3
Restrictive Relative Clause Formation
W NP [ X NP Y] Z
+def S +def S
SI:
SC:
1
1
2
2 4
+WH
+PR0
4
3
5
5
= >
(optional)
Condition: 2=4
3. Extraposition
X [NP S ] Y
NP +def NP
SI:
SC:
1
1
2
2
3
0
4
4 + 3
(optional)
205
4. Post-definitization Deletion
W NP [X NP Y] Z
+def S +def S
SI:
SC:
1
1
2
2
3
0
4 5 6
0 0 6
=>
(obligatory)
Condition: 2=4
or> 2 = 3 4 5
5. Non-restrictive Relative Clause Formation
[X NP Y] ; [ W NP Z]
S ±def S S +def S
SI:
SC:
or
SC:
1
1
2
2 5
+WH
+PR0
5 6
4 6
^ (optional)
0 0 0
6 + 213
6. Conjunct Incorporation
[X NP
S
Y] ; [W NP Z]
S S +def S
SI:
SC:
1 2 3
1 2 [4 5 6] 3
Relative Clause Reduction
X [ NP be VP] Z
S +def S
+WH
+PR0
SI: 1
SC: 1
2
0
3 4
3 4
Condition: 2=5
or 1 2 3 = 5
last cyclic
Condition on
second SC: 2 =
+def
\
Ip (optional)
Condition: 2=5
last cyclic
= >
(optional)
Condition: 2^5
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
A Descriptive Syntax Of King Alfred'S Soliloquies
PDF
The Grammar Of English Causative-Transitivity
PDF
Plutarch On The Glory Of The Athenians: A Reassessment
PDF
A Pedagogical Model Of Discourse Structure
PDF
A case grammar of the parker manuscript of the "Anglo-Saxon chronicle" from 734 to 891
PDF
Theme, Rheme, And Focus As Grammatical Universals
PDF
The Appositive In English: A Transformational Analysis
PDF
A Sociolinguistic Study Of Selected Vowel Changes In Los Angeles English
PDF
A Syntactic Analysis Of The Writing Of Some Fourth-Grade, Fifth-Grade, And Sixth-Grade Caucasian Children In The Los Angeles Schools
PDF
Alfred Of Sareshel'S Commentary On The 'Metheora' Of Aristotle. (Latin Text)
PDF
Structure, Character, And Theme In The Plays Of Arthur Miller
PDF
Introduction To The English Edition Of Meinong'S "Ueber Annahmen" With Selections From The Translation: "On Assumptions"
PDF
The Feminine World View Of Elizabeth Cleghorn Gaskell
PDF
Drayton And Decorum: A Stylistic Study Of The Relation Of Ornament To Subject In 'Peirs Gaveston' (1593) And 'Englands Heroicall Epistles' (1619)
PDF
Double-Entendres In 'The Canterbury Tales'
PDF
The Development Of Method And Meaning In The Fiction Of 'Saki' (H. H. Munro)
PDF
Variant Forms Of English And Scottish Popular Ballads In America
PDF
Facies Variation And The Miocene-Pliocene Boundary In Southern California
PDF
Temporal Structure Of Spoken Korean: An Acoustic Phonetic Study
PDF
Joyce-Bergson Correspondences In The Theory And Time Structure Of 'Dubliners,' 'A Portrait,' And 'Ulysses'
Asset Metadata
Creator
Nicholas, James Karl
(author)
Core Title
Relativization, Complementation, And Deep Structure Conjunction In The Grammar Of English
Degree
Doctor of Philosophy
Degree Program
English
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
language, linguistics,OAI-PMH Harvest
Language
English
Contributor
Digitized by ProQuest
(provenance)
Advisor
Finegan, Edward (
committee chair
), Brown, William H., Jr. (
committee member
), Downing, Bruce T. (
committee member
)
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c18-554959
Unique identifier
UC11362032
Identifier
7206090.pdf (filename),usctheses-c18-554959 (legacy record id)
Legacy Identifier
7206090.pdf
Dmrecord
554959
Document Type
Dissertation
Rights
Nicholas, James Karl
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA
Tags
language, linguistics