ESD TR 68 419_Integration_of_Transformational_Theories_on_English_Syntax_Vol_2_Oct68 419 Integration Of Transformational Theories On English Syntax Vol 2 Oct68
ESD-TR-68-419_Integration_of_Transformational_Theories_on_English_Syntax_Vol_2_Oct68 ESD-TR-68-419_Integration_of_Transformational_Theories_on_English_Syntax_Vol_2_Oct68
User Manual: ESD-TR-68-419_Integration_of_Transformational_Theories_on_English_Syntax_Vol_2_Oct68
Open the PDF directly: View PDF 
.
Page Count: 536
| Download | |
| Open PDF In Browser | View PDF |
ESD1 ACCESSION LIST.
ESTi Call No.
ESD-TR-68-4I9, Vol. II
Copy No.
^025
J-
of
*2^
cys
£J&^UQ*fP%~
INTEGRATION OF TRANSFORMATIONAL
THEORIES ON ENGLISH SYNTAX
Robert P. Stockwell
Paul Schacter
Barbara H. Partee
October 1968
COMMAND SYSTEMS DIVISION
ELECTRONIC SYSTEMS DIVISION
AIR FORCE SYSTEMS COMMAND
UNITED STATES AIR FORCE
L. G. Hanscom Field, Bedford, Massachusetts
f
ESD RECORD CO:"^
RETURN TO
SCIENTIFIC & TECHNICAL INFORMATION DIVlSlOf
(ESTI), BUILDING 12U .
This document has been
approved for public release and
sale; its distribution is
unlimited.
,.-'~--*i v-i-'-v
(Prepared under Contract No. AF f9(628)-6007 by the University of
California, Los Angeles, California.)
AD7023OO
LEGAL NOTICE
When U.S. Government drawings, specifications or other data are used for any
purpose other than a definitely related government procurement operation, the
government thereby incurs no responsibility nor any obligation whatsoever; and
the fact that the government may have formulated, furnished, or in any way supplied the said drawings, specifications, or other data is not to be regarded by
implication or otherwise as in any manner licensing the holder or any other person
or conveying any rights or permission to manufacture, use, or sell any patented
invention that may in any way be related thereto.
OTHER NOTICES
Do not return this copy.
Retain or destroy.
TABLE OF CONTENTS
VOLUME ONE
I.
II.
III.
IV.
V.
VI.
VII.
VIII.
Page No.
GENERAL INTRODUCTION
1
BASE RULES
29
CASE PLACEMENT
37
DETERMINERS
85
PRONOMINALIZATION
183
NEGATION
253
CONJUNCTION.
319
RELATIVIZATION
1»1*5
VOLUME TWO
IX.
NOMINALIZATION AND COMPLEMENTATION
527
INTERROGATIVE
625
IMPERATIVE
659
GENITIVE
699
CLEFT AND PSEUDO-CLEFT
797
PASSIVE
8Ul
XV.
RULE ORDERING
851
XVI.
SAMPLE LEXICON
931
BIBLIOGRAPHY
1027
X.
XI.
XII.
XIII.
XIV.
XVII.
NOMINALIZATION AND COMPLEMENTATION
Contents
I.
BIBLIOGRAPHY
1
II.
INTRODUCTION
1
A.
Claims of the Present Analysis
1
B.
Previous Scholarship
k
1. Chomsky's 1958 Analysis
5
2. Lexicalist vs. Transformationalist .... 6
3. The Distinction between Nominalization
and Complementation
7
k. IT + S
23
5. Extraposition, IT-Replacement, and
Second Passive
2k
6. The Erasure Principle
28
III.
THE PARAMETERS OF NOMINALIZATION
A.
Pactive/NonFactive
1. Syntactic Justification of the
Distinction
2. Criteria for Factivity
3. The Abstract Instrumental
31
31
31
3k
37
B.
Sentential/NonSentential
C.
Emotive/Non Emotive
39
1. The Sources of Complementizers
39
2. Classes of Emotive/NonEmotive Predicates UO
D.
Infinitivalization
1*2
1. Conditions for TO-REPLACE-AUX
k2
2. Derivation of Infinitivals with [+EMOT]
Predicates
k3
3. The Fact -> it?
U7
k. Conditions for EQUI-NP-DEL
U9
5. Derivation of Infinitivals with EQUINP-DEL
50
6. Conditions for Raising Subject to
Subject, or Subject to Object. ... 57
527
37
Contents
£.
IV.
Gerundive/NonGerundive
1. The Relation between Factivity and
Gerundives
2. Gerundives after Prepositions
3. Generic Gerundives
U. Adverbial Gerundives
5. ing-of Gerundives
58
58
60
6l
62
63
F.
Non-Action Infinitival Tense Constraints. . . 6U
G.
Deep Structure Constraints
1. Tense/Aspect Constraints on the
Sentential Object
2. Case Constraints between Matrix and
Constituent
68
H.
Indirect Questions
69
I.
Miscellaneous Exception Features
1. TO-DEL
2. TO-BE-DEL
3. EXTRA
U. RAIS-OBJ-TO-SUBJ
71
71
71
72
73
66
66
THE RULES OF NOMINALIZATION
Ik
A.
GER
71*
B.
FACT-DEL
79
C.
FOR-INSERT
80
D.
EQUI-NP-DEL
8l
E.
RAIS-OBJ
85
F.
RAIS-OBJ-TO-SUBJ
86
G.
RAIS-SUBJ
87
H.
TO-REPLACE-AUX
87
I.
TO-DEL
88
J.
TO-BE-DEL
89
K.
ONE-DEL
90
528
Contents
L.
THAT-INSERT
91
M.
EXTRA
91
N.
THAT-DEL
92
529
NOM - 1
I.
BIBLIOGRAPHY
Bolinger, D. (1967) "Entailment and the Meaning of Structure"
Bowers, F. (1968) "English Complex Sentence Formation"
Chapin, P. (1967) On the Syntax of Word-Derivation in English
Chomsky, N. (1958) A Transformational Approach to Syntax
(1968) "Remarks on Nominalization"
Fillmore, C. (1967a) "The Case for Case"
Kiparsky, P. and C. Kiparsky (1968) "Fact"
Lakoff, G. (1965) On the Nature of Syntactic Irregularity
(1966c) "Deep and Surface Grammar"
Langendoen, T. (1966a) "Some Problems Concerning the English
Expletive it"
Lees, R. (i960) The Grammar of English Nominalizations
Rosenbaum, P. (1967a) The Grammar of English Predicate Complement
Constructions
(1968) English Grammar II
and T. Langendoen (l96Uc) "On Sentential Adjuncts of
Transitive and Intransitive Verbs"
and D. Lochak (1966) "The IBM Core Grammar of English"
Ross, J. (1967c) Constraints on Variables in Syntax
Wagner, K. (1968) "Verb Phrase Complementation: A Criticism"
II.
A.
INTRODUCTION
Briefly, the Claims of the Present Analysis
Except for the relative clause, all embedded sentences
in this grammar are directly dominated by the node NP. The node
NP itself only appears directly dominated by some case, a case
determined by a head verb, adjective, or noun. Given this frame
of reference, therefore, all sentential complements, whether on
nouns as in (l.a), on verbs as in (l.b), and (l.c), or on adjectives as in (l.d), are nominalizations of the S dominated by an
NP which is dominated by the Neutral case which has undergone
objectivalization (l.a,b,d) or subjectivalization (l.c). If the
head noun is the deletable noun fact, the nominalization may
appear in the surface structure to be dominated by some case other
than Neutral, but (l.e), where the sentential subject might be
assumed to be dominated by a deep structure Instrumental (or Means),
is derived from (l.f), where the item that would be dominated in
a deeper structure by Instrumental case is fact.
(l)
(a)
(b)
(c)
The fact that he left early was annoying.
He demanded that she leave early.
It appeared that he was stupid.
530
NOM - 2
(d)
(e)
(f)
He is anxious that she understand his motives.
That he has blood on his hands proves that
he is guilty.
The fact that he has blood on his hands proves
that he is guilty.
Traditionally, grammarians have divided simplex sentences
into three large classes (sometimes with a fourth — exclamations
or assertives), the classification being determined by the form
or mood of the verb that is characteristic of each type: declaratives (indicative mood), imperatives (subjunctive mood), and
interrogatives (inversion of subject and auxiliary, or special
verb forms in some languages). All three types of simplex sentences can be embedded. When embedded, they undergo transformational mapping into surface structures that differ considerably
from the surface structure of the simplex form, the form they
would have as the topmost S, to which last-cyclic rules would
apply (e.g. inversion of the interrogative, deletion of the second
person subject of imperatives). The nominalization rules provide
an account of these differences in form, describing in particular
their clausal form, their infinitival form, and their gerundive
form.
Derived nouns like proposal, insistence, inference, denial,
or claim, which have been taken as transformationally derived,
by some grammarians, are here taken as lexically derived, for
reasons set forth in the general introduction and in part recapitulated in the annotation (Section B) below. The class of nominals
that have been labeled "Action Nominals" (e.g. by Lees, i960),
having the form V-ing of OBJ,- as in the killing of the rats,
the several bombings of civilians that we witnessed, the eliminating
of deadwood from the ranks, - are likewise taken here as lexically
derived, for the same reasons (the fact that they have such nounlike qualities as taking a full range of determiners, relative
clauses, singular/plural contrasts, and so on). It is necessary
to distinguish these action gerundives, which are lexically
derived, from transformationally derived gerundives, either factive
(the fact of his having given money to John), generic (hunting
polar bears is fun), or verb complements of a highly restricted
type (he avoidedTeaving).
The description of nominalization is set forth in terms
of a set of parameters, some of which are quite general in that
they partition the predicates which govern nominalizations into
large sets each characterized by a definable range of general
syntactic properties, and others of which are essentially exception
features that set off small classes exhibiting syntactic irregularities.
531
NOM - 3
One important parameter is the distinction between
FACTIVE and NON-FACTIVE first set forth in detail by the
Kiparskys (1968). They proposed that many of the differences
in the form and meaning of nominalizations depend not on essentially
arbitrary syntactic features but rather on semantic features in
the governing items. Factive predicates can only occur when the
speaker presupposes that the sentential object or subject of the
predicate is true, or factual; non-factive predicates occur when
the speaker merely asserts or believes the predicate to be true,
but does not presuppose its factuality. The distinction is
clearest under negation, since the presupposition remains constant
in both the negative and positive forms of the sentence:
(2)
(a)
(b)
(c)
(d)
It is odd that the door is closed.
[Kiparsky MS, p.8]
It isn't odd that the door is closed.
[Kiparsky MS, p. 9]
I regret that the door is closed.
[Kiparsky MS, p. 8]
I don't regret that the door is closed.
[Kiparsky MS, p. 9]
But with a non-factive predicate, the assumption about the factuality of the sentential object is polarized by negation of the predicate :
(3)
(a)
(b)
It is likely that the door is closed.
It isn't likely that the door is closed.
(c)
I believe that the door is closed.
(d)
I don't believe that the door is closed.
To anticipate later details, factive nominalizations have the deep
structure "the fact that S", non-factive nominalizations have the
deep structure "that S". More precisely, the structures of (k):
(h) (a)
(b)
CASE,.
PREP^
NEUT
NP
PREP
NP
NEUT
the
fact
PREP
NP
FACTIVE
NON-FACTIVE
532
NOM - k
Note that (b) is identical with (a) beginning with the lower
right-hand node NEUT. That is, factive nominalizations appear
in a case-frame with the head item fact, non-factive nominalizations appear in a case-frame with any head item except fact.
Qua nominalizations, they are alike, and the differences between
them depend on the head item. The relevant claim made by the
differentiation of these structures is that so-called factive
predicates do not have sentential objects. They have an NP
consisting of the fact as object. The noun fact in turn does
have a sentential object. The sentences (2.c) and (3.c) have
the same surface structure by virtue of a rule which deletes
the fact in (2.c). The deep-structure prepositions are retained
or deleted by entirely general rules that operate also with
non-sentential NP's throughout the grammar.
A second general parameter in the description of nominalizations, also first set forth by the Kiparskys (1968), is the
distinction between EMOTIVE and NON-EMOTIVE predicates. Predicates
which express the subjective value of a proposition rather than
knowledge about it or its truth value are said to be emotive.
This class of predicates takes for in infinitival nominalizations,
as in It is important for us to solve the problem.
Infinitival nominalizations are taken to be a secondary
consequence of several distinct processes which have the effect
of leaving the verb without a subject with which it can undergo
agreement: either marking the subject with an oblique surface
case (as when for is inserted with emotive predicates), or
deleting it (as when it is erased by an identical NP in the
matrix sentence), or raising it out of its own sentence. In
the general lines of this analysis, details aside, we again
follow the Kiparskys (1968).
Not all gerundives are best analyzed as nominalizations.
One class which was historically adverbial remains clearly adverbial
in sense, although the deep structure of the underlying adverbial
is not clear. But the gerundive in He went hunting, earlier
He went a-hunting, still earlier He went on-hunting, and others
of the same type (He kept (on) working, He saw me fishing, He
continued questioning her) cannot naturally be related to
deep-structure nominals unless these nominals are themselves
part of an adverbial.
B.
Previous Scholarship
1.
2.
3.
k.
5.
6.
Chomsky's 1958 Analysis
Lexicalist versus Transformationalist
The Distinction between Nominalization and Complementation
IT + S
Second Passive, IT-Replacement, and Extraposition
The Erasure Principle
533
NOM - 5
1.
Chomsky's 1958 Analysis
In his early writings on transformational grammar Chomsky
mentions various types of nominalizations. The rules he proposed
were offered as illustrations of certain properties of transformational grammars rather than as full-scale accounts of nominalizations in English. Chomsky has since changed his position on
several aspects of nominalization. The following account of his
early sketch of complementation and nominalization is mainly of
historical interest only, though Chomsky's sketch of complementation, at least, was sufficiently satisfactory that Lees (i960)
kept most of the same classes and for several parts of the analysis
made no attempt to go any deeper.
The 1958 paper distinguished ten classes of verbs that
take different types of complements. In the examples below
(Chomsky 1958) the complements have been underlined:
(5)
(a)
consider, believe,... They consider the
assistant qualified.
(b) know, recognize,... We know the assistant
to be qualified.
(c) elect, choose,... We elected him president.
(d) keep, put,... We kept the car in the garage.
(e) find, catch,... We found him playing the flute.
(e1) persuade, force,... We persuaded him to play
the flute.
(f) imagine, prefer,... We imagined him playing
the flute.
(f) want, expect,... We wanted him to play the flute.
(g) avoid, begin,... We avoided meeting him.
(g1) try, refuse,... We tried to meet him.
Some of these verbs can obviously be assigned to more than one of
these classes. Chomsky derived these sentences from separate
underlying sentences, the matrix containing a dummy complement
which was replaced by part of the constituent sentence in a transformational mapping:
(6)
(a)
(b)
(c)
They consider COMP the assistant. MATRIX
The assistant AUX be qualified. CONSTITUENT
They consider the assistant qualified. DERIVED
SENTENCE
The 1958 account contained a separate transformational rule for
each of the above complement types. The rules are all very similar,
and it is obvious that Chomsky was not attempting to achieve much
generalization. His main point was that each of the above complements differed by at least one condition, and that this condition
depended on the classification of the matrix verb.
53U
NOM - 6
Besides these rules for complementation, Chomsky proposed rules for various types of nominalization. The various
types are underlined in the following examples:
(7)
(a)
(b)
(c)
(d)
(e)
(f)
John's proving the theorem was a great
surprise.
To prove the theorem is difficult.
John's refusal to come was a great surprise.
The growling of lions is frightening.
(Cf. Lions growl.)
The proving of theorems is difficult.
(Cf. Theorems are proved.)
The country's safety is in danger.
In his derivation, Chomsky provides a dummy nominal which is
replaced by the appropriate form of the constituent sentence,
with one rule for each type of nominalization. E.g., in Chomsky
(1958) the sentence (7-a) has the analysis
T-it-C + be + a+ great + surprise c
John - C - prove + the + theorem
3
{:
-^
John + S - ing + prove + the + theorem C + be + a + great + surprise
This is equivalent, in the model of Aspects (Chomsky, 1965), to
a tree of the following form:
John prove
the theorem
a great surprise
The 1958 paper nowhere discussed the distinction between
nominalization and complementation, apparently simply assuming
its validity, an assumption subsequently shared by Lees (i960).
2.
Lexicalist versus Transformationalist
The general arguments which led the UCLA research group to
adopt the lexicalist position with respect to such nominals as
proposal, safety, insistence, claim, etc. have been presented in
GEN INTRO under the heading Theoretical Orientation. The lexicalist
535
NOM - 7
position leads in a natural way to the adoption of Fillmore's
Case Grammar. The properties of nouns like proposal, insistence,
killing, ... are, in this frame of reference, in no way specific
to a discussion of nominalization, since their expansion in the
deep structure is quite parallel to that of verbs and adjectives,
and the rules of nominalization which apply to sentences embedded
within verbal case frames apply equally to sentences embedded
within nominal case frames.
Chapin (1967) has presented arguments which suggest that
neither position, lexicalist nor transformationalist, is entirely
correct, but the areas of his research are not developed in this
grammar and did not lead us to modify our position. For example,
he shows that -able in general presupposes a passive underlying it:
"John is pervertable" should be related to "John is able to be
perverted". He claims this must be a transformational relationship
since there is no apparatus in the lexicon as presently conceived
to utilize the passive within a lexical derivation. He goes on
to argue that -ity must also be transformationally derived, since
it is added to adjectives in -able. But nouns with -ity are highly
idiosyncratic in their semantic and syntactic properties, not
predictable in these respects from the underlying verb or adjective.
This kind of evidence suggests that transformational processes
somehow belong within the part of the grammar traditionally known
as "derivational morphology"; and of course Lees (i960) presented
a vast range of similar evidence.
3.
The Distinction between Nominalization and Complementation
Inspection of Chomsky's (1958) examples and rules indicates
that his "complements" appear in object position, and his "nominalizations" in subject position. His complementation rules contain
conditions which mention the verb in the matrix sentence, but his
nominalization rules do not. These observations are purely fortuitous, since nominalizations are not confined to subject position,
and even in that position they obey constraints in respect to the
matrix verb:
(8)
(a) *John's refusal to come is difficult.
(b) *John's refusal to come is in danger.
(c) He tried to anticipate John's refusal to come.
(d) He was annoyed by the fact of John's proving
the theorem.
Lees (i960) takes (9.a) to be a typical complement construction,
and (9«b) to be a typical infinitival nominalization:
(9)
(a)
(b)
I force him to go. [Lees (i960), p. 7*+]
I plead for him to go. [ibid]
536
NOM - 8
He points out that these constructions differ in several ways
(p. Ik): "... (l) for him is deletable in nearly all cases:
"I plead to go", while from the Comp sentence him is omitted only
after a special subset...: "I try to go", but not: *"I force
to go"; (2) there is no passive: "He is forced to go by me",
but not: *"He is pleaded for to go by me"; (3) the sentences in
question seem to be parallel to others with an abstract object,
not an animate object: "I force him to go" parallel to: "I
force him", but "I plead for him to go" parallel to: "I plead
for it"; (U) there is no WH-transform of an internal noun:
"Whom do I force to go?", but not *"Whom do I plead for to
go? ...
Lees' arguments demonstrate that (9.a) and (9.b) must
be distinguished, but of course they do not show that the distinction is one of category (NP vs. COMP). Rosenbaum (1967a),
originally written as his dissertation in 1965, argues that
complements and nominalizations, though they must be distinguished
in respect to the relation they have to other nodes of the
sentence, should not be distinguished in respect to their internal
structure. He argues further that they share a wide range of
common transformations such as complementizer specification,
deletion of subjects, and the like. The sentence underlying
him to go in (9.a) and (9.b) is itself a nominalization in both
examples, but the structure of the predication is different
because of the presence in (9.a) of an additional node (details
omitted).
(V)
Equi-NP-deletion applies to (9.a') to derive I force him to go.
If the constituent subject of (9.b*) were identical with the
matrix subject, the same deletion would apply to derive I plead
to go.
537
NOM - 9
The most important virtue of Rosenbaum's analysis
is that it provides an account of the relation between verb
complements and nominalizations. This it does in two ways:
first, by showing that many structures that had previously been
considered verb complements are in fact nominalizations functioning as objects of verbs or objects of prepositions; second,
by arguing that nominalizations are themselves derived from
noun-complement constructions (the IT + S analysis), and that
the same complementizers that operate in verb complementation
(that, for...to, POSS...ing, etc.) operate in noun complementation.
In collapsing the two putatively distinct structures ,
Rosenbaum takes complementation as primary. By "complement"
he means an S introduced into the structure as right sister of
some head item:
NP
N
N-COMP
VP
S
V
S
V-COMP
ADJ
ADJ-COMP
The analysis developed subsequently in the present paper takes
nominalization to be primary, by which we mean that there is no
S involved in these rules which is not directly dominated by NP.
The difference is by no means purely notational, since a number
of quite distinct substantive claims are involved. For the
differences to be made clear, Rosenbaum's views must be summarized
in some detail. However, Rosenbaum's 1965 dissertation views
are clearly not the same as his current views, and we infer
from the Preface to Rosenbaum (l967a)that at least some of his
current views are quite similar to ours. In the Preface he
writes: "First, the number of clear cases of verb phrase complementation [i.e. V-COMP, above] has diminished to the point where
their general existence becomes questionable" (p.ix). The verb
complementation paper of UESP (1967) was devoted largely to
providing evidence against the existence of verb phrase complementation. In view of Rosenbaum's retraction above quoted, the
present paper merely summarizes some of the problems inherent in
Rosenbaum's earlier view, since we agree that the distinction
between VP and NP complementation is not fully viable.
Two other investigators independently (Wagner (1968) and
Bowers (1968)) take a position like that of UESP (I967), arguing
that many of the passive and pseudo-clefted examples cited by
Rosenbaum are not totally out if the appropriate prepositions
are assumed: e.g. What she condescended to was to talk with us
is better than *What she condescended was to talk with us;
and What Bill tended to was to think big is better than *What
Bill tended was to think big (Wagner, 1968). But we certainly
538
NOM - 10
do not agree with Wagner, as will appear in detail below, that
if these prepositions are correctly inserted in the ordering of
rules, then "Rosenbaum's arguments come to nothing" (Wagner,
1968, p.91), since we still reject such examples as Wagner's
(3*0, To drink beer is condescended to by nine out of ten people,
or even worse, ...is tended to... , which he would, on the
arguments presented, have to accept. The question of where one
draws the line of grammaticalness is touchy, and presumably
subject in these cases not so much to dialect variation as to
genuine uncertainty on the part of native speakers being faced
with examples of a type so rarely met in normal discourse that
they simply have no clear intuition about them. It becomes, we
shall argue, a question of strategy in handling data of a type
where decisions about grammaticalness are so shaky.
Rosenbaum's (1967a)classes of VP-Complementation are
illustrated in (10)-(12) [classes and predicates from Appendix
of Rosenbaum (1967a)]:
(10)
(a)
(b)
Intransitive Verb Phrase Complementation
The doctor
[(101)] [A.5.1]
The doctor
[A.5.2]
(ll)
condescended to examine John,
finished
examining John.
Transitive Verb Phrase Complementation
S
(ADV)
(a)
(b)
(c)
to examine John. [A.6.1]
examining John. [A.6.2]
examining John. [A.6.3]
commanded
found
imagined
539
NOM - 11
(12)
Oblique Verb Phrase Complementation [like transitive
except that the object of the matrix verb is introduced by a preposition]
(a)
(b)
I rely on the doctor to examine John. [A.7]
We prevail upon the doctor to examine John. [A.7]
NP complements are characterized by a configuration in which the
node NP immediately dominates N + S,
(13)
(DET)
N
so that any of the NP's in (lU) may have this internal structure and
be instances of NP complementation:
(1U)
(ADV)
PREP
Rosenbaum's classes of NP-Complementation are illustrated in (15)
(l8) [classes and predicates from Appendix of Rosenbaum (1965a)]:
(15)
Subject NP complementation
(a)
(b)
(c)
(16)
That the doctor examined John does
not matter. [A.2.1.1]
For the doctor to have examined John
seems awful. [A.2.1.2]
The doctor's examining John mortified
the whole family. [A.2.2]
Object NP complementation
(a)
(b)
(c)
(d)
Everybody thinks that the doctor examined
John. [A.1.1]
We prefer for the doctor to examine John.
[A.1.2.1]
They believe the doctor to have examined
John. [A.1.2.2]
They remembered the doctor's examining
John. [A.1.3]
5U0
NOM - 12
(17)
Intransitive oblique NP complementation [The
constituent S is part of a prepositional object
of a verb which has no other object. The
preposition is deleted before that-S and
infinitivals.]
(a)
(b)
(c)
(18)
They hoped (for) that the doctor would
examine John. [A.3.1]
They arranged (for) for the doctor to
examine John. [A.3.2]
They approved of the doctor's examining
John. [A.3.3]
Transitive oblique NP complementation [The
constituent S is part of a prepositional phrase
which complements a verb that has another object.
The preposition is deleted before that-S and
infinitivals.]
(a)
(b)
(c)
Mary convinced Jean (of) that the doctor
had examined John. [A.U.l]
They forced the doctor to examine John.
[A.U.2]
They suspected the doctor of examining
John. [A.k.3]
To argue against the distinction between VP-COMP and
NP-COMP one must have in mind some alternative. An alternative
for which one might argue is that (10), (ll), and (12) are
analyzable as instances of NP-COMP, thereby eliminating the distinction between NP-COMP and VP-COMP. This is our understanding
of what Rosenbaum means by the sentence in his Preface (1967 a)
asserting that there appear to be few cases of VP-COMP. Such an
argument depends on showing that the criteria by means of which
Rosembaum distinguished the two are in some way faulty criteria.
His criteria were these:
(a)
(b)
(c)
Behavior of the COMP under the passive rule;
Behavior under the pseudo-clefting rule;
Behavior under the extraposition rule;
and we add
(d)
Behavior under pronominalization.
5U
NOM - 13
(a)
COMP and Passive
Consider first these examples from Rosenbaum (1967a):
(19)
(a)
(b)
Everyone preferred to remain silent.[15.a.l)]
To remain silent was preferred by everyone.
[(15.a.2)]
(c) John tended to play with his little brother
of ten. [(I5.b.l)]
(d) "To play with his little brother often was
tended by John. [(l5.b.2)]
(19.d) is unquestionably bad; but (19.b) is not impeccable, either.
By an oversight, though tend is a paradigm example of VP-COMP
in the text (p.lU), it does not show up at all in the lists of
Rosenbaum's appendix: presumably it belongs with A.5.1,
Intransitive Verb Phrase Complementation with for-to Complementizer.
With these examples, passivization is ungrammatical:
(20)
To examine John was
(a
(b
(c
(d
(e
(f
(g
(h
(i
(J
(k
(1
(m
(n
(o
(P
"begun
"ceased
"commenced
"condescended (to)
"continued
"dared
"declined
"endeavored
"failed
"gotten
"grown
"hastened
"managed
"proceeded
"refused
"started
by the doctor
This observation is significant as a test for a distinction between
VP-COMP and NP-COMP, however, only if there is a class of sentences
comparable to (20) in which passivization is grammatical. The
relevant class is presumably A.1.2.1 (Object NP Complementation
with for-to complementizer), since that class includes prefer,
which is cited in (19) as a viable example of passivization:
(21)
To examine John was
(a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
(i)
(J)
?preferred
"borne
"demanded
?desired
?disliked
?expected
?feared
?hated
"intended
?liked
5U2
by the doctor
NOM - lU
(k)
(1)
On)
(n)
(o)
(p)
(q)
?loathed
?loved
?promised
^prescribed
?requested
?required
*wanted
One cannot easily convince himself that these are fully grammatical.
One can much more readily convince himself that if the verbs of
(20) and (21) are different in respect to the structure of their
complements, the test of passivization certainly does not provide
satisfactory motivation for the distinction.
It appears in general to be true that an infinitival,
in particular a subjectless one, cannot become subject under the
passive rule. If true, this is an interesting fact, and one which
requires explanation: e.g. it suggests that if, in the deep
structure of The doctor prefers/demands/desires...to examine John,
there is motivation to assume a deep structure dominance of
to examine John by a node NP, then somehow in the reduction of
that deep structure to the surface infinitival either the NP
node must be removed, or some other device must prevent passivization. We provide an account below of what such a device might
be. But first consider these examples further: some sentences
in (21) can be improved by retaining a subject and seeking a
semantic content that is somehow - though it is not clear how more appropriate to the structure: e.g.,
(21')
(a)
(b)
For the comprehensives to be given after
the end of the term is generally preferred
by the slower students.
[with extraposition] It is intended for
the better students to finish their
degrees in three years.
The number of instances where passivization of for-to constructions
with subjects results in a fairly high-grade output is substantial;
if one finds the higher-grade examples persuasive, the conclusion
must be either that complementation and nominalization are distinct
structures, since no amount of tinkering with the sentences of
(20) will produce examples of the quality of (21'),or that there
is some other factor which permits passivization in just these
instances but in no instance where the subject of the infinitival
is deleted. Tinkering with sentences like those of (20) has been
claimed (by UESP 1967, Bowers 1968, and Wagner 1968) to produce
examples that are significantly better than some rejected by
Rosenbaum, and this claim is certainly correct. Rosenbaum, in
citing examples like *To think slowly was tended by me, neglected
the preposition that shows up in the slightly better pseudo-cleft
form (Bowers' (1968) example 33) What Bill tended to was to think
big; i.e., the passive, if it exists, is (?) To think slowly
5U3
NOM - 15
was tended to by me. But in fairness to Rosenbaum, it must be
acknowledged that the improvement, in this example and in the
others that can be modified in the same way, is not a startling
black-and-white up-grading to obvious grammaticality.
If one feels, as we do, that some of the extraposed
passives like (21'.b) are close to fully grammatical; that the
examples (21) are better with subjects supplied for the infinitivals,
but that they are about as bad as (20), taken as they stand; and
that the examples (20) are irreparably bad, - then one has a
problem in strategy (since the grammar one writes depends, in
this instance crucially, on one's conclusion about these examples).
One strategy would be to take a hard line on the question of what
is grammatical in these instances where the data is so fuzzy.
This would force the grammar to assert that It is intended for
students to finish in three years is as bad as For students to
finish in three years is intended, which is not true, or that
To finish in three years is intended is as bad as To finish in
three years is managed, which also is not true.
There is a gradation among these examples, however: one
might explain the relative persuasiveness of It is intended for
the students to finish in three years on the assumption that it is
derivatively generated (in the sense of Chomsky, "Some Methodological
Remarks on Generative Grammar", Word 17, 19^1) from It is intended
that the students finish in three years, i.e. an analogy which
associates for-to with subjunctive, since for-to corresponds
with subjunctive in a wide range of examples: It is important
for him to finish in three years/It is important that he finish
in three years; I prefer for him to finish in three years/ I prefer
that he finish in three years. But verbs like begin, manage,
continue, decline, fail,... , not having a corresponding that-S
subjunctive, should not, and do not, lend themselves to this
analogical extension at all.
This hard line strategy would require in the present
grammar that we allow passivization just in case there has been
no reduction to infinitival form. Without now anticipating our
subsequent detailed analysis of infinitivals, a device which would
block all moving of infinitivals into passive subject would be to
place the rules of infinitival reduction after the rule of passive
subject placement, formulating them in such a way as to exclude
reduction if the embedded sentence had been made subject of a
passive verb. This device would be unnatural, however, since
with some predicates such as tragedy, important, an infinitival
as subject is unobjectionable: For her to have married so young
was a tragedy that we all deplored; For them to wear a lifejacket
will be important to their survival if they get shot down. It
would also be ad hoc, since it would require repetition of the same
constraint in a number of rules determining infinitival reduction.
5UU
NOM - 16
Alternatively, a device which is also ad hoc but much less
unnatural, since passivization requires a number of special constraints not required by active subject placement anyway, would
be to constrain passivization so as not to move any sentential
NP into passive subject unless that sentence contained an AUX:
i.e. unless it were still a "real" sentence, not an infinitival
reflex of one. But there is independent motivation to place
the rule TO-REPLACE-AUX, which establishes infinitival form,
after the case placement rules, whereas the constraint just
suggested will filter out Just the right examples only if the
passive rule follows TO-REPLACE-AUX. Since we believe we have
fairly strong reasons to treat passivization along with case
placement in general, and since the case placement rules must
precede TO-REPLACE-AUX, the suggested constraint to "real
sentences" cannot serve to block passivization in these instances.
A third alternative is to block only subjectless
infinitivals from passivizing. As noted above, it is the
subjectless infinitivals which are consistently bad when passivization of the matrix verb puts them into subject position i.e. the examples (21), as distinct from (21*) where the
infinitivals have subjects. A compromise between a totally
"hard line" position, then, and the Bowers/Wagner/UESP (1967)
position, is to block passivization under the condition that the
would-be sentential passive subject is lacking its own subject,
thereby admitting (21'), but excluding (20) and (21). That,
after much discussion, is the consensus solution of the present
grammar. It is ad hoc in that the passive rule must have a condition that blocks passivization of subjectless infinitivals. It
is also unnatural in view of the fact that the rule does not
otherwise have to look at the internal structure of the NP that is
to be moved to passive subject. But it correctly reflects our
intuitions about the set of grammatical sentences.
(b)
COMP and PSEUDO-CLEFT, EXTRAPOSITION
Behavior of the complement under passivization, then,
turns out to be no satisfactory justification for the putative
distinction between VP-COMP and NP-COMP. Consider, now, the second
basis, pseudo-clefting:
(22)
(a)
1.
2.
(b)
1.
2.
I hate you to do things like that.
[Rosenbaum (1967a)(10.a.l)]
What I hate is for you to do things
like that. [lO.a.2]
We prefer you to stay right here.
[lO.b.l]
What we prefer is for you to stay
right here. [10.b.2]
5»»5
NOM - 17
(c)
1. I defy you to do things like that. [lO.c.l]
2. *What I defy is for you to do things like that.
[10.c.2]
(d)
1. We tempted you to stay right here.
2. *What we tempted was for you to stay right here.
The pseudo-clefting test depends on the assumption that what is
clefted is an NP, a claim which is supported by the third test,
extraposition, which indicates that (22.a) and (22.b) contain
NP's that can be extraposed, whereas (22.c) and (22.d) do not:
(221) (a)
I hate (it) very much for you to do things like
that. [(11.a)]
(b) I prefer (it) very much for you to stay right here,
[(ll.b)]
(c) *I defy (it) very much for you to do things like
that. [(12.a)]
(d) *We tempted (it) very much for you to stay right here.
[(I2.b)]
But of course pseudo-clefting also depends on the assumption that
what is clefted is a constituent; one of the surprising aspects
of Rosenbaum's book is that while he is the scholar who first
clarified the distinction between They expected the doctor to
examine John and They persuaded the doctor to examine John
(discussed by Chomsky (1965), pp. 22-23), he nonetheless fails
to note here that the fact about (22.c) and (22.d) which blocks
pseudo-clefting, and extraposition, is that neither for you to
do things like that nor for you to stay right here is a constituent.
The difference between (22.a-b) and (22.c-d), already noted as
the distinction between (9.b*) and (9.a*), is precisely that
between expect and persuade discussed by Chomsky. That is, for
these examples the question of VP-COMP vs. NP-COMP is simply
irrelevant. The distinction between expect and require, which is
even clearer than, and exactly like, the distinction between
expect and persuade, is the following:
The sentence (23.a) is cognitively synonymous with the
passive (23.b):
(23)
(a)
(b)
They expected the doctor to examine John,
They expected John to be examined by the
doctor.
But the sentence (23.c), identical with (23.a) in surface structure,
is not synonymous with (23.d):
(23)
(c)
(d)
They required the doctor to examine John.
They required John to be examined by the doctor.
5U6
NOM - 18
(23.c,d) are paraphrased by an explicit Dative in (23.d,f):
(23)
(e)
(f)
They required of the doctor that he examine
John,
They required of John that he be examined
by the doctor.
The examples with require (or persuade) have, minimally, a deep
structure that includes an animate object in addition to a
sentential object:
(23)
(C)
they
require the doctor
S
the doctor examine John
The examples with expect (or hate or prefer) have no such animate
NP object in addition to their sentential object:
(23)
(a')
the doctor examine John
In short, the pseudo-clefting argument that supports the circled
NP of (23.a') is irrelevant to the question of whether (23.c')
should have the circled NP or not.
5U7
NOM - 19
Although pseudo-clefting is not an argument appropriate
to the distinction between the examples (22), it is relevant to
the discussion of other examples of the putative contrast between
NP-COMP and VP-COMP, in fact to the same examples as those to
which the passive test was adduced. Bowers (1968) claims that
although *To see his friend was rejoiced at by him is not
grammatical, What he rejoiced at was to see his friend [(13) and
(lU)] is. Bowers is not quite so happy with What he tempted Bill
to was to be interviewed by the company [(17)], but he is not
willing to state categorically that it is ungrammatical; similarly
What they condemned him to was to die [(23)]. If grammatical,
such examples dispute the NP-COMP/VP-COMP distinction proposed
by Rosenbaum.
The problem with pseudo-clefting as a test is that there
are numerous examples which have no corresponding grammatical
non-clefted infinitival cognates: e.g.
(2k)
(a
(b
(c
(d
(e
(f
(g
(h
What I look forward to is for him to break
his neck.
*I look forward (to) (for) him to break
his neck.
I look forward to his breaking his neck.
What I would really enjoy is for people
to leave me alone.
*I would really enjoy (for) people to leave
me alone.
What I deplore is for idiots to be running
the country.
*I deplore for idiots to be running the country.
It is deplorable for idiots to be running the
country.
What I propose is that they quit sticking
their noses in
the department's affairs.
What
I
propose
is
for them to quit sticking
(J
their noses in the department's affairs.
[Perhaps not fully well-formed, but derivatively
related to (i).]
(k) *I propose for them to quit sticking their noses
in the department's affairs.
(i
5U8
NOM - 20
(l)
(m)
What I require is that he do better.
What I require is for him to do better.
[Perhaps not fully well-formed, but derivatively related to (l).]
(n) *I require for him to do better.
(o) I require him to do better.
(2k.&) seems impeccable, but (2l+.b) is totally out. (2k.d) is
good, but (2U.e) quite dubious. (2l*.f) is impeccable, but only
rarely is (2U.g) claimed to be grammatical (e.g. by the Kiparskys
(1968)). The remaining sets involve the possibility of a derivative relation to a subjunctive. It is hard to see how data like
these can be used to support or deny the NP-C0MP/VP-C0MP distinction. It is certainly legitimate to use evidence from pseudoclefting to argue for one or another element of content in the
deep structure of an infinitival: e.g., we claim that the
existence of (25.a) argues for a subjunctive in the underlying
form of (25.b), even though there is no corresponding form (25.c):
(25)
(a)
What I especially want is that my daughter
grow up to be a gracious lady.
(b) I especially want my daughter to grow up
to be a gracious lady.
(c) *I especially want that my daughter grow
up to be a gracious lady.
But to argue from the pseudo-cleft that there must be a certain
structural distinction in the available non-clefted cognates claims
that we understand the conditions under which pseudo-clefting is
permitted; the data of (2k) testify that we, at least, do not
understand these conditions.
(c)
COMP and PR0N0MINALIZATI0N
The fourth criterion, pronominalization, not proposed by
Rosenbaum, tends to support the circled NP of both (23.a') and
(23.C):
(26)
(a)
(b)
Mary expected the doctor to examine John,
and I expected it, too.
Mary required the doctor to examine John,
and I required it of him, too.
But pronominalization provides contrary evidence in other examples:
5U9
NOM - 21
( forced
(27)
(a)
)
*Mary J commandedi the doctor to
I ordered C
(told
)
K
f
(forced
)
examine John, and I ^commandedSordered \
told
)
him (into) it, too.
(b)
?The doctor condescended to examine John,
and the other specialist condescended to
it, too.
(c)
?I prefer to be examined by osteopaths,
and Mary prefers it, too.
(d)
?John tends to like blondes, and I tend
toward it, too.
The examples (27.a) are all bad, except perhaps force with into;
(27.b,c,d) are extremely questionable, only really acceptable in
the form A condescends/prefers/tends to do X, and B tends to do
it, too. It appears, in fact, that there are no very satisfactory
examples of it-anaphora where the item replaced is an infinitival
complement: this fact strongly suggests that the derivation of
infinitival complements is not a matter of simply replacing a
sentence by a cognate infinitival form - that several steps are
involved in the derivation, and that in the course of this derivation the underlying sentence is mutilated in such a way as no
longer to be recognizable as an NP, for pronominalization, or
else somehow the necessary conditions for pronominalization were
not present in the first place. Since the present grammar does
not attempt to deal with the PRO-ing of sentences , a solution to
this problem continues to be outstanding, nor do we have any very
clear notion of what solution might successfully be proposed.
Returning, now, to the main line of argument: Are there
solid syntactic grounds for the distinction between VP-COMP and
NP-COMP? The criteria which have been proposed fail to make the
distinction consistently. The claim that there are at least two
distinct structures, namely those with a dative (23.C*) and those
with only a sentential object (23.a*), is persuasively motivated
by both passivization and pseudo-clefting, but that distinction is
independent of the distinction in question. The fact that passivization is ungrammatical with subjectless infinitival complements
(20) and (21) may or may not be correctly analyzed as a function
of a condition on the passive rule, but if the facts are as we
have outlined, they do not support the distinction in question.
What, then, remains as a basis for the distinction between VP-COMP
and NP-COMP?
550
NOM - 22
It seems to us that there is one kind of argument for
VP-COMP, not raised by Rosenbaum, which is difficult to eliminate.
Consider the semantic interpretation of the following sets:
(28)
(a)
(b)
(c)
He forgot to study the lesson.
He forgot that he was to study the lesson.
He forgot that he (had) studied the lesson.
(29)
(a)
(b)
He avoided studying the lesson,
He neglected to study the lesson.
In (28), it seems clear that neither (b) nor (c) is entailed by (a),
but any derivation which assumes a deep structure NP-sentential
object of forget will encounter grave difficulty avoiding the claim
that something like (b) or (c) is indeed entailed by (a). In such
sentences as (28.a), involving a contrary-to-fact embedded sentence,
a way out, though not otherwise motivated, is to assign a subjunctive aspect to the verb of the embedded sentence, thus distinguishing between the deep structure of (28.a) and that of (28.b,c).
In some closely similar sentences, there is independent Justification for subjunctive: in particular, example (25) above. Although
(25.c) does not exist, (25.a) strongly suggests that (25.c) is
indeed the deep structure obligatorily reduced to (25.b): it would
otherwise be quite impossible to explain the subjunctive form of
the pseudo-cleft (25.a). Since there is not comparable pseudocleft form for (28.a) , the assumption of subjunctive to account for
the contrast within (28) can be argued only by analogy with (25).
The examples (29) contain the same problem of interpretation, but
they permit neither the non-subjunctive contrasts analogous to
(28.b,c) nor pseudo-cleft forms analogous to (25.a), although the
sentences (30) are at least readily interpretable:
(30)
(a)
(b)
*?What he avoided was that he study the lesson,
*?What he neglected was that he study the lesson.
Since there is at least a not-totally-unreasonable solution to the
problem posed by (28), and since there appear to be no other
persuasive arguments in favor of VP-COMP, we set this argument aside
also as insufficient to Justify the distinction between VP-COMP
and NP-COMP.
(d)
Nominalization versus Complementation:
Conclusion
We conclude that the distinction between NP-COMP and VP-COMP
is not a necessary or revealing one. The only alternative is not,
however, that all "complement" structures are what Rosenbaum (1967)
calls Noun Phrase Complementation. Our claim is that they are not
complements at all, but nominalizations: i.e., they have the deep
structure (31):
551
NOM - 23
(31)
NP
To argue that they are not complements, we must now consider
Rosenbaum's arguments that the structure of NP-Complementation is
(32):
(32)
NP
/f
(D)
N
it
IT + S
Rosenbaum's (1967a) arguments for assuming it in the deep
structure are these:
(a)
The rule of Extraposition moves sentential subjects
and objects out of their deep-structure position
and adjoins them at the end of the matrix sentence.
When moved out in this way, there is evidence that
such sentences are no longer dominated by NP but
rather are adjoined directly under the matrix S.
In the original position of the extraposed sentence,
the expletive it_ appears in the surface structure.
(b)
The it_ which appears in the surface structure is
not the same as the it_ of pronominalization, since
it can't be questioned or relativized; i.e., this
it is a dummy like the it_ of It's raining.
(c)
NP-Complementation and VP-Complementation share
most rules, in Rosenbaum's analysis, but not the
rule of extraposition. E.g., I hate (it) very much
for you to do things like that is NP-Complementation,
and grammatical under extraposition (from object);
but »I defy (it) very much for you to do things like
that is ungrammatical, a fact which Rosenbaum explains
by claiming that it is VP-Complementation, which is
not subject to extraposition.
(d)
Finally, the statement of complementizer transformations is simplified by making the complementizer a
feature on it^ and spreading it into the sentential
complement.
552
NOM - 2k
The four arguments above are reconstructed from Rosenbaum's
"Defense of the Phrase Structure Rules" (pp. 9-23). A fifth
argument, stated by Lakoff (1966c) is
(e)
If one argues that the it^ is introduced transformationally in the proper environments, it
is virtually impossible to define what is meant
by "the proper environments."
(a) is clearly a fact, but not an argument unless it is indeed
"virtually impossible" to state the proper environments for
transformational insertion of It. (b) is also a fact, but
equally statable of an it_ inserted by a non-anaphoric transformational rule, (c) is a valid argument, but it depends on
the validity of the NP-COMP/VP-COMP distinction, as stated by
Rosenbaum; it is not specific to IT + S, since the distinction
between NP-COMP and VP-COMP can equally well be made as between
S dominated by VP, and S dominated by NP. From (c) all that
is clear is that some basis must be provided to permit extraposition in the right instances, which is true of (e) also,
(d) is a weak argument because it depends on Rosenbaum's
preference for a particular formalism; if it turns out that the
Kiparskys (1968) are right, and that the complementizers come
from a variety of deep sources, the formalism (even if it were
the best possible) could not be employed anyway. So only (e)
is a real argument. Lakoff acknowledges that the environment
in which extraposition from subject occurs is readily statable;
the one that he finds "virtually impossible" to state is the
environment of "vacuous extraposition from object." But at the
time of presenting his arguments he was unaware that the only
instances of extraposition from object are factives. The notion
"factive" is independently motivated, and it provides precisely
the environment, fairly easily stated (although a few items must
be marked with exception features), that Lakoff found difficult
to state.
There appears, then, to be little solid Justification for
the IT + S analysis, and we have accordingly rejected it.
5.
Extraposition, IT-Replacement, and Second Passive
To account for the relationships between sentences like
(33),
(33)
(a)
(b)
(c)
That John will find gold is certain.
It is certain that John will find gold.
John is certain to find gold.
553
NOM - 25
(d) *That John found gold happened.
(e) It happened that John found gold.
(f) John happened to find gold.
a rule of Extraposition (deriving (33.b) from (33.a), and (33.e)
from (33.d)) has been widely assumed (e.g. Ross (1967c). Rosenbaum
(I9b7a)t and Lakoff (1965)); and a rule of IT-replacement (deriving
(33.c) from (33.b), and (33.f) from (33.e)) was proposed by
Rosenbaum (1967a)and appears to be generally assumed, though the
form of it varies (see, for example, discussion of the problem
in Kiparsky (1968), in particular footnote 6).
A class of sentences that require a similar derivation
(and incidentally thereby reduce the candidates in (20) for analysis
as VP-Complementation) is the class of so-called "transparent"
predicates (i.e. selectional restrictions determined by the verb
of the complement):
(3k)
(a
(b
(c
•That John got tired began.
*It began that John got tired.
John began to get tired.
(35)
(a
(b
(c
*That John was a tyrant continued.
*It continued that John was a tyrant,
John continued to be a tyrant.
(36)
(a
(b
(c
*That John worked hard ceased.
*It ceased that John worked hard.
John ceased to work hard.
Our derivation of (33) - (36) by a process of "raising to subject"
is discussed below.
Another class of sentences that seem to require a similar
derivation is that of (37):
(37)
(a)
(b)
(c)
They believe that Bill is intelligent.
They believe Bill to be intelligent.
Bill is believed to be intelligent.
Lees (i960) labeled (37.c) as the "Second Passive". He correctly
observed (p. 63) that "It is as though the passive transformation
could apply either to the whole That-Clause nominal as subject
[generating That Bill works hard is said (by someone)] or only to
the internal nominal subject of the That-Clause [generating (37.c)]".
Our analysis of such sentences in Section III is essentially the
same as Lees', with the additional observation about to-insertion
of the Kiparskys which provides a general account of why the form
of the that-clause is infinitival after the subject has been lifted
up into the matrix sentence by a process of "raising to object", and
then taken as the passive subject by the regular subject placement rule.
55U
NOM - 26
Rosenbaum (1967a) has claimed that there is no need for
a second passive rule, if the grammar contains rules for extraposition and it-replacement. His (excessively ingenious) derivation of sentences like (37), contrary to Lees' clearly correct
intuition, is the following:
(38)
(a)
(b)
*One says it-for Bill to work hard.
*It-for Bill to work hard is said.
[Passive of (a)]
*It is said for Bill to work hard.
[Extraposition on (b)]
*Bill is said for to work hard.
[It-replacement on (c)]
Bill is said to work hard. [For-deletion]
(c)
(d)
(e)
[Perhaps it should be noted, though irrelevant to these arguments,
that the subject of the matrix sentence cited as "one" above is
not used by either Lees or Rosenbaum; Lees uses "people" as the
deletable subject, Rosenbaum uses "they". Our arguments that
"one" is the deletable indefinite subject appear in Section III.]
If the other rules indeed worked as claimed by Rosenbaum—
e.g. if IT+S were well-motivated, if for-to infinitivalization were
well-motivated as the deeper structure of all to- infi-nitivals,
and if the distinction between VP-complementation and NP-complementation
were sound — then a counter-intuitive derivation like (38) might
still be justified, as Rosenbaum tried to Justify it, by the fact
that such rules are independently needed and might therefore just
as well be used to account for this apparently irregular construction.
Since none of these conditions appear to hold firmly, we have sought
a different analysis. Since we have a rule of subject placement,
both passive and active, the most natural solution is an optional
rule preceding subject placement which raises the subject of an
embedded sentence into the subject position of the matrix sentence,
in instances like (33.c), (33.f), (3^4.c), (35.c), and (36.c):
taking (33.c) as typical, these have the (simplified deep structure
(39):
(39)
NP-
(fohn)will find gold
"John is certain to find gold."
555
NOM - 27
Similarly, an optional rule can raise the subject of an embedded
sentence into the object position of the matrix sentence, in
instances like (3T.b), and then Subject Placement will move this
object into subject of the matrix:
(uo)
MOD
one
(gill) is intelligent
"One believes Bill to be intelligent."
(1+1)
one)
to be intelligent
"Bill is believed to be intelligent."
With all but one small set of verbs of this class, all steps in the
derivation are grammatical. The exceptions - say, rumor, repute have one ungrammatical step for which we have no account:
556
NOM - 28 a
(U2)
(a) They say - Bill is intelligent.
(b) *They say - Bill to be intelligent.
(c) Bill is said to be intelligent.
The details of this derivation are presented in Section III.D.6,7.
We anticipate them here in general outline to show how our treatment of this class of examples is related to other studies. In
particular, our analysis obviates both a second passive rule,
while formalizing precisely the intuition of Lees (i960) quoted
above, and relates the phenomenon of It-replacement to a general
set of conditions for subject placement.
6.
The Erasure Principle
It is a general principle of transformational theory that
deletions in the course of a derivation must be recoverable.
Otherwise any derivation with a deletion would be infinitely
ambiguous. The kind of deletion that commonly occurs in complement structures is erasure under an identity condition: e.g.
for a whole host of reasons the deep structure of a sentence
like He tried to leave is assumed to contain two occurrences of
the subject he_: He tried + He AUX leave. The subject of the
embedded sentence is erased by the higher identical subject,
in this instance. Rosenbaum (1967*0 found it necessary to develop
an erasure principle which would guarantee for his derivations
that there could be no ambiguity as to which was the erasing NP.
The principle cannot be simply that the first NP to the left is
responsible for the erasure, even though such a principle would
be a first approximation which would work well for such sentences
as (U3):
(1+3)
(a)
(b)
They tempted John
[Rosenbaum (l967si)
We forced John to
[Rosenbaum (19679)
to leave early.
ex. l8.a]
ignore his work.
ex. l8.b]
The consideration of purpose clauses eliminates this principle,
since it would require that "boat" and "car" be the erased
subjects in (kk):
(kk)
(a)
(b)
I sold the boat to save money.
[Rosenbaum (1967*} ex. 19.a]
She took the car to buy bread.
[Rosenbaum (1967a)ex. 19.b]
Rosenbaum sets forth a principle of minimum distance (measured by
counting the number of branches in the path connecting two nodes)
which eliminates the problem of (kk), since the subject of the
purpose clause is more distant from the matrix object than from
557
NOM - 28 b
the matrix subject (because in Rosenbaum's tree there is an additional Pred-Phrase and VP node dominating the object).
Consider, however, the status of the principle of minimum
distance as applied to Fillmorean trees:
(k5)
(a)
MOD
tempt
PREP
NP
PREP
NP
PREP
NP
John
John leave early
(b)
MOD
PROP
I leave early
For several reasons, the rule of EQUI-NP-DEL, which erases the
embedded S's in (^5), must apply fairly early - before the Case
Placement rules that move the appropriate NP into surface subject
558
NOM - 29
position: in particular, because it must precede raising of
the subject of embedded S*s to object of matrix as in (UO),
in order to allow normal reflexivization in (k6) but block it
in (U7):
(U6)
John believes himself to be intelligent.
(J+7) *John wanted himself to work. hard.
[in the sense of "John wanted to work hard."]
If this rule is prior to the Case Placement rules, then the
distance of the erasing NP is identical in (1+5.a), where the
Dative NP is responsible, and in (U5.b), where the Agent NP is
responsible. We have, therefore, stated the rule in such a way
that the erasing NP is identified by the case node dominating it,
and we have replaced the principle of minimum distance by the
principle that an identical dative has erasure priority over
an identical agent.
If it were not necessary for EQUI-NP-DEL to precede the
Case Placement rules, as we believe it is, there would be a very
natural way to capture Rosenbaum's principle within this Frame
of reference. The distances would come out right because of the
elimination of certain nodes in the objectivalization rule,
nodes which must be eliminated for totally independent reasons
(see discussion in BASE RULES). Consider the structures (U5):
these are the structures as they exist prior to the application
of the rules of subjectivalization and objectivalization early
in the cycle: after the application of those rules, the structures are as in (U51):
U5')
(a)
tempt
John
[+DAT]
PREP
NP
John leave early
559
NOM - 30
(b)
I
AUX
[+AGT]
I leave early
In these trees, by Rosenbaum's principle of branch-counting to
determine minimal distance, the subject of the embedded sentence
is one branch closer to the Dative than to the subject of the
matrix sentence. The principle therefore would make the right
decision in these cases.
A sentence that Rosenbaum's principle and our own
Dative/Agent principle both fail to explain is (U8):
(U8)
He promised us to leave at once.
The sentence is perhaps only marginally grammatical
anyway; if it, and others like it, are fully grammatical, then
the verb itself must be marked for the erasing condition which
it requires. Or some other general condition, different from
either Rosenbaum's or ours, must be found. But the example is
suspect on another score: if our formulation of the structures
(1+5') is indeed correct, where the principle of minimum distance
works really because the Dative has been objectivalized—which
in turn was motivated by the requirement of the passive form of
(l+5'.a) John was tempted to leave early, then it should be the
case that the passive of (U8) is We were promised to leave at
once, which is clearly ungrammatical. From this evidence, one
must conclude that the structure of {k8) is somehow radically
different from that of the examples that are relevant to the
principle of minimum distance. A possible conclusion is that
(U8) is a simple blend of the two constructions He promised us
that he would leave at once and He promised to leave at once
both of which are fully grammatical and are generated with no
special problem by the present grammar, in ways discussed subsequently under Section III.D.5.
560
NOM - 31
III.
THE PARAMETERS OF NOMINALIZATION
A.
B.
C.
D.
E.
F.
G.
H.
I.
A.
Factive/NonFactive
1.
2.
3.
1.
Factive/NonFactive
Sentential/NonSentential
Emotive/NonEmotive
Infinitivalization
Gerundive/NonGerundive
Stative Infinitival
Deep Structure Constraints
Surface Structure Constraints
Miscellaneous Exception Features
Syntactic Justification of the Distinction
Criteria for Factivity
The Abstract Instrumental
Syntactic Justification of the Distinction
The Kiparskys (1968) provide the following lists of factive
and non-factive predicates (MS pp. 1 and U):
(U9)
With factive subjects
With non-factive subjects
significant
odd
tragic
exciting
relevant
likely
sure
possible
true
false
matters
counts
makes sense
suffices
amuses
bothers
seems
appears
happens
chances
turns out
With factive objects
With non-factive objects
regret
be aware (of)
grasp
comprehend
take into consideration
take into account
bear in mind
ignore
make clear
mind
forget (about)
deplore
resent
care (about)
561
suppose
assert
allege
assume
claim
charge
maintain
believe
conclude
conjecture
intimate
deem
fancy
figure
know
realize
NOM - 32
[Know and realize are asserted to be semantically factive,
syntactically non-factive.]
The distinction is supported by the following kinds of syntactic
evidence:
a.
Only factives allow either that-S or Fact that S:
(50)
(a)
The fact that she solved the problem
is significant.
(b) #The fact that she solved the problem
is likely.
(c)
I regret the fact that she solved the
problem.
(d) *I believe the fact that she solved the
problem.
b.
Only factives allow the full range of gerundive
constructions:
(51)
(a)
Her having solved the problem is
significant.
(b) *Her having solved the problem is likely.
(c)
The professor's not knowing the answer to
that question was surprising.
(d) *The professor's not knowing the answer
to that question was true.
(e)
I regretted her having contemplated her
navel for so long.
(f) *I asserted her having contemplated her
navel for so long.
c.
Most non-factives allow raising the subject of the
constituent S to subject of the matrix S [Rosenbaum's
IT-Replacement; in the present grammar simply one of
the options permitted in the early subjectivalization
rule, governed by the rule feature (. RAIS-SUBJ ]
discussed under Section D below], but none of the
factives do: [Examples (52) from Kiparsky (1968)
MS p. 3]
(52)
(a)
(b)
It is likely that he will accomplish even
more,
He is likely to accomplish even more.
562
NOM - 33
(c)
(d)
It seems that there has been a snowstorm.
There seems to have been a snowstorm.
(e)
It is significant that he will accomplish
even more.
(f) #He is significant to accomplish even more.
(g) It is tragic that there has been a snowstorm,
(h) *There is tragic to have been a snowstorm.
d. Extraposition is optional with sentential subjects of
factives, but obligatory with sentential subjects of
non-factives: [Examples from Kiparsky (1968) MS p. k]
(53) (a)
(b)
That there are porcupines in our basement
makes sense to me.
It makes sense to me that there are
porcupines in our basement.
(c) *That there are porcupines in our basement seems
to me.
(d) It seems to me that there are porcupines in our
basement.
e.
"Vacuous extraposition from object" is optional with
factives, but disallowed with non-factives; it is obligatory
with a small sub-set of factives:
(5M (a)
(b)
I regret that she lives far away.
[Factive]
I regret it that she lives far away.
[Optional]
(c) *I hate that she lives far away. [Factive]
(d) I hate it that she lives far away.
[Obligatory]
(e)
I suppose that she lives far away.
[NonFactive]
(f) *I suppose it that she lives far away.
[Disallowed]
f. Only non-factive predicates allow what the Kiparskys
non-committally call the "accusative and infinitive
construction", which turn out to be infinitival reductions
like any others except that they must be stative:
563
NOM - 3U
(55)
(a)
We assumed the quarterback to be
responsible.
(b) *We ignored the quarterback to be
responsible.
(c) He supposes himself to be competent.
(d) *He grasps himself to be competent.
A number of the non-factives disallow this construction also — the Kiparskys note that charge is one
such: in our dialects intimate is another; and for
many speakers also anticipate, emphasize, and
announce, which are both factive and non-factive.
But in any case, none of the factives allow this
construction.
The deep structure proposed by the Kiparskys for factive
and non-factive nominalizations is (56):
(56)
(a)
NP
fact
(b)
S
NP
S
factive
non-factive
From the point of view of our "Fillmore-cum-Lexicalist" base, the
S in (56.a) is an NP-object of fact, as in (U.a).
2.
Criteria for Factivity
It appears that the full range of the Kiparskys' observations can be captured by a feature [+/-FACT], a strict-subcategorial
feature specifying that the predicate is compatible with the noun
fact as a realization of the case NEUT in its case frame. All
items which disallow factive objects but accept sentential objects
are marked [-FACT], [+/-S]. This is the class of non-factive
predicates. All items which allow factive objects are marked
[+/-FACT], [-S]. This is the class of factive predicates. They
do not accept sentential subjects or objects at all: those surface
structures in which embedded sentences appear to occur really
occur as objects of the noun fact, which is deletable (as proposed
by the Kiparskys) by the rule of FACT-DEL. Finally, those items
which allow both factive and non-factive objects are marked
[+/-FACT], [+/-S] — e.g., listed by the Kiparskys, anticipate,
acknowledge , suspect, report, remember, emphasize , announce,
admit, deduce. But there is no need, as they propose, to list these
each as two different verbs (though not, they agree, unrelated),
56U
NOM - 35
since we can redundantly specify that [+FACT] -> [-S], and
[+S] -» [-FACT]. Under the convention of obligatory specification in our lexicon, and these redundancy rules, only the
permitted clusters of features will emerge.
The remaining problem is to find a diagnostic for nonfactivity. Those predicates which should be marked [+/-FACT]
are easily diagnosed simply by testing whether or not they
allow "the fact that S" as subject (or object, as appropriate).
Those which should be marked [-FACT] are also easily diagnosed,
by the converse of the test for factivity. But how does one
determine that a clausal object of a verb which also allows
"the fact that S" is not an instance of deleted "the fact"?
That is, given (57),
(57)
(a)
(b)
He reported the fact that she had committed
the crime.
He reported that she had committed the crime.
how does one determine that report is [+/-FACT, [+/-S] rather
than simply [+/-FACT], [-S]? The Kiparskys point to a subtle
semantic contrast between the factive and non-factive interpretations of sentences like (57.b). They claim that factive gerundives
derive only from deep structure "fact that", and infinitivals
only from deep structure non-factives, resulting in the contrasting
interpretations of (57.b):
(57')
(b)
FACTIVE:
He reported her having committed
the crime.
NonFACTIVE: He reported her to have committed
the crime.
The gerundive is said to imply that the report was true in the
speaker's mind, while the infinitival is said to leave open the
possibility that the report was false, or at least non-substantiated.
We find this distinction over-subtle, and believe we can read
either sentence either way; but in any case it is impossible to
perceive a corresponding distinction with other verbs claimed to
be of the same class:
(58)
(a)
(b)
(b')
He acknowledged the fact that she had committed
the crime,
He acknowledged that she had committed the
crime.
FACTIVE:
He acknowledged her having committed
the crime.
NonFACTIVE: He acknowledged her to have
committed the crime.
565
NOM - 36
Testing the same distinction with anticipate, suspect, remember,
emphasize, announce, admit, deduce suggests that the distinction
is, at best, transitory. There are other contrasts between
otherwise identical factive and non-factive objects; these are
viable, but they cannot be hinged on the gerundive/infinitival
contrast. Thus the Kiparskys• example (59):
(59)
(a)
(b)
I explained the suspect's inching doorward.
I explained that the suspect inched
doorward.
where (59.a) is derived from "I explained the fact that the suspect
inched doorward", requires distinct meanings of explain; "to give
reasons for" in (a) and "say that S to explain X" in (b). But
since explain does not allow infinitival reduction in the non-factive
instance (b), this example in no way supports the contrast claimed
for examples like (57). It shows only that explain requires two
distinct lexical entries, which happen in this instance to
correlate with [+/-FACT], but that correlation does not appear to
exist in general for those verbs that take both factive and
non-factive objects.
A diagnostic which works for most of the factivityindifferent verbs cited by the Kiparskys is reduction of sentential
objects to stative-infinitival form, which is consistently
disallowed by factives:
(60)
The professor
(a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
anticipated ?
acknowledged
suspected
reported
remembered
emphasized ?
announced (?)
deduced
/
/
>
\
*
Bacon to be the real
author.
There are dialect differences about the viability of examples (a,f,g).
As noted above, it is not universally true that non-factive predicates are compatible with this structure (e.g. charge, intimate) ,
but perhaps all the factivity-indifferent ones are. In the present
analysis, at any rate, it has been assumed that predicates are
factive or non-factive in accord with the test of whether they
allow "the fact that S"; and if they allow it, and also allow
stative-infinitival reduction, they are marked as factivity-indifferent
(i.e. [+/-FACT, [+/-S] with obligatory specification of these such
that if one feature is plus, the other is minus).
566
NOM - 37
3.
The Abstract Instrumental
One fringe benefit of the Kiparskys• analysis of factive/
non-factive nominalizations is that a slightly messy aspect of
nominalization within the Case Grammar frame of reference is
cleaned up. At one point in the development of this grammar it
was assumed, almost by default, that at least two distinct
underlying cases must be allowed to dominate nominalizations, for
sentences like (6l):
(6l)
(a)
(b)
That he broke out of jail proves that he
was guilty,
Her leaving early suggests that she was
bored.
Fillmore suggested that the subject nominalization of these sentences
should be dominated in the deep structure by the Instrumental Case
(or conceivably some case like "Means" that does not now appear in
the grammar). The problem with that suggestion was that there was
then no way whatever to limit the range of cases under which the
feature [+/-S] could appear, though it was clear that we did not
want sentential objects under Datives, for example. But if all
sentences of the type (6l) involve only factive nominalizations
(in the subject), as appears to the case, then Fillmore's suggestion
can be adopted, but not with Instrumental case directly dominating
the nominalization: rather it dominates a factive of the structure
specified in (U.a), since clearly the sentences (6l) are reductions
of (6l»):
(6l')
(a)
(b)
B.
The fact that he broke out of jail proves
that he was guilty,
The fact of her leaving early suggests that
she was bored.
Sentential/NonSentential
The noun fact is itself a non-factive predicate. If any
predicate is [-FACT], it may or may not take a sentential NP in its
case frame. It must be marked [+S] if its only possible realization
of the case NEUT is sentential, or [-S] , if it cannot take a
sentential realization of NEUT. If it takes either, then it is
marked [+/-S] and specified one way or the other under the
convention of obligatory specification.
If a predicate allows a sentential realization of NEUT,
it must still be marked for the kind of sentence permitted or
required. Predicates which are constrained to indicative sentences
are marked [-IMPER], [-WH-S]; those which are constrained to
imperative sentences are marked [-INDIC], [-WH-S]; and those which
are constrained to interrogatives are marked [-IMPER], [-INDIC].
These features are hierarchically related to the feature [+S] such
that there is a lexical redundancy rule (62):
567
NOM - 38
(62)
[+IMPER]
[+INDIC]
[+WH-S]
-»
[+S]
The kinds of constraints that are provided by these features are
illustrated in (63):
(63)
(a) They demanded that she leave.
(b) *They demanded that she left.
(c) *They demanded what she was doing.
They expected that she would leave.
(e) *They expected that she leave.
(f) *They expected who arrived late.
(d)
(g) They knew that she left.
(h) *They knew that she leave.
(i) They knew who left.
(J)
They asked that she leave.
(k) They asked who left.
(1) •They asked that she left.
(m)
(n)
(0)
They insisted that she leave.
They insisted that she left.
*They insisted who left.
The features [FACT], [S], [INDIC], [IMPER] , and [WH-S] are strict
subcategorial features in the hierarchy (6U), with the definitions
(65):
(6U)
FACT
[+]
[-]
[-S]
[INDIC]
(65)
(a)
[FACT] = [
(b)
[S]
= [
[IMPER]
NEUT[NPthe
S
[WH-S]
fact
-NEUTUP^ J^
568
NEUT[NP[S^
] ]
NOM - 39
(c)
[INDIC] = [
NEur^ptst-SJC]]]]
where -SJC means that the predicate
of that S does not contain the
morpheme SJC ("subjunctive")
(d)
[
[IMPER] = [
NEUT
NP[S[+SJC]]]]
where +SJC means that the predicate
of that S contains the morpheme SJC
(e)
[WH-S]
[
= [
NEUT[NP
[WH]]]]
S
where WH means that the S contains
the feature [+WH]
A predicate which allows only a non-sentential NP as realization of
the case NEUT, and does not allow the noun fact with its potential
complementation, would be marked [-FACT] [-S] in the lexicon. No
provision is made here for those predicates that allow only cognate
objects other than sentential ones, like dream:
(66)
C.
Emotive/NonEmotive
1.
2.
1.
He dreamed that he had solved the problem.
He dreamed a pleasant dream.
The Sources of Complementizers
Classes of Emotive and NonEmotive Predicates
The Sources of Complementizers
Rosenbaum (l967a)proposed that that, for-to, and POSS-ing
were essentially idiosyncratic features on the heads of sentential
complements. It is still hard to find satisfactory generalizations
to account for the gerundive complements , but at least that and
for appear to be redundant on semantic and/or configurational
facts. The item that can be inserted by an extremely general rule,
given the conditions that there is an embedded sentence dominated
by NP and that subject-verb agreement has applied; it is subsequently deletable by an optional rule which applies to all such
structures provided that they are not subjects, and are non-factive.
The item for appears to depend, as claimed by the Kiparskys , on
a class of head items which have the feature [+EMOT]. As is
demonstrated in Section III.D of this paper, the independent
insertion of for in the presence of the feature [+EMOT] has numerous
syntactic consequences in conjunction with several other processes
which all result in the formation of infinitivals.
569
NOM - kO
We therefore reject, along with the Kiparskys, the
spurious introduction of for, as done by both Lees (i960)
and Rosenbaum (l967a),in the derivation of infinitival nominalizations. Instead we insert for in the presence of the feature
[+EMOT] on the head item. This label "emotive" refers to "all
predicates which express the subjective value of a proposition
rather than knowledge about it or its truth value" (Kiparsky,
1968).
2.
Classes of Emotive and NonEmotive Predicates
Depending
dominated by NEUT
objectivalization
are from Kiparsky
on the case-frame of the predicate, a sentence
may undergo either subjectivalization or
in the early rules of the cycle. These lists
(1968).
[+EMOT] [+FACT]
subjectivalization
important
crazy
odd
relevant
instructive
sad
suffice
bother
alarm
fascinate
nauseate
exhilarate
defy comment
surpass belief
a tragedy
no laughing matter
The Kiparskys list three factive predicates which require objectivalization of the sentence under NEUT, but these are ungrammatical
with for-to constructions in all dialects we have checked. Their
examples are regret, resent, and deplore. We find the examples
(67) ungrammatical, but evidently the Kiparskys do not:
(67)
(a)
(b)
(c)
*We regretted for her to do it.
*We resented for her to do it.
*We deplored for her to do it.
For us there appear to be no [+FACT], [+EMOT] examples of verbs with
which the NEUT would undergo objectivalization — i.e. there are
no sentences of the type (67) with factive predicates. The one
apparent counter-example has been analyzed correctly by Lees,
Rosenbaum and others as containing a preposition with the verb
which deletes the for-complementizer, and it is non-factive in any
case:
(68)
(a) We hoped for them to do it.
(b) We hoped for a solution to the problem.
(c) *We hoped for the fact that they would do it.
In contrast with the [+EMOT], [+FACT] class of predicates
with subjectivalization, there is a non-factive class; there is a
corresponding class with objectivalization:
570
NOM - 111
[+EMOT] [-FACT]
subjectivalization
[+EMOT] [-FACT]
objectivalization
improbable
unlikely
nonsense
a pipedream
[+FUT]
intend
prefer
reluctant
anxious
willing
eager
[+FUT]
urgent
vital
The feature [+FUT] is a deep structure constraint discussed in
Section G of this paper. It requires that the tense of the predicate of the embedded sentence refer to a time posterior to that
of the matrix predicate.
To show that the feature [EMOT] is on a parameter orthogonal to that of the feature [FACT], the Kiparskys list [-EMOT]
examples of each type:
[-EMOT] [+FACT]
subjectivalization
[-EMOT] [+FACT]
objectivalization
well-known
clear
(self)-evident
goes without saying
be aware of
bear in mind
make clear
forget
take into account
[-EMOT] [-FACT]
subjectivalization
[-EMOT] [-FACT]
objectivalization
[+FUT]
probable
likely
turn out
seem
predict
anticipate
foresee
[+FUT]
[+/-FUT]
imminent
in the works
say
suppose
conclude
571
NOM - k2
D.
Infinitivalization
1.
2.
3.
h.
5.
6.
7.
1.
Conditions for TO-REPLACE-AUX
Illustration of l.a: Derivation of Infinitivals with
[+EMOT] Predicates
The fact — it?
Conditions for EQUI-NP-DEL
Illustration of l.b: Derivation of Infinitivals with
EQUI-NP-DEL
Conditions for Subject Raising
Illustration of l.c: Derivation of Infinitivals with
Subject Raising
Conditions for TO-REPLACE-AUX
Following the Kiparskys' view of the matter (1968) ,with
minor modifications, the infinitive is taken to be simply the
form of a verb that has not undergone agreement with a subject,
always marked by to_ unless deleted by the exception feature
[+TO-DEL]. The list of [+TO-DEL] verbs includes the verbs of
sense perception see, hear, feel (but not taste, smell), and such
verbs as help, make, have, let. The conditions under which a
verb does not undergo agreement with a subject are the following:
a.
When the subject is marked with an oblique (surface)
case, as when it is in construction with a preposition
for inserted with the [+EMOT] verbs.
b.
When the subject is erased from the clause of the
verb, e.g. by EQUI-NP-DEL, where the erasing node
will be either a deep structure dative, or it will
be a deep structure agent in the absence of a dative.
c.
When the subject is raised from its own clause into
the next higher S; it may be raised to object of the
next higher predicate by the regular objectivalization
rule if it is marked [+RAIS-OBJ], or it may be
raised to subject of the next higher predicate by
the regular subjectivalization rule if it is marked
[+RAIS-SUBJ].
Given any instance, then, of a verb that has not undergone agreement
with a subject, for any of these reasons, the rule of TO-REPLACE-AUX
applies to insert the form to_ in the position of the Auxiliary:
more precisely, to_ replaces tense and modal, retaining Perfect
and/or Progressive and inserting Perfect in case the tense was
Past:
572
NOM - U3
(69)
2.
(a)
He expected — She would have done it,
He expected her to have done it.
(b)
He supposed — She did it.
He supposed her to have done it.
(c)
He ordered her — She SJC do it.
He ordered her to do it.
(d)
He believed — She is working on it.
He believed her to be working on it.
Illustration of l.a:
Predicates
Derivation of Infinitivals with [+EMOT]
The derivation of infinitival nominalizations with [+EMOT]
predicates proceeds roughly along the following lines: given a
structure like (70.a) with a factive predicate, the optional rule
of FACT-DEL yields (70.b), after the usual rules of objectivalization and subjectivalization have been applied:
(70)
(a)
MOD
AUX
[+PAST]
regret
PREP
NP
PREP
NP
\NOM
the
fact
N
I
John
NEUT
NP
PREP
I
S
Mary [+PAST] leave early
573
NOM - kk
(b)
NP
[+DAT]
I
John
MOD
PROP
AUX
[+PAST]
V
regret
Mary [+PAST] leave early
It might be noted in passing that if only the most general transformations had operated on the structure (TO.a), the output would
be (70.c); and if the optional GERUNDIVE transformation had been
applied, the output would be (70,d), with the preposition of_
being retained as the marker of the deep structure NEUT.
(70)
(c)
(d)
John regretted the fact that Mary left early,
John regretted the fact of Mary's having
left early.
If FACT-DEL has been applied to derive (70.b), that structure is
then subject to THAT-INSERT, yielding (70.e):
(70)
(e)
John regretted that Mary left early.
Since FACT-DEL follows GERUNDIVE, the output could also be (70.f):
(70)
(f)
John regretted Mary's naving left early.
In those dialects like the Kiparskys' in which regret is a [+EM0T]
verb that allows objectivalization, the rule of FOR-INSERT applies
to the structure (70.b), of which the output is (70.g):
(70)
(g)
John regretted — for — Mary PAST leave early
This is subject to obligatory TO-REPLACE-AUX, with the output (70.h)
(70)
(h)
John regretted for Mary to leave early.
Since (70.h) is ungrammatical in the dialects we have had access to
(see discussion in Section III.C.2 above), the generalization
about for being dependent upon [+EM0T] predicates is immediately
suspect. One almost wonders if the generalization would have been
57U
NOM - »+5
noticed at all by speakers of a dialect for which regret, resent,
and deplore, which are obviously emotive in semantic content,
are ungrammatical in constructions like (70.h). But the generalization is valid for such a wide range of examples (Section III.C.2)
that these three items must be marked simply as exceptions:
i.e. they are semantically [Emotive] but syntactically [-EM0T].
Illustrating further, this time with an example that is
not dialectally tainted, consider (71):
(71) (a)
[-PAST]
will
suffice
[+EM0T]
[+FACT]
[-S]
the
N
we [-PAST] have a solution
575
NOM - U6
After the usual early rules have been applied, (71.a) has the
structure (71 .b):
(71)
(b)
PROP
the
N
PREP
NP
fact
of
S
will
suffice
[+FACT]
[+EMOT]
[-S]
[-PAST] have a solution
(71.b) is the structure underlying (71.c) to which PREP-DEL and
THAT-INSERT have been applied:
(71) (c)
The fact that we have a solution will suffice.
If, instead, the optional rule of FACT-DEL is applied, and then
PREP-DEL and THAT-INSERT, the sentence is (71.d):
(71)
(d)
That we have a solution will suffice.
But if FACT-DEL is applied, and then the rule of FOR-INSERT
is applied, followed by the then obligatory TO-REPLACE-AUX, the
sentence is (71.e):
(71)
(e)
For us to have a solution will suffice.
EXTRAPOSITION can optionally be applied either to (71.d) or (71.e):
(71)
(f)
(g)
It will suffice that we have a solution.
It will suffice for us to have a solution.
The mention of extraposition brings us to a proposal of the Kiparskys'
which we reject, namely the source of it in (71.f,g).
576
NOM - 1*7
3.
The fact -» it?
Consider the sentence (70.c), John regretted the fact
that Mary left early. The Kiparskys claim that the fact may
be pronominalized as It, thus deriving the sentence (70.c'):
(70)
(c')
John regretted it that Mary left early.
The sentence is certainly grammatical. But the Kiparskys' claim
that It derives here from pronominalization of the fact is
dubious in the extreme, for the following reasons:
(a)
(72)
(b)
(70)
(c)
(73)
Definite pronominalization cannot be so construed
as to end up with a definite pro-form followed
by a modifier/complement/sentential object of
any kind. Only the "whole NP", a notion that is
not totally clear (see PRO), is subject to
definite pronominalization. This fact explains,
e.g., the ungrammaticality of (72):
*The belief that the world was round replaced
it that the world was flat.
Even if there were no general fact such as (a),
derivation of It by pronominalization of the fact
would run into grave difficulty in the face of the
grammaticality of (70.c) when pronominalized as
(70.c'), but the ungrammaticality of (70.d) if
a similar pronominalization is attempted to yield
(70.d'):
(d') *John regretted it of Mary's having left early.
The assumption of the Kiparskys that there really is
a head noun in sentences like (73.b,d) but not in
sentences like (73.a,c),
(a)
(b)
I take it that you all know the answer.
I resent it that you all know the answer.
(c)
(d)
I would hate it for anyone to reveal the secret.
I would resent it for anyone to reveal the
secret.
would be greatly strengthened if Ross's Complex NP
Constraint (see REL) held for (b) and (d), which are
putative pronominalizations of the fact, but not for
(a) and (c), which are assumed to come from "vacuous
extraposition from object" (Rosenbaum (1967a),
accepted by Kiparsky (1968), with the qualification
"perhaps"). But in fact relativization on answer
and secret is equally good in either member of the
pairs:
577
NOM - hti
(T31)
(a/b)
This is the answer which I take/resent
it that you all know.
(c/d)
This is the secret which I would hate/resent
it for anyone to know.
That the Complex NP Constraint should hold in these
examples (not cited by the Kiparskys) follows from
their claim that the ungrammaticality of (73.e,f),
which are cited by them, is accounted for by the fact
that the Complex NP Constraint disallows relativization
across a lexical head noun, namely the fact whether
pronominalized or not:
(73)
(e)
(f)
*This
John
*This
that
is the book which you reported it that
plagiarized,
is the book which you reported the fact
John plagiarized.
But (73.e,f) prove nothing, since (73.g) is ungrammatical
anyway:
(73)
(g)
*You reported it that John plagiarized the book.
This entire argument may be with a straw man, since in the preliminary
version (the only one we have seen) there is a footnote #7 in which
the Kiparskys point out that "It appears now [i.e. presumably at
some time after completing the main body of the manuscript] that
questioning and relativization are rules which follow fact-deletion."
Their other observations about the blocking of movement transformations
(the Complex NP Constraint) by virtue of the presence of the head
noun fact (as in NEG-raising, which occurs only with non-factives,
and RAIS-TO-SUBJ, which also occurs only with non-factives) may be
correct; they do not depend on pronominalization.
Thus while there is no doubt that the Kiparskys' observation
that the surface form it-that-S is generally acceptable with factive
predicates and unacceptable with non-factive predicates is a correct
observation, and while it is appealing to explain this on the basis
of pronominalization of the fact, the explanation is unsatisfactory.
In this analysis, then, the fact is treated as deletable by the
rule FACT-DEL; once deleted, then vacuous extraposition can apply:
(7U)
(a)
(b)
I hate it that she dresses so conservatively.
[Factive, Obligatory extraposition from object]
I regret it that she dresses so conservatively.
[Factive, Optional extraposition from object]
578
NOM - U9
There is a redundancy relation between extraposition from object
and factivity. The rule for such extraposition can be framed
only given a statable environment, and that environment is statable
only by mention of the feature [+FACT] on the governing predicate.
But there are indubitably factive predicates like grasp which do
not permit extraposition from object (and must be marked with an
exception feature):
(75)
(a)
He grasped (the fact) that the project was
almost over,
(b) *He grasped it that the project was almost over.
There are factive predicates like hate which require extraposition
(so that the rule is not always optional):
(76)
(a) He hates it that the project is almost over,
(b) *He hates that the project is almost over.
and there are the great majority of factive predicates with which
extraposition is optional:
(77)
U.
(a)
He regrets that the project is almost over.
He regrets it that the project is almost over.
Conditions for EQUI-NP-DEL
In outlining the derivation (71) and (72) we were illustrating
the operation of the first of three conditions under which a verb
does not undergo agreement with a subject, namely when for is inserted
under government by the feature [+EMOT], thereby assigning an
oblique surface case (whether actually labeled accusative, or
blocked from participating in subject-verb agreement by some other
device: see the analysis of subject-verb agreement and pronoun form
in PRO) which cannot participate in subject-verb agreement rules,
in turn forcing the verb into the infinitive form by the rule
TO-REPLACE-AUX.
The second condition under which a verb does not undergo
agreement with a subject is when the subject has been erased by some
coreferential node in the matrix. There are two classes of such
coreferential nodes: the transformation of EQUI-NP-DEL must inspect
a structure and determine whether the subject of the embedded
sentence is identical with a dative, or if there is no dative then
with an agent in the matrix sentence. If there is such a coreferential
node, the subject of the embedded sentence is erased.
579
NOM - 50
5.
Illustration of l.b:
EQUI-NP-DEL
Derivation of Infinitivals with
The first of the two classes of coreferential nodes to
which EQUI-NP-DEL applies, erasing the subject node of the sentential
object, is a dative node governed by the same head item as the one
which governs the sentential object, as in (78.a):
(78)
(a)
MOD
I
I
AUX
TE
NEUT
require
[DAT = of]
DAT
ACT
^ NP
PREP
PREP
NP
you
PRES
you SJC solve the problem
The position of the dative after the object is its normal position:
(78)
(b)
I require the answer of you.
I gave the book to you.
Its position before the object in the clausal nominalization (78.a)
is presumably the result of a late reordering rule having to do with
the length of the constituents, which is supported by the order
of elements after extraposition:
(78)
(c)
I require of you that you solve the problem.
I require it of you that you solve the
problem.
Recall now that the objectivalization rules of this grammar make
the realization of the NEUT case into the object unless the verb is
marked for objectivalization of a different case. Thus a sentence
like He aimed the gun at John is an instance of objectivalization
of the instrumental case, and He filled the pool with water is an
instance of objectivalization of the locative case. Ordinary
datives, in sentences like I gave him the money, are instances of
optional objectivalization of the dative. Consider now the sentence
(78.c): in it,we have objectivalized NEUT, not DAT. If we had
chosen Passive Subject Placement in the early rules, the sentence
would be (78.d):
(78)
(d)
That you solve the problem is required
of you (by me).
580
NOM - 51
Now, the sentence which illustrates EQUI-NP-DEL with the verb
require is (78.e):
(78)
(e)
I require you to solve the problem.
But this sentence can only be derived from (78.a) if EQUI-NP-DEL
has applied, and then objectivalization, since the passive is (78.f):
(78)
(f)
You are required to solve the problem (by me).
In short, then, the deep structure (78.a) underlies both (78.c) and
(78.f), and EQUI-NP-DEL is optional for this verb.
The two derivations from (78.a) resulting in (78.c) and (78.e)
are possible only if EQUI-NP-DEL is optional for this verb. Besides
require, the verbs ask and request are of this type. More frequently
the verbs which share the derivation from structures like (78.a)
have obligatory EQUI-NP-DEL if the coreferential NP appears in an
embedded imperative. Such verbs are force, allow, implore, permit,
persuade, want, warn, encourage, instruct, and remind. If it were
not obligatory, the starred examples of (78.g) would result:
(78)
(g)
I
*I
*I
*I
forced
forced
forced
forced
him to solve the problem.
that he solve the problem.
him that he solve the problem.
to/of/for him that he solve the problem.
The condition of obligatory EQUI-NP-DEL depends on embedding of an
imperative, since remind, persuade, warn, and instruct take both
indicative and imperative embeddings = I reminded hinu that he
was leaving at one = I reminded him to leave at one.
*
A different set of verbs which also shares the derivation of "I
require you to solve the problem" is differentiated from the require
class only by the fact that its case frame has Dative optionally,
as require does, but if Dative is present then EQUI-NP-DEL is obligatory.
Examples are command, order, advise, urge, and desire. The constraint
just stated provides for the grammatical examples of (78.h) while
blocking the ungrammatical one:
(78)) (h)
I commanded that he solve the problem. [No dative]
I commanded him to solve the problem.
*I commanded him that he solve the problem.
There is a small class which, like those above, takes embedded
imperatives, but this class disallows EQUI-NP-DEL:
(78)
(i)
I insist/demand/suggest that you solve the problem.
*I insist/demand/suggest you to solve the problem.
581
NOM - 52
Since this class disallows EQUI-NP-DEL (if it allows Dative in its
case-frame at all, as in "I insist that you solve the problem for me",
which may better be analyzed as a Benefactive case), there is no
infinitivalization of the preceding type. Demand, however, allows
infinitival!zation of the type discussed below, as in (79):
(79)
I demand to see a doctor.
The second class of coreferential nodes to which EQUI-NP-DEL
applies in the derivation of infinitival nominalizations is those in
which there is no dative directly dominated by the governing item,
but the relation of coreferentiality holds between the matrix and
constituent agents. Agent-agent coreferentiality may be obligatory,
as with a verb like learn, condescend, or try:
(8l)
(a)
(b)
(c)
He condescended to resign when he came of age.
He tried to do his homework.
He learned to analyze sentences.
(d) *He condescended Mary to resign.
(e) *He tried Bill to do his homework.
(f) *He learned Mary to analyze sentences.
Or agent-agent coreferentiality may be optional as with expect, intend,
want, forget, remember,...:
(82)
(a
(b
He expected Mary to leave early.
He expected to leave early.
(c
(d
He intended for Mary to leave early.
He intended to leave early.
(e
(f
He wanted Mary to leave early.
He wanted to leave early.
A single rule of equi-NP-deletion handles both instances like (78.e)
and (8l)-(82), since the rule applies first to a coreferential
dative, and if it finds none it applies to a coreferential agent.
In either instance, the subject of the sentential object is erased,
leaving the conditions necessary for infinitivalization with to,
namely a verb without a subject to which the agreement rules would
apply.
In addition to the two classes of equi-NP-deletion, there is
an indefinite subject one which is deletable, but such deletion
applies after such rules as for-insertion with [+EMOT] predicates
and therefore provides no new basis for infinitivalization:
582
NOM - 53
(83)
(a)
(b)
For one to see her is for one to love her.
To see her is to love her.
(c)
In order for one to get good grades, it is
necessary for one to study hard.
In order to get good grades, it is necessary
to study hard.
(d)
(e)
(f)
6.
John's proposal for (some)one to end the war
in Viet Nam fell on deaf ears.
John's proposal to end the war in Viet Nam
fell on deaf ears.
Conditions for Raising Subject to Subject, or Subject to Object
The third and final condition under which a
have a subject remaining to provide for finite-verb
the subject of the sentential object is raised from
into the next higher S. There are two main classes
a.
(8U)
verb may fail to
agreement is when
its own clause
of raising:
Raise the subject of the sentential object to subject of
the matrix verb by the rule RAIS-SUBJ, governed by the
feature [+RAIS-SUBJ]. This rule precedes the regular
subjectivalization rule early in the cycle. From the
structure underlying (oU.a) it provides either for (8U.b),
where the entire neutral case is subjectivalized, or for
(81*.c) where the subject is raised.
(a)
(b)
(c)
Is unlikely - He will solve the problem.
That he will solve the problem is unlikely.
He is unlikely to solve the problem.
This analysis eliminates the spurious IT-replacement
rule of Rosenbaum, since (8U.c) is generated directly
from the underlying structure (8U.a), not from the
extraposition of (8U.V):
(8U)
(b1)
It is unlikely that he will solve the problem.
583
NOM - 51*
The rule of RAIS-SUBJ (read "raise subject to subject")
is obligatory with verbs like begin, continue, start
blocking (8U.f):
(8U)
(d) Began - He ran.
(e) He began to run.
(f) *That he ran began.
Sentences like (8U.e), analyzed as Intransitive Verb
Phrase Complementation by Rosenbaum (1967a), have a number
of special properties which argue that they belong with
the other RAIS-SUBJ verbs. The most striking such
property is the occurrence of the expletive there as
surface subject of the matrix verb in Just those
instances where it is possible as surface subject of
the embedded verb:
(8U)
(g)
(h)
There began to be rumblings of discontent,
There were rumblings of discontent.
A counterargument to this analysis, pointed out by
Perlmutter (1968b) is that with verbs that appear to
require deep structure subject identity, like try,
condescend, a verb begin must have a deep structure subject in order to be able to state the constraint that
blocks (8U.i):
(Qk)
(i) *I tried to begin to like Jazz.
Perlmutter concludes that the verb begin must be permitted
to occur in both configurations: i.e. with abstract subjects, as in (8U.d,e), and with concrete subjects and
complements, as in (8U.J):
(8U)
(j)
He tried to begin to do his work.
He began to do his work.
There are, however, difficulties in the notion "deep
structure constraint" on subject identity. If (8U.k) is
well-formed, as we believe,
(81+)
(k)
John tries to be difficult to please.
it must have a deep structure in which John is object of
please: i.e., To please John is difficult. The constraint
that the subject of try and the subject of its complement
must be identical cannot here be stated as a deep structure
58U
NOM - 55
constraint, only as a mid-derivation constraint, or
conceivably as a surface structure filter of some kind.
If (8U.k) is Judged not to be fully well-formed, then
it appears that begin will indeed have to be permitted
in both configurations, as Perlmutter claims. But
then there will be unexplained derivations of Perlmutter's
John began to read the book, which stands as an unsolved
problem. The data on which the case rests is not entirely
clear, since (8U.i), rejected by Perlmutter, is acceptable
to most speakers.
b.
Raise the subject of the sentential object to object of
the matrix verb by the rule RAIS-OBJ (read "Raise subject
to object") governed by the feature [+RAIS-OBJ], This
rule is optional for most verbs, but obligatory with a
few like consider which disallow clausal nominalization:
(85)
(a)
(b)
They expected that he would solve the problem.
They expected him to solve the problem.
(c)
(d)
He believes that she is intelligent.
He believes her to be intelligent.
(e) *He considers that she is intelligent.
(f) He considers her to be intelligent.
Like the rule RAIS-SUBJ, this one precedes the regular
objectivalization rule early in the cycle, thus providing,
in those instances where it is optional, for either the
clausal or infinitival nominalization of (85).
Consider now the motivations for claiming that the
subject of the embedded clause in (85.c) is raised to
object of believe in (85.d). If the analysis did not
raise the clausal subject she to object of believe, there
would be no natural explanation of the fact that reflexivization is possible in this position:
(85)
(g)
She believes herself to be intelligent.
Reflexivization is not normally possible down into a
lower sentence:
(85)
(h) "She persuaded John to like herself.
This argument is not totally convincing, perhaps, in view
of the fact that verbs like expect require EQUI-NP-DEL
under these circumstances, so that one cannot argue for
RAIS-OBJ on these grounds, with these verbs:
585
NOM - 56
(85)
(i) *He expected himself to solve the problem,
(j) He expected to solve the problem.
Nonetheless the RAIS-OBJ analysis, proposed by the Kiparskys
(1968), serves well to bring together all instances of
infinitivalization under a single principle of to-insertion and is adopted here. It is quite analogous to the
RAIS-SUBJ principle illustrated in (8U), which has been
accepted in some form by virtually everyone who has
examined sentences of this type. In the present analysis,
it is extended to cover the so-called "second passive" of
(86):
(86)
(a) One says — He is intelligent.
(b) *One says — him — to be intelligent. [RAISOBJ objectivalization]
(c) He is said to be intelligent. [Passive
subjectivalization]
(d)
(e)
(f)
(g)
One says — He is intelligent
One says — that he is intelligent. [Regular
objectivalization]
That he is intelligent is said. [Passive
subjectivalization]
It is said that he is intelligent. [Extraposition]
It is true that this derivation creates one ungrammatical
intermediate stage for the verbs say, rumor, and repute; but
all the others that are commonly analyzed as second passives
have no ungrammatical intermediate stage under this derivation — suppose, think, consider, believe,...—and there is
no reason to set up a different derivation for the verbs
say, rumor, and repute when all that is required is either
to make the passive obligatory with subject-raising in these
sentences, or to claim that some special surface constraint
filters out (86.b), since these verbs are idiosyncratic in
a number of ways.
There is one strong reason to maintain this derivation
of the 2nd passive even in the face of the ungrammatical
intermediate stage generated for say, rumor, and repute•
The only alternative derivation is by some form of ITreplacement after extraposition:
(86)
(g)
(h)
It is said that he is intelligent,
He is said to be intelligent.
586
NOM - 57
But, although this avoids an ungrammatical stage in the
2nd Passive derivation with say, rumor, and repute, it
provides another path for the comparable 2nd Passive
derivation with think, believe, suppose, etc.:
(86)
(i)
(j)
It was thought that he was intelligent,
He was thought to be intelligent.
But (86.J) can also be derived through the regular
passive from They thought him to be intelligent; since
(86.j) shows no trace of structural ambiguity, we believe that the general RAIS—OBJ solution is correct
and that IT-replacement should be rejected for 2nd
Passive derivations.
7.
Illustration of l.c:
Raisings
raising.
Derivation of Infinitivals with Subject
We consider now in detail one example of each type of subject
The deep structure of (8U.a,b,c) is shown as (8V):
(8U-)
MOD
AUX
unlikely
[+V.+ADJ]
[+RAIS-SUBJ]
He will solve the problem
The general rule of BE-INSERTION inserts be in front of the adjectival
predicate. The rule of RAIS-SUBJ, an alternative to the general subJectivalization rule, governed by the feature [+RAIS-SUBJ] on unlikely
(which is marked plus/minus this feature in the lexicon, since the
raising is optional), applies to move the subject of the sentential
object out; this leaves the usual configuration for the rule TO-REPLACEAUX, and the result is the structure underlying (8U.c).
587
NOM - 58
Alternatively, given a structure identical with (Qk') except
for negative specification of the feature [RAIS-SUBJ], the entire
sentential object will be subjectivalized, with the output being the
structure underlying (&k.b).
The deep structure of (85.a,b) is shown as (85')
(85')
MOD
AUX
expect
PREP
[+V.-ADJ]
[+RAIS-OBJ]
[+FUT-REDUC]
He will solve the problem
The rule of RAIS-OBJ, an alternative to the general objectivalization
rule, governed by the feature [+RAIS-OBJ] on expect (which is marked
plus/minus this feature in the lexicon, since the raising is optional),
applies to move the subject of the sentential object out, this time
into object position where in (81**) it was moved into subject
position; this leaves the usual configuration for the rule TO-REPLACEAUX, and the result is the structure underlying (85.b). Alternatively,
given a structure identical with (85') except for negative specification of the feature [RAIS-OBJ], the entire sentential object will
be objectivalized, with the output being the structure underlying (85.a).
E.
Gerundive/NonGerundive
1.
2.
3.
k.
5.
1.
The Relation between Factivity and Gerundives
Gerundives after Prepositions
Generic Gerundives
Adverbial Gerundives
ing-of Gerundives
The Relation between Factivity and Gerundives
It is now possible to consider in detail the proposal of the
Kiparskys that infinitival nominalizations derive from the sentential
objects of non-factive predicates only, and that gerundive nominalizations derive from the sentential objects of factive predicates. The
588
NOM - 59
question is, does there have to be a parameter [+/-GER] orthogonal
to the [+/-FACT] parameter? If there are gerundive nominalizations
that are factive, then the orthogonality of these parameters prevents
us from accepting without reservation the claim of the Kiparskys (1968)
that there is a redundancy relation between factivity and gerundive,
and between non-factivity and infinitival. We have seen examples
which violate the latter claim: The fact that she died so young was
a tragedy —• For her to have died so young was a tragedy; but such
factive infinitivals are, it is true, restricted to the [+EMOT]
constructions, so that there is indeed a correlation between the
infinitivals from RAIS-SUBJ and RAIS-OBJ transformations and nonfact ivity.
The correlation between factivity and gerundives is also high.
There are some verbs with which the gerundive is obligatory, as the
form of any sentential object: e.g. avoid, stop,
(87)
(a) She avoided leaving early.
(b) *She avoided to leave early.
(c) She stopped typing at 2:00 a.m.
(d) *She stopped to type at 2:00 a.m. [Ungrammatical
in the intended sense; grammatical as
Purpose ADV]
The Kiparskys do not deal with these, other than to eliminate them
from the class of gerundives that they claim are restricted to factive
predicates. It is clear that they are non-factive, since the fact of
cannot be construed with them. But it throws no special light on them
to assert merely that they "refer to actions or events" (Kiparsky,
1968). The point, rather, is that among all the predicates that
accept gerundive nominalizations, only the factive predicates accept
non-action gerundives (where non-action means that the embedded S
contains a [+STAT] predicate, or that the AUX includes PAST, PERF, or
PROG); and that, in turn, is equivalent to the assertion that only
the noun fact is compatible with non-action gerundive nominalizations
of sentential objects. That is, gerundive nominalization is restricted
to actions except when the governing item is fact. In support of this
view, consider (88) and (89):
(88)
(a)
(b)
(c)
(d)
He
He
He
He
hated to leave so early.
hated leaving so early.
hated having left so early.
disliked understanding the problem.
(89)
(a) He continued
(b) He continued
(c) *He continued
(d) *He continued
to work hard.
working hard.
having worked hard.
understanding the problem.
589
NOM - 60
Both dislike (factive) and continue (non-factive) are compatible
with either infinitival or gerundive nominalizations. But only
dislike is compatible with the non-action gerundive (88.c), and
the fact of can be construed with both (88.b) and (88.c), even if
infelicitously because of the semantic incongruity of disliking
the fact of anything.
The number of predicates which are compatible with gerundive nominalizations, outside of the [+FACT] class, is very small,
and they should be marked as exceptions. Since all factives allow
gerundive nominalizations, there must be a lexical redundancy rule
of the form (90):
(90)
[+FACT]
[+/-GER]
where [+GER] is a rule feature governing gerundive nominalization
of the sentential object of fact, with these predicates. Those
predicates with which gerundive nominalization is obligatory must
be marked [+/-GER], and all others are redundantly [-GER] by the
rule (91):
(91)
[-FACT]
[-GER]
By the general lexical convention that marked features cannot be
over-ridden by redundancy rules, the exceptional items marked
[+/-GER], if they have been selected with positive specification,
remain unchanged by (91). This is equivalent to a marking device:
(91')
2.
[-FACT]
GERj
[-GER]
Gerundives after Prepositions
The remaining instances of gerundive nominalizations are
of two types: those which appear after prepositions, and generics.
There is one more, largely problematic, type which we characterize
as adverbial.
After prepositions, two distinguishable situations exist:
(l) the preposition is a case-marking (transformationally-inserted)
preposition; or (2) the preposition is a deep structure lexical item.
In the former instance, the question of gerundivization is determined
by the head (see CASE PLACE II.B), since the head may also govern
a that-S embedding:
(92)
(a)
(b)
He insisted on her leaving.
He insisted that she leave.
590
NOM - 61
That is, insist is lexically marked [+/-GER], and if [-GER]
is chosen, then (92.b) is the result, with on deleted late by a
general rule deleting PREP before that-S. With other aberrant
prepositions - e.g. upon in rely upon - it must be assumed that it
is the verb which is governing gerundivization even though in this
instance there is no that-S possibility:
(93)
(a) He relies upon her working late.
(b) *He relies that she work late.
On the other hand, deep structure lexical prepositions allow only
gerundives:
(93)
(c)
(d)
He went out without her hearing him.
On considering the problem further, he decided
to rewrite the paper.
In factive examples the question of the deletion of prepositions is
irrelevant, since the rule of FACT-DEL determines the surface
structure of sentences like (93.e,f), with the object-marking preposition of retained after nouns and deleted after verbs:
(93)
(e)
(f)
He appreciated (the fact of) her working so hard,
His appreciation of (the fact of) her working
so hard.
A corresponding non-factive example demonstrates clearly that either
the fact of, or some preposition, must be present to protect gerundivization - otherwise the rules that govern infinitialization will
operate.
(9^)
3.
(a)
(b)
He intended to leave early.
His intention of leaving early was thwarted
by too much discussion.
Generic Gerundives
Generic gerundives are always subjectless in their surface
form:
(95)
(a) Taming lions is dangerous.
(b) "John's taming lions is dangerous.
(c) Climbing mountains is fun.
(d) *John's climbing mountains is fun.
They are paraphrases of for-to (i.e. emotive infinitival) constructions
with deleted indefinite subjects:
(95')
(a)
(b)
It is dangerous (for one) to tame lions,
It is fun (for one) to climb mountains.
591
NOM - 62
The existence of this paraphrase relationship suggests that generic
gerundives have an underlying indefinite/impersonal subject one
which is obligatorily deleted in the derivation from for-to to gerundive.
This assumption accounts for the fact that only animate subjects
are normally "understood" in subjectless gerundives. Given a verb
that will not accept an animate subject, subjectless gerundives cannot
be formed:
(96)
k.
(a) *Elapsing is dangerous.
(b) Time's elapsing is dangerous.
Adverbial Gerundives
These are essentially a residue class.
"intransitive" types:
(97)
Consider first the
He began/ceased/continued/finished/quit/started working.
If the general principles of to-insertion proposed by the Kiparskys
and elaborated in Section III.D. above have any validity - and they
do seem to generalize a number of othewise apparently idiosyncratic
facts - then (97) cannot be said to involve the normal processes
of nominalization at all, since EQUI-NP-DEL would remove the subject
of the sentential object, and TO-REPLACE-AUX would be obligatory,
yielding ungrammatical strings like (97'):
(97') *He finished/quit to work.
(The other examples of (97) would be grammatical because they do indeed
also operate under the normal rules of infinitivalization.) To claim
that these -ing forms are adverbial, as they were historically, is
difficult to Justify on syntactic grounds. In the absence of any
well-motivated analysis, we mark these "gerundive infinitives" by
the feature [+GER], the same exception feature used for avoid and deny,
and generate them accordingly, ordering the rules with the l+GER]
rule preceding all the rules having to do with infinitivalization and
thereby guaranteeing that such consequences as (971) cannot arise.
For lack of a better explanation, we handle the gerundives in "transitive" constructions in the same way:
(98)
I saw/felt/perceived/watched...him moving.
All of these have a corresponding infinitival form generated in the
normal way (except with [+TO-DEL]). It is at least possible that
they should be generated as normal embedded progressives with
[+TO-BE-DEL]:
592
NOM - 63
(99)
(a)
(b)
I saw him (to be) moving,
I felt him (to be) moving.
The obvious disadvantage of this proposal is that [TO-BE-DEL]
normally applies only when the predicate is adjectival:
(100) (a)
(b)
I considered him (to be) intelligent.
I believed him (to be) intelligent.
(c) *I considered him moving.
(d) *I believed him moving.
The semantics of this proposal are also rather bad in some instances:
(101) (a)
I heard him talking.
This does not imply "I heard that he was talking"; rather it implies
something much closer to "I heard him in the act of talking" or some
similar adverbial paraphrase. Similar semantic observations can be
made for most of the verbs in this class.
In short, we have no satisfactory analysis for the adverbial/
progressive gerundives. There are various ad hoc ways to generate
them, but none seem to shed any light on the way they are interpreted,
semantically.
5.
ing-of Gerundives
Constructions like The shooting of the lions, labeled "action
gerundives" by Lees (i960), are considered to be lexically derived,
like the proposal of a solution, his insistence on that answer, in
this grammar. That is, shooting is lexically available as a noun,
related derivationally to the verb shoot, and as a noun it may take
an object (i.e. it has its own case frame). Such nouns cooccur less
freely with a full range of determiners than do the proposal,
insistence types, but such constructions as Every shooting of lions
that we witnessed was unpleasant are so much better than similar
attempts to attach quantifiers and relative clauses to true gerundives,
as in *Every shooting lions that I witnessed, that no alternative to
lexical derivation is appropriate, given prior decisions in this
grammar about the kinds of relationships that lexical derivation may
be supposed to characterize.
593
NOM - 6k
F.
Non-Action Infinitival Tense Constraints
One set of the predicates discussed in III.D.6 permits RAISOBJ only if the verb of the sentential object is a non-action predicate (i.e. is marked [+STAT], or has PROG, PERF, or PAST in the AUX):
(102) (a) I believe that he works very hard.
(b) *I believe him to work very hard.
(c)
(d)
I believe that he is working very hard.
I believe him to be working very hard.
(e)
(f)
I believe that he has worked very hard.
I believe him to have worked very hard.
(102.f) is ambiguous between simple past tense, and perfective aspect:
(103) (a)
(b)
(c)
(d)
I believe that he worked hard yesterday.
I believe him to have worked hard yesterday.
I believe that he has worked hard all his life.
I believe him to have worked hard all his life.
The only constraint which differentiates these structures from the
RAIS-OBJ structures with verbs like expect is this restriction to
non-action predicates when they undergo infinitival reduction:
(10U) (a)
(b)
(c)
(d)
I expect that he will work very hard.
I expect him to work very hard.
I expect that he will be working very hard.
I expect him to be working very hard.
What is needed, then, in order to bring these verbs like believe (a
substantial list, including acknowledge, assume, imagine, judge, know,
maintain, suppose, think... and others which Lees (i960) analyzed as
permitting 2nd Passive" constructions, and which Kiparsky (1968)
refers to as accepting "the accusative with infinitive" construction)
into the basic pattern of infinitival derivation is some constraint
which will subject them to the same rules that expect conforms to
except that RAIS-OBJ can be permitted to occur with them only if the
conditions for stativity are met in the embedded sentence. Their
derivation is otherwise like that of "They expected him to solve the
problem" in (851). The problem is to find a way to say that with some
verbs (like expect) the rule RAIS-OBJ is optional provided that the
tense of the sentential object is future, and with other verbs (like
believe) it is optional provided that the verb of the sentential object is non-action (in the sense defined above).
59U
NOM - 65
A device which succeeds in stating the correct generalization is
for the rule of RAIS-OBJ to apply only if the matrix predicate is
not marked [+STAT REDUC] or [+FUT-REDUC]. Thus a verb of the
believe class is [+STAT-REDUC] and [+/-RAIS-OBJ]; if under the
convention of obligatory specification, the positive value is
chosen, the rule of RAIS-OBJ will apply because the verb is marked
[+STAT-REDUC]. There is no constraint on the verb of the embedded
sentence, but infinitival reduction will only occur if the predicate
is a non-action one, since RAIS-OBJ operates only on non-action
predicates if governed by a [+STAT-REDUC] verb. This is, however,
an ad hoc condition on the rule, which suggests that some insight
into the nature of the similarity between the believe class and
the expect class has been missed in this analysis. If the matrix
verb is marked [-STAT REDUC] the rule of RAIS-OBJ cannot apply.
Similarly, a verb of the expect class is [+FUT-REDUC] in the lexicon,
and [+/- RAIS-OBJ]. If the positive value is chosen, and the
matrix verb is marked [+FUT-REDUC], the rule of RAIS-OBJ will apply;
if the negative value is chosen, the rule of RAIS-OBJ cannot apply.
The verbs believe and expect differ only in the exception features
[STAT-REDUCJ and [FUT-REDUCJ.
A small subclass of the [+STAT-REDUC] predicates permits
only infinitival reduction, and only non-action complements:
e.g. consider;
(105) (a) *I consider that he is intelligent,
(b) I consider him to be intelligent.
These are marked [+/-S] (i.e. they don't have to take a sentential
object), but [+STAT-REDUC] and [+RAIS-OBJ], so that a sentential
object is always infinitivally reduced.
As noted earlier, the verbs say, rumor, claim, and repute
are like the believe class except that passivization is obligatory
after RAIS-OBJ:
(106) (a) Someone says that he is intelligent.
(b) "Someone says him to be intelligent.
(c) He is said to be intelligent.
595
NOM - 66
G.
1.
Deep Structure Constraints
1.
Tense/Aspect Constraints on the Sentential Object
a. Future
b. NonFuture
c. Stative
d. NonStative
2.
Case Constraints between Matrix and Constituent
a. Agent Identity
b. Dative Identity
Tense/Aspect Constraints on the Sentential Object
Earlier in several places (III.C.2, III.D.6, III.F) mention
has been made of the necessity to specify the tense of the sentential
object, for some predicates. Since we have a parameter already
having to do with the mood of the predicate in the sentential object
(Imperative, Indicative, Interrogative), it must be shown that the
present constraint in respect to tense is orthogonal to that one.
Consider the verb insist;
(107) (a)
(b)
(c)
(d)
(e)
(f)
I
I
I
I
I
I
insist that she take the medicine.
insisted that she take the medicine.
insist that she takes the medicine.
insisted that she takes/took the medicine.
insist that she will take the medicine.
insisted that she would take the medicine.
(107-a,b) are imperative embeddings. (107.c-f) are all indicatives;
the verb insist is factive in these instances and is compatible with
any tense or modal: all factives are, since the head item fact is.
We must consider, then, non-factive examples. Most of the predicates
that the Kiparskys (1968) label with the feature [+FUT] in fact
require embedded imperatives (Section III.C.2 above). We do not
view these as containing a future auxiliary (should, according
to the Kiparskys). But three items on their list are incompatible
with imperatives: predict, anticipate, foresee. Others with the
same property are expect, promise, stipulate, prophesy. They are
incompatible with subjunctive, and therefore [-IMPER]; but among
indicative possibilities, they are compatible only with future:
(108) (a) *I predict that he go bankrupt.
(b) *I predict that he went bankrupt.
(c) *I predict that he goes bankrupt every day.
(d) I predict that he will go bankrupt.
(e) I predicted that he would go bankrupt.
596
NOM - 67
These verbs, unlike the [+STAT-REDUC] non-action verbs above (III.F),
which are compatible with action sentential objects unless they
are infinitivally reduced, are compatible with future sentential
objects only, regardless of whether they are infinitivally reducible.
In order to take this distinction into account, then, two features
are needed with respect to stativity (a strict subcategorial feature
[+/-STAT], and a second feature [+/-STAT-REDUC] to provide for
reduction);
and two features are needed with respect to futurity,
a strict subcategorial feature [+/-FUT], to provide for the
correct selection, and [+/-FUT-REDUC] to provide for reduction.
There are, then, predicates like predict, anticipate, foresee, expect, promise, stipulate, and prophesy marked with the feature
l+FUT], which is an abbreviation, in the form of the features
[INDIC] and [IMPER] (65.c,d),requiring that the tense of the predicate
in the sentence dominated by NEUT contain the auxiliary will (present
or past, in accord with rules of tense sequence). Some of these
are also marked [+/-RAIS-OBJ], and therefore permit infinitialization—e.g. expect, for most dialects, and predict, foresee, and
prophesy for some dialects. Promise is [+FUT], [+IDENT], [-RAIS-OBJ],
as in (109):
(109) (a) I promise that I will leave.
(b) I promise to leave.
(c) *I promise Mary to leave.
(d) *I promise that Mary left.
(e) I promise that Mary will leave.
It is not clear whether there are predicates that must be
marked [-FUT]. Consider recollect, recall, remember:
(110) (a) ?I recollect that she will finish the paper
tomorrow.
(b) I recollect that she finished the paper
yesterday.
(c) I recollect that she said she would finish
the paper tomorrow.
The sense of (110.a) is that of (llO.c), suggesting that perhaps
(110.a) is a blend that should not be directly generated. There
are, however, no syntactic consequences of the type associated with
[+FUT] constraints (infinitival reduction), and the negative feature
[-FUT] is therefore not marked in the lexicon.
The predicates with adverbial ("action") gerundives, as in
(97), for which in any case we have no satisfactory analysis, appear
to be constrained to tense identical with the matrix tense:
597
NOM - 68
(ill) (a)
(b)
He
He
He
He
will continue — He will work/be working
will continue working,
continued — He worked/was working
continued working
No provision is made for this fact in the present analysis.
The feature [+/-STAT] is redundant on the strict subcategorial
feature [+/-[
AGT]] (see LEX). It is included here because of its
relation to the feature [STAT-REDUC], which constrains infinitival
reduction to non-action predicates in the sentential objects of the
believe class. Except for this syntactic consequence, stativity
would be treated in this grammar like such features as [+/-LIQUID],
a selectional feature that accounts for the unacceptability of (112):
(112) (a) ? The water broke in two.
(b) ? He chewed on the milk.
We would, then, generate (113) without the stativity feature:
(113) (a)
(b)
(c)
(d)
(e)
2.
*He was believed to depart.
*I considered him to solve the problem.
*I thought him to run the race.
*He tried to know the answer.
*He refused to be certain of the analysis,
Case Constraints between Matrix and Constituent
One feature of this type that plays a role in nominalization
is identity between the agents of the matrix and constituent
sentences. The predicates of (llU) require agent identity; those
of (115) require agent non-identity.
ink)
(a)
He tried to do it.
(b) *He tried Mary to do it.
(c)
He began to do it.
(d) *He began Mary to do it.
(e)
He continued to do it.
(f) •He continued Mary to do it.
(115) (a)
He yelled for Mary to do it.
(b) *He yelled to do it.
(c)
He advocated for Mary to do it.
(d) *He advocated to do it.
The feature [+/-AG IDENT] marks this requirement of agent identity,
and EQUI-NP-DEL applies at the appropriate point in the derivation
to erase the coreferential agent of the constituent sentence.
598
NOM - 69
A second feature, like [AG-IDENT] except that the matrix
dative is required to be identical with the constituent agent,
provides for examples like (116):
(116) (a)
(b)
I forced John to go to prison.
I commanded the sergeant to organize the troops.
This feature, [+/-DAT-IDENT], guarantees that sentences like (117)
will not be generated:
(117) (a) *I forced John that Mary leave.
(b) *I persuaded Mary that Jane go to prison.
It is possible that such nonsentences can be blocked without this
feature, since force requires EQUI-NP-DEL, a rule which would not
apply to a string like (ll7.a). But since EQUI-NP-DEL is not a
boundary-erasing rule, it is not obvious how (117.a) would be blocked
merely by the failure of this rule to apply. What the feature
[DAT-IDENT] does is guarantee identical dative and agent so that
EQUI-NP-DEL will always apply in such cases. With sentences like
(118), where [DAT-IDENT] is optional, the positive value of the
feature provides for infinitival reduction, and the negative value
for the clausal form:
(118) (a)
(b)
(c)
I warned Mary to leave.
I warned Mary that she must leave.
I warned Bill that Mary must leave.
Sentences like (119) are only apparent counterexamples to the deep
structure identity conditions [AG-IDENT] and [DAT-IDENT] because
they are derived (though the rule is not provided in this grammar)
as optional variants of the "get-passive":
(119) (a)
I
(I
(b) I
(I
H.
tried to be examined by the doctor.
tried to get examined by the doctor.)
forced Bill to be examined by the doctor.
forced Bill to get examined by the doctor.)
Indirect Questions
In section III.B we set up a feature [+/-WH-S] for embedded
interrogatives. It is necessary to distinguish, in respect to the
diagnosis of this feature, between true embedded interrogatives and
pseudo embedded interrogatives, the latter deriving from relative
clauses on deletable head nouns. The following are true indirect
questions:
599
NOM - 70
(120)
All
the
the
are
/(a)
(b)
I didn't < (c)
(d)
(e)
who left early
know
care (about)
what happened
' where they went
remember
realize
when they arrived
take into account; why they did it
.how they did it i
such sentences may be paraphrased by inserting "the answer to
question" in the blank between the column of predicates and
column of questions in (120). The following, on the other hand,
pseudo embedded interrogatives:
(121)
[{&)
(b)
I didn't (c)
(d)
1(e)
V
like
hate
recognize
suspect
deny
/
what happened
/ where they went
why they did it
The pseudo embedded interrogatives of (121) appear to involve
deletable head nouns (with appropriate morphophonemic changes) of
the form shown in (121'):
(121')
I didn't
(a)
(b)
(c)
(d)
(e)
like
Jthe thing that happened
hate
recognize / (the place to which they
suspect
went
deny
j the reason for which they}
did it
There are little-understood restrictions on the formation of pseudo
interrogatives, such as the impossibility of *I didn't like who
left early from I didn't like the person who left early, but it is
clear that their interpretation is quite different from the interpretation of true embedded interrogatives, and only the latter may be
derived as nominalizations.
The true indirect questions, but not the pseudo ones, are
subject to infinitivalization under the same conditions as other
nominalizations, namely whenever the subject of the embedded sentence
is removed from the possibility of subject-verb agreement. The only
condition that will remove it, since there is no possibility of
RAIS-SUBJ or RAIS-OBJ or FOR-INSERT with such structures, is EQUINP-DEL:
600
NOM - 71
(122) (a)
(b)
I don't know — What will I do
I don't know what I will do.
I don't know what to do.
I didn't take into account — How would I do it
I didn't take into account how I would do it.
I didn't take into account how to do it.
For all such infinitivalizations, the indirect question must be
future in its auxiliary, a constraint which is handled exactly as
with verbs like expect (Sections III.F, III.G.l). For reasons which
remain mysterious, clauses with why disallow infinitival reduction:
*I don't know why to do it.
I.
Miscellaneous Exception Features
1.
2.
3.
U.
5.
1.
TO-DEL
TO-BE-DEL
EXTRA
RAIS-OBJ-TO-SUBJ
SUBJ-SUBJ-IDENT
TO-DEL
The analysis provided for infinitialization in a wide range
of cases (e.g. those with raising of subject to object, like expect;
those with the dative erasing the embedded subject, like force;
those with the matrix agent erasing the embedded agent, like try;
those with raising of embedded subject to matrix subject, like
likely) also provides for predicates like see, watch, observet make,
help, hear... except that an ungrammatical intermediate stage is
generated:
(123) (a)
(b)
(c)
I saw — He dug a hole in the ground.
[Like expect]
*I saw him to dig a hole in the ground, [by
RAIS-OBJ, TO-INSERT]
I saw him dig a hole in the ground, [by TO-DEL]
I made him — He dug a hole in the ground.
[Like force]
*I made him to dig a hole in the ground, [by
EQUI-NP-DEL, TO-INSERT]
I made him dig a hole in the ground,
[by TO-DEL]
I helped — I dug a hole in the ground. [Like
try]
I helped to dig a hole in the ground, [by
EQUI-NP-DEL, TO-INSERT]
I helped dig a hole in the ground, [by optional
TO-DEL]
601
NOM - 72
These are analyzed, then, as perfectly normal infinitival nominalizations with the single peculiarity of to-deletion (obligatory in
most instances, optional at least with help).
2.
TO-BE-DEL
"To be" is optionally deletable in infinitival nominalizations with verbs like consider, believe, think, and obligatory with
the verb elect:
(121+) (a)
(b)
I consider him (to be) intelligent,
They elected him president.
The predicates which allow or require this deletion must be marked
with the exception feature [ +TO-BE-DEL], since it is not deletable
on any general or configurational basis:
(125) (a) I want him to be president.
(b) *I want him president.
(c) I expect him to be intelligent.
(d) *I expect him intelligent.
3.
EXTRA
Extraposition, as discussed in Section III.D.3, is a
dimension orthogonal to factivity. It is, nevertheless, a highly
redundant feature and needs to be marked as an exception feature,
either plus or minus, in only a small number of instances. All
the factive predicates that have subjectivalization of the sentential
object or instrumental allow extraposition optionally:
(126) (a)
(b)
It is significant/odd/tragic/exciting/
irrelevant...that she can't solve the
problem,
It doesn't matter/count/make sense/suffice/
amuse me/annoy me/amaze me...that she
can't solve the problem.
All the non-factive adjectival predicates with subjectivalization
of the sentential object require extraposition:
(127) (a)
It is likely/sure/possible/true/false that
she solved the problem.
602
NOM - 73
All of the non-factive verbal predicates with subjectivalization of
the sentential object require extraposition:
(128) (a) *That she solved the problem seems/appears/
happens...
(b) It seems/appears/happens that she solved the
problem.
With all examples of the types (126-128), then, extraposition is
predictable from other features. That is, extraposition from subject position is an ungoverned rule.
But extraposition from object position is governed by an
unpredictable exception feature [+/-EXTRA]. The evidence that it
is governed is cited above (II.B.5). This is a surprising fact, for
which we have no general explanation. Somehow, extraposition from
object is a dubious rule.
k.
RAIS-OBJ-TO-SUBJ
Consider now the famous examples always cited in demonstration of the distinction between deep and surface structure:
(129) (a)
(b)
(c)
(d)
(e)
(f)
John is eager to please.
John is eager — John will please one.
John is easy to please.
One pleases John — is easy.
For one to please John is easy.
It is easy to please John.
(129.a) is a straightforward instance of obligatory EQUI-NP-DEL,
and deletion of the indefinite/impersonal object one. But nothing
in the analysis so far will derive (l29.c). We can derive John is
certain to learn the secret, which depends on an early RAIS-SUBJ
rule, as discussed in section III.D.6. But here we have an otherwise similar instance, except that it is the object of the embedded
sentence which is raised to subject of the matrix sentence. (The
same distinction between easy and certain would hold under any
other analysis — IT-Replacement (Rosenbaum), or a version of the
present analysis in which (129.f) is taken as an intermediate
stage between (e) and (c).) It appears, then, that a feature
[+/-RAIS-OBJ-TO-SUBJ] must appear on adjectives like easy, difficult,
hard..., governing the same early rule of raising to subject that
is governed by [RAIS-SUBJ]. What is curious, however, is that in
other instances where an NP is raised out of a lower sentence, infinitival iz at ion is automatic because no subject remains to agree
with the verb; in this instance, the subject remains, but since the
603
NOM - Ik
only predicates which have this feature also have the feature
[+EMOT], infinitivalization takes place anyway, and provided that
the subject is indefinite/impersonal and therefore deletable, the
sentence (l29.c) turns out, by a very abstract derivation of
several steps, to have the same surface structure as (129.a):
(130) (a)
(b)
(c)
(d)
(e)
IV.
THE RULES OF NOMINALIZATION
A.
B.
C.
D.
E.
F.
G.
H.
I.
J.
K.
L.
M.
N.
A.
Easy — One pleases John.
Easy — for one to please John.
[FOR-INSERT, TO-REPLACE-AUX]
John is easy — for one to please
[RAIS-OBJ-TO-SUBJ, BE-INSERT]
John is easy — for to please.
[ONE-DEL]
John is easy to please.
[PREP-PREP-DEL]
GER
FACT-DEL
FOR-INSERT
EQUI-NP-DEL
RAIS-OBJ
RAIS-OBJ-TO-SUBJ
RAIS-SUBJ
TO-REPLACE-AUX
TO-DEL
TO-BE-DEL
ONE-DEL
THAT-INSERT
EXTRA
THAT-DEL
GER (Factive), GER (Non-factive)
Factive gerundivization applies first, and appropriate conditions
exclude non-factives from participation in this rule. Non-factive
gerundives are assumed either to be governed by a feature [+GER] or
a preposition, or to be generic alternatives of for-to constructions
(see Section III.E.3) generated by late optional rules. Only the
factives and governed gerundives are provided for in the rules
below. Adverbial gerundives (III.E.U) are treated as governed.
This rule is strictly ordered in respect to a number of subsequent
rules: it must precede FACT-DEL because "the fact of" is part of
60U
NOM - 75
the environment essential to stating the permitted gerundization;
it must precede all the rules of infinitialization, since the
"tense" category of the embedded sentence is replaced by to
unless it has already been removed by gerundivization.
1.
Schematic of GER (Factive)
the
#
NP
MOD
PROP
#
AUX
TE
(H)
(PERF)
(PROG)
(PAST)
NP
D
the
NQM
N
PREP
NP
fact
.^rW^
#
NP
NP
MOD
POSS
^PROP
AUX
ing
2.
#
(PERF)
(PROG)
Mp[s[»,
NP ^ TE (M) (PERF) (PROG)] X
The rule GER (factive)
S.I.
,X
JE[the
fact PREP
1
2
605
3
k
5
6
7
NOM - 76
S.C.
(a)
(b)
(c)
(d)
Attach [+GENITIVE] to 2
If 3 • PAST and 5 - 0, attach PERF as left sister of 6
ing replace 3 + U
T^EQUI-NP-DEL] replaced by [+EQUI-NP-DEL]
3. Notes on the rule: for discussion see Sections III.E.l, III.E.2.
U. Examples: see (88), (93).
GER (Non-factive)
Since the factive gerundive rule depends on the presence of
fact as the head item governing the actant which dominates the
nominalization, structures to which the non-factive gerundive rule
applies do not meet the structure index above, nor do the factive
ones meet the structure index below, since it is a governed rule
requiring the feature [+GER], or a preposition.
5.
Schematic of GER (Non-factive)
MOD
MOD
PROP
V
NEUT
[+GER]
Ci
NP
6.
ing
PROP
The rule of GER (Non—factive)
x(TREP|
I
V,H\
([+GERy
12
S.C.
(a)
(b)
NP
,
S
-
x
•
AUX,
3
,
U
5
ing replaces k
PEQUI-NP-DEL] replaced by [+EQUI-NP-DEL]
606
NOM - 77
7.
Notes on the rule:
III.E.2, III.E.U.
for discussion see Sections III.E.l,
8.
Examples: see (87), (89), (9^.b).
Problem. There is a major unresolved problem not discussed earlier
nor handled in this rule,in connection with EQUI-NP-DEL in gerundive
nominalizations. Consider the following examples:
(131) (a) Bill
(b) Bill
(c) Bill
(d) *Bill
imagined
imagined
imagined
imagined
that he was leaving.
himself to be leaving.
leaving.
to be leaving.
Suppose imagine is marked [+/-RAIS-OBJ], [-EQUI-NP-DEL], and
[+/-GER]. It is, like consider, [+STAT-REDUC] also. Now, if
[+GER] is chosen (l31.c) is the output. If [-GER], then there
is no way to block (131.d), since EQUI-NP-DEL will apply and then
TO-REPLACE-AUX. If it is marked [-EQUI-NP-DEL], as is the case
for verbs of the consider class, then (l31.d) will not be generated,
but neither will (131.c). Clearly within this grammar some
important generalization has been missed, since we must enter
imagine twice in the lexicon: once with [+/-RAIS-OBJ], [-EQUINP-DEL], and [+STAT-REDUC], like verbs of the consider class; and
again with [+GER] and [+EQUI-NP-DEL], like avoid.
But the problem of EQUI-NP-DEL meets a much more difficult
obstacle when it appears that we have no effective way to state
EQUI-NP-DEL at all in gerundive nominalizations. Consider the
following examples:
(132) (a)
(b)
I told Mary about seeing John,
I asked Mary about seeing John.
In (132.a) the embedded sentence is "I saw John." In (132.b) it
is, in one reading, "Mary saw John." Probably (132.b) should be
explicated in a way parallel to the explication we propose for (133):
(133) (a)
(b)
(c)
I asked him what to do.
I asked him to tell me what to do.
I told him what to do.
That is, we claim that the peculiarity in the EQUI-NP-DEL of (133.a)
results from deletion of the underlined material of (133.b), which
is completely regular as to EQUI-NP-DEL:
607
NOM - 78
I asked
him
he
tell
me
I
do wh-something
But now, in order to provide for EQUI-NP-DEL, we are introducing
deletions of strings that are difficult or impossible to recover.
Consider a more extreme case of the same sort:
(131*) (a)
Mary told me
himself that
(b) *Mary told me
herself that
about the plans for shooting
John had been laying all summer,
about the plans for shooting
John had been laying all summer.
Why is (l3*».b) bad? Because we only discover in the final relative
clause that the subject of "plan to shoot herself" must be John,
not Mary. But how can EQUI-NP-DEL come about correctly in (l3U.a)
when there is no noun present to be deleted? It is only inferred
from the relative clause that the agent of plan would be "John," if
it were present. If it were present, it would correctly delete the
subject of "John shoot himself," but there would be nothing to
delete the John of "John's plan," unless there is some sort of
totally mysterious rule that permits deletion upward from a relative
clause.
A related problem in stating EQUI-NP-DEL in gerundive
nominalizations resides in the general fact that nouns have
subjects (i.e. AGT or DAT in deep structure) which often have to
be inferred at two or three removes, and yet which can bring about
EQUI-NP-DEL of noun subjects of clauses embedded as cases under
the head noun. Thus:
(135) (a)
(b)
He has no objections to studying French,
He spoke at some length about the various
objections to studying French that had prevented
him from doing it in high school
Clearly, even if the POSS of "objections" in (135.a) is relatively
accessible as the matrix subject, it is thoroughly buried in (135.b);
yet in both cases the deleted subject of the gerundive may be "he"
under one reading. It is possible, however, that such readings are
wrong: it may be in both examples that the correct reading is either
subjectless or perhaps one's (studying French). But the problem
remains in examples like (136), where the indefinite subject, or
608
NOM - 79
or subjectless, interpretations are hard to defend:
(136) (a)
(b)
(c)
The interest in visiting Las Vegas that Mary
displayed...
The addiction to smoking pot that caused John's
death...
The exhaustion from overindulging in sex that
eventually ruined his eyesight...
In sum, we cannot yet state the conditions for EQUI-NP-DEL
in gerundive nominalizations; ve have included the regular instances
("He avoided leaving") in the regular EQUI-NP-DEL rule, along with
the ones that produce infinitives, avoiding the problem of imagine
by a form of double-entry book-keeping; and we suggest, in our
discussion of the rule, a way to handle the almost-regular examples
like "She has no objections to studying French"; but examples like
(13M and (136) are beyond these rules.
B.
FACT-DEL
This rule deletes the noun fact, its determiners and any
prepositioned modifiers (e.g. very in The very fact of his having
crashed proves it), and the preposition of that marks its object.
The rule must precede FOR-INSERT in order to guarantee that those
predicates which are both factive and emotive can appear in either
that-S or for-to-S constructions (e.g. It was a tragedy that he did
that, It was a tragedy for him to do that); the latter possibility
would be blocked if FOR-INSERT preceded this rule. It must precede
EQUI-NP-DEL to guarantee getting I regretted solving the problem
but not #I regretted my solving the problem, since EQUI-NP-DEL does
not apply across an intervening head noun fact; from this it follows
that these rules claim that I regretted the fact of my solving the
problem is grammatical, but that *I regretted the fact of solving
the problem is not (unless it is from indefinite-NP-deletion).
1.
Schematic of FACT-DEL
NP
»
609
(NP) [Pruned]
NOM - 80
2.
The rule of FACT-DEL
S.I.
NP
X
[ the X Fact of]
II
l
I
S.C.
Np
[S] X
l
Erase 2
Condition:
The rule is optional unless 1 contains the
feature [-FACT-DEL], in which cast it cannot apply.
3.
Notes on the rule: A general convention prunes the NP which is
exclusively dominated by another NP. The condition on the rule
is to prevent deletion of fact with a small number of predicates
which do not permit it: *He contemplated that she was leaving/
He contemplated the fact that she was leaving.
k.
Examples:
C.
FOR-INSERT
see (50)-(51+), (57), (6l).
The rule must follow FACT-DEL, since a sentential object of fact
may become object of a [+EM0T] predicate after fact is deleted and
thereby subject to this rule, and it should also be ordered prior to
EQUI-NP-DEL in order to guarantee that "It scared him for Mary to
Jump" and "It scared him to Jump" will have parallel derivations—
i.e. both from [+EM0T], with EQUI-NP-DEL in the second instance,
giving "It scared him for-to jump", with for deleted by the general
PREP-PREP-DEL rule. The reverse order would derive "It scared
him to jump" by EQUI-NP-DEL, without FOR-INSERT applying at all, or
perhaps applying vacuously. It is convenient, but not mandatory,
to order the rule prior to the general case placement rules,
since with that ordering the governing item is to the left of the
sentential complement, whether that complement is subsequently to
be placed to the left of the predicate, as its subject, or to
the right, as its object.
1.
Schematic of FOR-INSERT
MOD
V
[+EM0T]
V
[+EM0T]
c
,*
(PREP)
NP
(
i
NP
NP
S
S
1
MOD
|
PROP
610
for NP
[+PREP]
MOD PROP
NOM - 81
C± = NEUT or INS
PREP present if non-factive; in the factive instances, it
has been deleted by FACT-DEL
[ (PREP)
S.C.
(a) 2
(b) Attach
for
[+PREP]
NP S
[ [# NP X
as left sister of k
3. Notes on the rule: The optionality of the rule is regulated in
the lexicon, so that desirable, e.g. is [+/-EMOT] to provide for
both "It is desirable that he do it'V'It is desirable for him to
do it."
U.
Examples: (70), (71).
D.
EQUI-NP-DEL
This rule must precede RAIS-OBJ, since that rule raises the
subject of the embedded sentence up into the object of the matrix,
where reflexivization would be expected (*He wanted himself to go)
rather than deletion (He wanted to go): i.e., EQUI-NP-DEL erases
the subject of a lower S on the basis of a coreferential NP in the
higher S. The rule must follow FACT-DEL in order to account for
He forgot about having done it, and it must follow GER to account
for He insisted on doing it. The rule operates with a set of priorities,
such that a coreferential dative in the higher S has first erasure;
in the absence of a coreferential dative a coreferential agentive
may bring about the erasure. This priority principle, for which
we can provide no explanation, implies that the derived structure
is always unambiguous, i.e. that the deleted item is uniquely
recoverable. With all instances that result in infinitialization
this appears to be true: such types as He persuaded me to leave,
He wanted me to leave, He told me to leave. He expected to leave,
He taught her how to do it, etc. are unambiguous. There are
611
NOM - 82
examples with gerundives in prepositional phrases, however, which
are ambiguous: He told her about solving the problem, where one
sense is factive ("He told her about the fact that he had solved the
problem"), the other sense apparently non-factive ("He told her how
to solve the problem"). In the first sense, the wrong item performs
the erasure (the agentive he, not the dative her): in the second
sense, the dative performs the erasure, and the sense is correct
if we assume a subjunctive in the embedded sentence ("He told her
about - she SJC solve the problem"). A priori, one feels that the
second sense has a dummy manner nominal that has been deleted:
He told her about - (a way of) - she SJC solve the problem - He told
her about (a way of) solving the problem, which provides some explanation of the fact that it paraphrases He told her how to solve the
problem. With this possibility of a source for the second sense
in mind, we may reexamine the problem of the first sense in an
example like He argued with her about reporting the accident,
which seems ambiguous as between "they report the accident," "the
fact that he had reported the accident," and "the fact that she
had reported the accident." If He argued with her comes from He
and she argued..., one reading would be explained, but the ambiguity
would not be, since He and she argued about reporting the accident
clearly does not have either of the other interpretations. From
such examples we conclude that the dative-agentive priority erasure
principle is valid, if at all, only for nominalizations directly
dominated by the actant NEUT in the same case frame as DAT and AGT.
This does not explain the difficult examples above with about: it
merely sets them aside for some different principle, or some modification of this one, to explain, (it sets them aside on the assumption
that about HP in tell about NP and argue about NP are instances of
some actant other than NEUT, perhaps Associative"; at any rate a
case can be made from "tell something about" and "argue the decision
about" that they are not ordinary neutral objects marked with about.)
A second problem has been alluded to above in the discussion
of the gerundivization rule: namely the fact that in some kinds of
sentences the rule of EQUI-NP-DEL seems to apply transparently
through noun heads which directly govern the embedded sentence.
(137) (a)
(b)
(c)
Mary has a certain fondness for telling lies.
I have no objection to studying French.
I take great pride in working hard.
It may perhaps be argued that "have fondness" = "be fond", "have
objections" = "object", and "take pride" = "be proud" or the like;
but there are grave difficulties in the way of such a proposal.
Assuming that such phrases are neither lexical units nor transformationally derived, the rule of EQUI-NP-DEL must see through them
to the subject NP: i.e. such nouns are "transparent" in some quite
unclear sense, for this rule - this fact is left unformalized in
612
NOM - 83
the rule as formulated below.
1.
Schematic for EQUI-NP-DEL with erasure by preferential Dative
(the circled NP's are coreferential)
SEP
NP PREP MH
I
ANCNOM
S
(N£) MOD
PREP
M
D
PROP
J
[+DAT]
MOD
NEUT
EP
NP
DAT
PREP
AGT
NP
D NOM
MOD
PROP
N
[+DAT]
613
PREP
HP
NOM - Qk
Schematic for EQUI-NP-DEL with erasure by coreferential agentive
(the circled NP's are coreferential)
MOD
NOM
(NP) MOD
PROP
N
[+AGT]
MOD
MOD PROP
2.
Rule for EQUI-NP-DEL
NP
S
S.I.
X
[
[ NP X]
S.C
Erase 2
Condition:
DAT
AGT
[X NP] X
[X NP]
2 • 3, or if 2 # 3 or if 3 is null, then 2 = k,
6lU
NOM - 85
3.
Notes on the rule: see discussion in Sections II.B.6, III.D.U,
III.D.5. Examples of the type He screamed to jump perhaps
should be taken as [+EMOT], i.e. He screamed for someone to
Jump - they may achieve infinitivalization by the [+EMOT]
route, rather than by the EQUI-NP-DEL route: this is borne
out partially by the fact that *He screamed to Mary to jump
is ungrammatical, whereas He screamed to Mary for her to jump
is well-formed.
U.
Examples: (79) - (83).
E.
RAIS-OBJ
This rule applies before the early objectivalization rule,
to which it is an optional alternative for most predicates, the
former rule being inapplicable if this one has applied. It takes
the subject of an S dominated by NF and attaches it as right
sister of the V in the immediately dominating proposition, i.e. it
makes it the object of the matrix verb. The optionality of the
rule is determined by the convention of obligatory specification
which permits the selection of either plus or minus on the feature
[RAIS-OBJ] except for a few predicates like consider which are
plus only.
1.
Schematic of RAIS-OBJ
2.
The rule RAIS-OBJ
S.I.
X
I
S.C.
S
[ » MOD
1
(a)
(b)
pROp[
**[
V PREP
I
2 3
I
,
[ 0
1
^T
NP X
5 6
Attach 5 as right sister of 2
Erase 3 and 5
Condition:
2 contains the feature [+RAIS-OBJ] and does
not contain the features [-STAT-REDUC] or
[-FUT-REDUC].
615
NOM - 86
Notes on the rule: for discussion see Sections III.D.c.b,
III.D.7, III.F. PREP (3) is erased because the general
objectivalization rule, which would have erased it, is no
longer applicable.
k.
Examples:
F.
RAIS-OBJ-TO-SUBJ
see (85), (85').
This rule is disjunctively ordered with respect both to
RAIS-SUBJ and the general case placement rules. It takes the
object of an S dominated by NP and attaches it as right sister
of the boundary of the next higher S - that is, it makes it the
subject of the matrix sentence. The optionality of the rule is
determined by the convention of obligatory specification which
permits the selection of either plus or minus on the feature
[RAIS-OBJ-TO-SUBJ].
1.
Schematic of RAIS-OBJ-TO-SUBJ
S
** MOD
2.
0
PROP
The rule RAIS-OBJ-TO-SUBJ
S.I.
t-
S.C.
c
&
X
(a)
(b)
1
NP
[ 0
MOD
I
L
2
pROp[X
-V,
[
L
g[X
V NP X
J
6 7
Attach 6 as right sister of 2
Erase 6
Condition: k contains the feature [+RAIS-OBJ-TO-SUBJ]
3.
Notes on the rule: for discussion see Section III.I.U.
k.
Examples: see (129), (130).
6l6
NOM - 87
G.
RAIS-SUBJ
This rule applies before the early subjectivalization rule.
It takes the subject of an S dominated by NP and attaches it as
right sister of the boundary of the next higher S—that is, it
makes it the subject of the matrix sentence. The rule is an
optional alternative to the general subjectivalization rule, the
latter being inapplicable if this one has applied. The optionality
of the rule is determined by the convention of obligatory specification which permits the selection of either plus or minus on
the feature [RAIS-SUBJ].
1.
Schematic of RAIS-SUBJ
2.
The rule RAIS-SUBJ
S
S.I. X
# MOD
L
I
1
S.C.
(a)
(b)
L
2
PROpt X "l 8[l
J . I
U
5
W X
6 7
Attach 6 as right sister of 2
Erase 6
Condition:
3.
Notes on the rule:
and III.D.7.
k.
Examples:
H.
TO-REPLACE-AUX
h contains the feature [+RAIS-SUBJ]
for discussion see Sections III.D.6.a,
see (8Uc), (8U').
The rules which set the stage for this rule—i.e. which
establish the conditions necessary for it to apply, namely the
617
NOM - 88
condition that there be no NP on which subject-verb agreement can
be hinged—have applied in the order presented above, except for
the rule which assigns accusative case to the NP's after prepositions
and verbs (see PRO paper), which applies also before this rule.
RAIS-OBJ has removed the erstwhile subject of the sentential
object of verbs of the expect class; RAIS-SUBJ has removed the
subjects of the sentential objects of predicates of the likely
class, and also of the "II Passive" class; FOR-INSERT has provided
the condition for assigning accusative to the subject of sentential
objects of the [+EMOT] class.
1.
Schematic of TO-REPLACE-AUX
r
(for NP)
CTE
=>
AUX
(M)|
(SJC
j
PROP
(PERT)
(for NP)
(PROG)
AUX
to
(PERF)
Rule for TO-REPLACE-AUX
NP S
CTE (M i)
S.I. X
[ [(forNP)
fsjC J (PERF) (PROG)
S.C.
(a)
(b)
PROP
(PROG)
X
to replaces 2.
attach PERF as right sister of 2.
3.
Notes on the rule: The rule must apply after subjectivalization,
since otherwise the subject with which the verb would agree
would still be under PROP. For further discussion, see III.D.
k.
Examples:
I.
TO-DEL
(69), (79), (8l), (8U), (85), (86).
1. Schematic of TO-DEL
PROP
PROP
»
V
[TO-DEL]
[+TO-DEL]
618
NOM - 89
Ride for TO-DEL
s.l.
x
i[lJ!0,jJEL]
PROp
NPr
ar
l
AUXr
Si
b
t # AUA[to X] V
1
1
1
S.C.
3.
2
1
3
Erase 2.
Notes on the rule: for discussion see III.I.l.
Examples:
J.
X
i
(123).
TO-BE-DEL
The be which is deleted by this rule comes either from
the base as a V (with a following NP), or is supplied by the early
rule of BE-SUPPORT (with adjectives). The rule does not delete
be from PROG (i.e. the auxiliary be)t which in fact is still simply
PROG at this stage in the derivation and therefore not available
for deletion.
1.
Schematic of TO-BE-DEL
PROP
PROP
NP
I
[+TO-BE-DEL]
[+TO-BE-DEL]
S
NP
PROP
I
X
2.
Rule of TO-BE-DEL
PROP
S.I.
1
NPr S
[+TO-BE-DEL]
1
S.C.
[X to
J
2
Erase 2 + h
3.
Notes on the rule: for discussion see III.I.2.
h.
Examples:
(12H), (125).
619
PROP
L
[
be X
I
3
k
5
NOM - 90
K.
ONE-DEL
The deletion of the indefinite/impersonal one can only
occur in for-infinitival or POSS-ing constructions derived from
them; and only when these are subjectivalized or essive. The
appropriate deletion in infinitives linked by the copula is not
provided for here, since the derivation of such nominalizations has
not been provided for in this grammar.
There is some reason to believe that sentences like "to
know her is to love her" are derived from conditional sentences.
In any case, they provide a special problem for this grammar,
since we have no natural way to explain why they are infinitives
at all, there being not [+EM0T] governing item in the fuller form
"For one to know her is for one to love her."
1.
Schematic of ONE-DEL
PROP
for
2.
Rule for ONE-DEL
S.I.
X
NP
[
S
[#
for NP
AUX
[to X
12
S.C.
Erase 2
Condition:
3.
3
The rule is optional.
Notes on the rule: the rule as it stands is useless for all
examples like "To know her is to love her", since no provision
is made for them. For examples like "It is amusing to collect
butterflies", however, the rule does provide. Since generic
gerundives are assumed to derive in turn from these infinitivals
(i.e. "To collect butterflies is amusing" is taken to be the
source of "collecting butterflies is amusing"), though not
provided for in these rules, there are necessarily no examples
620
NOM - 91
of one + POSS deletion; the ungrammatically of "One's collecting
butterflies..." is explained in this way. These infinitivals,
in turn, may derive from conditional sentences in ways we do
not yet understand.
L.
THAT-INSERT
This rule must be placed quite late in the grammar; at
least after relativization (for reasons see REL paper). The
conditions for its operation will obtain at any middle-to-late
stage in the derivation. All that is really needed is to be able
to identify an S dominated by NP, where the AUX of the S still
contains tense, and the S still has a subject.
1.
Schematic of THAT-INSERT
NP
0
that
NP
MOD
PROP
AUX
I
TE
2.
The rule of THAT-INSERT
S.I.
X
S.C.
M.
HP. Sr
•r[
[ # HP
AUX.
[TE
J
U
2
12
3
X
J
Attach that as right sister of 2.
EXTRA (from Subject and Object)
Extraposition is extremely general and applies not only to
nominalizations but also to relative clauses. The rules below are
specified only for nominalizations, since the conditions under which
extraposition is permitted for relative clauses are more restricted
than those for nominalizations, and not as well understood.
1.
Schematic of EXTRA (from Subject)
S
it
[-PRO]
621
NOM - 92
2.
Rule of EXTRA (from Subject)
S.I.
X S[
12
S.C.
(a)
(b)
NP
[S] MOD
3
1*
PROP ]
5
X
6
Attach 3 as right sister of 5.
it_ replaces 3.
Conditions:
(l)
(2)
Obligatory if 5 dominates [-TRANS,-FACT];
k 4 ing + X
3.
Notes on the rule: the it_which replaces the extraposed
sentence has the feature [-PRO] because it is non-anaphoric;
it is, however, still dominated by NP in order to participate
in verb agreement. The first condition stated is for non-factive
intransitives like seem, happen. The second condition blocks
extraposition of gerundives.
U.
Schematic for EXTRA (from Object)
V
[•EXTRA]
5.
Rule for EXTRA (from Object)
PROP.
NP. .
S.I. X
[V
[S]
[+EXTRA]
12
3
h
S.C.
(a)
(b)
X
5
Attach k as right daughter of 2.
It replaces k.
6.
Notes on the rule: this is "vacuous extraposition", obligatory
with verbs like hate, like, optional with factives like prefer,
regret. For discussion see III.D.3. Note that the rule feature
[+EXTRA] is redundant on the feature [+FACT] and does not have
to be lexically specified, except for hate, like, and the seem/
appear class.
N.
THAT-DEL
This rule optionally deletes the item that which was inserted
by the rule THAT-INSERT, but only if the NP dominating the S from
which that is deleted is not a subject, and only if the head V is
non-factive. That is never deletable after a Noun head.
622
NOM - 93
1.
Schematic for THAT-DEL
PROP
PROP
V
[-FACT]
NP
V
[-FACT]
NP
|
that.•..
2.
Rule for THAT-DEL
S.I.
X
PR0P
[
V
[-FACT]
NPrS,
'["[ that X
-i
1
S.C.
2
3
Erase 2.
Kay 1969
623
62k
INTERROGATIVE
Contents
Page
I.
BIBLIOGRAPHY
1
II.
INTRODUCTION
1
A.
B.
III.
Accepted Analyses
1. The Analysis of AUX
2. The Triggering of Interrogative (and
Other) Transformations
1
1
Analyses Not Generally Accepted
1. Q as a Separate Trigger
2. Q as the Only Trigger
3. WH in Questions and Relative Clauses as
One Morpheme or Two
h. Attachment Transformations
5. Indirect Questions
6. Alternative Questions
2
2
k
2
5
7
7
7
THE DERIVATION OF INTERROGATIVE STRUCTURES
10
A.
10
10
11
12
Ik
B.
Alternative Questions
1. Conjunction Spreading
2. WH Spreading
3. AUX-Attract ion
U. WH-Deletion
5. Reduced Alternative Question (including
yes-no questions)
15
WH Questions and Other Question Types
19
1. WH Question Words
19
2. WH Fronting
19
3. Tag Questions
20
k. Negative Questions from Tag
2h
5. Questioned Quote (Including Echo Question) . 25
6. WH-Questioned Quote
27
7. Declared Quote
31
625
INTERROG
I.
BIBLIOGRAPHY
Chafe, W. (1968) "English Questions"
Chomsky, N. (1957) Syntactic Structures
(1958) "A Transformational Approach to Syntax"
Elliot, D. (1965) "Interrogation in English and Mandarin Chinese"
Katz, J. and P. Postal (196U) An Integrated Theory of Linguistic
Descriptions
Klima, E. (196Uc) "Negation in English"
Kuroda, S-Y. (1965b) Generative Grammatical Studies in the
Japanese Language
(1966a) "Attachment Transformations"
Lees, R.(1960a) The Grammar of English Nominalizations
(l960e) Review of Interrogative Structures of American
English" (by D. Bolinger, 1957)
Malone, J. (1967) "A Transformational Reexamination of English
Questions"
Rosenbaum, P. (1966) IBM Core Grammar
Ross, J. (1967a) "Auxiliaries as Main Verbs"
(1967c) Constraints on Variables in Syntax
II.
INTRODUCTION
A.
Accepted Analyses
1.
The Analysis of AUX
Chomsky (1957) proposed the following analysis of the node
AUX:
(1)
AUX
-*
C (M) (have + en) (be + ing)
[where C = Tense, M = Modal]
This analysis, as Chomsky showed, allows for a simple and uniform
account of the behavior of auxiliaries in interrogative, negative
and emphatic structures.
Recently (e.g. in Ross (1967a) it has been suggested that
the material to the right of the arrow in (l) does not represent
the proper deep-structure analysis of AUX; but the general adequacy
of (l) as an account of the structure of AUX that is relevant to
the interrogative, negative and emphatic transformations has not
been seriously challenged. In the present grammar, we assume an
analysis of AUX similar to Chomsky's (cf. Base Rule 3), but leave
open the question of whether this analysis represents a deep or
a deepest, structure.
626
INTERROG - 2
2.
The Triggering of Interrogative (and Other) Transformations
Katz and Postal (196U) suggest that projection rules which
ascribe meaning to transformations can be dispensed with in the
grammatical theory if certain transformations that were considered
to be optional (cf. Chomsky (1957)) are instead obligatorily
'triggered' by an optional dummy node in the P-marker (pp. 79117). Katz and Postal support their suggestion with both semantic
and syntactic arguments. The semantic arguments have to do with
synonymity, paraphrase relations and the simplification of the
projection rules. The syntactic arguments are generally along
the lines of contextual restrictions which distinguish between the
products of certain transformations and their previously-assumed
sources (e.g. between interrogative and declaratives), and "explanation" of previously unmotivated rules.
The triggering of T-rules which change meaning by a dummy
node in the P-marker has been accepted by most generatively-oriented
linguists.
B.
Analyses Not Generally Accepted (or at least not incorporated
into this grammar)
1.
Q as a Separate Trigger
In the work cited already, Katz and Postal assume two
triggers for the interrogative: (l) Q, which is parallel to NEG
for negation and IMP for imperatives and (2) WH, which is a "scope
marker" for Q, and is a constituent of an Adverb (VH-either-or)
in the deep structure underlying yes-no questions , but a constituent of a Determiner in the Deep structure underlying WH questions.
It is the Q that, according to their analysis, triggers AUX
inversion (and WH fronting), carries the various features for contextual restrictions, and, in the semantic interpretation, accounts
for paraphrase relations.
In their Justification for the node Q, Katz and Postal
propose the following arguments:
a.
Semantic Argument:
Q accounts for the paraphrase relation that holds between
the questions in example (2) below, and the respective sentences
in example (3):
627
INTERROG - 3
(2) (a)
(b)
(c)
Did Bill see John?
Who saw John?
Who(m) did Bill see?
(3) (a)
(b)
(c)
I request that you answer:
I request that you answer:
I request that you answer:
"X Bill saw John."
"X saw John."
"Bill saw X."
"where X (in (3.a)) is one of a special class of sentence adverbs
including yes, no, of course, etc." (p. 85).
b.
Syntactic Arguments:
(i) There is a class of sentence adverbials that cannot occur with
yes-no questions, though they can occur in declaratives and in tagquestions: e.g.,
(U) (a)
[Certainly]
/Perhaps
> he is a doctor.
I Probably
(b) |*Certainly j
<*Perhaps
/ he is a doctor?
•Probably J
{certainly
perhaps
> a doctor, isn't he?
probably
(iij Some negative preverbs do not occur in questions: e.g.,
(5) (a) He hardly/scarcely eats.
(b) *Does he hardly/scarcely eat?
For some speakers, examples like (5.b) appear to be grammatical in a
suitable context.
(iii) Some preverbs can occur in questions but not in the corresponding
statements: e.g.,
(6) (a) *He ever eats.
(b) Does he ever eat?
(That is, some-any alternation, of which sometimes-ever alternation is
a special case, is tied to questions (and negatives, etc.).
(7) (a)
(b)
You have some bread.
Do you have any bread?
628
INTERROG - k
(iv) Katz and Postal also argue, although mostly by implication,
that the trigger nodes are in some sense an explanation for the inversion of AUX and the subject and for the fronting of WH, while an
optional question transformation gives no reason for such transformations. One could, that is, equally well expect any other kind of
operation in an-optional transformation, but the trigger nodes can be
said to "attract" both AUX and WH. In general however, the inversion
of AUX depends on the sentence-initial position of any [+AFFECT] morpheme (in the sense of Klima, 196U), including NEG and WH; and since
the fronting of WH-elements is common to both interrogatives and
relatives, it cannot be explained by the presence of Q.
There is one major problem with the analysis proposed by Katz
and Postal: if Q and WH can be independently chosen, strings containing
only a WH will not yield a surface structure. Katz and Postal propose
that such strings are, in any case, necessary for relative clauses and
indirect questions, (in our view, the WH in relative clauses not only
shows different syntactic behavior (cf. Section II.B.3 below) but is also
predictable, and should for the latter reason not be in the deep structure at all.) Presumably, then, some kind of "blocking" transformation
will be required in cases where an S dominating WH but not Q is
generated in non-embedded position.
2.
Q as the only Trigger
Malone (1967) proposes a trigger Q for both yes-no questions
and WH questions but no separate WH trigger. The difference between
yes-no and WH questions, according to Malone's analysis, depends on
where the Q is attached: if it is directly dominated by S, (i.e.
attached to the ART of the NP questioned) a WH question will result,
(in other words, Malone's Q is equivalent to Katz and Postal's WH.)
In addition, Malone has an "internal valence" and an "external valence",
the former to account for the re-ordering in the surface structure of
questions, the latter to account for interrogative intonation.
Leaving the problem of valences aside for the moment, it seems
certainly desirable to have only a single trigger. As was indicated
above, if Q and WH can be independently chosen, structures containing
only the latter will not yield a surface structure. Furthermore, the
semantic and syntactic characteristics that Katz and Postal attribute
to their Q may equally well be attributed to their WH (Malone's Q).
(in our analysis, which makes use of a single interrogative trigger,
we use the symbol WH for this trigger. We interpret WH as a feature
that may occur either on the conjunction or_ or on the Determiner of
an NP. In the former case, the resultant sentence is an alternative
question, which, under certain circumstances, may be reduced to a yesno question. In the latter case, the resultant sentence is a WH
629
INTERROG - 5
question. Where yes-no questions and WH questions show different
syntactic characteristics, the differences may be associated with
the position of the WH feature in the underlying structure.)
Turning now to the Internal and External Valences proposed by
Malone, it appears that an analysis that uses both Valences and Q
proliferates triggers needlessly. That is, Malone reduces the two
triggers used by Katz and Postal to one, but then introduces two more
of his own, Of these two, Internal and External Valences, the Internal
Valence provides for syntactic inversion and thus corresponds closely
to the Q of Katz and Postal. In effect, Malone's analysis is the same
as that of Katz and Postal with respect to Q and WH except for the
labels.
"External Valence" is intended to provide for intonation in
questions, specifically the differences between yes-no and WH questions,
and between echoic and non-echoic questions. Syntactically, however,
the assumption of a valence does not explain the differences in intonation, because the difference between the echoic and non-echoic questions
is due to the fact that the former are embedded in a sentence of the
form: 'did you say, "X?"'. Echoic questions are thus direct quotations
and behave syntactically and intonationally exactly like other direct
quotations. Malone's analysis however, cannot exhibit this parallel
in the behavior of echoic questions and other quotations. Because
Malone's analysis fails to capture this generalization, his positing
of an External Valence is not explanatory. If there is also a way to
explain the difference in intonation between yes-no and WH questions
without having to posit a valence (or a Q), then we could do without
valences altogether. The basis for such an analysis does, in fact,
exist in the form of alternative yes-no questions. Malone's analysis
with valences is insufficient for these in any case, because it would
have to show how alternative questions relate to both yes-no and echoic
questions (according to Malone, all three types have the same External
Valence.)
3.
WH in Questions and Relative Clauses as One Morpheme or Two
Katz and Postal (196U) and by implication Chomsky (1957) and
Lees (1960a), as well as others who have dealt with interrogation and
relative clauses, have analyzed the WH in questions and relative
clauses as the same morpheme. There are several factors that argue
against such an analysis, and thus for an analysis which describes
them as two different morphemes. The first of these can be summarized
by saying that the WH in Rel clauses is always predictable. That is,
given the configuration unique to a Rel clause, plus the requisite
identity (NOM, NP, or N, depending on the analysis), then the grammar
will obligatorily delete the identical head item and attach the feature
[+WH] under the ART node.
630
INTERROG - 6
The relative pronoun is thus derived in much the same way as are
other pronouns, i.e., by the syntactic process of pronominalization,
and thus need not occur in the deep structure at all.
The rest of these factors fall under the heading of "different syntactic behavior"; there are several of these which will
be discussed below.
a.
Pied Piping
Ross (1967c) notes that there is a constraint on Rel clauses
(Pied Piping) which does not apply to WH questions. It is for this
reason that we get sentence pairs like:
(8) (a)
(b)
...the table of which the leg was broken.
...*the table of which what was broken
where (8.b) is ungrammatical because Pied Piping does not apply to
interrogatives.
b. Ross also noted (op. cit.) that questions, but not Rel clauses,
may contain an "existential" there is phrase. Thus, we get:
(9) (a)
(b)
c.
Who is there in my bedroom?
I didn't know the young woman who there was in my
bedroom.
#
The WH-word in questions is normally analyzed as:
(10)
NP_
DET~~"
WH
"N
ART
some
j one
I thing
s
t reason ?
The configuration yields who, what, why, how, etc., in the
surface structure. Two facts about this analysis are noteworthy. The
first is that there are a number of question words, but only two relative pronouns (who and which). The second is that the noun in (10)
must be [+PR0], and the ART [-SPEC], in order to yield the proper
semantic interpretation of interrogatives. The ART in Rel clauses,
on the other hand, is only [+SPEC] in the NOM-S analysis (cf. REL
section). If the noun in the question configuration is [-PRO], then
the ART can be either plus or minus SPECIFIC to provide for the contrast
shown in (11):
631
INTERROG - 7
(11) (a)
(b)
Which boy did he see?
What boy would wear an outfit like that?
From the foregoing discussion it seems clear that the WH
in questions and in Rel clauses should indeed be two different
morphemes, and that the latter should be transformationally introduced.
k.
Attachment Transformations
Kuroda (1965b and 1966a) claims that certain sentence adverbials,
among them WH, can occur only once in each #S#. They are then placed
into the proper positions and attached to the proper node by what
Kuroda calls "attachment transformations." The merits of this analysis
with respect to adverbials like just, even, etc. do not concern us
here. What does concern us, is the fact that his analysis forces him
to ascribe the same deep structure to sentences like:
(12) (a)
(b)
(c)
Who saw some /thing I ?
Lone
J
What did someone see?
Who saw what?
Since we have tried to maintain wherever possible the KatzPostal hypothesis that semantic differences should correspond to deepstructure differences, the deep structure introduction of WH as a
feature on individual determiners seems preferable. Furthermore,
(l2.c) would appear to disconfirm the claim that WH is one of these
elements (if indeed there are any) which can occur only once per #S#.
In any case, WH is certainly not freely attachable to nearly any
constituent, as are, e.g., only and every.
5.
Indirect Questions
Katz and Postal (op. cit.) claim that one justification for
Q as a trigger lies in the fact that it "attracts" the AUX, and that,
therefore, the difference between direct and indirect questions can
be expressed by not having a Q in the latter, since they do not have
AUX attraction. It seems to us that this fact can be captured fairly
simply by having AUX attraction a last-cyclic rule, and hence there is
no need for the node Q with indirect questions.
6.
Alternative Questions
The existence of alternative questions such as:
(13) (a)
(b)
(c)
Are you coming or aren't you?
Will John eat fish or won't he?
Should I give her a present or shouldn't I?
632
INTEPROG - 8
has been recognized for some time. In fact, Katz and Postal utilized
the alternative question structure to derive indirect yes-no questions
of the type:
(lU) (a)
(b)
Does he know whether John is home?
He doesn't know whether John is home.
which they then analyzed as being related to the respective sentences
in (15):
(15) (a)
(b)
Does he know the answer to the auestion:
"X either John is home or John isn't home"?
He doesn't know the answer to the question:
"X either John is home or John isn't home."
We believe that the Katz and Postal analysis of indirect
questions (yes-no) is correct. In fact, we suggest that all yes-no
questions are derived from alternative questions. Such an analysis
has the following advantages:
a.
b.
c.
d.
It unifies the derivation of direct and indirect yes-no questions.
It automatically accounts for the intonation contour in yes-no
questions and thus obviates the need for Malone's External
Valence.
It eliminates any need for the trigger Q, since the difference
between yes-no and WH questions is accounted for by deriving
yes-no questions from alternative questions.
It makes yes-no questions part of a larger pattern of alternative questions like in (l6):
(16) (a)
(b)
(c)
Did John come to the party,or did he stay home?
Are you cooking dinner, or do we eat out?
Is Fred going to marry Abigail, or is he going
to stay a fool all his life?
This analysis of yes-no questions does not require the creation
of any new rule apparatus, since that part of the derivation that has
to do with two sentences is available in the conjunction rules, and the
part of the rules particular to questions is needed for WH questions in
any case. Rules deleting one of a pair of identical sentences, or
portions thereof, are also needed elsewhere in the grammar.
Lastly, it would appear that the analysis proposed here not
only fits the semantic analysis given in Katz and Postal, but extends
that analysis, since according to the analysis proposed here, the
sentence corresponding to (3.a) is:
(3') (a)
I request that you answer: "Yes, Bill saw John,
or no, Bill didn't see John."
633
INTERROG - 9
Turning now to the co-occurrence restrictions that Katz and
Postal ascribe to the node Q, we note that they are of three kinds:
a.
b.
c.
a class of sentence adverbials: certainly, perhaps , probably;
some negative preverbs: hardly,...
some preverbs: ever, and some-any alternations
The sentence adverbials do not really constitute a clear case,
because some of them (e.g. probably) are acceptable in questions,
while others (e.g. certainly) are not, as shown in the following:
(a)
Will he /Pliably
I
(b)
When will he I Probably
(^certainly ;
come?
\
C "certainly )
(c)
come?
Why did hejProbably
)come?
I "certainly )
For this reason, it seems to us that there is not a grammatical
co-occurrence at work here, as Katz and Postal think, but a semantic
incompatibility. In that case, we do not want to ascribe the
incompatibility to any one node, but we want to have the semantic
component declare the whole sentence as unacceptable.
As for the preverbs mentioned in (b) and (c) above, it appears
that the restrictions that were ascribed to Q hold true for all
questions, as well as for a number of other sentence types. Thus,
preverbs of the type ever, as well as some-any alternants, occur whenever a sentence is marked as containing [+AFFECT]. This feature is
part of negation and several other words having the negative in their
semantic interpretation, e.g. scarcely (cf. NEG), as well as being
part of interrogation. Preverbs of the type hardly, on the other
hand, are negative in the same way as scarcely as can be seen by
applying Klima's tag-question test:
(18)
,
He hardly ate,
£ld he
I "didn't he
These negative preverbs have various other co-occurrence restrictions,
e.g. they cannot occur in imperatives; for example:
(19)
"Hardly eat!
nor with some verbs taking an embedded imperative that ends up in the
surface structure predicate; as in,
63U
INTERROG - 10
(20) (a) *I persuaded him to hardly eat.
(b) I expected him to hardly eat.
In all, then, it seems to be as possible to ascribe the co-occurrence
restrictions of types (b) and (c) to the node:
(21)
CONJ
[+or]
[+WH]
as it is to ascribe them to the node Q.
III.
THE DERIVATION OF INTERROGATIVE STRUCTURES
A.
Alternative Questions
1.
Conjunction Spreading
WH spreading will be carried out in part by the Conjunction
Spreading schema (cf. CONJ section) since all conjunctions are spread
from the one which is the leftmost daughter of the top S. The Conjunction Spreading schema changes the deep structure tree of (22.a) to
(22.b):
(22) (a)
he always snores
he doesn't always snore
CONJ
[+or]
[+WH]#
CONJ
[+or]
[+WH]
he always snores
635
he doesn't always snore
INTERROG - 11
2.
WH Spreading
The WH must next "be brought into the lowest S's. This rule
must follow the one discussed above, but precede the Initial Conjunction Deletion rule.
SI:
[+WH] ff X #
#
CONJ
[+WH] # X
CONJ
k 5 6
'
8
9 10
# #
'
11
Attach 3, 8 as right sisters of k> 9 respectively.
Delete 3, 8 from complex symbols of 2, 7 respectively.
Insert CONT (trigger for continuing rising intonation
pattern) as left sister of 6.
SC:
1.
2.
3.
COND:
The rule is obligatory.
Notes:
This rule has the peculiar effect of introducing a feature
([+WH]) into a position not dominated by any lexical rule.
Perhaps ADV should also be inserted. Cf. next rule.
Example in Tree Format:
(23) (a)
§
CONJ
[+or]
[+WH]
he always snores
he doesn't always snore
[+WH] he always snores CONT
[+WH] he doesn't always snore
CONJ
[+or]
636
INTERROG - 12
3.
AUX-Attraction
M
SI:
(S C0NJ)« #,
v[x/[+WHl |x] X TNS ( HAVE» (NEG) (ADV) X #
IADV)
1[+NEG]J
BE
[HP J
1
23
U
5
6
SC:
1.
2.
Attach 5, 6, 7 as right sisters of 3.
Delete (original) 5, 6, 7.
COND:
1.
If 6 is null, 9
2.
3.
The rule is obligatory.
The rule applies last-cyclically.
Notes:
(i)
(ii)
(iii)
(iv)
(v)
(vi)
(vii)
789 10
V
- f|*
U+X
U-BE])
There appear to be no strong arguments for ordering
the Initial Conjunction Deletion rule prior to this
rule. It must precede the Reduced Alternative
Question rule. The trees in this section are drawn
as though the rule had already applied to remove the
initial conjunction.
The rule is intended to apply to WH questions (see
below), alternative questions and sentences with preposed negative adverbials (cf. NEG). In fact, the
rule will not apply to alternative questions unless
the WH-spreading rule were to insert a node ADV
dominating the feature [+WH]; alternatively, constituent 3 of the S.I. could be stated to be any
single constituent immediately dominated by S.
The X at h is probably tantamount to (NP).
Condition (l) blocks the derivation of such forms as
»Does he be going (or doesn't he be)?, *Where did he
have gone?
Condition (3) prevents [+WH] from triggering AUXattraction in Rel clauses and indirect questions.
This rule follows a number of rules which affect the
order of elements within MOD, e.g. Pre-verbal ADV
placement, Pre-verbal NEG placement (cf. NEG). The
application of these rules accounts for the discrepancy
between the order given here of elements 6, 7, and 8
and their deep structure order,
We accept Ross's (1967c) output condition (3.27) that
S's containing internal S's dominated by NP's are
unacceptable, as the explanation for the ungraramaticality of »Did that John showed up please you? and
therefore put no special condition on this rule to
exclude such sentences.
637
INTERROG - 13
(viii)
(ix)
The HAVE in 3 of the S.I. of the AUX-attraction rule
cannot be [+V]. Thus the WH-deletion rule generates
(25.d) but not (26.a) (which is grammatical in British
English). Since AUX-attraction is a last-cyclic rule,
NEG must already be in the position indicated in the S.I.
of this rule (i.e. following HAVE). Therefore, we would
derive Has he something to do or doesn't he? but not
(26.a), (cf. NEG p. 53).
Apparently the usual condition on conjunction constraining the conjoining of identical sentences (S, / So) does
not obtain in the case of alternative questions. Thus
sentences like (25.f), which achieve their effect by
seeming to offer a choice without actually doing so, are
both grammatical and common.
Example in Tree Format
{2k) (a)
# [+WH] NP
he [-PAST] always snore
CONJ
[+or]
he [-PAST] NEG always snore
(by applying AUX-ATTRACTION to each subtree dominated by S)
638
INTERROG - lU
(2U) (b)
PROP #
[-PAST] he always snore
CONJ
[-PAST] NEG he always snore
[+or]
Examples:
(25) (a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
(i)
(J)
(k)
Does he always snore or doesn't he always snore?
Could he have left yesterday or was he being detained?
Are you a man or are you a mouse?
Has he left or does he have something to do?
Can't you hear me or aren't you listening?
Is Chomsky right or is Chomsky right?
Was his doing that a surprise or had you expected it?
Was it a surprise for him to do that or had you
expected it?
Was it a surprise that he did that or had you expected
it?
Is it raining or is it snowing?
Is there a book on that table or isn't there one there?
Ungrammatical and disallowed:
(26) (a) *Has he something to do or hasn't he?
(b) *Does he be going or doesn't he be?
k.
WH-Deletion
SI:
# [+WH] TNS X
12
3
SC:
Delete 2.
COND:
The rule is obligatory.
639
INTERROG - 15
Notes:
This rule deletes the [+WH] that has been moved to
sentence initial position by WH-Spreading, after the
application of AUX-Attraction.
Example in Tree Format:
Tree (2U.b) is changed to (27) by this rule.
(27)
#
S
TNS
# TNS NEG NP
5.
MOD
PROP
0
Reduced Alternative Question (including yes-no questions)
SI:
(M
)
(M
#TNS( HAVEH (NEG) NP X CONT # OR #TNSUHAVE^) (NEG) NP X
BE )
(BE
1
2
3
U
5
6
SC:
1.
2.
3.
Delete 9 or:
Delete 6, 8, 9 (where 7 • NEG) or:
Delete 5, 6, 7, 8, 9, 10
COND:
1.
2.
1...3 * 6...10, except 2 i 7
The rule is optional.
Notes:
(i)
(ii)
789
§
10
The three SC's are all optional. Their products are
considered stylistic variants of each other and of
non-reduced alternative questions.
Yes-no questions are generated by SC (3).
Example in Tree Format:
The REDUCED ALTERNATIVE QUESTION rule operates on the tree of
(27) converting it by the three SC's into the respective trees of (29)
6U0
INTERROG - 16
-*
(29) (a)
[-PAST] he always snore CONT
CONJ
[+or]
[-PAST] NEG
he
(29) (b)
[-PAST]
he
always
snore
CONT
CONJ
[+or]
NEG
(29) (c)
CONT
[-PAST]
6Ul
INTERROG - IT
Examples
(30) (a
(b
(c
(d
(e
(f
(g
(h
(i
Does he always snore or doesn't he?
Does he always snore or not?
Does he always snore?
Doesn't he always snore or does he?
Doesn't he always snore?
Did you say he always snores?
Did you (just) say, "He always snores."?
Did you (just) say, "Does he always snore?"
Do you have a son or a daughter or don't you?
Ungrammatical and disallowed:
(31) "Doesn't he always snore or?
Grammatical but not generated by this rule:
(32) (a)
(b)
(c)
(d)
He always snores? (derived from (30.g) by T-ECH0QUESTION)
Does he always snore? (homophonous with (30.c) but
derived from (30.h) by T-ECH0-QUESTI0N)
Doesn't he always snore? (homophonous with (30.e)
but derived as stylistic variant of He always snores,
doesn't he? by T-TAG-QUESTION)
Do you have a son or a daughter ? (This is a simple
alternative question, with two simplex sentences in
its deep structure, as opposed to (30.j): Do you
have a son or a daughter ? (which is_ generated by
this rule and has the meaning 'Do you have a child?'.)
(30.j) has four simplex sentences in its deep structure. The intonation contours clearly differentiate
the graphically identical questions.)
Justification:
(i) The major Justification for deriving yes-no questions as
stylistic variants of (a subset of) alternative questions is semantic.
That is, sentences like (30.a,b,c) are perfect paraphrases of one
another, and all are perfect paraphrases of the underlying full
alternative question, Does he always snore or doesn't he always snore?
(ii) A further Justification is the fact that this derivation
automatically relates the rising intonation pattern of yes-no questions
to the rising pattern of the first part of alternative questions.
6U2
INTERROG - 18
(iii) This analysis agrees with Katz and Postal's analysis of yes-no
questions in having WH plus OR (in Katz and Postal, WH plus either-or)
in the deep structure of yes-no questions. It is not clear, however,
whether Katz and Postal consider yes-no questions to be reduced
alternative questions, or whether they would say that alternative
questions include an additional S in their deep structures that is
absent in the deep structures of yes-no questions.
(iv) Malone's (1967) analysis of yes-no questions, which distinguishes such questions from statements on the basis of interrogative
(vs. declarative) "sentence valences", cannot account for the
relations between yes-no and alternative questions, and is rejected
on these grounds.
(v)
The condition on SC (2) excludes strings such as (31).
Problems:
(i) There is some doubt about whether negative sentences such as
(30.e) are in fact yes-no questions. The present treatment assumes
that they can be, i.e. that (30.d,e) can be derived as alternative
stylistic variants of: Doesn't he always snore or does he always
snore? (This latter sentence, however, is itself rather peculiar
unless the auxiliaries are stressed: You said he doesn't always snore,
but now you seem doubtful. Well, doesn't he always snore or does he
always snore?)In any case, it seems clear that the usual interpretation of Doesn't he always snore? is a paraphrase of He always snores,
doesn't he?—see (30.c)
(ii) It is perhaps a problem for this derivation of yes-no questions
that the answers to such questions are different from the answers to
alternative questions:
(He does.
)
(33) Does he always snore, or doesn't he always snore?V He doesn't.)
(3U)
Does he always snore? ( *es,(»uhe„doe8!;, \
^No (, he doesn't).)
(iii) SC (l) retains only the pre-subject part of AUX, in the second
of the conjoined questions. Thus from Should he have been doing that
or shouldn't he have been doing that? SC (1) derives: Should he have
been doing that or shouldn't he? But the following are also grammatical: Should he have been doing that or shouldn't he have? Should he
have been doing that or shouldn't he have been? The same patterning
of AUX retention is found in other kinds of conjoined structures—He
should have been doing that and she should (have (been)), too.—so
perhaps the general conjunction-reduction rules are all that is
6U3
INTERBOG - 19
necessary to account for the sentences generated by SC (l). Similarly,
SC (2) seems only to be a special case of a more general phenomenon:
cf. He loves Jane and not Mary, Either he loves Jane or not.
B.
WH Questions and Other Question Types
1.
WH Question Words
Since the WH's which yield question words are introduced as
features on the determiner of the indefinite NP, there is no need
for a WH-ATTACHMENT rule with interrogative structures. The various
question words (and relative pronouns) are derived from the feature
complexes under the determiner node. The actual "spelling" of the
feature complexes takes place in the second lexical lookup. The
discussion and justification of this procedure, along with the rules,
are found in the DETERMINER section.
2.
WH Fronting
SI:
# X (PREP)
[D [+WH] X]
X
1 2
SC:
1.
2.
Attach 3 as right sister of 1.
Erase (original) 3.
COND:
1.
2.
2^X f+WH] X
The rule is obligatory.
Notes:
(i)
(ii)
The fronting of [+WH] will trigger AUX-ATTRACTION.
In some cases the constituent with WH may be fronted
from within a subordinate clause: When has he
decided to leave? Where did she tell him to go?
What did it surprise him that she did?
Fronting must be prevented, however, when the
constituent with WH occurs in a relative clause or
an indirect question. Rel clauses are one of the
configurations where the movement across a variable
is blocked by Ross's COMPLEX NP CONSTRAINT. The
fact that interrogation is also impossible out of
an indirect question suggests that the deep structure
of indirect questions should have a lexical head.
For example:
6kk
INTERROG - 20
(35) (a)
The man
S
came
the man killed who
(b) *Who did the man who kill came?
(36) (a)
You know
who came
(b) *Who do you know came?
(c) *Who did you know come?
(iii)
Condition (l) is needed to prevent the stacking of WH's.
(37) (a) *Why where when did you see him?
(b) Why, where and when did you see him?
(iv) A sentence with WH can be conjoined only with another sentence
containing WH:
(38) (a)
(b)
He died where and when?
Where and when did he die?
(39) (a) *He died here and when?
(b) *Here and when did he die?
3.
Tag Questions
There are certain requisites that any solution for tag questions should meet. First, they should not be generated as optional
variants of yes-no questions, since they are semantically distinct
from them. That is to say, they appear to be either negative or
positive statements with an appended question element. They do not
have the neutral disjunctive either/or characteristic of the alternative question. Tag questions are underlying suppositions, hopes,
fears,etc., for which the speaker is seeking confirmation. An alternative question seeks only information.
In addition, there is a co-occurrence restriction that holds
for yes-no questions but not for Tag questions. As pointed out by
Katz and Postal (196U), some sentence adverbials can not occur in
yes-no questions, but can occur in Tag questions (and in declaratives—
cf. II.B.2 above); e.g.,
(U8) (a) Certainly John is a doctor.
(b) Certainly John is a doctor, isn't he?
(c) *Is John certainly a doctor?
6U5
INTERROG - 21
This means, that if we were to derive Tag questions from
yes-no questions, we would have to constrain these sentence adverbials
so as to trigger the "optional" Tag transformations. Such a constraint
seems a very unlikely one.
Second, we would want the same rule for AUX ATTRACTION that
applies to alternative questions to apply to the AUX in the Tag.
Third, the obligatory occurrence of the oppositive value of
negation in the Tag to that in the main statement should be shown to
be a function of the value of negation in the supposition underlying
the tag question and not inherent to the tag in the deep structure.
For example, in (U9):
(U9)
John has left, hasn't he?
the NEG in the tag results only because there is no NEG in the main
statement. While in (50):
(50)
John hasn't left, has he?
the non-occurrence of NEG in the tag results from the NEG present in
the main S.
Previous analyses of tag questions have failed to meet one or
more of these requisites. Klima's analysis (l96^c) fails with respect
to the first requirement given above. The second and third are
recognized. Thus for Klima (51) and (52) are two sets of optional
variants:
(51) (a)
(b)
Has John left?
John has left, hasn't he?
(52) (a)
(b)
Hasn't John left?
John hasn't left, has he?
Rosenbaum (1966) fails with respect to the first and third of
the above requisites. For Rosenbaum all tag questions are optional
variants of negative yes-no questions. Tag questions with a negative
in the tag are derived by optionally moving the negative of a main
sentence negative into the tag. This results in the claim that
(53.a,b,c) are all optional variants:
(53) (a)
(b)
(c)
Hasn't John left?
John hasn't left, has he?
John has left, hasn't he?
6U6
INTERROG - 22
There are two possible analyses that we have considered.
They both present certain difficulties. For this reason we shall
not present specific rules in this section, but rather we shall
briefly outline the alternative analyses.
One possibility is to suppose that tag questions are the
result of a statement plus a following alternative question which has
been further reduced. This alternative question might originate in a
sentence adverbial. (5^.a) would be the deep structure for John has
left, hasn't he? The alternative question in (5^.a) would then undergo
CONJ SPREADING, WH SPREADING, CONJ DELETION, AUX FRONTING, WH DELETION,
and ALTERNATIVE Q RED, to yield (5**.b):
(5*0 (a)
CONJ
[+or]
[+WH]
John hasn't left
John has left
<5M (b)
John has left
hasn't John left
(5U.b) then undergoes the tag rule which moves adverb to post-position
and further reduces the question in the tag which results in (51*.c):
(5M (c)
John has left
hasn't he
6U7
INTERROG - 23
The principle difficulty with this analysis is the stating
of the identities in the tag reduction rule. We want to state that
the S of the tag (i.e. ADV) is identical to the main sentence S with
the exception of NEG. (This must be stated as a condition.) However,
since the tag S has undergone AUX FRONTING it is no longer formally
identical. As a result we must tortuously list the elements in both
S's and their identities. Thus, although it is possible to write
such a rule, it is rather complicated to state. A main virtue of
this approach is that it does not add any new symbols to the base
structure (except ADV S) and employs the mechanism needed for
alternative questions plus one additional rule.
A second possibility which we have considered is that tag
questions result from a copying rule which copies the subject NP and
the relevant parts of AUX after a sentence and makes the tag opposite
to the main sentence in negation. This, however, demands a separate
trigger in the base. It has been suggested that WH be generated as
a sentence ADV for this purpose. The copying rule would then operate
on (55.a) and convert it to (55.b):
(55) (a)
PROP
(55) (b)
John
has
CONJ
left
[+WH]
6U8
John
has
INTERROG - 2U
The WH, which has been post-posed, then serves as a trigger
for the AUX ATTRACTION rule (as it does in alternative questions) to
apply to the tag. There are technical difficulties with this solution,
too. First of all, WH coming from ADV may have to be restricted to
non-embedded sentences since tag questions, unlike alternative and WH
questions, do not appear to tolerate embedding, e.g. *I wonder whether
John has left, hasn't he? (This generalization is not entirely
correct since for many people the following sentences are grammatical):
(56) (a)
(b)
(c)
(d)
I think
I'm sure
I imagine 7 (that) John has left, hasn't he?
I suppose
• • •cLC •
(e) ?I know (that) John has left, hasn't he?
Note the presence of that which seems to indicate that tag questions
are really quite different from alternative and WH questions; e.g.,
(57) (a) »I know that who left
(b) *I know that whether he left or not
Yet there is a peculiar restriction on embedded tag questions which
we do not fully understand: they must have 1st person singular pronouns as matrix subject:
(58) (a) "John thinks that Mary has left, hasn't she?
(b) *They are sure that we have left, haven't we?
k.
Negative Questions from Tag
There is a type of negative yes-no question which resembles
tag questions in that it seems to involve an underlying supposition.
The supposition is positive, however. This is illustrated in (59):
(59) (a)
(b)
Didn't John write any poetry last year?
Didn't John write some poetry last year?
(59.a) is an ordinary alternative question, but (59-b) seems to mean
that the speaker supposes that John did write some poetry. We propose
that (59.b) has the same base structure as (60):
(60)
John wrote some poetry last year, didn't he?
If we were to choose one of the above alternatives (59-b) could be
derived as follows: a tree such as (5^.a) for the underlying structure
of (59.b) would be reduced by deletion of the main statement S and the
right sister S of the tag, to:
61*9
INTERROG - 25
(61)
i
ADV
/ Didn't John write some poetry last year? )
( Hasn't John left?
/
5.
Questioned Quote
(Including Echo Question)
SI:
# [+PAST] you sax
X CONT #
SC:
Delete 1.
COND:
This is an optional (stylistic) rule.
Note:
The SI characterizes a subset of the products of REDUCED
ALTERNATIVE QUESTION rule: viz., yes-no questions with the
subject you and the verb say. Say, which means "(just) say
in this linguistic context" is different from the ordinary verb
say in that it takes only quotes sentences or pro-forms
as objects. Its surface form, however, is homophonous with
that of the ordinary transitive verb.
Example in tree format:
[+PAST]
you
SAY
he's going
650
CONT
INTERROG - 26
(62) (b)
he's going
CONT
Examples:
(63) (a)
(b)
(c)
He's going? (cf. Did you (just) say: "He's going?")
Is he going? (cf. Did you (just) say: "Is he going?")
Where did he go? (cf. Did you (just) say: "Where
did he go?")
Justification and Alternatives:
(i) To date, Malone (1967) is by far the fullest treatment of
echo questions and other echoic sentences (see WH QUESTIONED QUOTE,
DECLARED QUOTE, below). The present analysis differs from Malone's
in that it relates all echoic sentences to deep structures that include the verb SAY (see Notes above). This analysis seems justified
by the interchangeability of echoic sentences and sentences with SAY.
(ii)
Examples like (63.b) are homophonous with yes-no questions.
(iii) Examples like (63.c) are distinguished intonationally from
two other sentence types with initial WH words: WH questions and WHquestioned quotes. The questioned quotes have a /233+7 intonation
pattern, the WH questions a /231+/ intonation pattern, and the WHquestioned quotes a /333+/ pattern:
(6U)
(6h)
2
33+
Where did he go?
(Echo question)
(65)
2
31+
Where did he go?
(WH question)
(66)
3
33+
Where did he go?
(WH-questioned quote)
651
INTERROG - 27
6.
WH-Questioned Quote
a.
Intonation Introduction
# you [+PAST] SAY # X
SI:
i
i
SC:
1.
2.
COND:
The rule is obligatory.
Notes:
(i)
(ii)
(iii)
(PREP) [+WH] X
<
.
#
•_i
Attach RAISING INTONATION ("t") as left sister of 2.
Attach CONT as left sister of 3.
See QUESTIONED QUOTE, Notes for SAY.
The "+" introduced by the SC is an intonation marker.
It represents a high pitch (Trager-Smith level 3)
on all material that follows it.
CONT is also an intonational marker. It represents
a final pitch rise.
Example in tree format:
(67) (a)
you [+PAST] SAY
he[+PAST] see
652
[+WH]
[+INDEF]
[+N]
[+PRO]
[+HUM]
yesterday
INTERROG - 28
(67) (b)
you
[+PAST]
SAY
he[+PAST] see
t [+WH
[+H
yesterday
[+INDEFJ UPRO
L+HUM.
Examples
(68) (a) You
(b) You
(c) You
(d) ?You
said
said
said
said
he saw + who(m) yesterday?
• who saw him yesterday?
he saw him t when?
+ what?
Ungrammatical and disallowed
(69) *Did you say he saw t who(m) yesterday? (Possibly
grammatical, but only as a reply to: Did I say he saw
(inaudible) yesterday?, in which case it is derived
from: You said, did I say he saw t whom yesterday?)
Related examples
(70) (a)
(b)
(c)
(d)
tWho(m) did you say he saw yesterday?
+Who did you say saw him yesterday?
tWhen did you say he saw him?
+What did you say?
653
INTERROG - 29
Grammatical but not Related to this Rule:
(71) (a)
Did you say he saw him yesterday?
(b)
2
3
1+
Who(m) did you say he saw yesterday?
(c)
2
3 1+
What did you say?
Justification
(i) The underlying structure of WH-questioned quotes is differentiated from that of other questioned quotes in two ways: (a)
the WH-questioned quotes are derived from declaratives, rather than
interrogatives, with you SAY in the matrix S; (b) the WH-questioned
quotes obligatorily include WH in the object of SAY. The reason for
(a) is that sentences like (68) and (71.a) are grammatical, while
sentences like (69) are not.
(ii) The ordinary WH FRONTING and AUX ATTRACTION transformations
operate optionally on (68.a,b,c) to yield (70.a,b,c) respectively.
In the case of (68.d) the WH QUESTION transformations perhaps
operate obligatorily to yield (70.d).
(iii) The need to distinguish SAY from the ordinary verb say becomes
clear through a comparison of (70.a) with (71.b) and (70.d) with
(71.c). (71.b,c) are simple WH questions, while (70.a,d) are WH
questions based on WH-questioned quotes.
b.
You-said Deletion
SI:
# you [+PAST] SAY
XtX [+WH] X
SC:
Delete 1
COND:
The rule is optional.
65U
INTERROG - 30
Example in Tree Format:
(72) (a) (The input tree equals the output tree for the
above Intonation Introduction rule, (67.b).)
(b)
NP
MOD
AUX
he[+PAST] see
t
[+WH]
[+INDEF]
[+HUM]
[+H]
[+PR0]
Examples
(73) (a)
(b)
(c)
(d)
He saw • who(m) yesterday?
• Who saw him yesterday?
He saw him + when?
+ What?
Related Examples
(7k) (a)
(b)
+ Who(m) did he see yesterday?
+ When did he see him?
Grammatical but not Related to this Rule
(75) (a)
(b)
2
3
1+
Who(m) did he see yesterday?
3 1+
What?
655
yesterday
INTERROG - 31
Justification
(i) Examples like (73) are derived by optional deletion of
'You said' from the examples (68) respectively given for IntonationIntroduction rule above. This derivation is justified on the grounds
of semantics as well as on the basis of intonation.
(ii) Examples like (7*0 reflect the optional operation of the
ordinary WH-QUESTION transformations upon (73.a,c) respectively.
(iii) (7k) may be contrasted with (75). The latter are simple WH
questions, while the former are WH questions based upon WH-questioned
quotes that have undergone 'you-said' deletion.
7.
Declared Quote
0
SI:
I [+PAST] SAY 0
12
SC:
Delete 2 and k
COND:
1.
2.
X
(CONT)
0
3
3 4 X + CONT
The rule is optional.
Example in Tree Format
(76) (a)
0
I
[+PAST]
SAY
[-PAST]
he
PROG
I
ing
656
go
INTERROG - 32
(76) (b)
Examples
(77) (a)
(b)
(c)
2
3
1+
Is he going?
He's going. (As reduction of I said, "He's going.")
Who's going? (As reduction of I said, 'Who's going?")
Grammatical but not Generated by this Rule
(78) (a)
(b)
(c)
2
3
3+
Is he going?
He's going. (As non-quoted statement.)
Who's going? (As non-quoted WH question.)
Justification
(i) Examples like (77) are derived by optional deletion of "I
said" from the sentences "I said (77)." Semantic and intonational
arguments for this derivation may be adduced.
(ii) When the declared quote is a yes-no question, it differs
intonationally from a non-quoted yes-no question—compare (77.a) with
(78.a). In other cases, declared quotes are homophonous with their
non-quoted counterparts—compare (77.b) with (78.b) and (77.c) with
(78.c).
(iii) Condition (l) on the rule guarantees that if CONT is indeed
present, it must be chosen as element h of the S.I. and hence must
be deleted.
December 1968
657
658
IMPERATIVE
Contents
Page
I.
II.
III.
BIBLIOGRAPHY
1
DISCUSSION AND ANALYSIS
1
A.
The Range of Phenomena Treated
1. Included in the UESP Rules
2. Not Dealt with in the Verb Rules
1
2
3
B.
The Underlying Subject of Imperatives
1. Constraints on Imperative Subjects in
Respect to Person
2. A Note on the Vocative
6
6
15
C.
Imperatives and Peremptory Declaratives. . . . lo
D.
Imperatives, Requests and Questions
1. Behaviour Common to Imperatives and
Requests
2. Differences between Imperatives and
Requests
3. Behaviour Common to Requests and Questions
k. Differences between Requests and Questions
5. Conclusion
17
IT
19
23
2U
2b
E.
The Underlying Auxiliary of Imperatives. ... 26
1. The Presence of a Modal
26
2. The Choice of a Modal
27
F.
Tagged Imperatives
29
G.
Blocking Problems
31
TRANSFORMATIONAL RULES
33
659
IMPERATIVE
I.
BIBLIOGRAPHY
Bolinger, D. (1967c) "The Imperative in English"
Boyd, J. and J. Thome (1968) "The Deep Grammar of Modal Verbs"
Chomsky, N. (1955) The Logical Structure of Linguistic Theory
esp. pp. 691-69U
Hasegawa, K. (1965) "English Imperatives"
Hornby, A.S., E.V. Gatenby and H. Wakefield (1963) The Advanced
Learner's Dictionary of Current English
Katz, J. and P. Postal (196^b)An Integrated Theory of Linguistic
Descriptions, esp. pp. 7^-79
Kiparsky, P. (1963) "A Note on the English Imperative"
Klima, E. (l96Uc) "Negation in English", esp. pp. 258-260
Lakoff, G. (1965) On the Nature of Syntactic Irregularity
Lees, R. (l96Ub) "On Passives and Imperatives in English"
Thorne, J. (1966) "English Imperative Sentences"
II.
A.
DISCUSSION AND ANALYSIS
The Range of Phenomena Treated
The UESP grammar provides rules for only a small proportion
of the constructions which have at various times been regarded as
imperatives or as closely related to them. In some cases this is
because too little is known about the construction in question.
However, in the case of forms like:
(l)
(a)
(b)
(c)
John, come here.
Will you come here!
You will come here!
all of which have been regarded by one or another transformational
grammarians as directly related to imperatives, there are good
arguments against postulating a direct transformational relationship
between any of these forms and true imperatives like:
(1)
(d)
Come here.
Immediately below are examples of the construction-types which
our rules account for, including embedded imperatives (i.e.
"subjunctives"). These are followed by examples of types not included in the rules. The question of possible constraints on the
deep structure subject of non-embedded sentences is then discussed.
660
IMP - 2
In the course of this discussion we separate vocatives from other
sentences which appear to be imperative. What we have called
peremptory declaratives are claimed to be declarative sentences
which in appropriate context may be interpreted as embodying a wish
or command, while requests are a kind of question open to a similar
interpretation. Vocatives, requests and peremptory declaratives
have been regarded as typical imperative forms in some earlier works.
The underlying auxiliary of imperatives is examined next, adopting a
position close to that of Lees (196U): the appropriate base rule
introduces an element, which we represent as SJC, disjunctive with
both modals and tense. Thus, we do not generate a modal such as will
in the deep structure of imperatives, but a separate form which behaves in certain respects like modals (in AUX-INVERSION) and in
certain respects like affixes ( in AFFIX-SHIFT and DO-SUPPORT ). In
connection with this argument, it is necessary to consider briefly
the significance of tagged imperatives, for which we do not provide
rules—in fact the grammar does not generate tags, for reasons set
out here and in INTERROG.
This treatment of imperatives may be open to the objection
that it fails to relate them to a number of constructions which
appear to be semantically or syntactically similar. For example,
the grammar does not provide directly for the fact that certain
readings of (l.a-c) are close paraphrases of (l.d) and that all
these, together with (2.a-c) may perhaps incorporate a common semantic
element, in contrast with declaratives and questions.
(2)
(a)
(b)
(c)
Go home now and I'll never see you again.
Let's go home.
May he go safely.
We claim that imperatives (like (l.d)) are syntactically
distinct from all the other examples in (l) and (2); it may be
possible in the future to give a more unified account of some of
the exemplified constructions, but we consider that any such treatment must recognize the syntactically distinct class of imperatives.
1.
(a)
Included in the UESP Rules
Plain Imperatives
These rules account directly for plain imperatives and subjunctives (which are here regarded as equivalent to embedded
imperatives).
(3)
(a)
(b)
(c)
(d)
(e)
(f)
Go there.
You go there.
Somebody go there.
Don't go there.
Don't you go there.
Don't anybody go there.
661
IMP - 3
(b)
Complements containing subjunctives
(U) (a)
(b)
They requested
\
They made the request [that
*
John be
Publicly
chastized.
(c)
(d)
He moved
I
He seconded the motion)
that tne goVernor
be
recalled.
(e)
(f)
It is desirable
"[
They talked about the necessity V*hat a bridge
'
be built.
The term subjunctive word, is used here to refer to those head
words that can take THAT-complements which contain SJC, the element
in AUX that distinguishes imperatives. Since there is no distinct
form in FOR-TO and POSS-ING complements for such embedded imperatives
(subjunctives) it is difficult to provide purely formal criteria
which would indicate when these complements are subjunctive. For
example, the insertability of please is not a criterion. Compare
(U.e.f) with (5).
(5) (a)
(b)
It is desirable to build a bridge.
They talked about the necessity of building a
bridge.
(c) *It is desirable to build a bridge, please.
(d) *They talked about the necessity of please
building a bridge.
Most subjunctive words are unmarked for the feature [IMPER] in the
lexicon since they may take either subjunctive or indicative
sentences as their complements. Words like know, which cannot take
a subjunctive complement are marked [-IMPER] in the Lexicon. (See
NOM and LEX.) Words like move, and perhaps propose, which can only take
a subjunctive in a complement clause are marked [+IMPER]. (See LEX .)
2.
Not Dealt with in the UESP Rules.
The following four types of constructions have not yet been
carefully investigated from a generative point of view. Wishes
have been totally excluded from the present treatment of imperatives.
Conditional imperatives, permission imperatives, and wish imperatives
are treated only in so far as their properties coincide with those
of plain imperatives.
662
IMP - U
(a)
(b)
Conditional imperatives.
(6)
(a)
(b)
(7)
Scratch a Russian and you will find a Tartar.
Permission imperatives.
(8)
(c)
Come here, and I'll give you a dollar,
If you come here, I'll give you a dollar.
(a)
(b)
(c)
Come home at 3:00 every morning (if you must).
Buy whatever you like.
All right, be miserable (I don't care).
Wish-imperatives.
(9)
(a)
(b)
(c)
Be happy.
Get well soon.
Sleep well.
(6)-(9) are all much like
ordinary imperatives but differ from them
semantically, and, to a greater or lesser extent, syntactically.
For example, they do not take tags comfortably. Please can occur
with none of the examples in (8).
(d)
Wishes
(10) (a)
(b)
May you be happy.
May you soon get well again.
In addition, modals of volition with their accompanying verbphrases have not been dealt with in detail. Such modals have been
treated by Boyd and Thorne as realizations of a performative proverb IMP. A grammar that treats auxiliaries as main verbs might
subsume these modals under the subjunctive words mentioned above
(H.A.l.b). This grammar does not treat auxiliaries as main verbs,
and the fact that a non-finite verb form follows both the modals
and the subjunctive words results from independent factors in the
grammar: modals have no affix with them in the deep structure so
there is nothing to move onto the verbs which follow, while subjunctive words on the other hand select, to follow them, an embedded
sentence containing SJC in the AUX. Since SJC is disjunctive with
TNS, there is once again no effect on the form of the main verb.
Some examples of modals of volition are:
663
IMP - 5
(11)
/shall
should
must
You / may
^ go.
\ might
1 could
I ought to
So far in this section we have been dealing with forms which
we exclude not on the basis of positive evidence but simply because
they have not yet been adequately dealt with from a transformational
perspective, or because we have been unable to incorporate them
into our treatment of the imperative. There is one more such construction, the let imperative, which has many points in common with the
true imperatives but which we do not attempt to deal with in detail.
(e)
Let imperatives.
(i.e. let used with first or third person subject to supply n
an indirect imperative)
(12) (a) Let's start at once, shall we.
(b) ?Don't let's start yet. (Let's not start yet.)
(c) Let us both have a try at it.
(d) Let there be no mistake about it.
(e) Let them leave as soon as they hear me call.
We do not have an analysis of these forms. They appear to be
closely related to ordinary imperatives but there are differences.
For example, quite a number of let imperatives do not admit a tag
with will you;
(13) (a)
(b)
(c)
(d)
*Let
*Let
*Let
*Let
them do their worst, will you. (defiance)
them all come, will you.
(defiance)
there be no mistake about it, will you.
AB equal CD, will you.
Moreover, let imperatives with a first person plural (inclusive)
subject differ formally from plain imperatives in which let is
followed by a complement with a first person plural (exclusive) subject: the let imperatives admit reduction of let us to let's and
some differ in the form of the tag:
(1*0 (a) Let us pass, will you)
(b) "Let's pass, will you J("allow us)
(c) Let us go in, shall we.) =I
(d) Let's go in, shall we. f
suggest that we...)
66U
IMP - b
We turn now to three forms which have been regarded by various
grammarians as imperatives. We shall devote the next three sections
to demonstrating that although they possess features in common with
imperatives, they must all be clearly separated from them. We do
not deal with these constructions in the imperative rules for the
reasons discussed below.
(f)
Vocatives
(15) (a)
(b)
(c)
(g)
Peremptory declaratives
(16) (a)
(b)
(c)
(h)
John, look at yourself.
Take off your coat, somebody.
Boys, come here, please.
You will leave immediately.
Shoes will not be worn in the gym.
You certainly won't do that.
Requests
(17)
f
Can
Could
Can't
\ Couldn't
\ Will
Would
Won't
?Wouldn't
^
j you leave immediately, please.
Finally, tagged imperatives which are described in detail in
section (F) are not dealt with in our rules since we do not have a
general Tag rule in the grammar.
(i)
Tagged imperatives
(18)
, will you \
ican you
j
would you I
could you f
won't you \
can't you j
(please),
B.
The Underlying Subject of Imperatives.
1.
Constraints on Imperative Subjects in respect to Person
Chomsky (1955), Klima (l9bUc), Kiparsky (1963), Katz and Postal
(l96Ub), Lees (l96Ub) and Hasegawa (1965) all agree that imperatives
665
IMP - 7
have you as underlying subject. This subject may (and in some
cases, must) be deleted. They support this claim by the following
arguments:
(a)
The reflexive in imperatives is yourself/yourselves:
(19)
Look at yourself,
but not:
(20) "Look at myself.
(b)
Tagged imperatives have you:
(21)
Go home, will you.
but not (as an imperative):
(22) *Go home, will he.
Thome, however, notes that there are certain kinds of imperatives in which it is less obvious that an underlying you is the
subject:
(23)
(a)
(b)
(c)
(d)
(e)
Nobody move.
Everybody get out as quick as he/you can.
Somebody pay the bill.
John, come here.
Sit down, boys.
He therefore admits nouns as the subject of imperatives, but requires
that the N-node contain the feature [+V0CATIVE]. This feature is
always realized by you either as a determiner on the noun, as in you
boys come here, or by itself. The feature [+V0CATIVE] (on you) may
be deleted in certain contexts, as in (23.d,e). Thome's disagreement with the conclusions the other investigators drew from sentences
(19-22) is thus less radical than it seems — apparently not radical
enough.
Thome fails to take into account, in any systematic way,
sentences (23.a-c) on the one hand and (23.d,e) on the other. In
the first place there is a major difference in intonation between
the two sets of sentences. (23.d,e) alone require a comma-intonation to set off what Thome considers the vocative subject of the
imperative, a fact which alone makes his analysis rather dubious.
Secondly, in sentences like (23.d) it is impossible to refer back
to John by a third person pronoun:
666
IMP - 8
(2U)
(a)
(b)
John, take off your coat,
John, take off his coat.
#
Sentences like (23.c), however, which do not require comma intonation after the subject, differ also from (23.d,e) in that they
admit third person pronominal reference. For many people, his in
(26) may refer to the person addressed, the subject of that sentence.
Thus, the subject of (26) is much more clearly third person than is
the subject of (2k).
(25)
Somebody take off your coat.
(26)
Somebody take off his coat.
Thome takes (26) to be ungrammatical; he considers it "an erroneous
form found among educated speakers", which replaces (25). he points
out that one says:
(27)
Take off your coat, somebody.
But not, with the same meaning:
(28) *Take off his coat, somebody.
However, the fact that (28) is not acceptable provides no support
for regarding (26) as having an essentially [+11 person] subject.
Even if (28) were transformationally related to (26), it would not
be enough to attribute the ambiguity of (26) to analogy or hyperurbanism. Such an "explanation" would give no account of why in
contrast with (26), (28) can never have third person anaphora to
its subject. In any case, (26) and (28) do not seem to be transformationally related.
It is in fact rather easy to relate (27) and (28) to vocatives
like (23.a,e). There are sentences parallel to (27), (28) but with
somebody in initial position, separated from the rest of the sentence
by comma intonation. Only that intonational difference separates
(27') and (28') from (25) and (26), on the surface.
(27')
Somebody, take off your coat.
(28') *Somebody, take off his coat.
Notice, however, that (28'), like (28), cannot occur if his is
understood to refer back to the subject.
667
IMP - 9
Furthermore, in forms which are unmistakably vocative, like
(29),
(29) *John, take off his coat,
(coref.)
his cannot refer back to the subject. We are not dealing in detail
with the derivation of vocatives in this report (but see B.2). It
is enough to suggest that (30') is a likely source for (30):
(30)
(30')
John, take off your coat.
John, you take off your coat.
Generalizing, we postulate that all the sentences above with
comma intonation have you as the underlying subject. You is, of
course, usually deleted. In this way, second person anaphoric
reference to vocatives, including those where the vocative NP is
indeterminate, is explained in the same way as the second person
reflexives and tags shown in examples (19) to (22). Thus, what
needs explanation is the fact that certain noun phrases, apparently
really the subjects of imperative sentences, can nevertheless select
third person anaphora. We take this to mean that those sentences
have [+III person] subjects.
It might be convenient if in fact it turned out that subjects
of imperatives could be quite freely generated. There is apparently
no natural way of constraining the subjects of topmost imperatives
so that they are second person NP's. Within the present grammar,
the only possibility is to block imperatives having subjects with
other features on the head N by, for example, leaving the SJC morpheme undeleted just in case the subject of a top imperative fails
to meet the relevant conditions. Not only does this necessitate
an otherwise unmotivated blocking transformation; it also introduces
a major and unexplained difference between (top) imperatives and
related sentences dominated by S, i.e. "subjunctives". (See NOM
and (U. a-f) above.) (Generally, we refer only to topmost sentences
as "imperatives".)
Apart from a few special cases like (26), however, where there
really does seem to be a third person subject in an imperative,
the restriction to second person subjects appears to be correct.
It is beyond question that the subject of an imperative is, in some
sense, being addressed by the speaker, even in cases where the subject NP appears to be third person. The impossibility of using in
these subjects any third person NP which intrinsically implies that
the referent is NOT being addressed makes this quite clear. All of
the following are non-sentences whether taken as vocatives or
imperatives.
668
IMP - 10
(31)
(a) *Your son come here.
(b) *My ambassador to you come back.
(c) *Me go away.
(d) *Her kiss John.
It is not only in imperatives that certain 3rd person NP's can
occasionally be used to refer to the person addressed. Consider
the sentence: The reader has undoubtedly noticed several errors
in this report. On one reading it can be paraphrased in certain
circumstances by, You have undoubtedly noticed several errors in
this report of which it seems to be a stylistic variant limited
(among other things) to cases where the writer or speaker is uncertain who in particular he is addressing.
In the light of this, consider the range of apparently third
person subjects occurring in imperatives. In the first place there
are a number of examples which include or could include an underlying second person partitive, either with of or with among. For
example:
(32)
(a)
(b)
The oldest of the girls (among you) sing a
lullaby.
One of the boys (among you) run ahead.
(c) ?A girl (among you) try to thread that needle.
(33)
(a)
Everyone of you pick up \ ?h*s I towel.
(^ yourj
(b)
Every one\ . ,
Everyone JPick
j his ]
* f?your j
u
,
towe1
'
(3*0
(a) None of you move.
(b) *None move.
(c) No-one move.
(35)
(a)
Somebody^ °^
I you run to the door.
(__ ?among J
(b)
Somebody run to the door.
It would be tempting to argue from (32)-(3^) that all superficially
third person subjects of imperatives come from NP's which dominate
a second person partitive. This would give a syntactically reasonable
source for both second and third person features in anaphoric
reference to the "third person" subjects—either to the features of
the top NP or to those of the partitive. As (33) shows, it seems
that second person anaphora in such cases is preferable when the
partitive is present while third person pronouns are more readily
669
IMP - 11
used when there is no overt partitive. However, (36) suggests that
there are cases (especially those that could NOT incorporate an of
partitive, but only one with among—see (36')) which vary rather
freely between second and third person anaphora when there is no
second person partitive present.
(36)
(a)
(b)
(36') (a)
(b)
The oldest of the girls put jyour \ purse
down and come here.
v.her ;
yourself
One of the boys test(
] while I wait.
(himself
J
The oldest of the girls I/"00* you \ ...
(*of you
J
aiaaa
you
One of the boys [
Z
] ...
(*of you
J
Unfortunately for any attempt to relate the second person
characteristics of third person subjects of imperatives to the presence
within the NP of an underlying and perhaps deleted second person
partitive, there is no independent evidence for setting up such a
partitive in sentences where it fails to appear at the surface.
Moreover, second person among partitives within third person
NP's (as in (36')) allow second person anaphora only in imperatives;
they can scarcely be used, therefore, to explain the fact that third
person imperative subjects are much like 2nd person NP's. Consider
the possibilities of using second person anaphora in the following
situations. When in a higher or conjoined NP, [+11 person] dominates
[+III person] in anaphora, the result is like (37):
(37)
(a)
(b)
John and you took 1 their ( shoes to the repair
(, your )
shop last month,
You of the men who are about to leave should
speak to I *their ! supervisors immediately.
^ your J
On the other hand, when [+11 person] is in a partitive with among,
dominated by [+III person], it is the latter feature that operates
in anaphora in indicative sentences:
(38)
The brightest boys among you have already finished
\ their)
(*your J homework.
670
IMP - 12
(Note that when the second person feature is within an of partitive
there appears to be a choice, as in, The brightest of you have already
Iyour \
their homework. This is irrelevant, however, since (36')
es that among partitives would have to be postulated for
at lease some third person imperatives.)
Thus, it is only in imperatives, like (39), that second person
anaphora can be attributed to an among partitive dominated by a
third person NP. But it was a peculiarity of imperatives that the
postulation of underlying partitives was supposed to explain
(39)
The brightest boys among you finish n your \ homework
.
|
C?theirJ
as fast as I ^ou > can.
\?theyj
There is another reason for rejecting such an explanation,
anyway. There are cases of third person NP's acting as imperative
subjects which cannot possibly include partitives. One instance of
a case where the partitive seems at least a little odd has already
been given, in (35.a,b). The following, all of which are acceptable
to many people, can not have second person partitives, as we show
in (hi).
(Uo)
(a)
(b)
(c)
(d)
(e)
(Ul)
The boy in the corner stand up.
All the children in the front row be quiet.
The oldest of the girls among the English
in this group sing a folk song.
Nobody move.
Everybody hurry up.
(a) *The boy in the corner \ of
I you stand up.
(. among;
(b) *A11 the children in the front row}°
) you
(among) *
be quiet.
(c) The oldest of the girls among the English in
this RT^PIifomone f
(d) "Nobody \
among
you
sinfi a
folk
son
8«
J you move.
(e) "Everybody jamongJ you hurry up.
(In some cases the starred forms of (Ul) may be possible but not
synonymous with the parallel sentences of (Uo).)
671
IMP - 13
It seems to be necessary to recognize that while the referent
of the subject NP of an imperative is addressed by the speaker,
constraining the NP basically to the second person, nevertheless
certain third person NP's can occur with second person reference.
If a third person NP occurs in this way in an imperative subject it
may apparently select either second or third person anaphora. We
have no way of representing these facts in the grammar. It seems
best to identify reference to the person addressed with the feature
[+11 person], to ignore second person partitives as irrelevant, and
thus to exclude (Uo.a-e) and (32)-(35) from the grammar until the
relationship between reference and the features on the noun can be
more adequately dealt with.
There is another possibility, which we have not explored in
detail. We have limited the imperative to a rather narrow set of
constructions. It is likely that these are related in various ways
to a number of the forms that are excluded from this treatment:
sentences with modals, Wish-imperatives, Let-imperatives and
vocatives, for example. Thus, there are sentences with third person
NP's separated from the rest by comma intonation which act like
vocatives but include a definite description.
(1*2)
(a)
(b)
Boys, come here.
The boy in the corner, come here.
(U3)
(a)
(b)
Boys, don't (you) break that.
The boy in the corner, don't (you) break that.
(a') *Don't boys, (you) break that.
(b') *Don't the boy in the corner, (you) break that.
It may
as the
as for
person
be that sentences like (U2.a) should be derived with you
deep subject and the third person NP outside the sentence,
vocatives (cf. B.2). By a later transformation the third
NP could replace you.
Let-imperatives would provide yet another source for third
person subjects. All the following are possible.
(kk)
(a)
(b)
(c)
Let the boy in the corner stand up now.
Let nobody move.
Let all the girls among you leave at once.
The deletion of Let (which is not understood here to mean allow)
would produce satisfactory third person imperatives. However, it
would be necessary to constrain Let-deletion in all sorts of unexplained ways to obtain:
672
IMP - lU
(1*5)
(a)
(b)
Let no-one be fooled by his explanation.
Let your son come to school properly
dressed in the future.
(c) Let John be the first to go.
(d) ?Let everybody not pay much attention to him.
While excluding:
(U6)
(a) *No-one be fooled by his explanations.
(b) *Your son come to school properly dressed in
future.
(c) *John be the first to go.
(d) *Everybody not pay much attention to him.
We therefore limit the grammar to second person imperative subjects.
Although it is quite clear that this will not account for all the
data, nevertheless it seems to be the nearest approach to a correct,
though limited, generalization that can be made at present.
Further evidence that all imperatives have, in some sense,
second person subjects may come from dialogs like the following.
We are not sure how to weigh this evidence. It appears to be
relevant to the question of their deep structure, since third person
anaphora from outside the imperative is apparently impossible, even
if it occurs within the sentence itself. It is assumed in (U7) and
(U8) that the second sentence of the dialog does not constitute an
explanation to a third party but is addressed to the same person.
(1+7)
The boy in the corner stand up. |You have I not done
^*He has )
{•his1" J h<>mework(1+8)
(a)
(b)
The eldest girl among you take off her shoes.
(?She ) Drou
) You 1
ght mud in on them.
The eldest girl among you take off her shoes.
Put them in the fireplace, will j J°u\ .
The following suggests that the same phenomena occur in tags:
(1+9)
(a) The boy over there stand up, will you.
(b) *The boy over there stand up, will he.
673
IMP - 15
2.
A Note on the Vocative
We have made no attempt to include vocatives in the formal
treatment presented here, but a suggestion of how they might be
included is perhaps in place. It may be observed that while we
must distinguish between imperative subjects and true vocatives,
the two cannot co-occur:
(50)
(a) *You boys come here, boys.
(b) *Some of you men help me lift this, men.
What may be involved in instances such as these is some
process of obligatory pronominalization, or deletion of identical
material. Compare the grammatical sentences in (51) with (50):
(51)
(a) You come here, boys.
(b) Some of you help me lift this, men.
(c) ?You come here, you boys.
(d) ?Some of you help me lift this, you men.
Notice that such second person pronominalization seems to apply to
all sentences that include vocatives, not just to imperatives:
(52)
(a) *Harry, Harry is wonderful.
[+V0C]
(b) Harry, you are wonderful.
[+V0C]
(c) You, Harry, you are wonderful.
If we assumed that all sentences could have a vocative, then
we could account for the second person pronoun as a result of
pronominalization which involved a vocative and any other NP in
the sentence which happened to be referentially identical with the
vocative. Under this analysis imperatives would be constrained so
that the subject of the imperative contained a copy of the vocative
NP. The advantage of this analysis would be that it used processes
(pronominalization and equi-NP-deletion) needed elsewhere in the
grammar.
Alternatively it is possible that the sentence to which a
vocative is attached always contains a second person pronominal NP,
marked in some way as co-referential with the vocative. Then
(52.b) rather than (52.a) would be the deep structure. This would,
of course, provide a somewhat more appropriate input to imperative
transformations if they demand, as we suggest, a second person subject. Either source would effectively exclude (50).
67U
IMP - 16
C.
Imperatives and Peremptory Declaratives
Katz and Postal observe that a sentence like:
(53)
You will go home.
may be interpreted in either of two ways: (a) as a predictive
statement or (b) as an order. Thome makes the same observation
about the sentences:
(5M
(a)
(b)
You, John, will come.
You will be examined by the doctor.
On the basis of such observations, these authors propose that
sentences like (53) and (5*0 are ambiguous and may correspond to
either of two different underlying P-markers: one with, and one
without, an imperative morpheme.
There are, however, a number of significant syntactic differences between such sentences involving the "peremptory future",
and true imperatives, which lead us to analyze (53) and (5*0 as
declaratives (with a possible special interpretation) and not as
ambiguously declarative or imperative.
(a) While the subject of a true imperative must include (in the
sense suggested above) a 2nd person feature specification, this is
not true of the peremptory futures in (55). (Note that though
peremptory declaratives are usually future, they may occur in the
present tense, e.g., such things are not done here.)
(55)
(a)
Trousers will not be worn by women in this
department.
(b) *Trousers, don't be worn by women in this
department.
(b) Sentence adverbs such as certainly may occur in sentences
involving the peremptory future but not in true imperatives:
(56)
(a) You certainly won't do that,
(b) *Certainly don't do that.
(c) While true imperatives can be conjoined with one another and
peremptory futures can be conjoined with one another, a true
imperative and a peremptory future cannot in general be conjoined.
675
IMP - IT
(57)
(a)
Be a good boy while I'm away and
touch any liquor.
(b) You will be a good boy while I'm
you won't touch any liquor.
(c) *Be a good boy while I'm away and
touch any liquor.
(d) *You will be a good boy while I'm
don't touch any liquor.
don't
away and
you won't
away and
(Sentence (57.c) is possibly grammatical as a conditional imperative:
i.e., in the meaning: "If you're a good boy while I'm away, you
won't touch any liquor".)
(d) A peremptory future can be conjoined with a declarative; an
imperative in general cannot be conjoined with a declarative:
(58)
(a)
I hate girls in trousers, and you won't wear
trousers again, my dear.
(b) You will not go to see that bloody war-picture,
and you know why.
(c) *I hate girls in trousers, and don't wear
trousers again, my dear.
(d) *Don't go to see that bloody war-picture, and
you know why.
((58.c-d) must be distinguished from conditional imperatives like
Step inside and I'll hit you, which can, and indeed must be conjoined to a declarative following them.)
On the basis of these observations, we conclude that sentences
involving the peremptory future are declaratives, and do not contain
an imperative morpheme. The imperative-like quality of such
sentences is, in our view, a matter of semantic interpretation:
any statement about the future—if its confirmation depends upon
the compliance of some persona other than the speaker with the
wishes of the speaker—may have this interpretation. It may be
best to refer to this as a "pragmatic" rather than a "semantic"
aspect of the sentence.
D.
Imperatives, Requests and Questions
1.
Behavior Common to Imperatives and Requests
(a)
AUX-attraction
Chomsky pointed out in 1955 that imperatives, like questions,
requests and wishes, undergo subject-auxiliary inversion (AUXATTRACTION ). Compare:
676
IMP - 18
(59)
(a)
(b)
(c)
Don't you drink brandy?
Won't you drink a glass of brandy, please?
Don't (you) drink any brandy, now!
In non-negated imperatives such as:
(60)
(a)
(b)
(You) have some brandy,
(You) be a good boy.
inversion was said to apply to a 0 auxiliary:
(61)
You 0 be a good boy =£ 0 You be a good boy.
This vacuous permutation of a zero element permitted a uniform
treatment of subject-auxiliary inversion for imperatives but
made it hard to account for You come here, as opposed to *Do
you come here.
Thus while AUX-ATTRACTION seems to apply to negative
and perhaps emphatic imperative sentences it is not a clear example
of a characteristic that is common to imperatives and requests,
because (a) the correct account of the presence of don't in negative
imperatives may not involve the general rule AUX-ATTRACTION and (b)
plain imperatives do not involve AUX-ATTRACTION (see Section E).
(b)
Co-occurrence Restrictions
Requests and imperatives share a number of co-occurrence
restrictions. For example:
(i)
Stative verbs:
Kiparsky (1963) and others have observed that a certain class
of verbs which Lakoff (1965) calls statives, occur neither in
imperatives nor in requests:
(ii)
(62)
(a) Understand the answer.
(b) *Want more money.
(c) *Hope it rains.
(63)
(a)
(b)
(c)
f *understand the answer,~)
Would you < *want more money,
> please?
(j*hope it rains,
)
Adverbials:
Kiparsky has also observed that certain adverbials fail to occur
in imperatives and requests alike. To repeat his examples:
677
IMP - 19
(6k)
(a)
You (will) learn this language surprisinglyfast.
[28]
(b) *Would you learn this language surprisingly
fast.
[29]
(c) *Learn this language surprisingly fast. [30]
(d) Learn this language fast.
[31]
(In the surface structure of examples (6U.a,b) surprisingly is a
modifier of fast.)
Katz and Postal, as well as Lees, have noted that certain
preverbs do not normally occur in imperative sentences:
(65)
(a) *Hardly
(b) *Scarcely > finish your work.
(c) *Almost
This observation also holds for requests:
(66)
(a)
(b)
(c)
' *hardly "j
Would you • "scarcely > finish your work, please?
*almost /
Chomsky (1955) makes the observation that imperatives do not
occur with a past time adverb:
(67) *Come yesterday.
Kiparsky notes that the same restriction holds for requests:
(68) *Would you come yesterday, please?
Please occurs in both requests and imperatives as in:
(69)
(a)
(b)
Won't you step in, please?
Step in, please?
On the basis of sentences like (69.a,b), Kiparsky proposed
that, in their underlying structures, requests include an IMP(erative)
morpheme, and that the underlying structures of requests and true
imperatives differ only in the auxiliaries involved.
2.
Differences between Imperatives and Requests
There are, however, a number of properties which are not
shared by requests and imperatives.
678
IMP - 20
(a)
Third Person Subjects
Imperatives and requests differ significantly with respect to
the apparently third person subjects which can appear in them.
Generative grammarians agree that in English the subject of an
imperative must correspond to the person (or at least one of the
persons) addressed in the sentence. Kiparsky claims that the subjects of requests (like imperatives) "are confined to the 2nd
person singular and plural" and maintains that (70) is ungrammatical:
(TO)
Would your son look at himself in the mirror, please?
The above sentence, however, is quite acceptable in the following
context:
"So your son, the prince, does not believe that Baby Jane
kissed him while he was asleep? Would your son look at
himself in the mirror, please? The rouge is still on his
left cheek."
The following also seem to be grammatical:
(71)
(a)
(b)
Would your son come over, please, and help
us with the planting?
Could your soldiers please help us build this
bridge, General Lee?
Sentences such as (70) and (71) where a request is made of a person
not addressed in the discourse, usually imply that the request should
be communicated to the person concerned. Sentence (70) perhaps
means: "Would you suggest to your son that he look at himself in
the mirror?" Sentence (71.b) means something like: "Could you
please get your soldiers to help us build the bridge, General Lee?"
In true imperatives as we saw above, it is crucial that the subject
be the person addressed. Compare the requests in (71) with the true
corresponding imperatives in (72):
(72)
(a) *Your son come over, please, and help us with
the planting,
(b) *Your soldiers please help us build this
bridge, General Lee.
This difference between imperatives and requests is exhibited rather
clearly by:
(73)
Would you and your guests please not make so much
noise?
679
IMP - 21
Conjoined NP's like you and your guests may occur as subjects of
requests. If such NP's are derived from two underlying sentences,
then one expects (7M to be grammatical, as it is:
(7*0
Would your guests please not make so much noise?
Notice however, that the imperatives corresponding to (73) and
(7*0 are ungrammatical:
(75)
(a) *Please don't you and your guests make so
much noise,
(b) *Please don't your guests make so much noise.
This we consider to be a significant difference between the two
sentence types.
(b)
Adverbials
The restrictions on sentence adverbs that may occur in
requests are not quite the same as those on sentence adverbs that
may occur in imperatives. Compare:
(76)
(c)
(a) Could you possibly come over please?
(b) Will you perhaps have a cup of coffee with us?
(c) *Possibly come over, please?
(d) *Perhaps have a cup of coffee with us. (cf.
D.l.b.ii above)
Passive Forms
There are passive requests formed with can, can't, could and
couldn't (but not with will, won't, would and wouldn't):
(77)
(a)
(b)
(c)
(d)
Can the soup be served after the hors d'oeuvre,
please?
Can't the curtains please be drawn?
Could the tables please be decorated with
flowers ?
Couldn't the piano be removed, please?
Passive imperatives are generally ungrammatical:
(78)
(a) *Be allowed to leave.
(b) ?Be flattered by what he will say.
(c) *Be elected chairman.
680
IMP - 22
In negative sentences it is apparently much easier to obtain
grammatical forms, such as:
(79)
(a)
(b)
Don't be hurt by what he says,
Don't be misled by his flattery.
We do not attach too much weight to the fact that imperatives differ
from requests in regard to the passive, since it would appear that
the imperative modal is more like will than, say, can, and, as we
observed, will does not occur in passive requests.
(d)
Negatives on Modals
Negatives associated with the modals in requests do not carry
negative force. Thus each of the following members of the pair
expresses roughly the same request:
(80)
(a)
(b)
Will you help me, please?
Won't you help me, please?
(81)
(a)
(b)
Can you please move over a little?
Can't you please move over a little?
Negatives associated with the imperative auxiliary, on the other
hand, carry negative force. Thus the members of the following pair
are obviously not equivalent:
(82)
(a)
(b)
Help me, please.
Don't help me, please.
Notice, also, that while (83.a) has a double-negative interpretation,
(83.b) is a simple negative.
(83)
(a) ?Please don't not come here any more.
(b) Won't you please not come here any more.
We do not know how much weight to attach to this observation.
It is not clear what the source of the additional semantically
rather empty negative is (cf. INTERROG, NEG) and consequently the
significance of its appearing in both questions and requests but
not in commands is still open.
We suggest, on the strength of most of this evidence, that
the underlying structures of requests and imperatives must be distinguished to an extent greater than Kiparsky allows. We believe,
in fact, that requests are probably best treated as a special subclass of (yes-no) questions, although this analysis, too, presents
681
IMP - 23
certain problems. Requests and yes-no questions have, in addition
to subject-auxiliary inversion, several other common characteristics,
which, unlike inversion, are not shared by imperatives.
3.
(a)
Characteristics Common to Requests and Questions
Negatives on Modals
Negatives associated with modals (and other auxiliaries) in yesno questions, may, like negatives associated with modals in requests,
lack negative force. Compare the following examples with (80) and
(81):
(b)
(8U)
(a)
(b)
Will he help me?
Won't he help me?
(85)
(a)
(b)
Can these people move over a little?
Can't these people move over a little?
Indirect Quotations
In indirect quotation, embedded requests, like some embedded
yes-no questions (which we do not deal with explicitly in INTERROG)
are introduced by if_:
(86)
(a)
(b)
He asked John if he would please play the
piano,
He asked John if he thought it would rain.
Embedded yes-no questions may also, however, be introduced by
whether, while embedded requests introduced by whether are
questionable for some speakers:
(87)
(a) ?He asked John whether he would please play
the piano,
(b) He asked John whether he thought it would rain.
Embedded imperatives, on the other hand, never are introduced by
if; they may start with that, which never introduces questions or
requests:
(88)
(c)
I demanded that he play the piano.
Tags
Neither yes-no questions nor requests admit tags, while
imperatives do.
682
IMP - 2k
(69)
(d)
(a) *Will John come in, will he?
(b) *Will you please come in, will you?
Intonation
Yes-no questions and requests both generally have rising
intonation:
(90)
(a)
(b)
Is it going to rain?
Would you please pass the salt?
But imperatives generally have falling intonation:
(91)
h.
(a)
Please pass the salt.
Differences Between Questions and Requests
Some-any suppletion
Yes-no questions can undergo SOME-ANY SUPPLETION while requests
cannot:
(92)
(b)
(a)
(b)
Will he give you some/any money?
Will you give me some/*any money,
Conjunction
Yes-no questions may be conjoined with other yes-no questions
and requests with other requests, but a yes-no question and a request cannot be conjoined very comfortably:
(93)
(c)
(a)
Is Mary going to do the dishes, and is John
going to take out the trash?
(b) Will you please do the dishes, and will you
please take out the trash?
(c) ?Is Mary going to do the dishes, and will you
please take out the trash?
(d) ?Will you please do the dishes, and is John
going to take out the trash?
Please
Notice, moreover, that although please can occur in certain
questions as well as in requests, in requests the word please can
be inserted after the subject while in questions this is not
possible. Compare the following:
683
IMP - 25
(d)
(9*0
(a)
(b)
Will you take the trash out, please?
What is the exact time, please?
(95)
(a) Will you please take the trash out?
(b) *What please is the exact time?
Negation
Although, as has been pointed out above, a negative on
the modal of questions and requests does not result in a negative
sentence, it appears that only a request (and not a question) must
have a clearly negative interpretation when the negative comes
after the subject. Thus, as questions the following can have
roughly the same meaning, (96.a) being more formal than (96.b).
On this reading neither differs significantly from (96.c).
(96)
(a)
(b)
(c)
Will John not be going to town?
Won't John be going to town?
Will John be going to town?
Compare, as requests:
(97)
(a)
(b)
(c)
Will you please not jump in before I get out?
Won't you please jump in before I get out?
Will you please jump in before I get out?
It is impossible to get readings of the requests, (97.a)
and (97.b), that are paraphrases. In requests, then, a negative
not directly associated with an auxiliary must have full negative
force, though in questions it may lack this. Such a difference
between requests and questions may constitute a rather serious
obstacle to the claim that the former are a special sub-type of
questions. This is consistent with our analysis of Yes/No
questions (see INTERROG) which, we argue, are conjuncts, differing only in that one is negative, the other positive. Either
the negative or the positive sentence is deleted on the way to
the surface, accounting for the lack of negative force in many
negative questions. However, requests cannot be regarded as
relatively uncommitted attempts to discover which of a related
pair of positive and negative statements is true. A request is
an endeavor to bring about one or the other of the two possible
states of affairs. For example, in (97.a and b) to bring it
about that the person addressed (a) refrains from jumping in, and
(b) Jumps in (respectively) before the speaker gets out. Only
(b) is at all similar in meaning to (97.c).
68U
IMP - 26
Thus, any attempt to associate requests and yes/no questions
will need to set up a separate semantic apparatus, presumably
working on only one of the related conjuncts. It is not clear
that this can be done economically or even consistently. This
does not, of course, constitute positive evidence for regarding
requests as a kind of imperative.
5.
Conclusion
In spite of the problems raised by these differences, it may
be possible to treat requests as a subclass of yes-no questions
with certain special syntactic properties, some at least stemming
from their peculiar semantic characteristics.
Just as there is no clear reason to posit an Imperative
morpheme, SJC, in the underlying structure of peremptory declaratives, so there is no clear reason to posit such a morpheme in
the underlying structure of requests. Requests do not undergo
any of the transformations, and do not obey any of the surface
constraints which are exclusively characteristic of imperatives.
(AUX-ATTRACTION in requests can be triggered by WH just as well
as it can by SJC.)
The analysis of requests as questions with a special interpretation receives further support from the fact that in addition
to examples in which the form of the request is that of a yes-no
question, we find such examples as:
(98)
Why don't you (please) leave me alone?
The suggestion is that any declarative or interrogative can be
interpreted as a peremptory declarative or request, respectively,
provided that it obeys appropriate selectional restrictions. It
is not clear how far such a device will make it possible to explain
the interrelationships between the various forms which we have
noted. However it is clear that the earlier assumptions, which
identified imperatives and requests, and failed to account for the
close ties between the latter and questions, leave too much of
the syntax unexplained.
E.
The Underlying Auxiliary of Imperatives
1.
The Presence of a Modal
Lees (196i+b), and Klima (l961*c), both make the following observation: do-support in non-imperative sentences depends on the first
element that follows TENSE in the auxiliary or in the verb phrase;
do-support does not occur if this element is be_, the auxiliary have,
or a modal.
685
IMP - 27
(99)
(a)
(b)
(c)
(d)
*He doesn't be nice.
*He doesn't have done it.
*Does he be nice?
*He does have done it.
In these cases EMPH or NEG moves to the right of be_, have, or a
modal. Emphatic and negative imperatives, however, require dosupport , even for the verb be:
(100) (a)
(b)
(c)
(d)
Do be
Do be
Don't
Don't
nice
there by five.
be silly.
be sitting there then.
They take this as evidence that all imperatives contain a modal
element which operatives in Preverbal Particle Placement, so that, for
example, we get (101) and then (102). (Note that in this grammar SJC
covers TNS+Modal but at this point we follow Klima's model.)
(101)
(102)
NEG you TNS will be -ing sit there then => (by PPPrule)
you TNS will not be -ing sit there then.
If imperatives did not have a modal in their underlying structure,
we would instead have a derivation from (101') to (102*) by Preverbal Particle Placement, which, on deletion of you would yield
the incorrect (101"), or (102") if AUX-ATTRACTION had also applied.
(101')
NEG you TNS be -ing sit there then =4> [by PPP-rule]
(102')
You TNS be not -ing sit there then.
(101") "Aren't sitting there then.
(102") *Be not sitting there then.
If on the other hand we accept Lees' and Klima's claim, appropriate
deletions after AUX-ATTRACTION will lead to the application of DOSUPPORT, giving (100.d) from something like (102).
2.
The Choice of a Modal
Chomsky (1955) postulated that imperatives are derived from
strings containing any one of those modals which never occur with
past time specifications. This would automatically ensure that
imperatives would only occur with non-past adverbials, but would
permit multiple derivations for apparently unambiguous sentences.
According to Klima (l96Uc) the modal will accounts for the formation
of the usual tag question by a copying rule which derives (10U)
from (103):
686
IMP - 28
(103)
(You will) close the door.
(10U)
(You will) close the door, won't you?
Kiparsky (1963), however, has drawn attention to the fact that
other tags occur after imperatives (cf. Section II.E.).
Lees (l96Ub) argues that the underlying modal element is a
zero morpheme, which he calls IMP, but which, in our analysis, is
taken to be identical with the subjunctive (SJC). This marker
functions as a modal in such rules as AUX-ATTRACTION and PREVERBAL
PARTICLE PLACEMENT.
Lees* analysis, incorporating a special zero modal that also
acts as an affix, is based on the observation that the ordinary
affirmative imperative of the verb be has the form (105) and not
(106):
~"
(105)
Be there by five.
(106) "Are there by five.
He points out that, morphologically, the imperative in (105) is
not the ordinary finite verb-form (resulting from the attachment
of the element TNS to the underlying verb-stem). He concludes that
the imperative is a verbal affix in its own right, parallel to TNS
but with no effect on the verb to which it is attached. No ad hoc
rule is then needed for deleting a postulated auxiliary in imperatives, since the auxiliary is a phonologically unrealized morpheme,
moved onto the verb or triggering DO-support in appropriate ways.
Were it not treated as an affix, but as an ordinary modal, it would
require special deletion and would never trigger DO-SUPPORT. As
(107) shows, D0-SUPP0RT must apply (as if SJC were TNS), when EMPH
or NEG has prevented it from moving onto the verb.
(107) (a)
(b)
Do come here,
Don't come here.
However, the situation is more complicated. Consider the
derivation of the following sentence, in which the subject, you,
has not been deleted.
(108)
You sit down.
After AUX-ATTRACTION has taken place, this sentence would have looked
something like (109).
(109)
SJC you sit down.
687
IMP - 29
Since the "affix", SJC, would be prevented by you from moving onto
the verb, it would trigger DO-SUPPORT, resulting in (110), which
is ungrammatical for most speakers.
(110) *Do you sit down.
To generate (108), as we must, we could either delete SJC just
in case neither EMPH nor NEG is present, or alternatively perform
AUX-ATTRACTION only when one of those morphemes is present. The
first solution is essentially the one rejected by Lees. Both involve ad hoc manipulation of the rules, but it appears that there
is simply a certain amount of untidiness in the data which Lees'
solution could handle no better than any other. In our rules we
have chosen another possibility. It is apparent that the rule of
AUX-ATTRACTION which is applying here is rather different from the
general rule of that name. Apart from possible constraints on
the application of the rule mentioned above, there is the fact that
we no longer have any motivation for an initial IMP morpheme, since
we have a special imperative form in the AUX—i.e. SJC. Hence there
is nothing parallel to WH or t+Affect] to attract the AUX. It is
possible that we are dealing with a different rule, and thus that
this IMPERATIVE-SUBJ-AUX-INVERSION can follow Affix-switching. Since
SJC acts as an affix it will then be available for inversion with
the subject only if there is a NEG or EMPH present to prevent it
from moving onto the verb. To prevent (110) it is necessary to
make Y0U-DELETI0N obligatory if do precedes it. This is well
motivated, though, as we show in discussing TOP SJC DELETION
(rule 3, below), it has some unfortunate consequences.
F.
Tagged Imperatives
Two proposals have been made to account for tags in a generative grammar: (a) a copying rule and (b) conjunction reduction. In
the copying-rule proposal, (cf. Klima, 196U) a sentence such as
(lll.b) is derived by copying the auxiliary and the (pronominalized)
subject of the input sentence (ill.a) and appending them as a tag:
(ill) (a)
(b)
Writers will never accept suggestions.
—>
Writers will never accept suggestions, will they?
Both Lees and Hasegawa have noted that this rule will not account
for the peculiarities of imperative tags. In previous analyses, in
which imperatives and requests were closely related, it seemed
reasonable to derive tag-imperatives from requests, but to do so
in fact introduces additional problems; not only is it hard to
see how tags such as those in (112) can be accounted for by copying,
it is also to be noted that requests do not admit any tags as shown
in (113) (cf. Section II.D.3.C, above).
688
IMP - 30
(112)
Do help me, won't you?
(113) *Will you please come in, will you?
A copying rule that derived tagged imperatives from requests would
require that a modal-deletion rule apply to the underlying request
whenever the copying rule has applied. Thus, imperative tags
would be the only case where tag-formation entailed an obligatory
deletion in the original sentence, for there are indicative
sentences with both occurrences of the auxiliary and subject, such
as John will come, won't he?
There are other forms which a copying rule can't handle.
As has previously been noted, passives may occur in requests
containing the modals can and could:
(llU)
Could the windows please be opened?
No tagged imperatives exist for such requests:
(115)
(a) *The windows please be opened, could they?
(b) *The windows be opened, could they please?
Hence if tagged imperatives are derived by a copying rule from
requests, an ad hoc condition must block the application of the
rule to passives. For these reasons it seems to us that the
copying rule proposal must be rejected for tagged imperatives.
In the second proposal for deriving tagged imperatives, the
conjunction-reduction proposal (cf. Lees, I96U), tagged imperatives
are derived in two steps: (a) sentence conjunction and (b) reduction of the second sentence, just in case it meets a certain set of
conditions. These conditions are: (a) the preceding imperative
must not contain NEG and (b) the modal in the tag is will, with or
without, not. We can easily extend this condition, however, to
include other tags as in the following:
(116)
(a)
(b)
(c)
/'can
Come here, / can't
\ could
you?
A derivation of a tagged imperative would begin with the following
two underlying strings. For the moment it is irrelevant whether
(117.a) and (117.b) must be conjoined in some way in the base.
(117)
(a)
(b)
You SJC come with us
CONJ [you will come with us]
[+or]
[+WH]
[NEG you will come with us]
689
IMP - 31
The first step in the derivation is the conversion of (llT.b) into
an alternative question and then to the yes-no question (llti.b):
(118) (a)
(b)
you SJC come with us .
WH you will come with us.
At this point a problem arises. (119) is ungrammatical and so,
it seems, is any alternative version with a different conjunction.
(119) *Come with us and will you come with us?
Hartung (I96U), pp. 1+3-1+5, has argued in favor of extending the
power of transformations to combine sentences in such a way that
a rule could reduce the two parts of (118) directly to (120).
(120) ?Come with us, will you come with us.
The repeated material would be removed by rules required
independently in the grammar, to give (121).
(121)
Come with us, will you?
We do not in fact provide rules to generate any tags in this
grammar. For further discussion see INTERROG III.B.3.
G.
BLOCKING PROBLEMS
It is necessary to block imperative sentences if they
(a)
contain a subject NP which is not [+IIperson] (but see
section B). This enables us to exclude
(122) (a) *Me stand up .
(b) *Your father come here.
(c) *Him try to run faster.
(b)
have, as subject, an NP which is not an Agent. (This
assumes that certain intransitives, such as run have agentive
subjects. See LEX for discussion.) In this way we exclude
stative verbs from imperatives, as in (123).
(123) (a) Understand this part of the book.
(b) *Be tall.
(c) *Hear all of the discussion.
These constraints do not apply to embedded imperatives, i.e. those
sentences that we refer to as subjunctives. Thus, the following are
690
IMP - 32
quite acceptable:
(12U) (a)
(b)
(c)
(d)
It is necessary that I stand up.
I demand that your father come here.
It is imperative that you understand this part
of the book.
I propose that we hear all of his arguments.
Consequently, the constraints on imperatives must be transformational rather than selectional or sub-categorial. Given our
assumption that subjunctives are just embedded imperatives (which may
be something of an oversimplification) it is necessary to use a
last-cyclic transformation to block imperatives containing subjects
which are either not second person or non-agentive. This will
recognize the SJC morpheme in the top S. (Recall that we arbitrarily
chose not to allow such [+III person] imperatives as (26)).
In subjunctives, it is necessary that SJC be deleted in order
to exclude such sentences as (125).
(125) *I insist that John does not be given that fellowship.
In embedded sentences SJC simply prevents the verb from acquiring
an indicative form such as:
(126) (a) *Bill demanded that John left.
(b) *Bill will demand that John leaves.
It can then be deleted. Since SJC and TNS are mutually exclusive in
our base rules, no other mechanism is required to prevent (126)
from being generated. As long as SJC has been generated in the base,
that is enough. There is one small problem in using SJC in this way.
To prevent (125) it is necessary that SJC be deleted before DOSUPPORT applies. But the deletion of SJC must be effected by the
higher sentence into which it is embedded. Consequently, it must
take place on a cycle higher than the sentence in which it appears.
If DO-SUPPORT (see NEG page 59) is always to apply later than
EMBEDDED-SJC-DELETE the former rule must be last cyclic yet apply to
embedded sentences. Although such last-cyclic rules have been
discussed (e.g. by Ross (1967)), we have generally assumed in this
grammar that last-cyclic rules apply only to the topmost sentence—
because of the convention that transformations do not in general look
below the sentence on which they are working. Nevertheless, for this
particular purpose we assume that DO-SUPPORT is last cyclic, yet
applies to all appropriate parts of the string.
The SJC of all embedded sentences will already have been deleted
by then, but EMBEDDED-SJC-DELETE only applies to embedded SJC's,
because of its form. Consequently, when DO-SUPPORT applies, SJC
can still be present in the topmost sentences and it, appropriately,
triggers that rule.
691
IMP - 33
We can now return to the problem of blocking third person or
non-stative imperatives like (122) and (123) respectively but not
subjunctives like (12U). If a non-terminal like SJC is left in any
output string it is reasonable to assume that that string should block.
We have deleted all instances of SJC in lower sentences, by EMBEDDEDSJC-DELETE. Consequently (12U) can be generated. We now propose a
last-cyclic TOP-SJC-DELETE to follow DO-SUPPORT, deleting SJC just
in case both (l) the subject is [+11 person] and (2) the subject is
[+Agent].
Thus, although like Lees (l96Ub) we have a single morpheme
acting as both modal and affix we do not specifically give it zero
phonological shape, allowing it to disappear, but use that same
morpheme to block unwanted sentences. The process, as we have
described it, is reasonably neat. (Compare discussion of Lees in
E.2 above.)
Now, since we no longer have an initial IMP morpheme there is
little motivation for having the general AUX-ATTRACTION rule apply
to imperatives. (See Katz and Postal (l96Ub); NEG p.57; and E.2
above.) We can account better for the data, especially examples
(108) - (110) above, if we postulate a late rule IMPERATIVE-SUBJAUX-INVERSION, which inverts subject and AUX. This must follow
AFFIX-SHIFT, to allow SJC to move onto the verb in (110), You come
here, leaving nothing dominated by AUX in that sentence. It precedes
TOP-SJC-DELETE, of course.
We are probably losing a generalization by completely separating
S-INITIAL-AUX-INVERSION and IMPERATIVE-SUBJ-AUX-INVERSION, and there
may well be some way of recapturing the fact that these two rules
possess much in common while accounting for all the data. However,
sentences like Hardly ever did he go, where TNS is prevented from
moving onto the verb solely by the presence of he_ to its right, indicate that S-INITIAL-AUX-ATTRACT must precede AFFIX-SHIFT.
III.
TRANSFORMATIONAL RULES
The following rules significantly affect the derivation of
imperatives but are given elsewhere in the UESP grammar:
1.
Reflexivization
PRO Rule (p.U6)
2.
Affix-Shift
NEG Rule 8.
3.
DO-Support
NEG Rule 10.
U.
NEG-Contraction
NEG Rule 11.
692
IMP - 3^
1.
Embedded SJC Deletion (Obligatory)
S.I.
x
1
S.C.
s[
X
SJC
X]
X
2
3
U
5
Delete 3.
Conditions:
1. Obligatory
2. 1 or 5 is not null
Notes:
1. Condition (2) is intended to ensure that the rule applies
to embedded instances of SJC. Depending on the analysis of adverbs
in such sentences as Come here immediately, it may be necessary to
change the form of this condition.
2. The rule must follow TO-REPLACE-AUX (see NOM) so that the
AUX is not empty at the stage when that rule applies. Then ve can
obtain either (127) or (128):
(127)
It is important for John to come soon.
(128)
It is important that John come soon.
3. The rule must precede DO-SUPPORT (see NEG), in order to
obtain (129) rather than (130). This distinguishes the rule sharply
from TOP SJC DELETION. (Rule 3, below).
(129)
I insist that John not come so often.
(130) *I insist that John do not come so often.
h.
The rule need not precede either AFFIX SHIFT or YOU DELETION.
Examples:
A.
Grammatical
(131) (a)
(b)
(c)
I insist that you not leave as early as John.
It is important that he understand the answer.
I demand to see Bill. (with TO REPLACE AUX)
Notes:
1. Example (131.a) is generated rather than (132.a) because
SJC is deleted before DO-SUPPORT applies (assuming, as we have not
693
IMP - 35
done elsewhere, that DO-SUPPORT is last cyclic).
2. Example (I31.b) is obtained unlike (l32.b) because SJC
has been deleted independently of TOP SJC DELETE - which would have
failed to delete SJC, blocking the sentence, because he_ is neither
second person nor Agent.
B.
Ungrammatical - excluded
(132) (a) *I insist that you do not leave as early as John,
(b) *He understand the answer.
2.
Imperative Subject - AUX Inversion (Obligatory)
S.I.
(S CONJ)*
1
S.C.
1)
2.
#
X
NP
X SJC (NEG), X
2
3
k
5
6
7
Add 6 as left sister of h.
Delete 6.
Condition:
1) The rule applies in the last cycle.
2) 5 does not contain [+V],
Note:
The rule follows AFFIX SHIFT. Condition (2) then prevents it
from applying to You come here, since SJC is to the right of come
when it would apply.
Examples:
A.
Grammatical
(133) (a) Do come soon.
(b) Please do hurry.
(c) Don't run.
(d) Don't you run.
(e) ?Do someone help him quickly.
Notes:
1) In (a), (b) and (e) EMPH prevents SJC moving onto the verb;
in (c) and (d), NEG does. Compare (a) and (b) with (lS^.a.b) which contain no EMPH.
2) We include (e) since, although questionable,
as bad as (135). The latter can be easily excluded by
obligatory application of YOU-DELETION (q.v.), and the
handled by more general rules if (133.e) is included.
69U
it is not nearly
a well-motivated
data can be
In fact we
IMP - 36
have no way of obtaining (l33.e) in this grammar because we do not
have a [+11 person] "someone", and our rules (see rule 3 below)
exclude third person subjects in imperatives. But if we could
get someone help me I we would generate (l33.e).
B.
Ungrammatical - excluded
(13U) (a) *Do come soon.
(b) *Please do hurry.
Note:
These must be understood to contain no EMPH. Consequently
SJC is to the right of the verb and condition (2) blocks application
of the rule.
C.
Excluded by Other Rules
(135) *Do you help him quickly.
Excluded by YOU-DELETION (rule U, below)
3.
Top SJC Deletion (Obligatory)
S.I.
X
)),
This makes them quite similar too to verbs like load and fill
which are specially marked to allow the Locative to move into
subject position, even though, as we remarked above (see example
(35)) there are no clear instances of parallel nouns and verbs
allowing preposing.
c.
The Notion of Case in Relation to Nominal Heads
The preceding discussion of Locatives raises an interesting
problem which is in a sense fundamental to this entire paper. What
is it that distinguishes a case from other kinds of complements
on a head? Fillmore (1967a ) discusses the question in a general
way in relation to verbs but at the one point where he suggests
that cases may appear on certain nouns does not consider whether
the notion of case can conveniently be extended to nouns. We
merely outline some of the problems here. From time to time we
shall return to it, especially when dealing with possessives
and above all in Section C.l.e, where we deal with the
suggested distinction between alienable and inalienable possessives.
It is essential to recognize that the notion of case which
has been developed within transformational theory, especially by
Fillmore, appears to be most centrally concerned with the subcategorical, selectional and other semantic behavior of ordinary
lexical verbs so that as soon as one attempts to extend the notion
to apply also to nouns at the head of a construction some sort of
modifications, on at least subsidiary criteria, seem to be required.
The head nouns of (2l)-(23) present little major problem for this
grammar. There are minor difficulties in maintaining that S and
NP are alike in the base, where a noun and the related verbs fail
to act alike (the noun sometimes taking a restricted set of cases,
for example) but where there are pairs like destruction-destroy,
proclamation-proclaim, death-die, an NP which is Agent, Neutral or
Dative on the noun appears to be in essentially the same relation
to the head and to other NP's in the construction as it would be
if Agent, etc., on the verb. The same core of meaning is involved;
on the whole, the same selectional restrictions apply in the
enemy's destruction of... and the enemy destroyed....
This relationship between the accepted, reasonably well
established cases and verbal heads is brought out by considering
the criteria used to distinguish one case from another. As we
point out in LEX, these are far from satisfactory; nevertheless
Agent, for example, is semantically distinguished from Dative by
720
GEN - 22
the degree to which the entity referred to by the NP under the case
node is responsible for initiating and carrying out some action
characterized by the verbal head. (There are a number of syntactic
consequences which need not concern us here.) If the verb
involves no action at all, like know, there will be no Agent in
the case frame, only Dative. On the other hand, a verb like give
selects both these cases since the notion of giving necessarily
involves an active giver and a relatively passive receiver. There
appears to be a very close relationship still little understood
between aspects of the central, essential meaning of a verb and
the case frame it selects.
For the present it is enough to show that extending the notion
of case to apply to the complements of nouns in the base makes it
necessary that there be aspects of the meaning of nouns, like those
relations in the meaning of the verb give which determine the case
framework selected by the head of a construction. Nouns like
destruction offer few serious difficulties, but for head nouns like
weather, edge, head, mother, house, etc. (in examples (lT)-(20),
(30)-(33)) it is necessary to determine whether the notion of
case can have any meaning comparable to that which it has in relation to verbs, and, if so, whether the cases that occur on nouns
are limited to those that occur on verbs. It is hard to see how
a noun like table or dog could be analyzed as possessing relational
aspects of meaning in any way comparable to that found for verbs.
Moreover, although it may be possible to isolate appropriate
aspects of the meaning of Father so that John in John's father
comes from a deep structure case on the head, it is by no means
obvious that the case involved is one that even occurs on
verbs.
Let us re-examine the putative Locatives of (30) and (31) in
the light of these observations. First of all consider verbs like
load or arrive which select a Locative case. In both there is some
specific aspect of the meaning of the verb which requires a location. Loading cannot be carried out without some place onto (into)
which things are loaded; in arriving it is necessary that one reach
a place—which may or may not be mentioned. Directly related to
this, there is probably an optional Locative on the nominal arrival,
as on arrive;
(36) (a)
(b)
John arrived at the airport,
John's arrival at the airport
The question is whether in the woods in (30.a) is a Locative case
at all. If, instead, it is a locative adverb, this alone would
explain why it failed to prepose, and we could make Locatives on
721
GEN - 23
NP's generally preposable. (30.a) was one of the paradigm examples
of complements within an NP leading to Chomsky's formulation of the
x convention. We are concerned with the phrase house in the woods
as it occurs in
(37)
John's house in the woods
Chomsky argues that this cannot be derived from the ungrammatical
(in most dialects)
(38) *John's house that is in the woods
He argues that there are, in effect, two sources for (30.a), the
house in the woods, one a relative the house that is in the woods,
the other a phrase structure expansion of NP that includes
complements. Only the latter derivation can yield (37). If we
paraphrase the central meaning of house by "something to live in"
then the two meanings of (30.a) seem to be, vaguely:
(39) (a)
(b)
something to live in that is in the woods (Relative)
something to live in the woods in (NP complement)
The question is whether the Locative, in the woods, is a case on
the noun house when it is a complement on it. There is an alternative. Certain adverbs clearly occur in noun phrases, as in (Ho).
(HO) (a) John's arrival yesterday
(b) *John's arrival which was yesterday
It is quite possible that the Locative of (30.a) occurs outside
the "proposition" (i.e. Nominal) of NP, as an adverb. The possibility that Locatives occur in more than one place in the phrase
structure has often been remarked on—for example by Chomsky (1965)
and Fillmore (1967a). Whether this represents an example of a
Locative occurring outside the proposition (or Nominal) like,
perhaps, the second Locative in "He keeps his money in the bank
in Chicago," or is selected as a case by the head of the construction, depends on how it relates to the central meaning of house.
It seems best to leave this as an open question, and although, for
visibility and intensity it is at least as likely as for house,
that the Locatives are adverbial we shall not propose formal criteria
at this point for distinguishing this class. Consequently we still
require nouns to be specially marked for Locative preposing just
as verbs are.
722
GEN - 2k
However, it was not our prime purpose in this section to provide a solution to the Locative problem. We wanted to introduce
in a general way the question of what it means for a deep structure
complement to be a case.
d.
(i)
Possessives are not Derived from a Special Dative
When we turn to the problem of genitives appearing on "pure"
nouns (i.e. nouns relatively unrelated to verbs; it becomes more
difficult to see how far deep cases underlie them—they are, of
course, roughly identifiable with the "possessives" tentatively
set out in example (10). There we referred to forms like:
(Ul) (a)
(b)
(c)
(d)
John's hat
the man's arm
the farmer's daughter
that hotel's entrance
Although it is not necessary that all these come from a single
source in deep structure, that was almost certainly assumed by
transformational grammarians at one time, when a relative clause
containing have seemed to provide a reasonable source for virtually
all "possessives." Smith (196U), for example, displays no awareness of any need to distinguish different kinds of possessive.
Although, as we have argued above, it is impossible to derive all
such genitives from relative clauses, there is still the possibility that they all derive from a single case, occurring on each
of the head nouns of (Ul). We shall very soon reject this possibility, but it is instructive to see how far it will take us„
If we take into account only such possessives as (Ul„a-d) it
seems reasonable, at first, to postulate a single source, the
most likely case being the Dative. We might try to construct an
argument for deriving all possessives from that case, in something
like the following way. First of all, whatever semantic relation
holds between genitive and head in (Ul.a-d) appears to hold between surface subject and predicate NP in the parallel forms of
(U2). Any strangeness in the simple forms of (U2.b) and (U2.d)
would be attributable to the fact that in these we are directly
asserting what is in general assumed to be the case. The presence
of a single main verb, have, in all these sentences would seem to
argue for deriving all the genitives of (Ul) (subjects in U2) from
one case.
723
GEN - 25
(U2) (a) John has a hat.
(b) ?The man has an arm. (The man has a sore arm.)
(c) The farmer has a daughter.
(d) ?The hotel has an entrance. (The hotel has a
fine entrance.)
We might then notice how (k2.a) is able to appear in (l*3.b)
a paraphrase, admittedly rather clumsy, of a sentence with give
as the main verb.
(U3) (a) Bill gave John a hat.
(b) ?Bill brought it about that John had a hat.
John in (U3.a) must be a Dative, and the same basic semantic relationships seem to hold between John, Bill, hat and whatever verbal
elements axe present, in both sentences. This constitutes a prima
facie argument for analyzing John in (1+2.a) as a deep structure
Dative. Further support for this analysis, and a possible way of
extending it to the other genitives of (Ul), is provided by the
following, in which appropriate NP's are more or less successfully
associated with the characteristically Dative preposition to in
sentences that seem to preserve the same essential semantic relations as were found in \kl) and (U2).
f
(kk) (a) The hat belongs to John.
(b) ?The arm belongs tp_ the man.
(c)??The daughter belongs to_ the farmer.
(c») That little girl belongs to the lady sitting over
there in a red dress.
(cM) She was always a good daughter to her old father.
(d)??The entrance belongs to that hotel,
(d') This entrance belongs to the hotel next door.
(dM) This is the entrance to_ the hotel.
However, even on the basis of the limited data given in (Ul)(kk) it turns out to be quite impossible to argue effectively for
a single deep structure case. In the first place, it is certainly
wrong to attribute too much significance to the appearance of
have in all the sentences of (U2). Have is associated at least
as strongly with Locatives as with Datives, as in the following,
from Fillmore (1967).
C+5) (a)
(b)
There are many toys in the box. [85]
The box has many toys in it. [86]
72U
GEN - 26
Cf. also:
(U6) (a)
(b)
That bicycle has a bell.
There's a bell on that bicycle.
(U7) (a)
(b)
That door has a key.
There's a key to that door.
(U8) (a)
(b)
That door has a lock.
There's a lock on that door.
Fillmore and Bach (1967b) have argued that have in such forms
represents a late insertion, and though their arguments for a
completely empty V in deep structure are not fully convincing,
their data seems to provide abundant evidence that the surface
subject of have does not necessarily come from a Dative—or any
other single case selected by a verb.
Moreover, one of the main criteria for a Dative case in this
grammar is that the dominated NP be animate. (See LEX and
Fillmore (1967a).) In the verbs this seems to be quite satisfactory, and it would certainly be hard to justify allowing an
inanimate HP like that hotel in (Ul.d) to fall under the Dative
just in case it occurred under a noun head, or in the underlying
structure that made it surface subject of have. Thus, in order to
derive just the possessives of (kl) from a single case it seems
that we should have to posit some case other than Dative.
We have noticed already that the surface subject of have (with
which possessives clearly have much in common even if they are not
derived from it) seems sometimes to be a Dative, sometimes a Locative, likewise, the preposition to, which occurs in (M.a-d"), is
found with both cases. Lyons (1967) has argued that Dative and
Locative must be identified at a deeper level, distinguished largely
by whether the dominated NP is animate or not. Thus, this particular distinction may disappear on closer investigation of the issues
involved. Thus far, then, there seems to be no clear evidence
against obtaining all possessives from some sort of Dative/Locative
case.
However, when we look at the head nouns more closely, we discover a number of distinct semantic classes each of which determines in a different way the possible semantic relations holding
between head and genitive. As we observed in the last section,
the relation between the head of a construction and an NP under a
dependent case appears to be highly relevant to determining what
725
GEN - 27
case is involved. We must examine the semantic classes into which
possessives can be divided before deciding whether to assign all of
them a single deep structure case as source.
d.
(ii)
Two Semantically Distinct Classes:
Possessives
Kinship and Part-Whole
If possessives (as so far separated from other genitives)
bear different semantic relations to the heads of their constructions, it is highly possible that they have different sources—
not all of which need be cases. This much seems clear from the
discussion of case-relations in subsection 3 above. The following
classes, which we make no attempt to justify in detail here, suggest
some of the possibilities. It should not be assumed that the
classification is exhaustive or that the genitives of each class
have a single source though on the whole that does seem to be so.
(1*9) Kinship terms
(a)
(b)
(c)
(d)
(e)
the man's father
John*8 sister
the colt's dam
his child
someone's parents
(50) Part-whole relations
(i)
(a)
(b)
(c)
(d)
(e)
(ii)
Animate genitive (Body-parts)
the man's leg
John's heart
this centipede's toenails
its paw
someone's eyebrows
Inanimate genitive
(f) ?the saucepan's handle (the handle of the saucepan)
(g) ?the book's pages (the pages of the book)
(h)??something's wheel (the wheel of something)
?the chair's leg (the leg of the chair)
j) that hotel's entrance (the entrance to/of that hotel)
151
A minor point of clarification is necessary. Examples (50.f-j)
range from near-acceptability to ungrammaticality with considerable
726
GEN - 28
variation from speaker to speaker. There is a close relationship
between genitives and NP of NP (no genitive marker), and a condition, already mentioned, tends to exclude genitives that dominate
inanimate NP's. For the present discussion we shall assume that
genitives are generated for (50.f-j). We certainly find sentences
containing definite pronouns that must have come from such forms:
"The book had lost nearly all its pages;" "I want that saucepan
because its handle is a little longer." We shall return to this
question in E.2, E.3, but for the moment assume that (50.f-j) are
generated as genitives and are rejected by an independent constraint.
In all the examples of (U John ( ^=*> John's book)
(b) *the car the cop
bvj
(=^ the cop's arrival)
If, as we have been assuming, the "picture" nouns like book,
portrait, statue take cases (for example an Agent in (5**.a)) which
can prepose to form genitives, then, since these nouns can occur
with Agent and "Possessive" cases present (for they can represent
at once both concrete and abstract entities), it is necessary at
least to modify the subject placement rule so that this "Possessive"
case moves into genitive (i.e. subject) position in preference to
Agent, to yield:
(55) (a) John's book by Mailer
NOT (b) *Mailer's book /of \ John
«
In fact, even this ordering would not be enough to obtain the
right output. There would have to be a separate rule distinct from
both Active Subject Placement and Passive Subject Placement, which
obligatorily preposed the "Possessive" case. In particular, this
rule could not be a sub-rule of the Passive one, since the latter
operates on objectivalized NP's only, and we must allow my father in
the following, after undergoing objectivalization, to move by Passive
Subject Placement into the genitive (56.b).
730
GEN - 32
(56) (a)
(b)
the portrait of my father
my father's portrait
However, if John is the possessor of this picture, only (57.a)
is possible, not (57.b).
(57) (a)
John's portrait of my father
[of]
(b) *my father's portrait Jto I John
Finally, in a structure like (58), and in fact for all alienable
possessives, it seems that there is a major I.C. break between the
genitive, the metropolitan museum's, and the rest of the construction. This is not so for (59), and case-derived genitives in
general, as far as we can determine.
(58)
the metropolitan museum's portrait of a duchess
by Rembrandt.
(59)
Rembrandt's portrait of a duchess
This last piece of evidence is based on superficial data and is not
altogether reliable. However, the earlier evidence makes it seem
most unlikely that a deep structure case underlies alienable
possessives and we must assume therefore that they are derived from
some other source.
b.
From Postposed Genitives
The arguments given in subsection (7.b) makes it unlikely
that there will be any "possessive" case. However, most of the
examples relevant to that argument depend on analysing the "picture"
nouns as selecting ordinary cases.
We shall see that there
is at least some doubt about the correctness of that assumption.
For all other nouns the main syntactic objection to postulating a
case origin for possessives is the lack of an overt source, and
thus the need to introduce obligatory preposing. It might seem
possible to overcome both objections by regarding the "postposed
genitive of NP's as immediately derived from the underlying case
form without preposing. Then both (60.a) and (60.b) would have
failed to undergo preposing, while (6l.c) would ambiguously result
from the operation of a Subject Placement rule on such forms as
(60.a,b).
731
GEN - 33
(60) (a)
(b)
(c)
A book by Mailer
a book of Mailer's
[Agent]
["Possessive"]
Mailer's book
[Ambiguous]
There are serious objections to this proposal. In the first
place, it would require two quite different accounts of postposed
genitives involving two unrelated sets of conditions accounting for
the same distribution of surface forms. Secondly, the "Possessive"
case would, as a result of these conditions behave quite unlike other
cases in at least two important respects. Notice also that the
general semantic objection to deriving alienable possessives from a
case put forward in C.l.d(iii) above, would apply, of course, to this
particular representation of the "possessive case." The semantic
argument is not further reviewed here, but both of the syntactic ones
are.
The first requires a somewhat complex argument, dependent in
part on the analysis of postposed genitives made in Section E.l.
We take it as well established that within this grammar certain
genitives come from cases; for example the enemy's in "The enemy's
destruction of the city." The genitive is formed by preposing a
case into the Determiner. (See C.l-b above and CASE PLACE)) But
some of these genitives formed by preposing a case can then appear
as postposed genitives, to the right of the head:
(61) (a)
(b)
(c)
a proposal of the president's to end the war in
Vietnam.
all the most recent stories of his that I have read
those eyes of Lucinda's!
There are two plausible ways of obtaining such postposed genitives: by deleting elements in a partitive construction (so that
(6l.a) would come from, roughly, a proposal of the president's
proposals...), or by postposing a preposed genitive from its position in something like a [the president's] proposal. In Section E.l,
we argue for the second of these derivations. For the moment it
is irrelevant, however, which is correct, since the important point
is that when the genitive is formed from a case, it is initially
made into a preposed genitive first. On this, rules must operate
to form the postposed genitives of (6l.a-c).
Now, all postposed genitives, whether they represent alienable
possession or obviously come from cases are subject to at least one
constraint in common: they cannot occur with the definite article
unless there is also a relative clause present. Thus, although
732
GEN - 3k
(6l.a-c) are grammatical, none of the following are, where (62.d-f)
are understood as ordinary alienable possessives.
(62) (a) *The proposal of the president's
(b) *the stories of his
(c) *the eyes of Lucinda's
(d) *the books of mine
(e) *the house of Peter's
(f) *the chair of my father's
It is a fairly straightforward matter to prevent (62.a-c) from
being derived from preposed genitives. Several possibilities are
discussed in Section E. A single set of rather natural constraints
on the appropriate rules will achieve the right effect. But if
(62.e-f) are themselves cases it is impossible (as far as we can
see) to block these in anything like the same way.
In fact (62.e-f) can only be blocked by either (l) preventing
the possessive case from appearing in a definite NP with no restrictive
relative, or (2) forcing the possessive case to go througn the preposing (subject placement) rule just in case it was contained in a
Definite NP having no relative. Thus (62.d-f) would be avoided in
the base, or turned into my books, Peter's house, my father's chair,
respectively by making the preposing (subject placement) rule
obligatory. The first alternative is not worth further discussion.
We have no evidence whatever for any similar restriction on the
generation of cases on deep structure nouns. The second way of
avoiding the objectionable forms is only Just a little less objectionable. Like the first, it would separate the "Possessive" from
all other cases since forms like the following are perfectly
acceptable yet break the condition that would have to be imposed
on "Possessives":
(63) (a)
(b)
(c)
(d)
the
the
the
the
arm of the man
distruction of the city
attack by the cavalry
books by Iris Murdoch
Although this would separate the "Possessive" from all other cases
it is conceivable that motivation could be found for turning an
optional rule into an obligatory one—though it is important to
remember that the facts could be easily accounted for in a completely
general fashion if the postposed genitives of alienable possession
came from preposed genitives like all others do.
733
GEN - 35
The second constraint that would have to be imposed on the
"Possessive" case is needed to avoid forms like
(6U) "Mailer's novels of John's
and instead obtain, for example
(65) (a)
(b)
John's novels by Mailer
the novels of his by Mailer that Bill was talking
about
though not
(c) *the novels of John's by Mailer
It was pointed out above (p. 31), that if there was a "Possessive"
case it would have to prepose rather than the Agentive case if both
were present on a noun, thus necessarily getting (65.a) rather than
any other output. At that stage we were not considering any overt
"case" form for the "Possessive." Now that we are, however, the
conditions on preposing the "Possessive" case become highly unsatisfactory. For example, if any case preposed it would have to be the
"Possessive." That would prevent (6U). But notice that this condition would have to be over-ridden by the one discussed Just above:
if the top NP was definite yet contained no restrictive relative,
the "Possessive" could not prepose. That would prevent (65.c).
The price, however, seems unreasonable.
Notice that Jackendoff (1967) has a number of arguments
directed against essentially the same position as that which we are
in the process of rejecting. They do not carry over immediately to
this discussion because of important differences in the rest of the
grammars.
e.
Alienable and Inalienable Possessives
In Section C.2 above we showed that certain genitives come from
the deep structure cases generated on nouns by this grammar. In
the sections after that we have argued that, in the light of the
semantics of "case-hood"—discussed in C.3—and for independent
syntactic reasons, there are some genitives which cannot be naturally derived from cases. We have thus made a fundamental distinction
within the class of nouns which have no case structure immediately
T3U
GEN - 36
relatable to cases on verbs. Some, like father, entrance seem to
select cases in the way that verbs do—though we have not yet
determined what cases are involved. Others, like hat and kennel for
example, do not. The genitives which occur on them come, presumably
either from adverbs (a possibility which we shall not consider in
detail here) or relative clauses, which we discuss in the next
section. As was implied above, there are some nouns, like arm,
which form genitives in both these classes. Thus (12.a), John's
arm, is ambiguous. Arm may take cases, or enter into the (alienable)
possessive construction. Fillmore (19o7a) and Chomsky (19o7 ) both
attributed this ambiguity to a syntactic distinction between alienable and inalienable possession. It therefore becomes relevant to
ask how far the distinctions which they have made (the making of
which in fact occupies a large proportion of the current literature on genitives) represent a genuine syntactic distinction in
English. We noted earlier that the notion of case developed by
Fillmore was particularly concerned with NP's dependent on verbs.
We did not mention there that Fillmore himself extended the notion
of case to include just those nouns which represented inalienable
possession so that he argued for a Dative case on the noun arm.
On this, the characteristic "inalienable" behavior could be made
to depend. Chomsky, in turn, tried to extend the notion arguing
that in some way the enemy and destruction in the enemy's destruction of the city was "inalienable," just like John and arm in the
sense of John's arm where the arm is a body part; and, further,
that this intuitive "inalienability" could naturally be represented
in the syntax by generating the respective genitives in the determiner of the head rather than later moving them in. Neither of
these arguments is highly persuasive. Chomsky's rather fanciful
and otherwise unmotivated assumptions about the grammatical representation of inalienability allow him to account for the fact that
a picture of John's can never be a paraphrase of a picture of John
(where the picture shows John). But this is achieved by a trick of
ordering which in turn depends on obtaining of John's by a postposing rule and moreover fails to account in any way for the fact
that one of John's pictures lacks the sense in which John's picture
is a picture showing John.
In fact, there is no reason whatever for associating the
"inalienability" of any relation with a syntactic structure of
this sort. There is no more reason for supposing that inalienability is associated with cases generated on the head in the
base, rather than with NP's introduced into a Determiner from a
relative clause. The examples used by Fillmore suggest that what
735
GEN - 37
may be important in setting off nouns like father from others is
that they have an obligatory complement in the base. There is
something strange about a sentence like (66) while (67) may have
undergone deletion of some sort.
(66) *A father was walking down the street.
(67)
The father walked ahead, a little apart from the
rest of his family.
It may very well be that those nouns which require some complement
in the base all obligatorily select cases as a result of their
semantic make-up, though there is no a-priori reason for assuming
this, rather than that they are obligatorily modified by a restrictive relative, for example. Fillmore cites "louse" in Arapaho
as an inalienable—it is at least as likely that this word has an
obligatory restrictive relative as that it is semantically so different in that language that it is capable of selecting a specific
case.
Notice that although friend and secretary take cases, there is
nothing inalienable in the relationship between John and his friend
or secretary in John's friend, John's secretary; the important
characteristic of these genitives is simply that the relationship in
question in each is fully determined by the head. Furthermore, as
the following examples show, secretary along with a number of other
nouns selecting cases (all those below come from (U9) and (50)) do
not obligatorily select them.
(68) (a)
(b)
(c)
(d)
As I reached the office a secretary emerged
carrying a pile of papers.
Those legs can be carved from various kinds
of wood depending on the design.
I don't know where that handle came from.
All I could see in the back of the police
truck was a lot of arms and legs.
Probably the extent to which cases are obligatory on nouns is
related to the possibility of recognizing the objects named, independently of the defining relationship which is represented by a
case, but we are not concerned with that here.
736
GEN - 38
The point is that the phenomena described by Chomsky and
Fillmore under the designation of "inalienability" do not correlate
with any independently definable criteria so that their observations do not achieve any explanatory adequacy. The distinct
syntactic behavior on the part of inalienables, observed by both
Chomsky and Fillmore may seem to demand a separate syntactic class
of inalienable possessives. However, the ambiguity of (12.a),
John's arm, can be represented by deriving it from both (l) a
case and (2) whatever source yields the "alienable" possessives.
Moreover, the apparent differences in syntactic behavior of these
two senses, observed by Chomsky and Fillmore, turn out to be unrelated both to case and to semantically defined "inalienability."
Take the ambiguity of the following sentence, first discussed
by Ross (1967).
(69)
John broke his arm and so did Mary.
[Chomsky: 33]
The interpretation which is hard to account for is that in which
Mary broke her own arm, rather than assisting in some way in the
breaking of John's. The problem is that material deleted to make
way for so_ must apparently include her arm, but then her is not
formally identical with anything remaining in (69). Chomsky claims
that this interpretation is only possible if the arms that John
and Mary break are parts of their own bodies and that in such
structures inalienable genitives might be generated with dummy
NP's in the determiner, features later being copied in. Then the
source of (69) would be something like (70).
(70)
John broke o's arm and Mary broke A's arm.
Assuming that the rule replacing the second verb phrase by s£
preceded the copying rule, deletion could be accomplished on the
basis of formal identity.
Aside from the fact that there is no other motivation for this
proposal, the data scarcely warrants it. Even if for some people
the interesting reading of (69) may be excluded if the arms in
question are Just gruesome possessions of John and Mary, in (71.a,b)
the normal interpretation has Mary lose her book and John play
with his toys—yet these are alienable possessions.
(71) (a)
(b)
Peter lost his math book and so did Mary.
Sue played quietly with her toys and so did John.
737
GEN - 39
There is evidence in the other direction, though not as clear.
Consider the following sentence:
(72)
Algernon went to visit his young aunt who lives in
Georgia and so did Maisie.
Without special stress, it is highly questionable whether this can
be interpreted to mean that Algernon and Maisie visited separate
young aunts living in Georgia. Yet aunt presumably takes an "inalienable possessive." In all these sentences there seem to be a
number of factors at work excluding or favoring one interpretation
or another. It is not clear that a class of inalienables is
significant.
There are two more, related sets of facts which Fillmore
noticed and regarded as favoring a syntactic distinction between
alienable and inalienable possession. Sentence (73.a) is
ambiguous.
(73) (a)
(b)
(c)
I burned my fingers. [131*]
I burned your fingers. [135]
I burned my draft card. [136]
Only the first is ambiguous in the intended sense. Under both relevant readings of (73.a) an inalienable relationship between my_ (l)
and fingers is intended. The two senses correspond, roughly, to
(Ik) and (75).
(7*0
I burned something (on purpose)—my fingers.
(75) (a)
I burned myself (accidentally).
or
(b)
My fingers (got) burned.
The reading of (73) corresponding to (75) would come from something
like (76).
738
GEN - kO
(76)
MOD
PROP
V
Locative (!)
PREP
NP
N
burn
~Agent
PREP
NP
Dative
fingers
me
This would be converted to (73.a) by the general rule moving
Agents into subject position; the same rule derives (73.b,c) from
a similar structure.
The other reading of (73.a), however, Fillmore proposes to
derive from a tree of the form:
(77)
MOD
burn
Under this analysis a special rule preceding the ordinary subject
placement rules (which would give (U8.b) could optionally copy the
Dative NP into subject position to give (73.a) at the surface.
Presumably (78), if it is ambiguous, like (73.a), would be obtained
by then applying the rule that Fillmore postulates elsewhere in
order to derive (79.a) instead of (79.b). (See below for discussion
of some of the implications of that rule.)
739
GEN - Ul
(78)
I burned nqrself on the fingers.
(79) (a)
(b)
Mary pinched John on the nose.
Mary pinched John's nose.
[152]
[1^7]
This seems right on the whole, though the rules must be highly
complex. However, as far as we can determine, it is not relevant
to the claim that there is a distinction between (semantically)
alienable and inalienable possession. Notice first that the reading of (73.a) resulting from the special raising rule cannot be
obtained for (80).
(80) (a)
(b)
(c)
(d)
*John burned his beard.
*John burned his tooth.
*John burned his heart.
*John had unknowingly burned his lungs by inhaling
those fumes.
In fact the raising rule appears in these examples to be limited to
those parts of the body capable of feeling the effect of an accidental burning. Especially compare (80.a-d) with (bO.e).
(80) (e)
John burned his tongue because the chocolate you
gave him was still boiling.
In considering these examples it is important to recognize
that the intended sense correlated with the possibility of applying
the raising rule is independent of whether the burning was accidental.
At least, the burning could be accidental, as in (80.d), without
involving the intended meaning, for we can get sentences like (8l):
(81)
I burned my new coat.
which are ambiguous, the two meanings related to the possibility of
continuing the sentence by (8l.a) or (8l.b), depending on whether
the burning was accidental or not.
(8l) (a)
(b)
..., which was awfully careless,
..., to spite my husband.
However, the meaning of (8l) related to (8l.a), "I" is a case
in the top sentence, presumably in the Dative. For the relevant
meaning of (73.a) and for (80.e), however, the body-part noun is
itself the Locative or Dative case on burn. The claim is that a
structure like (77) cannot yield (8l). Note that for example (80.d)
iko
GEN - k2
is perfectly acceptable with a reading parallel to a sentence
like (8l), i.e. John had unknowingly burned his coat by leaving
it on the boiler^ There is an additional sense of (73.a) parallel
to this, too. In both (73.a) and (82), however, this "accidental"
sense has the subject, "I," a Dative on the verb, not on fingers
or coat.
With different main verbs the conditions under which structures like (77) can yield surface forms like (73.a) varies in
interesting ways. For example (82) is ambiguous in exactly the
intended sense, even though (80.b) was not.
(82)
John hit his tooth on a stone.
In this case, an accidental blow to the tooth is conceivable and
moreover it would be perceived as a sensation in the tooth.
While these are no doubt horrifying difficulties facing any
attempt to write such relationships into a grammar, it seems clear
that the alienable-inalienable distinction is relevant only in that
all those genitives that can possibly be subject to the rule
represent inalienable relations. But additional restrictions must
obviously be placed on the rule. Apparently these are dependent
on fine (yet none the less quite clear) semantic distinctions unrelated to the alienable/inalienable separation proposed by Chomsky
and Fillmore so that the latter distinction is redundant to the
point where it becomes altogether irrelevant. It is Just as odd
to interpret (83) in the sense of (75) as to interpret (73.c) in
that way. Yet the relevant relation in (83) is inalienable.
(83)
I burnt my father.
The inalienability of a possessive seems not merely insufficient
to determine whether it can enter this putative subject-raising
rule, but quite irrelevant to it.
A related argument for the relevance to the grammar of a
distinction between alienable and inalienable possession correlated
with deep structure cases turns out to fall under similar objections,
Example (81+) is relevant.
(8U) (a)
I hit John on |*he ) cheek.
(a1)*I hit John's cheek.
(b) *I hit John on I
the
I chair with a ruler.
this )
(b') I hit John's chair with a ruler.
7Ul
GEN - k3
Fillmore, as we have remarked above, would obtain (8U.a) by
moving John from a Dative case on the noun cheek, optionally lea
ing behind a copy which ultimately would pronominal!ze to his.
(8U.a') would result if Fillmore's raising rule (which must be
optional) had not applied. The deep structure postulated for
(8U.a,a') would therefore look something like (85), the optional
movement of John being shown by the dotted line.
(85)
MOD
PROP
hit
Since chair in (8U.b') cannot take an appropriate case to underlie
the possessive (which must therefore be derived from a sentence, or
whatever), there is no way of getting (81t.b) if, say, the movement
rule operates before such non-case derived genitives have been
formed. So far so good for the attempt to explain the possibility
of raising certain genitives by deriving them from cases while
others come from relatives. But the rule raising the dative of
(85) into object position in the sentence (giving (8U.a) instead of
(8U.a')) would apply only to a limited subset of the inalienables.
Thus, for example, it would have to be prevented from applying to
John in I hit John's father, for it must never yield *I hit John on
the father from it. Furthermore, unlike the rule discussed previously (for raising NP's like my in (76)), it would apparently
have to apply to certain nouns which cannot be regarded as entering into an independently defined inalienable relationship with
the head—though they may represent cases on that head. For example, many speakers will accept both the sentences of (86).
(86) (a)
(b)
I touched John's sleeve lightly.
I touched John lightly on the sleeve.
But unless "inalienable" means simply "behaves thus and thus with
respect to rules X, Y, Z," sleeve presumably does not take
inalienable possessives.
1k2
GEN - kh
It seems, in any case, that the circumstances limiting the
domain of this rule are highly complex, varying considerably from
speaker to speaker; it may well be that the rule is governed not
merely by the verb but by some sort of relation holding between
verb and the head of the relevant NP, as the following suggest.
A number of examples are given because in several cases it seems
likely that there is no transformational relation holding between
the set, a fact which may lead eventually to abandoning the raising rule, but will not substantially affect the selectional
problems involved.
(87) (a) (i
(ii
*I hit Mary on the braids with a ruler.
I hit Mary's braids with a ruler.
(b) (i
(ii
(iii
(iv
•I
I
I
I
touched
touched
touched
touched
(c) (i
(ii
(iii
(iv
*I
I
I
I
hurt
hurt
hurt
hurt
(d) (i
(ii
(iii
(iv
(v
(vi
(e) (i
(ii
Peter on the shoelace.
Peter's shoelaces.
Peter on the sleeve as I passed.
Peter's sleeve as I passed.
Sue on the toenail with a baseball bat,
Sue's toenail with a baseball bat.
Sue in the eye with a piece of wire.
Sue's eye with a piece of wire.
?I wounded John in the right leg with a
carving knife.
I wounded John's right leg with a
carving knife.
I wounded John in the eye.
Ill wounded John's eye.
I wounded John in the spleen.
*I wounded John's spleen.
I hit Reagan's fender with my old M.G.
I hit Reagan on the fender with my old M.G.
To generate all and only those of this group that form the sentences
with the overt surface "Locative" as cases on nouns in the deep
structure would do violence to the notion of case-dependency, would
depend on no independent criteria, and would mean that apparently
similar constructions like the source of the man's right leg in
(d.ii) and of John's spleen in (d.vi), would have to be regarded
as quite dissimilar. Thus, a raising rule cannot depend solely on
whether the genitive comes from a case. The examples of (87) make
it even less likely that an independent alienable/inalienable distinction is relevant.
7»+3
GEN - U5
Notice that the question is not whether additional factors are
involved or not. Fillmore recognized quite correctly that not all
inalienables go through the rules. The question is simply whether
there is an independently defined alienable-inalienable distinction
which is in any way relevant. All the evidence suggests that there
is not. There are a number of constructions about which
we understand very little, which operate when a number of different,
though related, classes of head nouns are involved.
Summary
In this whole section, III.C.l, we have tried to show that
there are arguments for deriving some genitives from deep structure
cases on nominal heads. Some of these, like arm, are not related
to verbs at all. In addition there are clear semantic and syntactic
arguments for deriving other genitives from some other source, perhaps relative clauses. We have shown, too, that the arguments for
deriving some genitives from cases are independent of the putative
alienable/inalienable distinction—which seems to have little substance, in fact. Before dealing in detail with those genitives not
derived from cases, which we shall now call POSSESSIVES (dropping
the pointless "alienable"), it is necessary to examine some problems
with the distinction which we have been building up in this section.
2.
Problems with the Proposal
We must turn to some considerations which tend to break down
somewhat the distinction between case-derived genitives and those
originating in, perhaps, some sort of relative clause. Most of
the problems turn out to be serious only if particular relative
clauses provide the source for possessives, and thus are in some
sense more relevant to the argument developed in the next sub-section, where different relative clauses are considered as possible
sources for alienable possessives. However these problems are at
the same time highly relevant to the notion of case extended, as
in the preceding pages, to apply to nouns, and it is convenient to
deal with some of the issues which can be resolved in the next
section (III.D) at the same time as those which are apparently less
tractable.
It is, of course, important to the thesis that some genitives
are derived from cases on the noun, and others from relative clauses,
that there be independent criteria enabling us to distinguish these
7UU
GEN - U6
two classes. We have suggested in CASE PLACE that the subject placement rules are optional for nouns. Thus cases never obligatorily
form genitives. In all the clear instances, the other genitives,
i.e. possessives, never turn into a Prep-phrase following the noun:
*the book of the man, *the jewels of my mother, etc. (For further
discussion see E.3.) In all the problems that follow we shall find
a certain tension between this single (and admittedly not highly
motivated) syntactic distinction, semantic criteria, and the need
to avoid generating unambiguous genitives from more than one source.
a.
Have and Case-derived genitives
Assume, first, with Smith (196U), Chomsky (1967) and most other
transformational grammarians, that the relative clause underlying
possessives is roughly of the form of (88.a), the sentence underlying the relative being, of course, something like (88.c):
(88) (a)
(b)
(c)
the book that John has
t
John's book
John has a book
The first problem is that the ambiguity noticed in John's arm
of (12.a), which we ascribed to the origin of the genitive in
either a case or a relative clause, can appear in sentences like
(88.c) in form. For example:
(89) (a)
(b)
(c)
He has two hairy arms.
The baby has eleven fingers.
You have a dirty face.
Thus (89.a) may be continued by either (90.a) or (90.b) depending
on the reading.
(90) (a). ...so he can't be Jacob.
(b) ...which he took off a model gorilla.
Therefore (given a source of the kind assumed) a genitive of the
form John's dirty face will have a double derivation for the meaning related primarily to a derivation from cases, and three routes
from deep to surface structure altogether. This introduces a very
general problem. The word have is close in meaning to genitives
(both case-derived and possessives), at so many points providing
7U5
GEN - U7
a full paraphrase for genitives. Yet often, as we point out in
section D, it is inadequate as a source for possessives. Thus,
have gives us both too many and too few paraphrases.
The noun clothes provides an instance in which the addition
of on to have enables the latter to paraphrase what appears to be
correctly regarded as a genitive derived from a case. Example
(91.a) seems to be ambiguous in a way related to that noticed for
(12.a), disambiguated by the normal readings of (91.b) and (91.c)
which are in turn paraphrased by (92) or (93) respectively.
(91) (a)
(b)
(c)
John's
John's
Though
bought
clothes
clothes are scruffy today
he's not wearing any of them, John
most of his clothes in New York
(92)
The clothes that John has on are scruffy today.
(93)
Although he's not wearing any of them John bought
most of the clothes that he has in New York.
That there is a derivation of (91.a) from a case is suggested
not only by the meaning of the head but also by the existence of
such forms as (9*0 with of NP after the head. (See III.E.3 for
further discussion of of NP. Also see CASE PLACE.)
(91*)
The clothes of the old tramp were torn and dirty.
Other evidence is provided indirectly by (87), where particular
items of clothing probably act as if they selected cases. It might
seem possible to derive genitives from have while excluding have on;
but the general principles are far from clear. For example, if Mrs.
Smith, a schoolteacher, has a number of children with her, it is
probably acceptable to say that her children are misbehaving. (If
they are!) It seems that the genitive would have to be paraphrased
as "the children Mrs. Smith has with her," and not by the same
form omitting with her. Even if have relatives do not yield
genitives, or, if they do, if the unwanted forms can be excluded from
such a derivation 1
i* is disturbing to have sucn close parallels
to the case derived genitives contain a semantically rather empty
verb, without giving any account of the semantic relations between
the near paraphrases.
7U6
GEN - U8
The last example involving have is itself rather unclear
because the grammaticality of crucial forms is uncertain to many
speakers. However it bears an interesting resemblance to several
of the next batch of problems. If it is possible to get forms like
(95) ?The most recent interest of his uncle turned out
to be painting grasshoppers.
then we could safely regard interest and similar words as selecting cases which turned into genitives. To avoid the double generation of, for example John's interests we should then need to avoid
(or constrain) the generation of genitives from relatives containing have, since we could otherwise get that phrase either from a
deep structure consisting of interest and John in an appropriate
case, or from
(96)
the interests that John has
On the other hand, it is not absolutely clear that (95) or any
other form containing a Prep phrase on interest is fully grammatical
and a large number of examples are obviously bad:
(97) (a) *that interest of my friend
(b) *some interests of the chairman
(c) *an interest of that explorer
If we wish to maintain that this criterion separates casederived genitives from those originating as relatives, it is not
clear that the genitive of John's interests and so on can come
from a case. Then we should need to allow forms like (96) to reduce to genitives. Thus a decision either way, in this highly
inconclusive instance, might provide significant, almost crucial
evidence for or against deriving some genitives from a relative
containing have. In fact this example takes us rather deeper into
the problem of relating have to genitives, for despite the failure
of the prepositional phrase test—which would make preposing of the
case on interest obligatory and disturb the one slender syntactic
criteria for case-derived genitives known to us—the meaning of
that noun does indeed seem to incorporate the same semantic relations
as the adjective and verb in (98) below. Moreover the relation
between John and interest is constant and completely determined by
the meaning of the head in all of the following.
7U7
GEN - k9
(98) (a)
(b)
(c)
John is interested in mathematics.
Mathematics interests John.
John's interest in mathematics
Thus far, the semantic evidence tends strongly to support a derivation of John's in (98.c) from a case. However, there is a conflict
at this level too. In addition to (98.a-c) the following must be
taken into account.
(98) (d)
John has an interest in mathematics.
If have is a real verb in (98.d) John appears to be a case on it,
in such a way that the total meaning of (98.d) is essentially the
same as that of (98.a). Recall, however, that the origin and significance of have is far from clear (cf. Bach (1967b) and Fillmore
(1967a)). Once again we have reached something of an impasse, where
the interpretation of the evidence is not at all clear.
Moreover, notice that the relation between John and interest
in (98.d) is determined by the meaning of interest. (See p. 28.)
In a sentence like (99):
(99)
John has a fine home.
the relation between John and home is rather vague, reminiscent of
the indeterminacy of the meaning of his horse (52.c). Add to these
observations the fact that a sentence like (89.a), He has two hairy
arms is ambiguous in that the relation between he and arms may be
either that of possession or that which is determined by the meaning of arms. It then becomes clear that there must be some very
close tie between have and genitives in general—not just possessives. When the meaning of have is left undetermined or vague, the
meaning of the corresponding genitive tends to be so. When the
meaning of have depends on the meaning of its surface object,
genitives having that surface object as head are likewise constrained.
And where there is ambiguity in the have construction, there tends
to be the same ambiguity in the genitive. These observations do
not in any way suggest that have-relatives underlie all genitives.
The meaning and deep syntax of have is little understood and the
relation may well go in the other direction. We leave this as a
major unresolved problem.
7U8
GEN - 50
b.
Semantic Evidence for Extra Cases
(i)
House
There is another major problem which we can exemplify first
using the noun house. This noun may well call for some case or
other from which to derive certain genitives but for which the
syntactic criteria do not point unambiguously in that direction.
(100)
We're going to play at Billy's house today.
Here, assuming Billy is a child, the only possible relation between
Billy and house is that Billy lives in the house. It seems to be
the only relevant relationship in such a sentence, it
is not relevant whether Billy by chance owns the house or not. If
the meaning of house is basically something like a thing built for
someone to live in, it may be possible to argue that Billy fills
some sort of "slot" in the meaning in that it is he who lives in
this house. (But see discussion of (30.a).) Continuing for the
moment to assume that relative clauses with have provide the source
for (alienable) possessives we find it impossible to obtain such a
source for Billy's house in (100), despite the flexibility in meaning observed for have. Thus (101) cannot mean that Billy lives in
the house—what it can mean is not so clear.
(101)
Billy has a house.
As with so many of the forms derived from cases (cf. his dirty face)
we can get the right meaning from a have sentence if the noun is
further modified, in which case it is the modification that is
e.sserted. In (102) Billy may just live in the house.
(102)
Billy has a nice house.
(Billy's nice house.)
On the other hand, if we derive the genitive on house from
a case, where it has this meaning, there is apparently no form
like *the house of my mother or *a house of this child. So
we should have to postulate obligatory preposing of the case.
Moreover, the semantic argument is not compelling, and the significance of the evidence provided by (101), (102) is little understood;
in particular it is still an open question whether relative clauses
with have underlie any genitives. If not, or if there is an
alternative source for the genitive of (100) there is no compelling
argument at present for deriving that genitive from a case.
7U9
GEN - 51
(ii)
sort.
Table
The noun table will illustrate another problem of the same
The following seem to be possible paraphrases:
(103) (a)
(b)
John's table has turned out better than mine,
The table that John made has turned out better
than the one I made.
It is not possible to paraphrase this meaning of John's table by
a relative clause in which has is substituted for made. Does
this mean that some genitives come from relatives containing make
(create, produce...?), or is it the case that table—and all
artifacts—will inherently allow an Agent? In general we do not
get:
(10*0 (a) "The table by John (has turned out well).
(b) #a table /of V that carpenter
Ibyj
(c) *this bookshelf i of > my father
IbyJ
(d) ?that house by a Brazilian architect
though when the maker is famous in the right field such forms seem
quite acceptable.
(105) (a)
(b)
(c)
a house by Frank Lloyd Wright
the bowl by Leach
some chairs by Hepplewhite
It is worth noticing that there are resemblances between the
form of (105) and Agents found with picture, book, etc. There is
as yet no compelling semantic argument for deriving the genitive
of (103.a) from a case, and whereas it might prove feasible to
motivate a distinction between case-derived and relative-derived
genitives ((105) as against (10U)) it would be strange indeed to
find a condition on a preposing rule that made it obligatory or
optional according to the status of the person referred to by
the moved NP. If, as seems to be the case, (103.a) can be derived
from an alternative source, so much the better. In considering
relative clauses we shall consequently have to consider nouns like
table again.
750
GEN - 52
c.
Cases "Missing" from Certain Abstract Nominals
The last problem in this section concerns words like announcement . This exhibits some features in common with table, some with
interest. The problem is quite possibly crucial for a deeper understanding of the relationship between case and meaning. First, notice
that there are (at least) two different functions of the nominal in
question. It may be what Lees (1960a) called an "action nominal"
(106.a); it may on the other hand name an abstract or semi-concrete
entity akin to book (106.b). Any adequate account must be able to
show how the abstract entity, together with a semantically weak
verb ("make") paraphrases the related verb announce (108) in such
a way that the semantic relations and cases of the two sentences
eire essentially the same Just as for the noun interest used with
have, and sentences built around the related verbs or adjective.
(See examples (98.a-d) and discussion, above.)
(106) (a)
(b)
The announcement by the judge to the Jurors of
an adjournment to the following week caught them
all by surprise,
We heard that announcement some time ago.
(107)
The Judge made an announcement to the jurors.
(108)
The Judge announced something to the Jurors.
As with interest, we cannot be sure that none of these
sentences are derived from other structures; in particular, that
(107) is not derived from (108). Assume that they are independent.
Our main task is to explain why the action nominal seems, predictably,
to occur both as in (I06.a) with a by NP, and in the genitive form,
(106) (a1)
The Judge's announcement to the Jurors of
an adjournment to the following week caught
them all by surprise.
while the "abstract entity" form of the nominal occurs only as in
(109) with the Agent converted to a genitive and not as in (110);
though there is a relative clause paraphrase of (109), i.e. (ill).
(109)
We listened to the Judge's announcement to the jurors.
(110) *We listened to the announcement by the Judge to the
Jurors.
(ill)
We listened to the announcement made by the judge
to the jurors.
751
GEN - 53
Leaving aside other apparent discrepancies in the case framework of these various forms, recall that the noun interest, too,
occurs in the predicate of a semantically rather weak verb (have),
thus forming a paraphrase of the related verbs (98.d) and (98.a),
and there is some doubt about the acceptability of prep-phrase
forms after that noun: (97.a-c). It seems, particularly with
announcement, that the abstract entity nominal (of the pair) may
itself lack the Agent (Dative, if this applies to interest) which
the related verbs, in sentences and "action nominals" are capable
of appearing with. In other words in (107) the "dummy" verb make
adds this case to those of the nominal to make up the meaning of
the sentence as a whole, so that to get an Agent associated with
the nominal it is necessary to use a relative clause containing
this verb as in (ill), which will optionally reduce to the genitive
of (109). Such an account is so far adequate, dealing effectively
with the ungrammaticality of (110) or any other prepositional
phrase paraphrase: announcement takes no agent.
However, (112) is virtually a paraphrase of (107), while (113)
is of dubious grammaticality and certainly of different sense.
Assume some sort of equi-NP deletion to yield (107) (perhaps with
the instead of an) from (112).
(112)
The Judge made [his announcement to the Jurors]
yesterday.
(ll3)??The Judge made the announcement that he made to
the Jurors yesterday.
Then it would be necessary to postulate that nominals of this
kind had obligatory preposing of the Agentive case to form a
genitive. But otherwise cases do not obligatorily prepose to
form genitives. Since we understand so little about the difference
in internal and external behavior of different kinds of nominals,
having no specific motivation, for example, either for deriving
the one announcement from the other or for relating them in the
dictionary, and since it is not possible to distinguish the two
uses clearly, it is meaningless to pursue the question further
at this point. In the long run it may be that the relative merits
of the approach to nominals adopted in this grammar as compared
with that which has become known as "generative semantics," as recently
developed by Ross, Lakoff and McCawley, will be decided partly by
the facility with which they are able to handle relationships between constructions of the sort under discussion here. It would,
for example, be particularly interesting to examine in detail the
relationship between those deep structure nodes which ultimately
collapse under a lexical item inserted late in a derivation
752
GEN - 5U
according to recent proposals made by McCawley, and the cases which
in this grammar that "same" lexical item takes. None of this have
we undertaken and the problem of announcement must remain essentially
unsolved. For the purposes of this grammar we choose, quite
arbitrarily, to ignore the instances of obligatory pre-posing (e.g.
(110)) and to regard announcement as always selecting an Agentive
case, Just as the related verb does.
3.
The Cases Underlying Kinship and Part-Whole Genitives
There remains only one problem to be dealt with in this section: to determine if possible what cases are selected by the head
nouns to yield (l) Kinship, (2) Part-Whole and (3) Weather genitives.
It was convenient to postpone discussion of these until it had been
at least tentatively established that they were the only constructions quite unrelated to verbs in which the genitive came from a
deep case. In other words, that they might represent the entire
stock of cases selected by "real" nouns. They, together with the
(alienable) possessives (which are to be derived from some other
source) made up virtually all the "possessives" as these were
originally set up, and we have argued that (alienable) possessives
do not come from cases. Consequently it appears that all the nouns
that take cases yet are unrelated to verbs fall into one of these
three categories.
Fillmore (1967a, p. 66) regarded both kinship and body-part
genitives as coming from a dative on the noun itself; citing as
evidence for this particular case only the fact that the NP under
it is animate, and noting in passing the occasional appearance of
the typically dative preposition to, which we commented on above.
Although he does not deal in detail with the non-animate part-whole
genitives, he suggests later in the same paper that expressions
like (ilk) as well as behind the house, ahead of the cat, and
next to the tamer may come from locatives on the head "nouns" (i.e.
prepositions in the above instances).
(llU)
corner of the table, edge of the cliff,
top of the box
[l83]
The examples of (llU) are, of course, what we have referred
to above as purely relational part-whole genitives, distinct in
various ways from the other inanimate part-whole constructions
like key of/to the door, windows of the house and so on, with which
Fillmore does not deal. It seems likely however that he would have
analyzed those, too, as Locatives, while the relationship between
weather genitives and sentences like the studio is hot (Fillmore
753
GEN - 55
(1967a), example (8l) which Fillmore analyzed as having a Locative
subject, suggests that the animate case-derived genitives come from
Datives, and inanimates from Locatives, which is what, rather
arbitrarily, we assume in the lexicon of this grammar. (See LEX.)
Notice that Langacker (1967), dealing with French, analyzes forms
parallel to the ordinary inanimate part-whole constructions (e.g.
the door of the cathedral), as coming from a Dative rather than a
Locative (in a relative clause, as it happens, but that is irrelevant
here); but he does not offer any specific arguments for using that
case with the inanimates, beyond the possibility of making them
quite parallel to animates and in fact there do not seem to be any.
On the other hand there are no strong arguments for any other
particular case or cases. Recognizing this, and given our present
understanding (or, rather, lack of understanding) of the relationship between meaning and case framework, we generate only Dative or
Locative on these kinship, part-whole and weather nouns, relying on
factors other than case to account for the great differences in the
relationship between genitive and head in the three groups. In
fact, it is not even clear what kind of question it is to ask whether
the differences in the relationship between his and father and his
and arm in his father and his arm are of a sort that should be
represented by a difference of case. Nor is it clear whether we are
asking an empirical question if we query the appropriateness of
calling the cases Dative and Locative, thus associating them with
verb-related cases.
More important at the present time is the problematical fact
that the solution tentatively adopted in this grammar represents
a claim that differences in conditions on preposing (and other
rules) exhibited by the following (a) and (b) pairs are not directly
attributable to case differences. Needless to say there are other,
similar examples.
(115)(a) *the weather of Chicago
(b) the top of the mountain
(ll6)(a)??Everest,s top (?the mountain's top)
(b) Chicago's weather
•to]
(117)(a)
the weather < in> that city
lofj
to"|
(b) (i)
(ii)
the top ^inVthat box
of
•to!
the windows < •in > that house
ofj
75U
GEN - 56
[•to
(ll8) (a)
the house ) *of ^> the woods
I^
(b) *the woods' house
(The appearance of of-NP in the above is taken as evidence that preposing is not obligatory. See E.3.) An example like (118) was
discussed earlier. (See (38), etc.) The Locative may not represent
a case within NOM—but again it may. This question is open. Noun
compounds like table-top need to be taken into account, and these
we have not analyzed. (See Section F.) Some of these problems are
discussed further in CASE PLACE, and in section C and E, expecially
the problems of accounting for the appearance of of NP forms.
Summary
Summarizing section C in brief, we have shown above that there
are some nouns which, like destruction, take roughly the same cases
as the related verbs do; there are others, like arm which can
apparently take cases, though what cases are involved it is hard to
say; and, finally, there are nouns like kennel which take no cases.
Nouns from all three classes can appear with genitives„ For the
first two classes of noun mentioned, the genitive probably can
come from a case while for the last there must be some other source.
We have tried (though not with complete success) to suggest criteria
that will distinguish the three classes of genitive and have
discussed some of the problems that our analysis gives rise to.
In general it seems fair to claim that, so far, an x-case grammar,
such as this one is able to handle the problem of the source of
genitives at least as well as any other, and that it raises some
interesting and important questions about the semantics of the
genitive. For the rest, it is impossible to judge the analysis as
a whole without considering the source of possessives, to which we
now turn.
755
GEN - 57
D.
What Relative Clauses Yield Genitives?
In this section we can assume that kinship, part-whole and
weather genitives come from cases and, consequently, that the ideal
relative clause source for possessives will not yield these genitives
except to produce the desired ambiguity of such forms as Jane's eyes.
Thus, given the arguments in section C, for using cases for certain
genitives, it is absolutely necessary to avoid generating John's
father from a reduced relative clause, and if interest selects a
case which turns into his interest in mathematics we must avoid
generating this from a relative clause too. Since there is at
least some doubt about the case-frame of interest (see examples
(96)-(98)) it will be as well to avoid having to choose between
alternative sources for the possessive on the basis of their ability
either to generate or to exclude the genitive on interest. Obviously,
then, the special role played by cases on nouns in this grammar
places quite specific constraints on the relative clause source for
possessives.
Were it not for the fact that we are deriving a considerable
number of genitives from other sources than the relative, we should
have to impose very different, weaker constraints on that source.
It would, for example, have to yield the relevant examples of (17)
through (20), which, we have said, seems to be impossible to do in
any general fashion. (See also Jackendoff (1967).)
(17)
(a)
(b)
Chicago's weather
the weather in Chicago
(c) (i) *the weather that Chicago has
(ii) *the weather that is in Chicago
(d) (i) *Chicago has some weather
(ii) *some weather is in Chicago
(18)
(a) the lake's edge
(b) *the edge that the lake has
(c) ?the lake has an edge
(d) *the edge is to/of the lake
(19)
(a) the man's head
(b) *the head that the man has
(c) ?the man has a head
(d) *the head is to/of the man
756
GEN - 58
(20)
(a) Mary's mother
(b) *the mother that Mary has
(c) Mary has a mother
(d) *the mother is to/of Mary
We do not propose to deal further with the problem of deriving such
a wider class of genitives from relatives, but rather, assuming a
derivation from cases, to find a suitably constrained relative clause
source for possessives and to show the problems that this involves,
since those problems may well be crucial in considering the
theoretical claims of this grammar. This course of action demands
that we distinguish as separate, potential sources of possessives,
two forms that Smith (196U) assumed, without much discussion, to
be transformationally related stages in the derivation of possessives.
Underlying (119) were, successively, (120) and (121). We cite these
as Smith did, ignoring irrelevant differences in her framework, and
in particular the matrix sentences of (120) and (121).
(119)
...John's hat...
[37]
(120)
...the hat is John's...
[31]
(121)
...John has a hat...
[38]
ft
t
(120) and (121) are not synonymous; nor do they occur in the
same environments, as we shall show in the course of the rest of
this section.
Notice that Smith's argument for deriving (119) from a sentence
containing (121) as a relative clause via one containing (120)
depends in a large part on considerations of simplicity which turn
out to be quite irrelevant. Between (120) and (119) come the stage
(119').
(119') (*)...the hat of John's...
[36]
The genitive was then preposed. Superficially, the resulting
series of transformational steps resembles that through which
adjectives are taken: the book that is green »* *the book green *+
the green book. Just as for possessives the middle form, after
reduction of the relative is sometimes obligatorily reduced (as in
the above examples) and at other times may not be: the missing
10 pages book, *a John's hat vs. the book missing 10 pages, a hat
of John'8. However, clearly the conditions for preposing adjectives
757
GEN - 59
and possessives are quite unrelated. Moreover, as example (122)
shows, the genitive is moved into a very different position. Thus
there must be two quite separate pre-posing rules:
(122) (a) John's three green books
(b) *green three John's books
In (122), three is generated in Det to begin with; it is clear that
the adjective has to be placed to its right, the possessive to its
left. Thus the similarity between the derivation of genitives and
that of adjectives turns out to reside only in the fact that both
make use of the rule of relative reduction. Even that is suspect,
however. Observe that in general copular sentences containing
predicate nominals seem not to reduce.
(123) (a) The man that is a carpenter came later,
(b) *The man a carpenter came later.
If (123.b) is to be excluded, rather than becoming (See Bach (196Tb))
The carpenter came later, then it is not obvious that Smith's proposals would introduce greater generality into the grammar even in
this respect. Anyway, it is necessary to constrain the relative
reduction rule in various other ways that are
little understood but which make it hard to support any analysis on
the grounds that that analysis would increase the generality of
the reduction rule. For example, it is apparently necessary to
prevent the reduction of (123'.a) since there is no acceptable output:
(123') (a) The man that is ill wants to leave.
(b) *The man ill wants to leave.
(c) *The ill man wants to leave.
Moreover, Smith's proposal requires that the postposed genitive
(hat of John's) represent a stage in the derivation of the preposed
one, for those genitives which come from relative clauses. For those
coming from cases however, genitive marking takes place in the preposed form. There is apparently no non-arbitrary way of accounting
for the fact that the conditions for post-posing/pre-posing would be
essentially the converse of each other for these two sets if we
therefore consider the stages in Smith's derivation as alternatives,
weighing each against the criteria which must be met by the source
of possessives in this grammar. This must not be taken to mean
that we assume entirely independent sources for (120) and (121)
758
GEN - 60
since both may come from a single deep structure which is subject to
different derivational constraints below this level. For the
present purpose, however, we can ignore that possibility and assume
that the two structures differ in the base.
1.
Relative Clauses with Have
Sentences with have, like (121), are available to provide the
source of most possessives. The meaning seems to vary appropriately,
yielding very nearly the right semantic range. Nevertheless, as
the following examples show, there are semantic problems with such
a derivation.
(1210 (a)
(b)
(c)
Our dog has a kennel.
The kennel that our dog has is too small.
Out dog's kennel is too small.
(125) (a)
(b)
(c)
Billy has a house.
The house that Billy has is beautiful.
Billy's house is beautiful.
(126) (a)
(b)
(c)
I have a cold.
The cold that 1 have is growing worse.
My cold is growing worse.
(127) (a)
(b)
John has a horse.
The horse that John has belongs to the riding
school,
John's horse belongs to the riding school.
(c)
(b') The horse that John has is likely to win him
some money.
(c1) John's horse is likely to win him some money.
(128) (a)
(b)
(c)
(129) (a)
(b)
(c)
Mary has an interest in mathematics.
The interest that Mary has in mathematics is
surprising to her parents.
Mary's interest in mathematics is surprising to
her parents.
Mr. Smith has an idea.
The idea that Mr. Smith has is probably right.
Mr. Smith's idea is probably right.
759
GEN - 61
A number of these examples certainly seem to provide evidence
that have is very closely related to possessives. For example
(l2U.a) does not imply that the dog owns the kennel, while in (125.a)
ownership can he the relation between Billy and the house. In (127),
correctly, the favored reading of both the (b) and (c) sentences is
that John is simply borrowing, or riding the horse, while in (b*)
and (c') there is about the same degree of vagueness, for John may
own or have bet on or drawn the horse in question. The (a) sentence
includes all the right possibilities. It is unclear how some of
these are filtered out for (b) and (c), but notice that the underlying relatives of (b) and (b1) give just the right meanings for
(c) and (c') respectively.
It has already been pointed out that in general there are no
have relatives for kinship,part-whole and weather genitives (provided
they have no modifiers—see below).
(130) (a) *the mother that John has
(b) *the face that Mary has
(c) *the temperature that the room has
This is another point in its favor if these genitives come from cases.
On the other hand, there are a number of serious problems with
this derivation. First, have relatives unless arbitrarily prevented
from doing so, will yield a second derivation for any case-derived
genitive that has a modifier present in the NP:
(131) (a)
(b)
(c)
the rich uncle that John has
the lovely eyes that her son has
the awful weather that Chicago has
Moreover, for some kinship terms there appear to be viable relative
clauses containing have, though they are dubious paraphrases of the
corresponding genitives.
(132) (a)
(b)
The sisters that John has help him to understand
women.
John's sisters help him to understand women.
Another problem concerns examples (128) and (129). If interest
and idea do not allow cases, then the fact that there are have
relatives paraphrasing the genitives is indeed an advantage of deriving possessives from that source. However, semantically it seems
most likely that nouns like these will take cases, and in the discussion of the last section that obligatory preposing might have to
be postulated anyway for certain constructions if forms like
announcement are also taken into consideration.
760
GEN - 62
If so,
(128) and (129) must be regarded as counterexamples to
the proposal to derive possessives from have.
(125) raises a different problem with have as the source for
possessives. We argued, in connection vith examples (100) and
(101) that Billy has a house cannot simply mean that he lives in
one, and that the house that Billy has can't refer to one that he
lives in (as a child, without renting or owning it), but that
Billy's house as in, "We're going to play at Billy's house today"
can mean Just exactly that: a house in which Billy lives.
A further objection to this proposal is that it fails to provide a suitable source with the right range of meaning for the
following possessives, among others:
(133)
Peter's team
(13*0
That is Maria's chair so don't sit there.
(135)
John has Billy's ruler.
The first of these can be used to refer to a team that Peter is
associated with in that it is the team that he:
(136) (a
(b
(c
(d
(e
(f
(g
(h
(i
coaches
captains
owns
has placed a bet on
plays for; is playing for at present
works for
belongs to (though he doesn't play)
supports—in general
has Just favored, in an argument
but at most the team that John has can refer to (a)-(d). Both (e)
and (f) could conceivably come from cases but we can see no source
for the others.
The meanings of (131*) which concern us here vary roughly between
(137) and (138). (139) does not paraphrase either.
(137)
That is the chair that Maria will sit in.
(138)
That is the chair that Maria likes to sit in.
(139)
That is the chair that Maria<
761
GEN - 63
The last example of the group, (135), is similar.
phrased by (lHO), but rather by (l^l).
It is not para-
(lUO) (?) John has the ruler that Billy has.
(lUl)
John has the ruler that belongs to Billy.
(lUO) is a perfectly grammatical sentence. It just happens to be
a contradiction as it stands. Consider also a sentence like (lU2)
where the genitive represents a relation of (legal) ownership,
which is contrasted with (physical) possession.
(1U2)
John doesn't actually have any of his money himself.
The next (and last) two problems do not directly concern the
derivation from have, but represent difficulties which arise in
other areas if possessives are derived from have-relatives. In the
first place, it will be necessary to generate some genitives from
relative clauses containing a verb like make. We noted in the last
section that on the whole there seemed to be no good arguments for
deriving genitives like my table where this means
(1U3)
the table that I made
from, say, an Agentive case on table, but that they seemed rather
to fit into a peculiar sub-class of possessive. If so (and the
question is not really settled) it is presumably necessary to derive
my table, in this sense, from something like (l1+3). Certainly
have-relatives don't merely give awkward paraphrases, or present
neat derivations; in this instance they are altogether unsuitable.
Finally, if have (or, indeed, any construction other than the
predicate genitive) provides the source of possessives, it is
necessary to account in some way for predicate genitives like
That book is John's. These could, of course, be quite unrelated
to other genitives, but on both formal and semantic grounds (the
latter described in detail below) this seems unlikely. Alternatively, they could be derived from other genitives. The most
plausible method then involves deleting nouns in the predicate of
a copular sentence:
ilkk) (a)
(b)
That book is John's book.
v
That book is John's.
762
GEN - 6k
It might be argued that the rules needed are those required in the
grammar anyway, (l) NOUN REDUCTION TO ONE to reduce one of two
identical nouns to one and (2) ONE-DELETION to delete one in certain
environments. (See PRO, II.B.2 and III.C.) These rules do indeed
operate on genitives.
(lU5) (a)
I have my book and Mary has her book.
(b) *I have my book and Mary has her one.
v
(c)
I have my book and Mary has hers.
However, there are two problems in getting these rules to produce
the right predicate genitives. The first is exhibited in the
following:
(1U6) (a)
(b)
That is John's table.
That is Chomsky's book on politics.
(lU7) (a)
(b)
That table is John's.
That book on politics is Chomsky's.
It is surprising that while (lHT.a) can refer to a table that belongs to John or to one that he made (just as (lU6.a) can), (ll*7.b)
can only refer to a book that belongs to Chomsky, although (lU6.b)
is ambiguous between this reading and that in which he is the
author. Thus, if the deletion rule applies to (1^7.a) where the
genitive is a possessive, it will have to be restricted in a
peculiar way to prevent it from applying to the Agentive genitive
of (lVf.b). That it may not apply at all to such forms is suggested
by the following (see PRO, where in fact neither (lU8.a) nor (lU8.b)
is generated.)
(lU8) (a) John saw the blue book while I saw the green,
(b) *That book is the green.
The second problem with the deletion of predicate genitives
to yield (lVf.a) and (lU7.b) is closely related. In other positions
in a sentence the head noun deletes from such genitives as
Chomsky's books (where the genitive comes from an Agent) to give:
(1U9)
I read one of Conrad's stories this week and one of
Poe's last week.
763
GEN - 65
It is then irrelevant what case the genitive comes from. If the
case was Neutral it is possible to do this kind of deletion following the general ONES-DELETION rule mentioned in the last paragraph,
to give:
•
(150)
Mary's (recent) portrait (by Augustus John) isn't
as good as Arthur's.
However, it is quite impossible to get a Neutral reading (where
the portrait in question represents Mary) for the predicate
genitive:
(150') *That portrait is Mary's
i
[Mary]
Neutral
though it is possible, perhaps, to get this interpretation for that
portrait is one of Mary's recent ones. This observation
suggests a relationship between postposed genitives (see E.l) and
predicate genitives, but we are unable to pursue that possibility
here.
Finally, apart from the difficulties noticed above in defining
the domain of the deletion rule, notice that predicate genitives
should, by this derivation, imply that there is only one object of
the given kind in mind. So this chair is John's shoul be equivalent
to This chair is John's chair. However this does not appear to be
the case.
(151) (a)
This
z
(b)( This
|?This
(?This
chair is John's.
(So are five others in
the room.)
one is John's chair. 1 (?So are five others
chair is John's one. \
in the room.)
chair is John's chair.1
cf.
(152)
2.
This chair is green.
(So are five others in the room.)
Relative Clauses Containing Predicate Genitives.
Let us now consider the advantages over the have derivation of
deriving the predicate genitive in the base as the source of
alienable possessives. In the first place, not only do the plain
case-derived genitives then lack a relative clause source, but the
modified ones like kind old mother do too. They would not do so if
76U
GEN - 66
have provided the source.
(153) (a) *the kind, old mother that is John's
(b) *That kind, old mother is John's.
Yet those nouns like arm vhich have ambiguous genitives can appear
in such constructions. The meaning in that case is, in general,
limited to that of the possessive, which is as it should be.
(15U) (a) (?)the eye that is John's
(b) That eye is John's.
(15**. a), it is true, is somewhat infelicitous, but simple adjectives,
too, seem to require preposing; so (?)Bring me the book that is green
seems no less unsatisfactory than (l5H.a). In general, relatives
containing the predicate genitive, like those containing preposable
adjectives, are clumsy and bordering on the unacceptable. If, however, adjectives are derived by preposing, this similarity is, if
anything, in favor of our derivation.
Consider next the ability of the predicate genitive to provide
appropriate deep structures for (l25.c) and (I27.c,c'):
(125)
(a)
((b)
(c)
(127)
(a)
(b)
Billy has a house.
The house that Billy has is beautiful,
Billy's house is beautiful.
John has a horse.
The horse that John has belongs to the riding
school.
(c) John's horse belongs to the riding school,
(b') The horse that John has is likely to win him
some money,
(c') John's horse is likely to win him some money.
The following seem satisfactory, having the same range of meaning
as preposed genitives; (125') and (127')could certainly be used
to assert ownership, but. equally, to assert that the transitory
relationship implied by (127.c') holds.or to refer to the fact
that Billy lives in a particular house (125.c'), (For the moment
we ignore (127.c), as opposed to (l27.c').)
(125')
That house is Billy's.
(127')
That horse is John's.
765
GEN - 67
(1251) appears to be a more appropriate deep structure than the
comparable have sentence, though the restrictive relative based
on a predicate genitive is particularly bad: ?the house that is
Billy's....
This proposal does not fare as well for (121*) as have did.
(12U') ?That kennel is our dog's.
Predicate genitives are in general not very satisfactory with nonhuman predicates: ?This bell is that cow's, ?the ball is my
kitten's .... With these non-human predicates, the postulated
deep structures are semantically quite appropriate. They could
all be paraphrased (grammatically) by ...belongs to... sentences
like
(12U")
That kennel belongs to our dog.
Nevertheless the proposed deep structures seem syntactically
dubious and represent a very weak point in the proposal.
For reasons that have already been explained, it is impossible
to use interest as crucial evidence for or against the proposal.
Assuming that it selects cases, (128) has absolutely no paraphrase
that uses a predicate genitive:
(128') *The interest (in mathematics) is Mary's.
Since this could provide highly significant evidence in favor of
this proposal and against using have, the choice between have and
the present source may depend on answering a question that remains
open.
The evidence from idea is difficult to interpret.
hand, there are sentences like (155):
(155)
On the one
Those ideas are mine.
On the other hand sentences with an unreduced relative on idea,
containing a predicate genitive, seem altogether barbaric. For
example, as a paraphrase of (l29.c) the following seems to be
semantically wrong and not simply awkward, as many similar sentences
are.
(129')
(c) *The idea that is your father's is probably
right.
766
GEN - 68
This could veil be taken as evidence for generating a case on idea,
but there seems no other motivation for that and (l29.c) would therefore constitute a rather serious counter-example to using the
predicate genitive as the source of possessives—if idea took
possessives. But, again, there is at present no clear answer to
that more fundamental question.
The noun cold, as in (126), unless it occurs with cases,
also provides counter-evidence:
(126') *That cold is mine.
We leave this, too, as a counter-example, but it doesn't seem
serious at this stage, since we know very little about the behavior
of cold in this sense.
The evidence from (133) - (135) is unambiguously in favor of
the predicate genitive. (133'), (131*') and (135*) have precisely
the right range of interpretation:
(133')
That team is Peter's.
(131*')
That is the chair that is Maria's, so don't sit there.
(135')
John has the ruler that is Billy's.
Moreover, if this is the source of possessives in general, it is
unnecessary to provide a derivation from a make relative for
genitives like my table, where the speaker made the table in question, (cf. (lWTJ The predicate genitive allows this interpretation, as in (1^6).
(156)
That table is mine but I prefer the one John made.
Yet there is no comparable interpretation for (lUT.b), nor, correctly,
for other sentences like That book is Chomsky's, although it
may be marginally possible to use a sentence like 'lhat picture is
Picasso's to identify the painter rather than the owner.
In addition to the problems that arise in regard to (12V) and
(129'), there are two general problems with the proposed derivation from predicate genitives. In the first place it fails to
give any account of the close semantic relationship between have
and genitives. Within the framework of this grammar that is not
necessarily very serious. In the first place, we do not generally
expect to find that all paraphrases have the same deep structures.
Secondly, though the parallels are far-reaching, they are not
767
GEN - 69
universal. Moreover, there are at least two ways in which have can
be related to the predicate genitive. First, if have itself is a
lexical item with extremely little semantic content taking a
Dative and Neutral, this might well yield meanings largely parallel
to those of a copular sentence containing Neutral and Dative cases.
The latter is a reasonable deep structure for Predicate Genitives.
On the other hand it is possible (and, in fact, in line with
Smith's original proposal) to have predicate genitives result from
the preposing of the Neutral rather than the Dative on have itself.
We have remarked elsewhere on other instances where a difference in
the application of Subject Placement rules can result in a change
of meaning. (See CASE PLACE, with reference to load, for example.)
A special subject placement rule for obtaining predicate genitives
from the same base as have could be made to prepose Neutral
instead of Dative, deleting have and thus triggering BE-INSERTION
(see CASE PLACE). However, in order to derive only the right
predicate genitives, it would be necessary to impose some peculiar
constraints on this particular application of the subject placement
rule. For example, to avoid generating:
(157) *The book is a professor's,
but, instead:
(158)
A professor has the book.
it would be necessary for the Dative to prepose obligatorily if
indefinite. Although there are, as we show in the next section,
constraints of this sort on preposing cases on nouns (but in
reverse—for indefinites often do not prepose), there appear to be
no other examples for verbs.
Despite the problems involved in the predicate genitive, it
seems to be overall the most appropriately constrained source for
possessives (via relative clauses) considered so far. Before we
leave it there are two further points to be noted which tend to
argue against it, however. First if predicate genitives are base
forms, the morphological resemblance between these genitives and
those derived from cases has to be regarded as purely accidental.
All things considered that is highly unsatisfactory. Second, the
peculiar and highly constrained nature of this construction, on
which we have remarked from time to time, is not obviously any
easier to account for in the base then by constraining deletion,
etc.—we have simply shown that there does not appear to be a
rational way of dealing with it either by deletion or by constraining the subject placement rule.
768
GEN - 70
3.
Other Possible Sources
Two other possible sources for possessives deserve brief
mention. The verbal form belongs to acts in almost all constructions in a very similar manner to the predicate genitive, which,
in most instances, it paraphrases. Notice, however, that there
are viable base sentences and relative clauses in most cases. In
a few cases, (e.g. the kennel that belongs to our dog—cf. (12V))
the improvement in comparison to the predicate genitive is quite
striking. In others, however (e.g. the team/house/horse that
belongs to John), the resulting construction is considerably
narrower in meaning, which is undesirable. This source would
avoid the morphological problem referred to just above, but would
re-establish the need for a different relative source for my table
(my = Agent). If there is in fact a single source for all possessives it is unlikely to be belongs.
It may well turn out that within this grammar, and in all
others deriving certain genitives from cases on nouns, the most
appropriate source of possessives is within the Determiner in deep
structure, as an alternative expansion of Art. Now this was the
source proposed by Chomsky (1967) for inalienable possessives (a
subset of the genitives that we derive from cases). We questioned
the appropriateness both of Chomsky's classification and of his
syntactic representation of "inalienable" relations. There seems
to be greater prima facie justification for proposing such a derivation for those genitives which lack all but a vaguely "possessive"
relationship with the head. We have not examined this proposal
in any detail to see whether it is generally viable (though notice
that the correct predicate genitives might be obtained by a rule
of deletion—or whatever—operating prior to the introduction of
other genitives into the determiner).
To summarize the observations of section D: given the
constraints imposed by the rest of the grammar, there is no completely
satisfactory source for possessives. The predicate genitive
probably represents the most suitable sentential source but
creates a number of problems. It is possible that possessives
should be generated as Articles, but this possibility has not
been explored.
This is a convenient point for a brief review of the relationship between sections C and D in which the sources of the genitives
have been discussed. There are indisputably close relationships
769
GEN - 71
between many genitives and cases in deep structure. We have been
able to provide evidence for extending the sphere of such relationto forms like John's arm, while rejecting the relevance of a
notion of alienability. As a result we have been able to suggest
a number of fresh approaches to the question of the source of
possessives. Although we have rejected all currently proposed
sources this in no way constitutes evidence against deriving
some genitives from cases, since there is apparently no more
satisfactory way of deriving all genitives in a general fashion.
In fact, by establishing a clearer distinction then before
between the two classes of genitive, we have been able to pose a
relatively small number of crucial questions—though we have not
been able to answer them in this grammar. To the extent that
these questions prove relevant to the problem of deriving genitives,
they will provide support for the particular distinctions suggested
here. Some of those questions, such as those raised with respect
to the cases on idea, interest, house, etc., may well show that the
notion of case is in fact not adequate to answer the questions
that it has allowed us to raise in this area.
E.
The Derivation of Genitives
We turn now to the operations that derive surface genitives
and related forms from the deep structures proposed above. First
we discuss and develop Jackendoff's (1967) proposal to obtain
postposed genitives (like a book of John's) by a process of deletion, from partitive constructions such as one (book) of John's
books. Although in some respects that proposal is attractive, it
appears to have less motivation than Jackendoff claimed for it.
For reasons given below, we reject his solution and offer an alternative analysis involving a postposing rule. In the light of this
we deal next with a number of constraints on the subject placement
rules that form genitives and on the rule which derives possessive
genitives from relative clauses. These constraints may not need
to be separate conditions explicitly stated in the rules, but may
result from rule ordering and so on. But in this section we have
not aimed to do more than describe the facts. The third question
dealt with in this section is the origin of "postposed nominatives"
such as the man in "the arm of the man." Some people have tried
to relate these directly to postposed genitives but we provide an
alternative account. Most of the discussion is included in CASE
PLACE and we merely summarize the argument here. This section
ends with a brief discussion of how predicate genitives might be
derived if they were not generated (as here) in the base, and
some remarks on the rule for deleting the articles when there is
a preposed genitive.
770
GEN - 72
1.
The Derivation of the Postposed Genitive
We have not yet accounted for forms like (159), in which the
genitive, instead of preceding the head, follows it.
(159) (a)
(b)
(c)
The books of John's that you need are on the
table.
We talked for a long time about some proposals
of his to lease three new properties.
A new novel of Iris Murdoch's came out last
month.
Smith (1964) regarded such postposed genitives as a stage in the
derivation of preposed genitives. We have already argued (see (119)
et. seq.) that there is little motivation for this, and that it
complicates the statement of preposing and postposing rules since
such forms as a proposal of mine, which are derived from cases,
must be produced by postposing, whether possessives like a book of
mine are or not. Yet the same constraints apply to both constructions, and postposed genitives that are possessives appear to act
in every way like those that are derived from cases.
Jackendoff (1967) proposed a very different derivation for
postposed genitives, giving them roughly the same underlying structure as surface partitives like some of John's books, something
like (160). Rules required to account for partitive constructions
in general will yield (l6l.b): the first occurrence of books is
reduced to ones and then deleted. Compare: Some men of the men—
Some ones of the men — Some of the men. For (162) on the other
hand it would be necessary to reduce instead the second occurrence
of books to ones. This could be done by making the partitive rule
optional for genitives as it must be if sentences like ?Few men of
those that had been left behind were willing to help are grammatical,
(160)
some
books
John's
771
books
GEN - 73
(161) (a)
DET
John's
some
books
books
(162) (a)
PREP
NP
I
DET
I
NP
Some
books
772
of
John's
GEN - fk
All, or virtually all, the postposed genitives would be produced
in the same way. For a number of reasons, this is an attractive
proposal and one which would fit well into our account of both
pronominalization and partitives—in so far as we have an account
of the latter. (See DET and PRO.) Jackendoff provides several
arguments for it, though he is less definite about the origin of
one than we have perhaps implied here. However, these do not seem
to be adequate to motivate it, in the face of a number of serious
difficulties.
Jackendoff observes that there appears to be a restriction
on the top NP of a partitive construction. He cites the following to show that if that NP is indefinite a partitive is possible,
but that if it is definite it must contain a relative clause, too.
(For further discussion of these problems see DET.)
(163) (a) two of the men
(b) *the two of the men
(c) the two of the men that objected
strenuously
[U5a]
|>5b]
[ l*5c ]
In general, it is clearly necessary to prevent structures like
(l6U) from appearing, while allowing forms like (165), in which
there is an unreduced relative present:
(16U)
PARTITIVE
(165)
If postposed genitives were derived from partitives, as Jackendoff
proposed, a single set of constraints (however formulated) would
773
GEN - 75
prevent the derivation of (l63.b) and the starred forms of (l66)
by blocking (l6U).
(166) (a) *the brighter ideas of his
(b) *the ideas of his
(c) *the two sons of Mary's
(d) his brighter ideas
(e) his ideas
(f) Mary's two sons
The acceptable forms of (167) and (168) would all come from
partitives in which a genitive occurred in the lower (i.e. partitive)
NP, not subject to the constraint of (l6U).
(167) (a)
(b)
(c)
a book of John's
what book of John's
some books of John's that I have
(168) (a) the shoe of Mary's that I lost
(b) *Mary's shoe that I lost
On the other hand, the acceptable forms of (l66) would come
from a genitive dominated by a single, non-partitive NP, for in
these the genitive is preposed: his (brighter) ideas (l66.d,e),
Mary's two sons (l66.f). As in the U.E.S.P. grammar, these
genitives, his and Mary's, are formed by Jackendoff by pre-posing
elements originally to the right of ideas and sons. Provided the
adjective preposing rule precedes the rule forming genitives, it
is a simple matter to allow his and Mary's to be obtained in
(166.d-f), while blocking (l68.b), *Mary's shoes that I lost.
The preposing rule must require that there be no relative in the
top NP, but ignores preposed adjectives (and numerals).
We now propose an alternative analysis of postposed genitives
which, as far as we know, has not previously appeared in print.
Once this alternative has been described it will be possible to
compare it with Jackendoff's partitive analysis.
It will be recalled that Smith (19&M regarded the postposed
genitive as directly obtained from her relative clause source. In
certain environments the (postposed) genitive was then necessarily
pre-posed. It is possible that, as Smith assumed, the postposed
genitive comes from a structure essentially the same as that which
yields preposed genitives, but that instead of the preposed form
being derived from the postposed, there are rules which obligatorily
77U
GEN - 76
postpose the genitive, moving it from the Determiner and placing
it to the right of the head N under certain conditions. These
conditions will, of course, have to yield the same distribution
accounted for by the partitive analysis, and a choice between the
analyses will depend on a comparison of the degree of naturalness
and motivation of the conditions compared with the extent to which
the partitive analysis can account naturally for the facts.
The conditions for postposing will depend largely on the
contents of the Determiners of the top NP, and on whether that NP
contains a restrictive relative clause which has not been turned
into a preposed adjective. Assume that when a genitive is formed
(from a case or a relative clause), it becomes right sister of ART.
If ART is indefinite the genitive has to be postposed:
(l66') (a)
a ^[John's] (blue) book ^a (blue) book of John's
(b)
what
(c)
some
NP
[John's] book =^> what book of John's
Np[John's]
books that I have
=^ some books
of John's that I have
None of the forms given as output above can ever be paraphrased by
a plain, preposed genitive like John's book. Therefore, when the
Article is indefinite postposing is obligatory.
On the other hand, if that Article is definite but there is
no relative clause present, postposing may not take place. Instead,
there is no surface realisation of the Article. (For further discussion of the deletion or loss of the Article see E.U.b of this
paper.) For example:
*(l67') (a) The jjp^18] brighter ideas *£ *the brighter ideas of his
(b) the
(c) the
NP
=£•
his bright ideas (by loss of ART)
[his] ideas *• *the ideas of his
=#
his ideas
[Mary's] two sons &• *the two sons of Mary's
=^
Mary's two sons (by loss of ART)
If, however, the top NP contains an unreduced restrictive
relative, postposition of the genitive must take place, whether
the Article of that NP is indefinite or definite. (This fact has
775
GEN - 77
been used by Chomsky (1965), Jackendoff (1967) and others, to argue
that restrictive relatives originate in the Determiner, but that is
not relevant here.) (l68'.a) shows that postposing may take place;
(l68'.b) demonstrates that it must do so.
(168') (a)
(b)
the
[Mary's] shoe that I lost =* the shoe
of Mary's that I lost,
the
[Mary's] shoe that I lost ±& *Mary's
shoe that I lost (by loss of ART)
(N.B. There are some dialects that apparently do allow (l68'.b).
In the same way, if there are demonstrative elements in the top
Determiner, postposing has to take place. That this is so follows
from the fact that the output of (169) can never be paraphrased by
simple preposed genitives like Lucinda's dresses.
(169) (a)
(b)
those ^[Lucinda's] dresses =^ those dresses of
Lucinda's
which Np[my] proposals =^ which proposals of mine
To sum up, postposing has to take place unless the top NP is
definite and contains neither an unreduced relative nor a
demonstrative.
We can now compare the partitive analysis with this one just
proposed. Jackendoff contrasts his own final version with two
others that he considers and rejects. One of these is essentially
that of Smith (196U) which we have already rejected.
The other proposal involves a rule which optionally creates postposed genitives in_ situ out of the input to the preposing rule if
that preposing rule has failed to apply to certain of these. This
is an unintuitive, ad hoc solution which is rightly rejected by
Jackendoff, and which we shall not deal with in detail. What is
important from our point of view is that the advantages of the
partitive analysis over either of these, carry over, with few
exceptions, to the analysis proposed here. In addition, our analysis has several advantages over the one using partitives.
We shall deal first with the advantages claimed by Jackendoff
for his system. The most important of these, if correct,
is important. He claims that the condition on postposing a genitive
776
GEN - 78
from within a definite NP can be reduced to the constraint (whatever it is) that blocks partitives on definite NP's unless they
contain restrictive relatives. If so, there is much to be said
for an analysis that allows this to be done, since the relationship holding between postposed genitives and elements of the top
NP, as exemplified in (l66)-(l68) certainly requires explanation.
Notice that Jackendoff's suggestion that the peculiar distribution of postposed genitives is related to the restriction on
partitives appears to be supported by another, related similarity
between the two constructions. In general the relative clause on
the head noun of the partitive (or postposed genitive) may not
reduce and prepose. (in (170.a), those ^= the ones: See DET.)
(170) (a) those of his books that are blue
(b) *the blue (ones) of his books
(171) (a) the books of his that are blue
(b) "the blue books of his
The advantages of reducing both problems to a single constraint
on partitives are somewhat reduced by the fact that that constraint
itself remains altogether unexplained. Moreover, when we examine
other constraints on the two constructions there seem to be a number of significant differences between them. First, although in
general the relative clause allowing postposing on partitives may
not be reduced ((170) and (171)), when there is a preposed superlative adjective (and perhaps in other cases) partitives are
allowed but not postposed genitives—unless there is a restrictive
relative clause in the NP.
(172) (a)
(b)
the youngest of the men
the newest of John's cars
(173) *the newest car(s) of John's
(llh)
the newest car of his that I've driven
Similarly, there are a number of quantifiers that fail to uphold
the parallel in any simple fashion. For example, if a phrase like
(175) comes from a partitive like (176) as we have argued in DET,
(175)
all (of) John's books
(176) (*)all books of John's books
777
(cf. all (of) the men)
(cf. all men of the men)
GEN - 79
it is necessary to account for the unacceptability, in most
instances, of such forms as (177) which would be optionally derived
from (176) by the proposed rules.
(177) *all books of John's
Similarly:
(178)
relatively few of John's books
(179) ^relatively few books of John's
The last of such counter-examples to any claim that the restrictions on postposed genitives that have to do with the top determiner
can be explained by reference to partitives comes from demonstratives. When there is a deictic in the top determiner, the genitive
must always postpose, but there are not always acceptable partitive
parallels.
(180) (a) I like these pictures of Rembrandt's but not those,
(b) *I like these (ones) of Rembrandt's pictures, but
not those.
(181) (a) I only want to meet those friends of yours.
(b) *I only want to meet those (ones) of your friends.
It must be granted that all these examples involve relatively
controversial elements; they cannot of themselves provide strong
evidence against Jackendoff's proposal. Moreover, all could be
handled by specially restricting the derivation of surface partitives or postposed genitives. However, they would all be accounted
for by the following rather simple explanation of the distribution
of postposed genitives.
After the genitive is formed by moving an NP into the Determiner,
the genitive can remain there only if the Article dominates [+Def]
and nothing more. This is a rather natural condition ensuring
that preposed genitives are unambiguous, but it requires that relative clauses be represented in some way that will associate them
closely with the Article—unless they have been formed into preposed adjectives. Following Chomsky (1967), Smith (19b1*) and
others, Jackendoff himself wishes to derive relative clauses in
the Determiner, and although we have not tried to work out details
here, we may arbitrarily assume that in some way the article
acquires a feature [+Rel] if there is a restrictive relative, but
that the preposing of an adjective deletes that feature. (Note
778
GEN - 80
that the ART S analysis of restrictive relatives is discussed in
EEL.) The following would have to postpose:
(182) (a)
(b)
the [John's] book that is over there
[+Def]
[+Rel]
(the book of John's that is over there)
those [your] friends
[+Def]
[+Dem]
(those friends of yours)
On the other hand, since preposed adjectives do not trigger
postposition, the following would be generated instead of the ungrammatical forms (171.b) and (173):
(I71.b')
(173')
his blue books
John's newest cars
The ungrammatical examples involving quantifiers, (177) and
(179), although superficially like (173), result from a constraint
on preposing, leaving such forms as (177') and (179') as surface
structures and entirely excluding genitives from such NP's.
(177*)
?
all books that are John's
(179') ?relatively few books that
=•> *John's all books
?are John's
=*
"John's relatively few books
(Note that the starred derivation from (179') is grammatical but
only as a non-restrictive, which is irrelevant.)
Although this constraint is otherwise unmotivated it is not
counter-intuitive. Moreover, to avoid deriving the fully reduced
form of (175), all John's books, from two sources this constraint
would probably have to be incorporated in any partitive analysis
of postposed genitives. (If, as may be the case, the relative
clause sources for (177'), (179') are ungrammatical, this removes
the need for the constraint, of course.)
So far, in response to Jackendoff's main claim, we have tried
to show that there are a number of significant differences between
779
GEN - 81
the constraints on partitives and those on postposed genitives; and
that there is a very natural way of accounting for these differences,
as well as for the constraints that the partitive analysis handles.
We regard the apparent similarities with the partitive, as a chance
result and Jackendoff's proposed generalization as a false one.
The second advantage claimed for the partitive analysis is
that, unlike Smith's (or the straw man that Jackendoff sets up), it
makes all genitives dominated by the Determiner at the surface. All
that is important, however, is that genitive formation should take
place by a single rule, in one environment. But our rules derive
all genitives by preposing, too. (Gerunds and predicate genitives
excepted.) There seems no particular advantage, on this
score, to either analysis, as compared to the other.
Thirdly, it was claimed by Jackendoff that the partitive
analysis will "eliminate the problem of bringing in nominalizations and measure expressions where they are not wanted," citing
the following as examples. (To make them relevant they should,
of course, include relative clauses since they have a definite
determiner, or they should be indefinite. They are, in fact, just
as bad in either case.)
(183) (a) *the assassination of Bill's
(b) *the height of mine
[Ula]
[Hlb]
These are supposed to be automatically excluded by the fact that
neither assassination nor height can appear in a partitive construction:
(18U) (a) *one of the assassinations of Bill
(b) *one of my heights
[U2a]
[^2b]
However, there seem to be a number of more basic constraints
on these head nouns. Between them these go a long way towards explaining both the lack of partitives (otherwise unexplained by
Jackendoff) and the ungrammaticality of forms like (183). It is
impossible to do justice to this claim since there are a great
number of irregularities in this area, and we merely indicate
where, in general, the solution seems to lie, knowing that there
are counter-examples to the generalizations proposed here.
First, observe that few nouns in these classes can appear
with an indefinite article if they have cases on them—even if no
genitive has been formed. In fact Jackendoff cites some relevant
examples:
780
GEN - 82
(185) *a repudiation of a heretic
[l8a]
(186) *a width of a finger
[l8d]
It is not immediately relevant that we can get a sentence like (187)
unless we could also get (188)—which has a Dative case (animate
object) on execution, like (l83.a), and is, like the latter, ungrammatical.
(187)
I should not like to witness an execution.
(188) *I should not like to witness an execution of a criminal.
This effectively excludes one environment that must avoid postposing since these nominals must not be indefinite when they bear
cases. When the head noun appears with a definite article only,
as in the assassination of a president, this can only lead to the
president's assassination, and prevents postposing anyway; so the
remaining question is whether it is possible to obtain forms like
(l89)-(l92). If so, any genitive formed from them would be expected
to undergo postposing.
(189) *The assassination of the president that I witnessed.
(190) *The height of a building that was measured by this
architect.
(191) *The execution of a notorious criminal that took place
yesterday.
(192) *That destruction of a village that John saw on
television.
These vary in acceptability but for many speakers all are excluded.
Some find forms like (191), with an indefinite object, rather better.
But these are irrelevant anyway since they are presumably automatically prevented from forming genitives by whatever constraint is needed
to exclude (193.a) in favor of (193.b). (We discuss this again in
section E.)
(193) (a) *We were surprised by a new saint's canonization,
(b) We were surprised by the canonization of a new
saint.
Thus, there appear to be independently motivated ways of preventing postposed genitives from forming on these nouns: the necessary
781
GEN - 83
environments simply do not occur.
Coupled with restrictions on
pluralization which also apply (cf. ??the heights of the buildings,
*the canonizations of the saints), these constraints will derivatively
prevent the formation of partitives, but we are not specially concerned with that here.)
There seems to be a small class of examples which cannot be
explained in this way. In (19M Washington can be the object of
ortrait. This is an impossible interpretation for (195), both
a) and (b):
?
(19U)
Some of Washington's portraits show him as a
young man.
(195) (a) *Some portraits of Washington's show him as a
young man.
(b) *The portraits of Washington's that I like best...
We have no explanation of (l95tb) since portrait can be indefinite
and can take a relative clause even when it has cases. However,
(19*0 occurs, and the partitive analysis is no better able to
explain (195.a).
The fourth and last claim regarding the partitive analysis
is that it accounts for the fact that indefinite NP's cannot appear
in postposed genitives such as (196). This it does by relating
them to (197), the equivalent partitive, which is also ungrammatical.
It was proposed by Jackendoff that genitives within a partitive be
limited to those with definite articles.
(196) *a daughter of a farmer's
[kka]
(197) ?one of a farmer's daughters
[U8a]
However, as he notes, (197) is much better than (196) and in
fact there are numerous examples in which the correlation quite
breaks down. The following should be equally acceptable according to the partitive analysis, but they are not, and in general
partitives seem much better able to accommodate the indefinite
article in the genitive NP than are postposed genitives.
(198) * /those 1
lthe 1
DO
°^s °f
a
certain old man's that he had
kept since his youth.
782
GEN - Qk
(199)
those of a certain old man's books that he had kept
since his youth
While we do not know how these facts are to be accounted for, we
do not find that they provide any support for the partitive analysis. It is in general very much more difficult to find acceptable
genitives having an indefinite article, and this is not
limited to postposed ones.
We may summarize the preceding discussion thus: Jackendoff's
claims turn out to have far less motivation than he argued for.
Moreover, the most significant of the observations that were supposed to support his position (the first one dealt with above)
tends in fact to throw doubt on the partitive analysis since it is
possible to account for the constraints on postposed genitives
more naturally by means of an alternative explanation. We shall
now consider further evidence against the partitive analysis, which
increases the likelihood that the alternative derivation of postposed genitives discussed above is (in essence) correct.
Before introducing this evidence, however, we have
to admit that there is a rather strong argument for relating
partitives and postposed genitives which Jackendoff did not even
consider. When two morphologically and syntactically similar
forms are close paraphrases of each other, this constitutes good
prima facie evidence for deriving them from the same source.
Consequently, in so far as (200.a) and (200.b) mean the same, it's
likely that they have a common source.
(200) (a)
(b)
Some of our antiques were damaged in the truck,
Some antiques of ours were damaged in the truck.
These two sentences are indeed very close in meaning and we must
contimfe to regard this fact as rather serious counter-evidence
to our proposal. Yet there are aspects of the relation between
these sentences which should be interpreted in favor of deriving
(200.a) and (200.b) from different sources. There is a very
important difference between them. The first, a partitive, presupposes that it is common knowledge that we have antiques. The
second does not. In general we do not assume that transformations
never change meaning. (See CASE PLACE.) Therefore, a difference
in meaning as slight as this, may seem to be little justification
for arguing that (200.a) and (200.b) differ in deep structure.
Nevertheless, the difference observed here is exactly what would
be predicted if the former had a definite article on antiques
somewhere in the deep structure, while the second was essentially
indefinite. Our analysis provides precisely this distinction,
while the partitive analysis does not.
783
GEN - 85
This same difference appears in even more striking ways in
the following:
(201) (a) Some of Mr. Smith's teeth fell into the bath,
(b) ?Some teeth of Mr. Smith's fell into the bath.
Although (201.a) is always acceptable, (201.b) cannot be used
with the sense of the genitive derived from a case. Consider
(202) and (203) in the light of this. Great grandparents seems
unable, like teeth, to occur as an indefinite with a postposed
genitive, although friends can.
(202) (a) *two great-grandparents of his
(b) two friends of his
Cf.
(203) (a)
(b)
two of his great grandparents
two of his friends.
We can see no clear syntactic explanation of these facts.
However, in each of the unacceptable sentences the whole set of
relevant objects (teeth and great-grandparents respectively) is
quite clearly limited in extent. It is at least plausible to
argue that this requires the use of a partitive. The strangeness
of (201.b) and (202.a) would then be regarded as of the same
order as the strangeness of using the sentence ?0ne book on the
table is damaged, where the complete set of books in question is
a matter of common reference. The normal sentence (for the
intended meaning) would be One of the books on the table is
damaged. This, like (201.a) and (202.b), uses a partitive.
Notice that whereas (203.b) implies that "he" has more than
two friends, (202.b) does not. This, again, is what one would
expect if we were dealing with a partitive and an indefinite NP
respectively, since the partitive requires a set larger than that
to which immediate reference is being made.
Our last example of this type of meaning difference is (20U),
(20U) (a)
(b)
During the meeting we considered
of John's about widening various
During the meeting we considered
proposals about widening various
some proposals
roads,
some of John's
roads.
There are in fact a great number of subtle differences between
these sentences, depending in part on whether the about clause is
78U
GEN - 86
read non-restrictively or not, and on which occurrence of proposals in the partitive that clause is supposed to go with.
(See DET for some related problems.) Our main claim here, however, is that (20U.a) requires no assumption about whether John
made other proposals but that (20U.b) implies either that John
made other proposals about widening roads or that he made others
that were not about widening roads, depending on whether the
about clause is on the lower or higher NP (respectively) or the
partitive. It would take too long to show why, but this is
exactly what is predicted by deriving (20U.a) from one indefinite
NP containing proposal, (20U.b) from a partitive.
We turn now to the second kind of evidence against the
partitive analysis. There are good grounds for supposing that we
need a rule to postpose genitives in any case, and if that turns
out to be true it is better to generalize this rule than to add
to the grammar the extra mechanism required to obtain the other
postposed genitives from partitives. At one point, Jackendoff
mentions the phrase:
(205)
that nose of his
[21]
and points out that "we clearly don't want preposing to take
place" in such a phrase. This is presumably to avoid that his
nose, or something of the sort. The discussion of (205) precedes
the proposal to derive postposed genitives from partitives and in
fact no effort is made in the paper to incorporate demonstratives
into the general account. Ordinary deictics could well be
incorporated, as we have shown. (E.g. (l82).)
However (205) does not contain an ordinary demonstrative, and
there is no partitive with a meaning anywhere near that of (205).
Consider two similar examples:
(206) (a)
(b)
Those eyes of Lucinda's often lead her into
trouble!
I dislike that ill temper of his.
Just what those and that are in these forms we do not know, but
it is clear that (l) there is no related partitive like *those
eyes of Lucinda's eyes from which to derive (206.a), and (2)
whatever those is, postposing of the genitive could be made to
follow from the generalization proposed above (p. 79), that
genitives may remain preposed only if the article contains no
more than [+Def], provided those and that violated that condition. There is at least no evidence against those and that
being dominated by the Article. They seem like articles yet they
lire not just definite. Though this argument depends on few
forms, and though the latter are relatively little understood,
785
GEN - 87
the conclusions seem quite indisputable: we need a rule postposing
genitives. Once this is admitted it is necessary to Justify very
thoroughly any proposal to apply constraints that prevent the rule
from applying to similar constructions. The arguments for the
partitive analysis must as a result be that much stronger in order
to be accepted—and we have called them in question on a number
of counts.
2.
Constraints on the Formation of Genitives
Three rules produce genitives: the Active and Passive Subject Placement rules and the Possessive Formation rule. The first
two are described in detail in CASE PLACE; however there are a
number of conditions which must be placed on these rules when they
apply in NP's, and those conditions were not dealt with there in
any detail. The Possessive Formation rule is not to be taken too
seriously, as it stands, as we explained earlier, but whatever
form the rule takes, conditions of the sort discussed here must be
imposed. If it turns out that possessives should be generated
directly in the Determiner some may have to be stated as output
conditions; otherwise it is likely that the account that follows
would apply to any rule for obtaining possessives. Some of the
conditions discussed in this section may well result automatically
from such factors in the grammar as rule ordering, but we simply
impose them arbitrarily on the rules themselves.
It is worth pointing out that in so far as the constraints
discussed here apply specifically where nouns and not verbs are
at the head of the relevant construction, they represent one of the
ways in which surface dissimilarities between S and NP arise. In
this grammar there are, of course, no deep structure subjects; by
imposing conditions, like those dealt with here, on the transformational section of the grammar, we are able to represent S and
NP as highly similar in structure in the base. It is therefore
an interesting question (to which we have no answer) whether
constraints like those proposed here can be more adequately
motivated than constraints on desentential derivations of NP's
and on base rules for S and NP.
Finally, notice that when any string fails to meet a positive
condition it is anticipated that the original (case or relative
clause) form will turn up at the surface. For example, if,
contrary to fact, the top Determiner had to be [+Definite] then
the rule would fail to apply to a case structure like an arm of
the man, but we should nevertheless expect this string to appear
at the surface. In other words, the conditions discussed here
786
GEN - 88
are simply conditions on the formation of (preposed) genitives.
(They do not apply to the formation of gerunds.) Because the
conditions inhibit the formation of preposed genitives, they
affect the distribution of postposed genitives, too.
(a)
Conditions on Determiners
(i)
The Determiner of the top NP
If we had accepted the partitive analysis of postposed
genitives we should need to impose on all three rules for forming
genitives a condition allowing them to apply only if the "top"
Determiner contained a [+Def] Article and neither demonstrative
nor relative clause appeared in that NP. An arm of the man and
a book that is John's would never form genitives. Something like
John's book would be a simple genitive, allowed by this partitive
condition, while a book of John's would introduce the genitive by
means of the lower (partitive) NP which would itself be John's
book.
We have given our reasons for rejecting the partitive solution and therefore need not consider this condition in detail.
However, notice that if we had needed to impose it, this would
have introduced a new argument against using the predicate
genitive as the source for possessives. We suggested that ?The
book that is John's is odd for the same reason that ?The book that
is green is unsatisfactory: there is a shorter, preposed form.
But whereas ?a book that is green becomes a green book, ?a book
that is John's would have to remain un-preposed if the top
Determiner had to be [+Def], and our "explanation" of the oddness
of the un-preposed forms would fall away. However, since we prepose all genitives and later postpose the genitives from a John's
book this particular objection falls away.
(ii)
The Lower Determiner
At least in true action nominals there seems to be a requirement that the lower determiner be definite:
(207) (a) *The girls were disturbed by a man's sudden
appearance on the balcony,
(b) The girls were disturbed by the sudden appearance
of a man on the balcony.
787
GEN - 89
(208) (a) *A young vandal's destruction of the fence
annoyed Mr. Jones,
(b) The destruction of the fence by a young vandal
annoyed Mr. Jones.
(209) (a) *A little child's canonization surprised us.
(b) The canonization of a little child surprised us.
As (209) shows, this condition applies to the Passive rule, as
well as the Active one.
It does not apply to the same extent to other nouns, neither
the Subject Placement rules nor Possessive Formation being
inhibited in this way for them, as the following show. It
is interesting that in some cases a generic rather than indefinite
reading tends to be given to the genitive.
(210) (a)
(b)
(c)
(d)
(e)
(f)
A student's mother came to see me.
A little girl's arm had just been hurt.
An old man's portrait of his daughter was
accepted for the exhibition.
A dark-skinned chinaman's portrait hung near
the door.
One boy's interest in astronomy took him as far
afield as Mt. Wilson.
A little girl's candy had spilt on the floor.
As far as this condition goes, it is necessary only to extend
whatever condition tends to prevent indefinite NP's from forming
the subject of a sentence, so that it applies also to true action
nominals—which of all the constructions falling under this discussion are most like sentences. We do not in fact incorporate
that condition in CASE PLACE and consequently generate (207.a), (208.a)
and (209 .a).
(b)
Conditions imposed by Definite Pronouns
Just as with sentences, NP's do not easily tolerate a definite
pronoun in the by NP Agentive phrase:
(211) (a) ?the execution of the criminals by him
(b) ?the criminals were executed by him
(212) (a) ?the portrait of swans by him
(b) ?the portrait was painted by him
788
GEN - 90
This, like the constraint discussed in connection with (207) and
(208), nay veil be connected with the conditions under which
topicalization is allowed, but we have not tried to account for
such conditions in this grammar and therefore do not deal with
this one in the rules.
There is another constraint which, if it applies to genitiveforming rules, must apply only to the Active Subject Placement
rule, and only when it operates within NP. Consider the following:
(213) (a
(b
*The arrival of him pleased the others.
His arrival pleased the others.
(21U) (a
(b
*The arm of him was broken.
His arm was broken.
(215) (a
(b
(c
The denunciation of him by Cicero.
His denunciation by Cicero.
Cicero's denunciation of him.
(216) (a It The portrait of him (by Rembrandt)
A
(b
His portrait (by Rembrandt)
(c
Rembrandt's portrait of him
Judging only by (213) and (2lU) it would seem that a Dative
must necessarily prepose if it is a definite pronoun and the only
case on the noun. However (215) suggests that the condition is more
complicated. In (215.a) an animate object has been formed,
presumably from the Dative case again. The first preposing rule
that could apply to this string is the Passive. If it applies,
(215.b) is produced. But the Passive is optional. If it does
not apply, (215.a) is left. Usually in NP's the Active rule is
also optional, but in this instance it must obligatorily apply, to
yield (215.c). Thus there seems, in fact, to be a condition
on the Active rule, which makes it obligatorily apply to whatever
NP it would normally move, just in case there is a Definite pronoun
under a case (perhaps necessarily Dative case).
It is not clear how far this constraint extends. (2l6.a) does
not seem too bad, for example. In the other direction, it would
be easy to have the rule cover examples (211.a) and (212.a). However, at least in this grammar, it is necessary that there be a
suitable condition on the rules of pronominalization, since
obviously those rules follow the genitive forming rules. Consequently the latter would have no way of recognizing derived pronouns at the stage when genitive formation takes place. To
789
GEN - 91
achieve the right effect it might seem possible to block pronominalization of the NP immediately after the head of a nominal
construction (probably excluding (2l6)). Since, however, we
generate definite pronouns in the base, we cannot simply have a
condition on the Pronominalization rule, but would have to
formulate an output condition. This may well be an artifact of
this particular grammar and we do not take the trouble to propose
a precise formulation of any condition that would account for
(213) - (216).
(c)
Conditions Depending on Animous
It has often been observed that animate NP's form genitives
far more easily than inanimates do. In some way it is necessary
to block:
(217) (a) *our house's picture
(b) *the picture's destruction by a maniac
(c) *the table's leg
Instead, we get the un-preposed case forms:
(218) (a)
(b)
(c)
the picture of our house
the destruction of the picture by a maniac
the leg of the table
This constraint is not absolute and seems to vary from speaker to
speaker. For example, speakers seem to vary considerably in their
judgments of the grammaticality of (219):
(219) (a) the water's edge
(b) ?the building's height
(c) ?the food's distribution
Whatever form the conditions may take in order to account adequately
for this data, they must be such that the previous condition,
which requires preposing of an NP if it is a definite pronoun,
can take precedence over the present condition:
(220) (a)
(b)
(c)
I estimate its height at about 200 feet.
Although you have the book back, many of its
pages are now torn.
It's destruction by a maniac surprised us all.
790
GEN - 92
Not:
(221) (a) ?I estimate the height of it at about 200 feet.
(b) *Although you have the book back, many of the
pages of it are now torn.
(c) *The destruction of it by a maniac surprised
us all.
On the other hand, there is no genitive relative pronoun for inanimates, and ve find (222) rather than (223).
(222)
The book, the cover of which had been torn, was
found outside.
(223)??The book, whose cover had been torn, was found
outside.
Inanimates never form possessives. (That much seems semantically
clear, at least.) And we have argued in CASE PLACE that for nouns
the Passive rule applies only to Datives (= Animate Objects).
(d)
A Condition on Plural Subjects
We noticed, in connection with (b) above, that indefinite
IP's do not easily form genitives. When the genitive NP is plural
but the head is singular, the result is very considerably worse:
(22U) (a) *some men's racehorse
(b) *those books of some expatriate English authors'
It is not, however, impossible to find a plural indefinite genitive
on a plural head, or to find plural definite genitives, as in the
following examples (respectively).
(225) (a)
(b)
some men's racehorses
the children's go-kart
We do not know anything more about this singularly odd constraint.
(e)
Length Constraints
There is some kind of constraint imposed by the length of the
potential genitive:
(226) *The man who lives on the corner's books
791
GEN - 93
It is not clear how this could be stated but it is presumablystylistic in origin. Notice that the constraint applies equally
to predicate genitives:
(226') •That book is the man who lives on the corner's.
The fact that in general all constraints apply equally to predicate
genitives and other genitives—including those derived from cases—
makes it seem likely that if the predicate genitive is the source
of possessives the constraints on genitives are all output conditions, sensitive only to the genitive and its dominating NP if
relevant.
3.
The Origin of of NP
That there is a relationship between genitives and
following the head noun has often been noticed. It has
generally been very clear what sort of relationship was
since for many common genitives the corresponding of NP
ungrammatical.
of NP
not
involved,
form is
(227) *a book of the boy
We have proposed, in CASE PLACE, a number of ways of deriving of NP
but only two concern us here. On the one hand, this form may
represent an "object" of the N, coming from a deep structure
Neutral or Dative case by a rule inserting of after objectivalization has deleted the original preposition. On the other hand, it
may result from the rule which changes the preposition of any single
case left to the right of the head N to of. In both instances,
of NP originates in a case and has never been a genitive—though
the deep structure from which it has been derived may be eligible
to form genitives which will paraphrase it.
This represents a claim that any of NP (of the classes we have
been dealing with) comes from a case rather than an embedded relative clause (i.e. passive). In the clear instances this seems to
be correct.
There are at least three other possible sources for of NP.
We do not discuss these in detail here, but there appear to be
good arguments against the following:
792
GEN - 9>*
1.
Certain genitives postpose to form of NP losing the
genitive marker as they move.
2.
The genitive marker deletes from certain postposed
genitives.
3.
The form of NP is a partitive of some sort (e.g., in
the arm of the man) from which genitives can form.
Jackendoff's account of of NP does not follow any of (l)-(3)
above, but runs into difficulties which appear to be quite typical
of any account that fails to distinguish between possessives and
other genitives. In order to exclude (227) he has to make genitive
formation obligatory.
For further discussion of the origin of of NP see CASE PLACE.
k.
Miscellaneous Problems
a.
The Predicate Genitive
If the predicate genitive is basic, it is necessary, as we
mentioned above in section D, to constrain it in complex ways. If
it is not basic, suitable conditions must be placed on deletion
and/or subject placement rules in order to secure the right output. The fact that many predicate genitives (e.g. Those books are
John's) appear not to be definite but in some way generic (cf. He
is a carpenter) is not easy to imagine handling under any deletion
analysis.
b.
Article Deletion
It might seem that given our analysis of genitives we require
a rule to delete the article just in case the genitive remains preposed, for then we have a definite article which, on the preliminary
breakdown given in example (167') precedes the genitive, thus:
(228)
the (John's) book
John's book
793
GEN - 95
However, it will be seen from CASE PLACE that the subject placement
rules attach the genitive NP to ART, leaving the existing feature(s)
still attached to that node as well. If the genitive NP is postposed, only the ART features remain, to give (by second lexical
lookup) such forms as a book of John's, and the book of his that I
lost. If, however, the genitive remains attached to ART, the
resulting structure looks something like (229).
(229)
DET
ART
[+D^f]
NP
A
The second lexicon is unable to read the feature [+Def] and there
is no surface form.
c.
Pronoun Suppletion
Consider the following two sentences:
(230) (a)
(b)
John took his book and Mary took hers,
Mary took her book and John took his.
In PRO it is argued that hers in (a) comes from:
(her book «• ) her one
«^ hers .
[+PRO]
Now her has come in a sense from her, itself [+PRO], which arises
by a similar process form:
(Pet one «^)
+III
+Fem
she one
she ,
[+PRO]
It is not at all clear how we can distinguish her and hers (and
similar suppletive pronominal forms) unless the second lexicon is
sensitive to structured sets of features or to the number of
occurrences of a feature on a node. Thus, at present, her and hers
are distinguished by the time of the second lookup simply by the
79U
GEN - 96
fact that hers dominates two occurrences of the feature [+PRO],
acquired by the two processes of ONE-DELETION. Nowhere else have
ve made use of such a device and we are unwilling to do so here.
We do not, however, have an alternative to propose.
P,
Problems not Discussed
1.
The relation between genitives and true compounds like:
(229)
table-top, chair-leg, river-bank, door-handle
2. The relation between the genitives discussed in the paper and
such compound genitives as:
(230) (a)
(b)
(c)
(d)
3.
(new) [gentlemen's clothing]
a big [boy's bicycle]
some [butcher's aprons]
a ladies' man
The following genitives:
(231) (a)
(b)
(c)
a summer's day
the journey's end
yesterday's paper
It is probable that (c) at least has an adverbial origin.
It is interesting that there are sentences having such adverbs as
yesterday in surface subject position, such as Yesterday saw the
beginning of a new quarter at school. These facts may be related.
IV.
TRANSFORMATIONAL RULES
All the rules of CASE PLACE are relevant. They are assumed,
and not repeated here. In addition, the following are required:
1.
Possessive Formation (Optional)
B 1
» -
irot wptX ART X N Xl [NP BE
ro
1
S.C.
NP
S
NP
] ]
[+Dative]
23U5678
(1)
(2)
Attach 9 to 3
Erase 7, 8, 9
(3)
Add [+Genitive] to 3
Conditions
(1)
\l)
(2)
53 does not dominate NP
7 dominates [+THATJ
795
9
GEN - 97
Note:
2.
(1)
The genitive output of this rule is quite parallel to
that of the Subject Placement rules (q.v.).
(2)
Must precede Rule 2, genitive postposing.
(3)
The five conditions of E.2 are relevant but they appear
to be output conditions rather than rule specific.
Genitive Postposing (Obligatory)
S.I.
.__[ NP
]
**"• [+Genitive]
1
S.C.
(1)
(2)
2
X
N
X
3
h
5
Attach 2 as right sister of k
Delete 2
Conditions
(1)
(2)
M [+Def]
(Note:
This is understood strictly: if 1
dominates anything in addition to
[+Def] the rule does not apply.)
3 does not directly dominate NP
Note
(l)
of_ will be inserted between the resulting N and NP by
the very general of-insertion rule (see CASE PLACE).
Thus it is assumed that ?the book by Mailer of John's
that I am reading is generated, if at all, by a later
scrambling rule (which we do not give). The output of
GEN-POSTPOSE would be the book of John's by Mailer that
I am reading.
Examples
A.
Grammatical
(232) (a)
(b)
(c)
B.
a book of his
the proposal of his that you
that nose of his
are thinking of
Ungrammatical - excluded
(233) (a) *the bicycles of hers
(b) ?that [announcement to the creditors] of the
chairman's
August 1969
796
CLEFT AND PSEUDO-CLEFT
Contents
Page
I.
II.
III.
BIBLIOGRAPHY
1
A.
Cleft
1
B.
Pseudo-Cleft
1
INTRODUCTION
1
CLEFT SENTENCES
2
A.
B.
IV.
Dsta-Oriented Observations
1. Constituents Which Can Be Clefted ...
2. Restrictions On Cleftable Constituents .
3. Dubious Restrictions On Clefting ....
k. Constituents Which Can Not Be Clefted. .
5. Phrase Structure Implications
6. Ordering Of Clefting With Respect To
A Few Other Transformations
7. Distinction Of Cleft-Like Constructions.
2
2
k
7
9
10
12
13
Review Of Analyses
1. Simplex Analysis
2. Predicate Relative Clause Analysis . . .
3. Cleft Complement Analysis
h. Impersonal Inversion Analysis
5. Subject Relative Clause Analysis ....
lU
lU
16
17
18
19
PSEUDO-CLEFT SENTENCES
21
A.
21
21
Data-Oriented Observations
1. Delimiting The Pseudo-Cleft Construction
2. Restrictions On Pseudo-Cleftable
Constituents
3. Dubious Restrictions On Pseudo-Clefting.
k. Constituents Which Can Not Be PseudoClefted
5. Phrase Structure Implications
6. Ordering Of Pseudo-Clefting With
Respect To A Few Other Transformations
797
26
28
29
30
31
Page
B.
V.
Review Of Analyses
1. Simplex Analysis
2. Extracting Analysis
3. Something Analysis
k. Parallel IT-S Analysis
5. Parallel NP-S Analysis
SUGGESTED APPROACH
32
32
33
36
39
ko
Ul
798
CLEFT AMD PSEUDO-CLEFT
I.
BIBLIOGRAPHY
A.
Cleft
Lees, R. (1963)"Analysis of the So-Called 'Cleft Sentence' in
English"
Moore, T. (1967) The Topic-Comment Function, Chapter U
Polutzky, H. (19607"*'Cleft Sentences1'
B.
Pseudo-Cleft
Chomsky, N. (1968) "Remarks on Nominalization"
Jacobs, R. and P. Rosenbaum (1967a) Grammar I
Kuroda, S-Y (1965a) "A Note on English Relativization"
(1968) "Notes on English Relativization and Certain
Related Problems"
Moore, T. (1967) The Topic-Comment Function, Chapter k
Peters, S. and E. Boch (1968) "Pseudo-Cleft Sentences"
Rosenbaum, P. (1967a) The Grammar of English Predicate Complement
Constructions
II.
INTRODUCTION
This presentation has four primary objectives: (l) to provide an elucidation of the syntactic restrictions of the cleft
and pseudo-cleft constructions, (2) to demonstrate the many
similarities of the two constructions, (3) to survey critically
the generative analyses thus far proposed, and (k) to suggest a
new approach to the analysis of cleft and pseudo-cleft constructions in light of (l)-(3) above.
As one can infer from the bibliography, very little has
been written on clefting and pseudo-clefting from a generativetransformational point of view. For this reason and because
there are some questions about when constructions should be
considered cleft and pseudo-cleft, a section on the question
of what constitutes an occurrence of each of these constructions
has been included (III.A.1-U and IV.A.l-U).
The two phenomena are first presented in separate, parallel
sections to allow their independent study while simultaneously
799
CLEFT - 2
facilitating their comparison. Comments peculiar to one or the
other construction follows the parallel comments in each section.
Following the introductory orientation, annotation and critique
of previous generative proposals are given. Our suggested approach, along with what we consider to be evidence for that
approach, concludes the presentation.
III.
A.
CLEFT SENTENCES
Data-Oriented Observations
The following remarks and examples are intended to give an
awareness of the various structures which undergo clefting and
those which must be restricted and excluded.
1.
Constituents Which Can Be Clefted
Non-constituents can not be clefted. The following examples
illustrate the constituents which can be clefted.
(a)
NP's can be clefted.
(1)
(a)
(b)
(c)
Rachel cried. «^ It was Rachel who cried.
Mark saw Rachel. *• It was Rachel who Mark saw.
Mark saw Rachel. => It was Mark who saw Rachel.
(b) The structures which the clefted NP dominates are practically unlimited, i.e., they have little effect on the clefting
operation.
(2)
(a)
(b)
(c)
(d)
(e)
(f)
(g)
NRRel:
It was Bill, who seems anemic, that I
was worried about.
RRel:
It was the man with the red coat who
stopped me.
ADJ:
It was that big oaf who stepped on my
foot.
POSS:
It was Sam's book that got torn up.
POSS-ING: It was John's coming home early that
caused problems.
FOR-TO: ?It is to come home late and not find
dinner ready that bugs me.
THAT-S: ?It was that Bill was prejudiced that
I ignored.
[Cf. Section III.A.3 on dubious clefts for those examples with
question marks.]
800
CLEFT - 3
(c) The head of a postposed genitive phrase may generally be
clefted.
(3)
(a)
(b)
It was a handkerchief that Mary wanted of Sue's.
It was a hammer that John took of mine.
This suggests that the head of a genitive phrase is in some way a
separate NP.
(d) When a prep accompanies the NP, there are restrictions on
clefting (cf. see III.A.l (i)) but many NP's can be pulled out of
prep phrases.
(U)
(a)
(b)
(c)
(d)
(e)
It
It
It
It
It
was
was
was
was
was
Bill that John relied on.
the exam that Sue cried about.
Bill that John gave the money to.
Sam that Evelyn came with.
a hammer that Ruth broke the window with.
(e) Whole prep phrases may be clefted.
quite diverse.
(5)
(a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
Their functions may be
It was about Esther that Marcia gossiped.
It was with a stick that Bill killed the rat.
It was to the store that Peter went.
It was for fun that Bonnie and Clyde held up
the bank.
It was for 3 years that Bill lived on that
island.
It was at 3 o'clock that school let out.
It was in school that Harry learned to succeed,
It was with anticipation that Martha waited.
Some clausal constructions (which might be analyzed as prep-phrases)
undergo clefting while others don't. Perhaps a difference in PS
configuration or a semantic restriction is responsible.
(6)
(a)
(b)
(c)
(d)
It was only while his boss watched that John
worked fast.
It was only after prolonged prodding that the
calf moved into the chute.
It was because John begged that I approved his
petition.
It was in spite of John's begging that I
rejected his request.
801
CLEFT - k
(7)
(a) *It was although John begged that I rejected
his petition,
(b) *It is if he comes that I'll scream.
(f) Some adverbials having neither NP nor prep phrase structures
may be clefted in some dialects.
(8)
(a)
(b)
(c)
It was suddenly that the ghost appeared.
It is eagerly that I await your arrival.
It was yesterday that he decided to quit.
In short, NP's, prep phrases, and a few single word adverbials
may undergo clefting.
2.
Restrictions On Cleftable Constituents
There are numerous restrictions which must be placed on the
constituents which can undergo clefting. Some of the following
are mentioned in Lees (1963).
(a)
One of a series of conjuncts can not be clefted.
(9)
(a) *It was John that I saw and Bill.
(b) *It was Elizabeth that and Norma went home.
Likewise, NP's and prep phrases within a conjunct can not be
clefted. (cf. Ross's (1967c) Coordinate Structure Constraint)
(10) (a) *It was Sam that slept and Bill ate.
(b) *It was Ruth that Mary slept and ate.
(c) *It was with his wife that Bill danced and Mark
wrote a letter.
(b)
A preposed genitive may not be clefted.
(11) (a) *It was Sue's that Mary wanted the handkerchief,
(b) *It was the airplane's that the landing gear
stuck.
(c) In case grammar terms, the cases following N can not be split
off when the head is clefted (in contrast to contiguous locative
modifiers of the VP which can not be juxtaposed to a clefted NP).
(12) (a) He read the preface to the book.
(b) It's the preface to the book that he read.
(c) *It's the preface that he read to the book.
(d) *It's to the book that he read the preface.
802
CLEFT - 5
(13) (a) He read the preface to his wife.
(b) *It's the preface to his wife that he read.
(c) It's the preface that he read to his wife.
(d) It's to his wife that he read the preface.
(d) Indefinite pro-fonus are not usually clefted. It seems likely
that this may be a semantic disqualification since one of the
functions of clefting is emphasis of an item.
(lU) (a) *It was something that John wanted,
(b) *It wasn't anything that Mike saw.
(as clefts)
Definite pro-forms (i.e. pro-forms which are overtly indefinite
but in some way definitized semantically) are quite all right as
clefted items.
(15) (a)
(b)
It was you who said that.
It was something new that Sue wanted.
(e) Sentences containing even, scarcely, only, etc. can not be
clefted.
(16) (a) *It is even John who likes old cars.
(b) #It is John who even likes old cars.
(c) *It was even old cars that John sold.
It seems to be the case that the discourse function of these
adverbs is mutually exclusive with the function of clefting.
(f) The subject within a THAT-S construction may not be clefted,
but if the that is absent, clefting is permissible.
(17) (a) John believes that Bill likes tea.
(b) *It is Bill that John believes that likes tea.
(c) It is Bill that John believes likes tea.
The object in a THAT-S construction is not so restricted however.
(18) It is tea that John believes that Bill likes.
This fact could be accounted for in the rules by placing a restriction on the cleft transformation that the variable preceding AUX
in an embedded complement S not contain X + that (cf. NOM IV. and
REL VI.A. for a fuller discussion of this type of restriction.)
803
CLEFT - 6
(g)
Sentence adverbs can not be clefted.
(19) (a) *It was obviously that the theorem was true,
(b) *It is probably that Mary went ice skating.
(h)
Intensifying adverbials can not be clefted.
(20) (a) *It was very that John was tired.
(b) *It was very that Ramon noticed the groundhog
quickly.
(i) When the NP to be clefted has a prep accompanying it, there
are three positions in which the prep is found when the NP is
clefted.
First, the prep may accompany the clefted NP.
(21) (a)
(b)
It was about marriage that Sue was discouraged,
It was on the davenport that Sue slept.
Second, the prep may remain while only the NP is clefted.
(22) (a)
(b)
It was marriage that Sue was discouraged about,
It was the davenport that Sue slept on.
Third, the prep may be fronted in the S from which the NP is
clefted and then it precedes the WH-linker.
(23) (a)
(b)
It was marriage about which Sue was discouraged,
It was the davenport on which Sue slept.
[Note that the latter two possibilities are also present with RRel
clauses.]
Various adverbial uses of prepositional phrases place restrictions on which of the above positions are possible. The examples
given above allow all three positions. Many prep phrases disallow
the second and third prep positions, i.e., they require the prep to
remain with the NP when clefted.
(2U) (a) *It was the morning that I got up in.
(b) *It was the morning in which I got up.
(25) (a)?*It was 3 years that Bill lived on the island for.
(b) *It was 3 years for which Bill lived on the island,
(26) (a) *It is Chicago that they hold the meetings in.
(b) *It is Chicago in which they hold the meetings.
80U
CLEFT - 7
(27) (a) «It was hand that I climbed the rope by.
(b) *It was hand by which I climbed the rope.
(28) (a) *It was fun that Bill held up the bank for.
(b) *It was fun for which Bill held up the bank.
(j) Prep phrases which can optionally delete their prepositions
(i.e. Datives and Benefactives which undergo objectivalization)
require the preposition when clefted. The preposition need not,
however, move in all cases.
(29) (a) *It was John that Bill gave the pencil.
(b) It was to John that Bill gave the pencil.
(c) It was John that Bill gave the pencil to.
(d) It was John to whom Bill gave the pencil.
(30)
(31)
#
It was Mary that Sue bought the book.
It was John who was given the pencil (by Bill).
(32) ?It was John who was bought the book (by Bill).
(k) Clefting out of the predicate of a copular sentence has idiosyncratic restrictions.
Predicate nominals can not be clefted.
(33) (a) *It is a conductor that John is.
(b) *It is Mary who the salesgirl is.
Some NP's in a predicate prepositional phrase can be clefted
if the preposition is not clefted too.
(3*0 (a) It was the train that John was on.
(b) *It was on the train that John was.
Some prep phrases can not be clefted at all.
(35) (a) *It was on time that Bill was.
(b) *It was time that Bill was on.
3.
Dubious Restrictions On Clefting
(a) We have seen that a preposed genitive can not be clefted.
It appears that some postposed genitives can be clefted while others
can not. While there is a special linker whose operating for animate
805
CLEFT - 8
NP's, there is none for inanimates. This may be the source of
greater queasiness about the second pair of sentences.
(36) (a)
(b)
It is the big man with the laundry bag whose
shoes are dirty,
It was Hannibal whose men rode in high style.
(36')(a) ?It was the car with blue pinstriping whose
hubcaps Bill liked,
(b) ?It was the tennis racket whose handle broke.
(b) It may be questioned whether NP's may be clefted without a
RRel which accompanies them. If grammaticality does not disallow
it in the following example, unclarity of interpretation does make
it somewhat unacceptable.
(37) (a) The kid who has long sideburns passed out.
(b) ?It was the kid that passed out who has long
sideburns.
(c) It is rather uncertain whether FOR-TO and THAT nominalizations
can be clefted. Some sentences seem definitely ungrammatical while
others are better. One might explain this phenomenon by invoking
Ross's (1967c) "Completely Enclosed S" output condition (p. 57,
3.27) in which he states that grammatical sentences containing an
NP (l) which is both preceded and followed by non-null parts of
that sentence and (2) which exhaustively dominates S, are unacceptable. Thus, a structure of the form (38) is unacceptable, though
grammatical.
(38)
where X and Y are non-null
In dialects which find all of 39-^1 unacceptable, such a solution
would account for all but (39.c-d) where the embedded S is presumably
pruned before surface structure because it does not branch. However
there is considerable disagreement about the data and we have no
explanation for those dialects which accept anything but (39.c) and
(39.d). The relevant
[S] structures are underlined in the
NP
examples.
806
CLEFT - 9
(39) (a)
(b)
(c)
(d)
(e)
(1*0) (a)
(b)
(c)
(d)
FOR TO: *It is for you to come early that
everyone prefers.
*It is for John to represent us that
I intended.
?«It was to pay the bill that Sam
wanted.
??It is to come home early and find
myself locked out that irritates me.
?It is for you to find me this way that
embarrasses me.
THAT:
»It was that John should represent us
that I decided.
*It was that Bill liked tea that John
believed.
?It was that you came early that surprised
me.
?It was that you would go that they doubted.
(1*1) (a) *It was the boy that that Bill is 65 amazed,
(b) *It was that she was guilty that that she left
proved.
1*.
Constituents Which Can Not Be Clefted
There are some constituents which are very clearly restricted
from undergoing clefting. Among them are the following.
(a)
(b)
Elements which occur in the AUX.
(1*2) (a)
(b)
Preverbs: *It is almost that the theorem is true,
*lt is scarcely that Bill has a chance.
(1*3) (a)
(b)
Modals: «It is must that Bill try harder,
*It is may that Bill finish early.
(1*1*)
NEG: *It is not that Sue baked a cake,
cleft)
(1*5) (a)
(b)
Have, Be: *It was was that John running all day.
*It was has that John run all day.
Particles
(U6) (a) *It was back that he sent the letter.
(b) *It was down that the man hosed the deck.
(c) *It was up that the woman ran the bill.
807
(as a
CLEFT - 10
(c)
Conjunctions, Articles, Postarticles, Adjectives, ...
(d)
Nouns
(U?) *It was pencil that I gave Bill a.
(e) Standard American English does not allow V's or PROP's to be
clefted.
5.
(U8) (a)
(b)
(c)
(d)
*It
*It
*It
*It
was
was
was
was
hit that John did (to) Bill.
know that John ? the answer.
sleep that John did through supper.
be that John tired.
(U9) (a)
(b)
(c)
(d)
*It
*It
*It
*It
was
was
was
was
hit that Harry did.
know the answer that John did.
sleep through supper that John did.
be tired that John ?.
Phrase Structure Implications
We turn now to a consideration of the mutual implications
of clefting and the PS rules.
(a) The structure underlying clefting must in some way provide
for the possibility of two Models, one in the THAT-S and one with
the COP.
(50) (a)
(b)
It may have been Dick who couldn't make up
his mind,
It might be the tent that we should leave
behind.
(b) The relationship of clefting to negation is a complex one.
We only draw the outlines here.
First, clefting seems to be a way in which NP's can be unambiguously negated in English (functioning somewhat like contrast ive stress). Viz.,
(51) (a)
(b)
Bill didn't steal the light bulb.
It wasn't Bill who stole the light bulb.
Second, when the clefted item is negated, there is always
either (A) an implied affirmative S following or (B) an implied
affirmative S preceding the negative cleft. Let us look at case
(A) and then (B).
808
CLEFT - 11
The implied S may be (a) a declarative sentence, (b) another
cleft sentence, or (c) but followed by a structure parallel to the
constituent clefted in the first sentence. Viz.,
(52)
It wasn't John who spilled the milk.
(53) (a)
(b)
(c)
Mary did it.
It was Mary (who spilled it).
but Mary.
When the affirmative S precedes the negative cleft, it may
be of either type (a) or (b) above.
(5>0 (a)
(b)
Mary spilled the milk. [It wasn't John (who
spilled it).]
It was someone else who spilled the milk, [it
wasn't John .../]
Note that the constituent of the declarative S's paralleling the
clefted constituent has special stress.
Third, there may be a series of negated S's tied to a series
of affirmative S's (either preceding or following). The reader
can reduce the following examples:
(55) (a)
(b)
(c)
(d)
It wasn't Mary who spilled the Juice and it
wasn't John who broke the glass. It was Sue
who did them both.
It was Sue who typed my papers. It wasn't
Sally and it wasn't Jane.
It wasn't the man's tie that bothered Bill.
It was his shoes and it was his coat.
It was at Luigi's that the spy met the blond
(and it was at Celso's that he met the brunette).
It wasn't at the bridge and it wasn't at the
museum.
[Note that a mixture in ordering of affirmative and negative S's
suggests a pair of cleft constructions or an afterthought.]
In order to make this implicational relationship explicit,
one might postulate deep structure sentences of the type specified
above for each cleft sentence. The underlying structure would thus
consist of a pair of sentences, one affirmative and one negative
and the transformation could choose one, erasing the other in the
clefting operation. In a sentence-grammar, such a solution would
be unacceptable, as the contrasted deleted item could not be
809
CLEFT - 12
recovered. What we are dealing with, apparently, is a case of
implication of the type which can only realistically be handled
by a grammar adequate to handle presupposition and contextual and
intentional reference, as well as implication and other facts of
true discourse. For the purposes of such a grammar, it appears
that the implicational relationships here discussed would have to
be taken into account.
(c) The clefting operation must not be allowed in imperatives.
(Note our treatment of Imp as SJC. Cf. IMP) That is, SJC in the
top S excludes the possibility of clefting.
(56) (a) Keep the change.
(b) *It was the change that keep.
Clefting may occur lower in the tree, however.
(57) (a)
(b)
Promise that it will be the scissors that
you'll return,
Notice that it is clefting that is operating.
(d) There must be TNS agreement between the copula and the verbal
of the WHAT-S.
(58) (a)
(b)
(c)
(59) (a)
(b)
(c)
(d)
(60) (a)
(b)
It was John that you saw.
It is John that you see.
It will be John that you'll see.
*It
*It
*It
*It
was John that you see.
was John that you'll see.
is John that you saw.
will be John that you saw.
It is John that you'll see.
It will be John that you see.
Note that (60.a,b) are found in most dialects. Though both "present"
and "future" implications appear in their surface sentences, [-PAST]
underlies both of these.
6. Ordering Of Clefting With Respect To A Few Other Transformations
(assuming that cleft is not on a higher cycle)
(a) Clefting may be after conjunction reduction since conjoined
NP's may be clefted. Alternatively, conjunction reduction might
operate on pairs of clefts.
810
CLEFT - 13
(61) (a)
(b)
It was Harry and Sam that tipped over the
outhouse,
It was on Monday and Tuesday that I had salami
sandwiches.
(b) Clefting is after reflexivization since reflexive (and
perhaps reciprocal) NP's may be clefted, thus placing the reflexive
NF to the left of its antecedent.
(62) (a) It was himself that John was concerned about,
(b) ?It was each other that Bill and John respected.
(c) Clefting is before questioning since clefted S's may undergo
both yes/no and WH questions.
(63) (a)
(b)
Is it a toothbrush that you need?
Who was it that said such a terrible thing?
(d) Clefting and pseudo-clefting are mutually exclusive on the
same cycle.
(6U) *It was John that what Bill did was hit.
7.
Distinction Of Cleft-Like Constructions
To avoid confusion, it is important that we sort out the superficially similar structures which appear much like cleft sentences.
(a) The anaphoric it may appear in a string having exactly the
same morphemes and order as the cleft it_, viz.,
(65)
It is money that I need.
The two different constructions which merge in (65) are easily distinguished, however. Lees has noted (1963, p. 382) that in the
sentence with the anaphoric it_ (66.a) the primary stress is on need
while in the cleft usage (66.b) the primary stress is on money.
Note that this result is obtained from the Nuclear Stress Rule
(Chomsky and Halle, 1968) since in (66.a) money that I need is a
single constituent and primary stress is applied to the rightmost
constituent, whereas in (66.b) money and that I need are separated
by a constituent break. The difference between the two structures
is clarified by context, as in (66) below:
811
CLEFT - Ik
(66) (a)
(b)
What money is that? It's money that I need,
What do you need? It's money that I need.
Moore (1967, p. 137) has added the fact that only the anaphoric It
can be replaced by the deictic pronouns this and that.
(67) (a)
(b)
What money is that? That is money that I need,
What do you need? *That is money that I need.
Furthermore, only the anaphoric it_ sentence can have non-agreement
in the TNS, viz.,
(68) (a)
(b)
What money vas that? It was money that I need,
What did you need? *It was money that I need.
•It is money that I needed.
(b) Under all of the following analyses except Klima's, there is
a second it_ which may also be distinguished from the cleft It.
This is variously called the "expletive it" (Langendoen), the
"impersonal it" (Lees), the "Pronoun it" (Rosenbaum), the "anticipatory it" (Curme), and the "introductory it" (Kruisinga). This
it replaces an extraposed NP. For further details see NOM. Notice
that the impersonal it construction does not undergo the WH
transformation.
(69) (a)
It worried [+WH,-DEF] ONE that John left.
(Impersonal it)
(b) *Who worried it that John left?
(c) *Who was it worried that John left?
while the cleft it^ construction does.
(70) (a)
(b)
It was [+WH.-DEF] ONE that John left.
Who was it that John left.
B.
Review of Analyses
1.
Simplex Analysis
(Cleft)
The simplest type of analysis of the cleft sentence would be
one involving only a simplex sentence. Both Lees (1963) and Moore
(1967) suggest and reject simplex analyses. Lees suggests the
following operations such an analysis might involve.
812
CLEFT - 15
(a)
Select a nominal or adverbial constituent Z and attach WH to it.
(b)
Front the WH-Z combination.
(c)
Introduce the sentence by a main clause consisting of IT-BE-Z.
(d)
Allow morphophonemic rules to change WH-Z to who, which, where,
etc. [Lees, 1963, p. 375]
A sample derivation follows:
(71) (a)
(b)
(c)
(d)
You want WH- the book.
WH- the book you want.
It is the book WH- the book you want.
It is the book which/that you want.
And from a sentence like:
(72)
Sam read the review in the train.
[Moore, 1967, p. 123]
one could derive:
(73) (a)
(b)
(c)
It was Sam that read the review in the train.
It was the review that Sam read in the train.
It was in the train that Sam read the review.
There are several problems that such an analysis raises.
First, there is evidence that the AUX's and preverb modifiers
of the two verbals of cleft S's are independent (except for TNS).
This demands a dual sentence source.
(71*) (a)
(b)
(75)
It wasn't John who didn't turn in his reg packet,
It is not the wife who never decides.
It must have been the wife who could always decide.
Second, the semantic component would be required to give the
same reading to all three sentences of (73) since they have a common
deep structure source unless this were regarded as another case of
attachment transformations determining meaning, or unless the structure underlying (73.a-c) were claimed to be three structures identical
to the deep structure of (72) except for some kind of emphasis, focus,
or topic marker.
813
CLEFT - 16
Third, the it_ which is inserted by the transformation must be
dominated by NP to allow the tag question transformation to operate
after the clefting. This would build structure in a way that we
would not like to permit.
(76)
It was John that read the book, wasn't it?
Furthermore, the entire result of clefting must be an S to allow its
operation in embedding.
(77) (a)
(b)
It surprised me that Bill read the review,
It surprised me that it was Bill that read
the review.
The remaining four analyses propose two sentences underlying
each cleft sentence. The first two assume that IT is introduced in
the PS rules; the last two that IT appears as the pro-form of an N.
2.
Predicate Relative Clause Analysis (Polutzky)
According to Lees recapitulation, H. J. Polutzky (i960) suggested
that the two sentences underlying the cleft sentence (78.a) are (78.b)
and (78.c).
(78)
(a)
It is the wife who decides.
(b)
It is the wife.
(c)
The wife decides.
(78.c) is embedded in (78.b) by the relative clause transformation.
Against this analysis Lees has raised three types of objections.
(a) The underlying sentences are not always available as sources.
Thus, the matrix sentence of:
(79)
It was of him that I asked it.
would have to be:
(80) *It was of him.
There are also sentences in which the constituent S can not undergo
the relative transformations as it stands. Thus,
8lU
CLEFT - 17
(81) (a)
(b)
It was in the drawer,
I put it in the drawer.
do not combine by the relative transformations to give:
but (82.b):
(82) (a)
(b)
(82.a)
It was in the drawer that I put it.
It was in the drawer that I put it in.
(b) There are strong ties between the two sentences involved in
clefting which suggest a more than casual relationship. Lees'
first tie—number agreement—is spurious since that would also be
required in the relative source. The second tie is tense agreement.
Thus,
(83) (a)
(b)
It is_ the boys who are naughty,
It was the boys who were naughty.
but not:
(8U) (a) *It is_ the boys who were naughty,
(b) *It was the boys who are naughty.
Obviously this is not an argument applicable only to the predicate
relative analysis. Lees' own analysis (below) requires a special
condition to capture the fact of tense agreement. The third tie is
the correlation between preceding reflexives and following nouns.
(85) (a) It was for himself that he did it.
(b) *It was for himself that they did it.
Assuming that the relative clause could be extended to handle prep
phrases, this third tie would also be handled by reflexivization
in the embedded S. In sum, the objection (b) is practically
weightless.
(c) It is difficult to consider that a relative pronoun since in
many clefts, there is no obvious antecedent of the type found in
relative clauses.
(86)
3.
It was only by dint of great effort that he proved it.
Cleft Complement Analysis (Lees)
The analysis which Lees proposed (1963) involves an ad hoc
phrase structure rule acting as a trigger for the cleft transformation.
815
CLEFT - 18
Converting his two-sentence framework into the present formalization,
two sample derivations follow.
(87) (a)
It AUX BE
g[the
wife decides]
(b)
Cleft =• It AUX BE
(c)
Equi-NP Del + morphophonemics **}
who decides.
(88) (a)
It AUX BE
S[I
s[the
wife WH-the wife decides]
It is the wife
saw him there]
(b)
Cleft s* It AUX BE
g[there
WH-there I saw him]
(c)
Equi-NP Del + morphophonemics «fr It was there that
I saw him.
Thus, a special phrase structure configuration is generated,
which only the cleft transformation will operate. The cleft
formation follows WH-attraction and fronting and operates to
cate the attracted constituent and change WH-X into a proper
upon
transduplipro-form.
Most of the difficulties Lees had noted in the Predicate
Relative Clause analysis (above) are avoided in the Cleft Complement
Analysis; however, new difficulties arise.
Moore (1967) suggested that an It derived in the base failed
to capture the intuition that it is a replacement for an NP.
More seriously, the addition of a unique PS rule to trigger an
obviously language-specific transformation has little to commend it.
Third, the idiosyncratic nature of the solution disallows the
possibility of relating cleft and pseudo-cleft.
k.
Impersonal Inversion Analysis (Klima)
According to Lees (1963), Klima proposed extending Lees' "Itinversion" transformations to account for cleft sentences. The Itinversion transformation (Lees (1960a), p. 9*0 corresponds to
Rosenbaum's extraposition transformation and provides for sentences
like: "That John left bothers me" =»* "It bothers me that John left".
Thus, underlying the cleft sentence of (89) is a sentence like (90)
employing a factive nominal as subject.
816
CLEFT - 19
(89)
It is the wife who decides.
(90)
Who decides is the wife.
This proposal has the drawback of requiring ungrammatical
source sentences. The sentences in (91.a-c) must come from the
corresponding forms in (92).
(91) (a)
(b)
(c)
It was for kicks that she rode the roller coaster.
It was to him that I gave the book.
It is very frequently that she shows up late.
(92) (a) *Why she rode the roller coaster was for kicks.
(b) *Where I gave the book was to him.
(c) *How often/when she shows up late is very
frequently.
Second, in cases where the initial interrogative-like clauses
are possible, they seem to be reduced relative clauses and not factive
nominals. Compare (93) with (9U).
(93) (a)*?Where I found the knife was near him. =^
(b) It was near him that I found the knife, (cleft)
(9h) (a)
(b)
Where I found the knife was obvious. =a^
It was obvious where I found the knife, (ertrapos)
(93) suggests (95) as its source while (9*0 does not.
(95)
The place in which I found the knife was near him.
Third, it is not obvious that (90) and (92) can even be considered
factive nominals.
5.
Subject Relative Clause Analysis (Moore)
Taking Lees' observation about reduced relatives as a starting
point, Moore (1967) suggests that all cleft sentences have as a
source a copulative sentence with an NP, including a restrictive
relative, as subject and a nominal as predicate. The restrictive
relative has a pro-form as head. The relative transformation operates
on the subject, which is then extraposed and replaced by the proform it. A special cleft transformation then inserts that be as left
daughter of the transposed subject. Assuming that the TNS's are
identical one may then optionally delete that BE D + Pro. Some
sample derivations follow. [Moore, 1967, p. 120J.
817
CLEFT - 20
(96) (a)
(b)
_[the #one read the review# one][CQpwas][
Sam]
rel =$ _[the one who/that read the review]
[
(c)
C0PWasHNPSam]
extra =^ [copwas] [Sam [Npthe one who/that read
(d)
the review]]
IT repl =» [NpIt][copwas][NpSam tNpthe one who/that
(e)
read the review]]
THAT BE =* ^ItH^vasH^Saa tNPthat
(f)
(97) (a)
was
the
one who/that read the review]]
THAT BE N DEL *» [NDIt ] L^wasJt Sam [ who/that
nr
Lur
up
up
read the review]] (optional)
j_[the #Sam read the review in the place# place]
[pQpWasltjjpthe train]
(b)
rel ^
[the place (in) which Sam read the review
NP
(in)][C0pwas][ the train]
(c)
extra =>* f
was] [.—the train [„the place (in)
COP
NP
NP
which Sam read the review (in)]]
(d)
(e)
IT repl =* ^p11^ [COPWas]^KPthe train fNPthe plaCe
(in) which Sam read the review (in)]]
THAT BE =-* LpIt][
was] [the train [jjpthat was
(f)
the place (in) which Sam...]]
THAT BE N DEL =* t^It] [copwas] ^Fthe
train
tNP(in)
which Sam read the review (in)]]
Moore contends that this analysis avoids the previous problems
since (l) It replaces an NP (and therefore allows interrogation without node building), (2) the whole sentence continues as an S (allowing embedding without node building), and (3) two separate AUX's are
provided (allowing double negatives and models). Further, the claim
is made that It is the regular replacement of an extraposed [NP]g
when the NP dominates an S. That is, the same process is at work in
factive nominals and cleft.
It is difficult to see how the first contention stands since
the whole NP (including the node NP) is extraposed. It_ easily replaces the NP but how does it get dominated by NP?
818
CLEFT - 21
Note also that the insertion of THAT BE is immediately followed
by its deletion. Its insertion seems motivated simply by the desire
to provide a source for a paraphrase.
One attractive feature of the proposal is its incorporation of
the pseudo-cleft construction. The pseudo-cleft is simply one early
stage in the derivation of the cleft. Thus, (99) may be reduced
immediately to the pseudo-cleft of (100) or may by the steps above
become the cleft of (101).
(99)
The thing #Sam read something^ vas the review
(100) What Sam read was the review.
(101) It was Sam that read the review.
We shall note when considering the pseudo-cleft analyses that this
derivation of the pseudo-cleft raises its own problems. However,
Moore's solution makes the first (and only) step toward combining
two constructions having many similarities.
IV.
A.
PSEUDO-CLEFT SENTENCES
Data-Oriented Observations
A major difficulty in working with the pseudo-cleft is the
delimitation of its domain. Some writers have restricted the constituent which can be pseudo-clefted to NP. Others have opened the
sluice gate wider, apparently not aware of the deluge which follows.
The first five analyses we will consider have dealt solely with
NP*s. We shall wade tentatively into the deeper waters as we note
representative data.
1.
Delimiting the Pseudo-Cleft Construction
This section will present three constructions which seem on
the surface to be pseudo-clefts.
819
CLEFT - 22
(a)
Pseudo-clefted NP's
This first
which an NP has
accepted by all
characteristics
construction, featuring WHAT plus an S out of
been moved to predicate nominal position, is
writers as pseudo-clefting. Let us note several
of the construction.
(1) The constituents under an NP seem to have little bearing on
its ability to undergo pseudo-clefting.
(102) (a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
(i)
(k)
What Bob saw was the book, which had
a tear in it.
What Bob saw was the book with the red
R Rel
cover.
ADJ:
What Liz admired was the huge Gutenberg
Bible.
What hit John was his stupidity.
POSS:
What the lecturer resented was a criticism
of Bill's.
POSS-ING: What frustrates Sara is his failing the
POSSlanguage exam.
What proved that she was guilty was her
leaving.
?What frustrates Sam is for him to come
FOR-T
home and not find a check in the mail.
?What would prove that she was guilty
would be for her to leave.
THAT?What Bill hated was that John left.
?What proved that she was guilty was
that she left.
NRRel
[Cf. the section IV.A.3 on Dubious Pseudo-Clefts for those questioned
above.]
(2) The element functioning as head of a genitive phrase with a
postposed genitive may be clefted.
(103)
What John took of mine was a hammer.
This would suggest that the whole genitive construction is an NP.
Note that the preposed genitive can not be pseudo-clefted.
(lOU)*What John took my was a hammer.
(3) Just as in the cleft construction, there is TNS agreement between the copula and the verbal of the S. This fact is most obvious
with the nominalizations.
820
CLEFT - 23
(105) (a) *What Bill hated is that John leave.
(b) *What Bill hated is for John to leave.
(c) *What Bill hates was John's leaving.
(106) (a)
(b)
(c)
What he saw was a book.
What he sees is a book.
What he will see will be a book.
Some apparent counterexamples follow in (107), where (a), (e), and
(i) are included for comparison.
(107) (a
(b
(c
(d
(e
(f
(g
(h
(i
What he
What he
What he
What he
What he
What he
What he
What he
What he
now).
saw was a book (three days ago).
saw is a book (new).
saw will be a book (three days from now).
sees was a book (three days ago).
sees is a book (now).
sees will be a book (three days from now).
will see was a book (three days ago).
will see is a book (now).
will see will be a book (three days from
Note that the cleft counterparts to (106) are fine but the counterparts to (107) (except (a), (e), and (i)) are ungrammatical (as
clefts_).
The sentences in (107) are only apparent counter-examples for
this reason: pseudo-cleft sentences have an identificational function. In (107) a predication is being made about the state or
condition of the item referred to in the subject. That is, the
sentences have much the same force as:
(108)
What I touched quivered.
quivered.]
[= The thing that I touched
In other words they contain subjects on which there are relative
clauses, and are not the result of pseudo-clefting at all.
(U) When the pseudo-clefted NP is negated there is an implied
affirmative sentence. If the affirmative sentence precedes, it may
be a simple declarative (a), a cleft sentence (b), or a pseudocleft sentence (c).
(109) (a)
(b)
(c)
The bat hit Sam. What hit Sam wasn't the ball.
It was the bat that hit John. What hit John
wasn't the ball.
What hit John was the bat. What hit John wasn't
the ball.
821
CLEFT - 2k
If the affirmative sentence follows, all of the above three mayappear and in addition but followed by an NP parallel to the pseudoclefted NP may occur. As in the cleft, here we are pointing out
facts, not making the obviously false claim that in a sentence
grammar these implied sentences are in the deep structure. (Cf. Sec.
III.A.5 for further discussion).
(110)
What hit John wasn't the ball.
(111) (a)
(b)
(c)
(d)
The bat hit him.
What hit John was the bat.
It was the bat that hit John.
but the bat.
(5) When nominalizations are pseudo-clefted, their surface forms
may differ in grammaticality from corresponding unclefted forms.
(112) (a) What I want is that he leave,
(b) *I want that he leave.
(113) (a) What I want is for him to leave,
(b) *I want for him to leave.
We have not accounted for this phenomenon.
(b)
Pseudo-clefted Cases (prep phrases)
A second construction, which is of dubious grammaticality, is
similar to the first but has various WH forms and a wider range of
predicates (including prep phrases). TNS agreement is maintained.
Their negation also implies a juxtaposed affirmative sentence.
Some examples follow.
(llU) (a) Where John slept was downtown/in a haystack.
(b) ?When I saw him last was at 3 o'clock.
(c) ?How he escaped was with a hacksaw.
(d) ?Why Sam read the review was because he was
interested in it.
(e) ?Who she wants to be seen with is the right
people.
(f) ?Whose house is on fire is theirs.
(g) Whether John left was the issue.
Note that whether will require a reduction from something like
the question whether.
822
CLEFT - 25
These sentences seem to be structurally parallel to the NP
pseudo-cleft constructions. We note only their dubious grammaticality
and the fact that their cleft counterparts are perfectly grammatical.
(c)
Pseudo-clefted PROP's
A third set of sentences which are normally not considered a
part of pseudo-clefting but which seem to undergo a very similar
operation involve what appears to be the pro-ing of the PROP. There
is no restriction on the number of NP's in the PROP, i.e., "transitives" and intransitives work equally well.
(115) (a)
(b)
What John did was throw the paper through
the window,
What Carol did was sleep.
It is possible to leave a copy of parts of the PROP unextracted,
viz.,
(116) (a)
What the mouse did was eat the cheese with its
paws.
(b) What the mouse did with the cheese was eat it
with its paws.
(c) ?What the mouse did with the cheese with its paws
was eat it.
Note that some constituents must obligatorily be copied into the
pseudo-clefted PROP while others are only optionally copied there.
Pronominalization can then take place too.
(Il6') (a) What the
(b) *What the
(c) ?What the
paws was
mouse did with the cheese was eat it_.
mouse did with the cheese was eat.
mouse did with the cheese with its
eat it (with them).
The specification of the preposition of cases which are left
in the unclefted PROP requires special consideration, as is suggested
by (117).
(117) (a)
(b)
What the mouse did with the cheese was eat it.
What the mouse did to_ the cheese was eat it.
There are restrictions on the PROP's which can be pseudoclefted. PROP's which have passive subjects, stative verbs, and
verbs not having an "effectum" relationship (cf. Fillmore, 1967a)
to their object may not be pseudo-clefted. Presumably this is because no suitable pro-form of the verb is available.
823
CLEFT - 26
(118) (a) *What the paper was done was thrown through the
window by John.
(b) *What Dick did was know the answer.
(c) *What John did to the table was build it.
A PROP which has a passive subject can not be pseudo-clefted,
apparently because the pro-form DO may not be passivized.
(119) (a) The back of the chair was fixed by Bill.
(b) *What the back of the chair was done was fixed
by Bill.
Likewise, a pseudo-clefted NP can undergo the Inversion transformation while a pseudo-clefted PROP probably can not. Compare:
(120) (a)
What was bothering Susie was the spider.
(b) The spider was what was bothering Susie.
(c) What Sam did was put another notch in his gun.
(d) ?Put another notch in his gun was what Sam did.
This seems to be nothing more than a simple restriction on the Inversion transformation.
We must also account for the ungrammaticality of (l21.f).
(121) (a)
What the mouse did was eat the cheese. =^
(b) What
(c) What
(d) What
(e) What
(f) *What
2.
(a)
the mouse did with the cheese was eat it.
the mouse did to the cheese was eat it.
Bill did was give the book to John. =»^
Bill did with the book was give it to John.
Bill did to the book was give it to John.
Restrictions On Pseudo-Cleftable Constituents
An NP which is a conjunct can not be pseudo-clefted.
(122) (a) *What I noticed and doves was parakeets,
(b) *What I noticed parakeets and was doves.
This obeys Ross's (1967a) conjunct movement constraint. Similarly,
a constituent of a conjunct can not be pseudo-clefted, but it isn't
clear that Ross's constraint will apply here, since it is not obvious
that anything is moved out of the first conjunct.
(122') (a) *What John bought was a watermelon and Bill bought
a canteloupe.
(b) *What went flat was a tire and the radiator leaked.
821*
CLEFT - 27
(b)
A preposed genitive may not be pseudo-clefted.
(123) (a) *What the landing gear was stuck was airplane's,
or (b) *What landing gear was stuck was the airplane's.
(c) The cases on a head noun may not be left behind when the noun
is pseudo-clefted.
(12U) (a) *What he read to the book was the preface.
(b)??What he read by James was a book, (as pseudocleft)
This suggests that nothing lower than NOM may be pseudo-clefted.
(d) The pseudo-cleft operation does not apply to some sentences
containing even, scarcely, only, etc.
(125) (a) *What Bill collects is even U.S. stamps,
(b) *What Bill only collects is U.S. stamps.
Note the following however, in which the relevant word is in the
what clause.
(125') (a)
(b)
What even Bill collects is trading stamps.
What John scarcely passed was the French exam.
(e) Within THAT-S constructions, the subjects can not normally be
pseudo-clefted while objects can.
(126) (a) *What John believes that causes waste is machines,
(b) What John believes (that?) machines cause is
waste.
(127) (a) *What John believes that is caused by machines is
waste,
(b) What John believes (that?) waste is caused by is
machines.
Note the same phenomenon with subject and locative.
(128) (a) *What John believes that grows on trees is tea.
(b) What John believes that tea grows on is trees.
As in the cleft, when that is deleted, any of the HP's may be pseudoclefted.
825
CLEFT - 28
(129) (a)
(b)
(c)
What John believes causes waste is machines.
What John believes is caused by machines is waste.
What John believes grows on trees is tea.
[Cf. Sec. III.A.2.(f) for discussion of parallel examples in clefting,
as well as an account (of sorts) for the phenomenon.]
(f)
Animate NP's can not be pseudo-clefted with the pro-form who.
(130) (a) *Who Bill saw was John.
(b) *Who Nancy stole for was her mother.
(c) *Who was seen by the police was the criminal.
All of these sentences must be prefaced by the one to make them
grammatical.
As Peters and Bach (1968) point
times be pseudo-clefted, e.g.
(131) (a)
(b)
out, animate NP's can some-
What Bill saw on the horizon was Mary,
What concerned John was Mary.
One might suppose that in sentences of this type the animate NP is
considered an object, i.e., it is treated as inanimate. Thus, much
the same thing is happening as in:
(132)
James Bond broke the window with the Russian.
where the Russian is an instrument, but while there is some motivation
for presupposing instrumentals are [-Animate], there appears to be no
independent motivation for doing so in the case of (131).
A second explanation is proposed in the consideration of Peters
and Bach's analysis. Cf. III.B.2.
3.
Dubious Restrictions On Pseudo-Clefting
(a) We have noted above that a preposed genitive can not be pseudoclefted. There is some uncertainty however regarding postposed
genitives.
(133) (a) ?What Bill liked the hubcaps of was the car with
the blue pinstriping('s).
(b) ?What the handle of broke was the tennis racket('s).
826
CLEFT - 29
The restriction is enhanced when an animate NP is pseudo-clefted
(contrast the cleft parallel).
(13**) *Whose shoes are dirty is the big man('s).
(b) There is some uncertainty about whether a R Rel can be split off
it head when the head is pseudo-clefted.
(135) (a) ?What John shot which had a magnificent tail was
a coon,
(b) ?What Billy likes which has black spots is a
snake.
(c) There are curious restrictions on pseudo-clefting complements
which have FOR-TO and THAT nominalizations as deep structure subjects.
(136) (a)
(b)
(c)
(d)
(137) (a)
(b)
(c)
(d)
k.
„„[For her to leave] would prove „_[that she
NP
NP
was guilty].
[That she left] proved [that she was guilty].
[Her leaving] proved [that she was guilty],
[That she left] proved nothing.
=£ *What
that she
=j* *What
guilty.
=^ What
guilty.
=£ What
for her to leave would prove would be
was guilty.
that she left proved was that she was
her leaving proved was that she was
that she left proved was nothing.
Constituents Which Can Not Be Pseudo-Clefted
There are a number of constituents which can not be pseudoclefted. Some of them are: Preverbs, Modals, NEG, HAVE, BE,
Particles, Conjunctions, Articles, Postarticles, Adjectives, Nouns,
etc.
(138) (a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
*What the fish is rotten is almost.
*What a table is shiny is should.
*What the paper tore easily is not.
*What the water running was was.
*What the pencil broken was has.
*What/where/how the woman ran the bill was up.
*What the pencil was was green,
*What I gave Bill a was pencil.
827
CLEFT - 30
5.
Phrase Structure Implications
There are a few relationships pseudo-clefts have which hear
directly on the PS rules.
(a) Provision must he made to allow modals to occur with either
or both verbals in the pseudo-clefted S. Note the difference in
meaning in pairs (139) and (l^O).
(139) (a)
(b)
What she wants may be spinach,
What she may want is spinach.
(lUo) (a)
(b)
What she caught may have been a trout,
What she may have caught was a trout.
(lUl)
What she may need can't be LSD.
(b) Sentences with pseudo-cleft NP's must be permitted to have
negation in one or both verbals; however, a contiguous affirmative
sentence or a following contrastive but phrase seems to be implied
in the discourse when the negative is on the clefted element.
(lU2) (a)
(b)
What John didn't like was the applesauce,
What John liked was not the applesauce.
(lU3) (a)
(b)
What John didn't talk about was not the taxes,
What Sue didn't like was not the applesauce.
[Cf. Sec. III.A.5 for a more detailed consideration.]
(c) Imperative constructions must be made mutually exclusive with
pseudo-clefting on the last cycle. That is, when SJC is present in
the top S (cf. IMP) pseudo-clefting can not operate on the last
cycle.
(lUU)
Buy the sled =#^
*What buy is the sled.
Peremptory sentences (cf. IMP) can, however, be pseudo-clefted.
(1^5)
You will buy the sled ^^ What you will buy is the
sled.
Clefts may be embedded in an IMP sentence however.
(lU6) (a)
(b)
Forget that what you bought is a white elephant,
Remember that what you seem to be is an honest
politician.
828
CLEFT - 31
(d) There must be TNS agreement between the copula and the verbal
in the WHAT-S. Cf. (105)-(108) for a discussion of this phenomenon.
6. Ordering Of Pseudo-Clefting With Respect To A Few Other Transformations (assuming that Pseudo-cleft is not on a higher cycle)
(a) Pseudo-clefting may follow conjunction reduction since pseudoclefted NP's may dominate conjuncts. Alternatively, conjunction
reduction might operate on a pair of pseudo-clefts.
(lVf) (a)
(b)
(b)
What I noticed was parakeets and doves.
What 1 was fed up with was parakeets and doves.
Pseudo-clefting comes after Reflexivization
The reflexive transformation must precede the pseudo-cleft
transformation unless reflexivization occurs on a lower cycle.
(lU8) (a)
(b)
What Bill saw in the mirror was himself,
What the missile damaged was itself.
(c) Pseudo-clefting precedes the copula switch transformation since
the copula switch transformation can apply to apparently all pseudocleft sentences. It puts the clefted item first in the sentence
followed by the copula and the remainder of the S.
(1^9) (a)
(b)
What fell was the book.
=^
The book was what
fell,
What she likes is applesauce. => Applesauce
is what she likes.
Note the lack of restriction on transposing in comparison to simple
copular sentences.
(150) (a)
(b)
This is the apple. =^ *The apple is this,
This was what I saw =4 What I saw was this.
(151) (a)
The secretary is a fool. =9^ *A fool is the
secretary,
His secretary is what concerns her. ••#• What
concerns her is his secretary.
(b)
829
CLEFT - 32
(d)
Pseudo-clefting precedes Relative formation.
The relationship to Rel formation is tentative. Since no NP
may be pseudo-clefted out of a relative clause, there does not seem
to be any reason to have relativization precede pseudo-clefting. On
the other hand, relativization can not occur using a pseudo-clefted
NP as the identical N.
(152) (a)
(b)
John wanted the pie [what was in the refrigerator
was the pie].
=^ *John wanted the pie which what was in the
refrigerator was.
This suggests that we put relativization before pseudo-clefting.
However, if the copula switch T has applied, the previously pseudoclefted NP is free to be relativized.
(153) (a)
(b)
John wanted the pie [the pie was what was in the
refrigerator].
=^ John wanted the pie which was what was in
the refrigerator.
Thus relativization must come after pseudo-clefting. A restriction
possibly related to that which prevents predicate nouns from being
relativized as in *the teacher that Bill is must then be placed on
the relativization rule.
(e)
The question transformations follow pseudo-clefting.
(15*0
Was what John wanted a match?
(f) Clefting and pseudo-clefting must be made mutually exclusive
within the same cycle.
(155) *It was John that what Bill did was hit.
B.
Review Of Analyses
1.
Simplex Analysis
This first analysis, which we shall call the simplex analysis,
we have not been able to find in print. [Jacobs and Rosenbaum
(1967a), p. 20, imply such a source but give no analysis.] Apparently
a single transformation would apply to an S, converting its P-marker
into one like (156).
830
CLEFT - 33
(156)
what
An appropriate NP is chosen in S, for pseudo-clefting. A new
sentence SQ is created dominating S-^ via an NP. The VP of SQ
dominates COP and the NP removed from S.. . What is made the left
sister of S,.
The simplex analysis meets numerous difficulties, the first
of which is the need to use powerful structure-building transformations .
Second, the Models of the sentences are apparently independent.
(157) (a)
(b)
What she can drink may he goat's milk.
What he might have said, not done, may have been
the faux pas.
Third, double negation may provide evidence for a two sentence
origin.
(158) (a)
(b)
What he won't eat isn't apples.
What he didn't do is not the issue.
Fourth, there is the difficulty of choosing only NP's that
can actually be pseudo-clefted. Cf. Peters and Bach's (1968)
second attack on Extracting Analyses (see IV.B.2) for a full discussion of this point.
Fifth, no specification was made in the deep P-marker as to
which NP would be pseudo-clefted, a criticism which applies equally
to the extracting analysis (following).
2.
Extracting Analysis (Chomsky, Peters and Bach)
Employing Peters and Bach's (1968) nomenclature, we next consider the extracting analysis of the pseudo-cleft. Peters and Bach
831
CLEFT - 3U
characterize this analysis as assigning pseudo-cleft S's deep
structures which (a) include exactly the deep structure of the
corresponding unclefted sentences and (b) form the pseudo-cleft
sentence by extracting the pseudo-clefted NP from its position
in the unclefted S.
Chomsky (1968) suggests in passing one form an extracting
analysis might take. He proposes a deep structure such as (159)
for pseudo-clefts. [Chomsky, MS, 1968 (1*7)]
(159)
IT
John read a book about himself
A
Some NP in Sp is chosen to replace the unspecified Pred and
that NP is replaced by the pro-form IT. The relativization transformation then operates on the identical IT's and a new rule
changes IT-THAT to what.
Peters and Bach propose two slightly different forms of
extracting analyses. The deep structure in both is identical to
(159). The operations performed on it in the first form are the
following. Some NP in S2 is chosen to replace the unspecified
Pred by attaching WH- to it. That NP is pro-ed by something. WHsomething is then attracted to the front of the S and a morphophonemic T changes IT-WH-something into what. Peters and Bach's
first attack on extracting analyses (pp. 2-3) is actually an objection to blocking symbols and is misleading and irrelevant. They
argue that blocking symbols allow an incorrect (descriptively inadequate) account of the post-determiners main, chief, etc. and
hence should not be used in the pseudo-cleft analysis and should be
excluded from the metatheory. However, Peters and Bach do not
show conclusively that blocking symbols cannot be appropriately
constrained in use.
832
CLEFT - 35
Peters and Bach's second argument against extracting analyses
also runs into problems. They first (pp. 5-6) present evidence
that the restriction on which NP's can be pseudo-clefted is not
tied simply to animacy. They note sentences such as the following:
(160) /Mary
"j
1 The news )) concerned John ^ What concerned John
was (Mary
1.
(^the news)
(161)
John saw {^73hipj -» What John saw was {^ghip}.
(162)
John amazed iMary
] «^ What John amazed
|«the shipj
was f *Mary
\
[_*the ship j
'
Their observation is that only an NP which can be replaced by
something in the unclefted S may be clefted.
They consider this devastating to all extracting analyses since
after transformations like subject raising and passive have applied,
it is impossible to tell if the NP extracted could have originally
been replaced by something. However, some verbs allow, for example,
both subjects which can be pro-ed by something and those which can't.
Viz.,
(163) (a)
(b)
(l6U)
The mouse ate the cheese,
Mary ate the cheese.
Something ate the cheese.
(165) (a) What ate the cheese was the mouse,
(b) *What ate the cheese was Mary.
The diagnostic something fails in all such cases.
The third problem which Peters and Bach point out for the extracting analysis concerns the place of semantic interpretation. If
semantic interpretation is on the base then there is no way of
indicating the difference in interpretation between the following
pairs of sentences all of which presumably have the same base P-marker.
833
CLEFT - 36
3.
(l66) (a)
(b)
What struck the house was lightning.
What lightning struck was the house.
(l6T) (a)
(b)
Was it lightning that struck the house?
Was it the house the lightning struck?
Something Analysis (Peters and Bach, Kuroda, Moore)
A second type of analysis presented by Peters and Bach we
shall call the Something Analysis. It assigns pseudo-cleft sentences
a deep structure similar to the following.
(168)
the
thing
John counted something
the pigeons
Something concerned John
The relativization transformation applies to S2 giving the thing
which/that which is optionally converted into what.
This analysis avoids the problem of determining the pseudocleftability of the NP and shows the paraphrase relationship of
what and the thing which/that.
Kuroda (1965a) implies an analysis very similar to the above.
He refines the pro-forms used, however, and correlates what with
that which. The basic pro-form in S2 is then SOME PRO. The two
possible derivations from (l68) are as follows.
(169) (a)
(b)
(c)
(d)
(e)
THAT PRO #WH SOME PRO John counted* was the pigeons
Def =s» THAT PRO #WH THAT PRO John counted* was
the pigeons
Pro Del —* THAT PRO #WH THAT John counted* was
the pigeons
WH amalg —* THAT PRO #which John counted* was
the pigeons
Pro-ing =^ That which John counted was the pigeons
83U
CLEFT - 37
[The R Rel identity condition is met with PRO.]
(170) (a)
(b)
(c)
(d)
THAT PRO #WH SOME PRO John counted* was the pigeons
Regr Del =* #WH SOME PRO John counted! was the
pigeons
WH araalg =^ What PRO John counted was the pigeons.
Pro-ing =^ What John counted was the pigeons.
Under Kuroda's analysis the thing which has the following derivation.
(171) (a)
(b)
(c)
(c)
the thing #WH SOME thing
Def =* the thing #WH THAT thing
Prog Del => the thing #WH THAT
WH amalg ^ the thing which
(It is not obvious how Kuroda plans to get the thing that.)
Kuroda's analysis, in contrast to the Something Analysis, distinguishes the thing which from what and that which. This accords
well with the difference in paraphrase relations possible when the
clefted item varies in abstractness. For example:
(172) (a)
(b)
What/that which they laid aside was the
tissue [-Abst],
What/that which they laid aside was the
issue [+Abst].
(173) (a)
The thing which they laid aside was the
tissue [-Abst].
(b) *The thing which they laid aside was the
issue [+Abst].
Moore extends this analysis to include the various PRO's which
Katz and Postal (196U) proposed. Viz.
(Ilk) (a)
(b)
(c)
(d)
(e)
(f)
the
the
the
the
the
the
thing that =* what
place that =^ where
time that ^> when
way that ^ how
question that »* whether
reason that «•*• why
He thus includes a great deal more under pseudo-cleft than any of
the proposals have.
835
CLEFT - 38
Against the Something and SOME PRO analyses, Peters and Bach
point out the complication they entail for numerous transformations
which require knowledge of the position of NP's in the S (e.g.,
reflexive, reciprocal, pronominalization, case marking, and number
agreement). These analyses do not indicate the position of the
clefted NP in its correlated unclefted S. Thus the reflexive transformation, to use Peters and Bach's example, must be extended in
such a way as to allow reflexives outside a single S, for example,
in order to generate (175):
(175)
What the missile damaged was itself.
The Reflexive transformation has to operate on the marked NP's in
the tree structure (175'):
(175')
the
the
missile
damaged
something
was
the missile
A second disadvantage of the Something Analysis is the difficulty
of stating co-occurrence restrictions across the copula. Thus, in
example (168), the thing S BE the pigeons is all right regardless
of the verb used, but the thing S BE Mary must be excluded if S contains the verb noticed but not if S contains the verb concerned.
Viz.,
(176) (a)
(b)
What John counted was the pigeons,
What concerned John was the pigeons.
(177) (a) *What John noticed was Mary.
(b) What concerned John was Mary.
The third point against this analysis is the failure of the
something diagnostic exemplified in sentences (l63)-(l65) and discussed immediately above them.
836
CLEFT - 39
U.
Parallel IT-S Analysis (Ross, Peters and Bach)
Having noted the deficiencies of the first three analyses,
especially their failures to (a) have the pseudo-clefted NP indicated in the deep structure, and (b) derive the pseudo-cleft NP in
a single S identical to the unclefted S, Peters and Bach suggest
the following deep structure (suggested to them by Ross) which
meets both of the requirements. Let us call this analysis the Parallel IT-S Analysis.
The deep structure source for (178) is (179).
(178)
What John counted was the pigeons.
(179)
the thing
John counted something
John counted
the pigeons
The relative T applies to NP, yielding a P-marker which the following pseudo-cleft T operates on.
(180)
SD:
.the thing
SC: ==
a[that,
X Y] AUX BE.IT #
e[X»
NP Y'] #
1
23
It
5
6789
1
2 3
k
0
0
7
0
0
There are several things to be said against this analysis as
it stands. First, as in the Something Analysis there is an optional
rule permitting the thing that to become what. The above arguments
against this and how it can be corrected can be repeated here. (See
sentences (172) -(173).
Second, the base structure is not a plausible one for semantic
interpretation (cf. "the thing that John counted was (it) that John
counted the pigeons"). NPp does not naturally provide a reading
837
CLEFT - kO
the pigeons. As it stands, semantic interpretation must wait until
the application of the pseudo-cleft transformation—a position Peters
and Bach rejected for the Extracting Analysis.
Third, from the parallelism of the thing S and the one S,
(as Kuroda (l°68) points out) the Parallel IT-S analysis will not
account for sentences like the following.
(181)
The one John told Mary to shave was himself.
(182)
The one who shot John was John himself.
The problem with (l8l) is the inability to reflexivize under the
IT-S, viz.,
(183)
^[IT _[*John told Mary to shave himself] ]
sir
o
And the problem with (182) is accounting for the presence of N plus
the reflexive (intensive?).
5.
Parallel NP-S Analysis (Kuroda)
Kuroda's proposal (1968) to correct the Parallel IT-S Analysis
simply involves replacing IT by the clefted NOM and making the following S non-restrictive. The deep structure for the tired out pigeon
sentence thus looks like (l8U).
(18U)
the thing
John counted something
the pigeons
John counted
the pigeons
This provides a much more plausible source for semantic interpretation.
We fail to see however how this analysis accounts for Kuroda1s
problem sentences. (185) will have the deep structure of (186).
838
CLEFT - 1*1
(185)
The one who John told Mary to shave was himself.
(186)
,„Jthe °ne se[John "told Mary 8
-jMary shave someone]]] BE
UP
NP_[John
[John told Mary
q[Mary
shave John]]]
All of the elements in NP2 will be deleted except those underlined.
What kind of rule will reflexivize the second John remaining in
NP2 and delete the first here but not do the same to the pigeons in
(loU) ?
V.
A SUGGESTED APPROACH
Previous analyses have treated clefting and pseudo-clefting
as two separate unrelated operations. We should like to suggest
that they are related in that they have a common deep structure and
that, in essence, the cleft transformation has an input the result
of pseudo-clefting. This suggestion is a highly tentative one, with
problems which we have so far been unable to resolve. The major
unresolved problem is that of deriving both the it_ of the cleft and
the what of the pseudo-cleft from a common deep structure in a wellmotivated way.
We seek to relate these two transformations, not only because
of an intuitive feeling that they are two different ways of doing
the same thing, but also because of a rather large number of
properties which they share, including:
a.
Both constructions share essentially the same constraints on which constituents may be focussed on
(cf. sections III.A.1-1* and IV.A.l-U) except that NP's
whose referential pronouns are who cannot be pseudoclefted and PROP'S cannot be clefted.
b.
Both constructions have the same tense restrictions on
the main and embedded verbs (cf. section IV.A.1.(a)(3)).
c.
Both constructions allow independent occurrence of
models and negation in the main and embedded verbals
(HI.A.5.(a)-(b); IV.A.5.(a)-(b)).
d.
Both exclude the occurrence of the imperative in the
top S (HI.A.5.(c); IV.A.5.(c)).
839
CLEFT - k2
e.
They are mutually exclusive within the same cycle
(HI.A.6.(d); IV.A.6.(f)).
f.
They have the same ordering relationships with conjunction reduction, reflexivatization, and question (ill.A.6;
IV.A.6.).
g.
Under negation, they share the same type of implicational
discourse structure (III.A.5.(b); IV.A.l.(a)(l+); IV.A.5.
(b)).
h.
Whenever the same elements can be either clefted or
pseudo-clefted, the cleft and pseudo-cleft appear
synonymous, as in:
(187)
It was the cheese that the mouse ate.
(188)
What the mouse ate was the cheese.
Because of differences in the two constructions, however, there
are difficulties in a sequential derivation of the type that we
propose. For example, we must generate some ungrammatical pseudocleft structures in the derivation of their corresponding grammatical
clefts, as in:
(189) (a) *Who lost his contact lenses was Alfred,
(b) It was Alfred who lost his contact lenses.
(190) (a) *Where they bought those bracelets was (in)
Solvang.
(b) It was in Solvang that they bought those
bracelets.
In addition, we must prevent the cleft operation from applying to
pseudo-cleft PROP'S:
(191) (a) What he did was fasten down the carpet,
(b) *It was fasten down the carpet that he did.
For the former cases, we must make clefting obligatory and for the
latter, restrict the cleft-transformation from applying to PROP's.
Obviously, too many such examples would begin to make the analysis
suspect, but the major problem, as mentioned above, is with the
pronouns. Because of this major unresolved difficulty, we present
no specific analysis here.
December I968
8U0
PASSIVE
Contents
Page
I.
BIBLIOGRAPHY
1
II.
INTRODUCTION
1
III.
A.
Chomsky
(1957)
1
B.
Katz and Postal
(1964b)
2
C.
The Two-Sentence Passive
4
THE CASE-GRAMMAR ANALYSIS
8
8Ul
PASSIVE
I.
BIBLIOGRAPHY
Chomsky, N. (1957), Syntactic Structures
(1965), Aspects of the Theory of Syntax
(1968), "Remarks on Nominalization"
Fillmore, C. (1966a), "A Proposal Concerning English Prepositions"
(1967a), "The Case for Case"
Hasegawa, K. (1967), "The Passive Construction in English"
Katz, J. and P. Postal (1964b), An Integrated Theory of Linguistic
Descriptions
Lakoff, G. (1965), On the Nature of Syntactic Irregularity
Lees, R. (1960), The Grammar of English Nominalizations
Perlmutter, D. (1968b), Deep and Surface Structure Constraints in
Syntax
II.
INTRODUCTION
In the development of generative transformational grammar,
there have been four basic analyses of the passive in English so far
proposed.
A.
Chomsky (1957)
In this analysis, the passive was an optional transformation which could be performed on the structure underlying an active
kernel sentence as follows:
(1) NP1-Aux-V-NP2 —)
NP2-Aux+be+en-V-by+NP1
Chomsky claimed that the passive was not always synonymous with the
active to which it was transformationally related and cited his wellknown example:
(2) Everyone in the room knows at least two
languages.
(3) At least two languages are known by everyone in
the room.
where the claim was, essentially, that in the "normal" interpretation
of (2), different languages may be known by different people, but in
(3) the languages must be the same for all of the people in the room.
8U2
PASS - 2
B.
Katz and Postal (1964b)
In order to support the hypothesis that slngulary transformations do not change meaning, It was necessary for Katz and
Postal to deal with Chomsky's claim. They argued both possible
alternatives. First, they stated that while there was considerable
disagreement about the data, to them both sentences could have the
same two Interpretations. Further, they argued that even if Issue
were taken with their interpretation of the data, the underlying
structures of (2) and (3) should differ by the presence or absence
of a manner adverb in order to capture the relations between manner
adverbs and passivizability. That is, it was claimed that the verbs
whose sentences could be passivized were also the verbs which
allowed manner adverbs. The underlying forms of passive sentences
thus contained the ADVERBjjanner constituent dominating by. plus a
passive dummy marker, whereas the underlying forms of active sentences did not. In the phrase structure rules one could choose
either the by_ plus PASSIVE or an actual manner adverb. Since dummy
morphemes were regarded as having no semantic content, the semantic
interpretations of active and passive sentences could be the same,
although the underlying forms were distinct.
However, as Lakoff (1965, Appendix F) has pointed out,
there are a number of exceptions to the manner adverb-passivizability
correlation, such as know, consider, think, perceive which do not
allow manner adverbs, but which can be passivized and resemble, owe,
have (as main verb), which allow manner adverbs but are not passivizable, as in (A) and (5):
(4)(a) John knows Canada very well,
(b) *Canada is known by John.
(5)(a) John was owed some money by his friends,
(b) *His friends owed John some money very well.
In addition, Lakoff also observed that in sentences such
as (6),
(6) 100 soldiers shot two students.
[F-9-25]
(which is parallel to sentence (2)), there is indeed an ambiguity,
but it is not the same ambiguity found in the corresponding passive
(7),
(7) Two students were shot by 100 soldiers.
8^3
[F-9-26]
PASS - 3
(which is parallel to (3)). He argued that while (6) and (7) share
one interpretation, namely (8)
(8) A group of soldiers, who were 100 in number,
shot a total of two students.
[F-9-26a]
that the two sentences have distinct second interpretations, (9) for
(6) and (10) for (7):
(9) 100 soldiers (perhaps out of a larger group)
shot two students apiece, though not the same
two students.
[F-9-26b]
(10) Two particular students (out of all of those
who were shot) were each shot by 100 soldiers
(though not necessarily the same 100).
[F-9-26c]
An additional argument for the Katz-Postal treatment over
the earlier treatment was that it avoided structure-building by transformation. As Chomsky pointed out (1957, pp. 73-74), one wants to
know that be by + AGENT phrase in the passive is a prepositional
phrase. The operation of his passive transformation, however, merely
attached the two constituents by_ and NP^ to the VP in the configuration (11):
(11)
In order to "be a" prepositional phrase, the two constituents would
have to be dominated by a common node, PREP PHRASE, which in turn
is dominated by VP. But if transformations are allowed to build
structure in this manner (particularly since other instances of the
need for such mechanism are rare and rather special cases), it is
difficult to see how to limit such structure-building power. In
order to constrain the grammar as tightly as possible, the structurebuilding mechanism was to be avoided if possible. The Katz and
Postal solution to this problem (independently motivated) was to
derive manner adverbs and a number of other adverb types from prepositional phrases. Note that in a case-grammar framework the
structure-building problem does not arise, as all NP's are introduced as part of an actant structure consisting of a case dominating a preposition and a noun phrase. Consequently, it is possible
for the Passive Rule to move the entire actant structure, simply
replacing an underlying preposition with by_. A third alternative,
suggested by Lakoff (1965) is to have no prepositional phrases in
the base.
Qkk
PASS - 4
A third advantage of the manner adverb formulation over
his earlier analysis was pointed out by Chomsky (1965). In his
earlier treatment, based on transitivity of verbs, he was not able
to include what he called psuedo-passives, such as (12):
(12) The proposal was vehemently argued against.
and therefore had to have a separate transformation to account for
them. Under the manner adverb analysis sentences of this type could
be handled by the regular passive rule by using the presence of the
manner adverb dominating by_ + Passive as the condition required to
determine passivizability rather than the presence of an NP immediately after the verb (a condition not met by the psuedopassives,
since a preposition intervened between V and NP) and so stating the
transformation that it made the first NP after V subject rather than
the NP immediately after the V.
C.
The Two-Sentence Passive
Schachter (UESP, 1967) and Hasegawa (1967) independently
proposed a two-sentence analysis which would provide a deep
structure of the form (14) for the sentence (13):
(13) John was killed by Mary.
(14)
NI
I
John
Mary
killed
John
Where, if the subject of the top S is identical to the object of the
bottom S, then passivization will take place. This analysis allows
the blocking of passive reflexives, such as (15):
(15)*
John was killed by himself.
Given the deep structure (16):
8U5
PASS - 5
(16)
John
John
killed
John
reflexivization (being an obligatory, cyclical rule) will operate
upon the embedded S, yielding (17):
(17)
John
John
killed
himself
The subject of the matrix sentence will then not be identical to the
object of the embedded sentence, so the passive transformation blocks.
At least two questions arise here, however. First, it is
not clear that we want to make it impossible to generate passive reflexives. While only a few people find (15) acceptable, there appear
to be many examples, such as (18) which receive widespread acceptance:
(18) A person who is not respected by himself will
not be respected by others.
Second, the claim that the reflexive and non-reflexive
realizations of the same referent do not meet the identity condition
required for the passive transformation (cf. example fl.7)) is questionable. Ross (at the 1967 UESP Conference) has suggested that there are
cases where the reflexive and non-reflexive realizations of the same
referent do meet identity conditions. For example, on one reading of
(19), there is an obviously deleted wash John.
8U6
PASS - 6
(19) John washed himself before I could.
Assuming, once again, that reflexivization is obligatory and cyclic, it
appears that reflexivization will have taken place in the main clause
before deletion of wash John (whatever the structure of subordinate
clause-sentences such as (19) is), so that wash himself is identical to
wash John in the sense (whatever it is) required for deletion. This
particular example is not quite so forceful as it would at first appear,
since, on the other reading of (19), wash myself has been deleted, so
that whatever is going on in this type of deletion probably is not the
same strong sense of identity characteristic of processes such as definite
pronominalization, where this strong sense of identity seems to be required. An additional parallel argument against using non-identity of
reflexive and non-reflexive forms to block passive reflexives is
Chomsky's argument (1965, pp. 176-182) that only inherent features are
relevant for the determination of identity. Additional discussion of
this issue is to be found in Ross (1967c), pp. 348 ff. and Chomsky
(1968).
This analysis also provides a means of explaining the fact
that while (20) is ambiguous as to who was willing, the corresponding
active sentence (21) is not.
(20) John was willingly sacrificed by the tribe.
(21) The tribe willingly sacrificed John.
The two-sentence passive allows the presence of the verb phrase adverb
willingly in either the matrix or the constituent sentence, whereas
for the corresponding active sentence, there is only one such node and
therefore only one such possible position for the adverb. Notice that
this ambiguity cannot be explained by postulating a complement-type
structure for the two readings of (20), such as (22) and (23):
(22} #The tribe was willing it The tribe sacrificed John #0
(23) 9 John was willing I The tribe sacrificed John //0
since the same ambiguity occurrs with such adverbs as on purpose which
cannot occur in structures such as (22) and (23) in the same way as
willing.
It has been suggested (by Ross at the 1967 UESP Conference)
that the "agentive" interpretation of (20) the reading associated with
(23)) comes not from an ordinary passive, but from a get passive roughly
of the form (24):
8U7
PASS - 7
(24)
ngly
sacrificed John
While such an approach clearly has problems of its own (it is not clear
how one would avoid getting the adverbials on either VP, or even on
both), perhaps a more careful formulation of the proposal would offer
a solution, in which case the ambiguity of the adverbs would no longer
support the Schachter-Hasegawa proposal. Their proposal suffers from
a related difficulty in that there seems to be no non-ad hoc way of
avoiding the generation of such adverbs on both the lower and upper
VP's simultaneously, resulting in a sentence such as (25):
(25) *John was willingly sacrificed by the tribe
willingly.
A related difficulty for this analysis is that not all such
passive sentences with these adverbials are ambiguous, as in (26):
(26) They were willingly allowed to leave.
It is quite unclear how such phenomena would be accounted for in the
two-sentence analysis.
An additional argument against the two-sentence passive,
credited by Ross (at the 1967 UESP Conference) to Chomsky, concerns
idioms such as keep tabs on, take heed of, etc. , as in (27) '•
(27) Careful tabs were kept on the whereabouts of John.
Nouns such as tabs and heed do not occur freely. They normally occur
only as a part of the above idioms, but in the two-sentence passive,
they would have to be generated freely like ordinary nouns or would
have to be limited by some strange constraint to occurring only in the
upper S of a two-sentence passive structure whose lower sentence
included the correct idiom. Neither of these alternatives is very
desirable. It is not clear, however, how much significance should be
attached to arguments based solely on idioms.
8U8
PASS - 8
We consider that a restricted two-sentence passive may
eventually prove most fruitful in explaining these ambiguities, but
we do not make such a proposal here.
III.
THE CASE-GRAMMAR ANALYSIS
Adopting as a basis Fillmore's (1967a) proposals concerning
subject and object placement in active and passive sentences, this
grammar handles passivization as an integral part of the early "Case
Placement Rules." (see CASE). These rules first objectivalize the
proper actant. Passive subject-placement may then optionally move
the objectivalized NP to the beginning of the sentence and mark the
preposition of actant which will be the passive (surface) agent [+by],
If passive subject-placement is not chosen, then active subjectplacement obligatorily occurrs. These same rules operate on NP
genitives in a parallel way insofar as genitives are parallel to
sentences. For a more complete discussion of this subject, see CASE.
May 1969
8U9
850
RULE ORDERING AND LIST OF TRANSFORMATIONAL RULES
The ordering of rules is summarized on the next page, with
rule numbers used to indicate those ordering relations for which
there are arguments presented in the list that follows.
The rules are then presented, usually one per page, with rule,
example, and ordering arguments for each. Where no number appears
to the right of the rule in the ordered list, this means only that
we have no argument about what it must precede, only arguments
about what it cannot precede (i.e. what it must follow), and these
arguments are stated with the relevant preceding rules. The same
convention applies in the ordering arguments: the arguments are
couched in terms of what other rules each rule must precede.
The conjunction schemata have not been included or ordered
in this list; and there are a few obvious minor rules which are
not formulated at all, though referred to in discussion.
851
RULES - 2
RULE ORDERING
Rule Name
Ordering:
Must Precede the Following
1.
GERUND [+/-FACT]
2, 3, 5, 19
2.
PREP SPREAD
7
3.
FACT DEL
*,5
k.
FOR INSERT
18, 22, 62, (5?)
5.
EQUI NP DEL
6
6.
RAIS TO OBJ
7, 12, 1U, 18, 19, 20, 21
7.
OBJ
10, 11, 12, 13, 17
8.
SINGLE-ACTANT-of
10, 11, 12, 13
9.
SOME-ANY
11, 13, 26, U3, U6
10.
RAIS OBJ TO SUBJ
13, 18
11.
RAIS TO SUBJ
13
12.
PASS SUBJ
13
13.
ACT SUBJ
Ik
1U.
REFLEX
22, UO
15.
PARTITIVE POSTPOSE
16, 18, U7
16.
PARTITIVE REDUCE
18, U7
17.
OF-INSERT
18
18.
ACCUSE MARK
19
19.
TO REPLACE AUX
20, 21, 22, 60, 61
20.
SJC DEL
21.
THAT INSERT (NOM)
23
22.
ONE DEL (GENERIC)
Uo
23.
EXTRA (FROM OBJ)
2U
Order Number
852
RULES - 3
2k.
EXTRA (FROM SUBJ)
25.
THAT DEL (NOM)
26.
SOME-ANY (REL)
UU
27.
NOUN FEAT TO ART
28, 1*0
28.
WH ATTACH
29
29.
WH FRONT
31, 31*. 38
30.
CLAUSE POSITION
38, 39
31.
REL -» THAT
32, 31*
32.
POSSESSIVE FORMATION
33
33.
GENITIVE POSTPOSING
3U.
THAT DEL (REL)
35.
ELSE
36
36.
ART ATTACH
37. 39
37.
ATTACH BLOCK
38.
REL REDUCE
39
39.
ADJ PREPOSE
1*0
1*0.
NOUN REDUCE TO ONE (SELF)
1*1.
NOUN -> 0
1*2.
PROPER NOUN THE DEL
1*3.
S INITIAL ADV PLACE
1*1*. 55
ui*.
NEG ATTRACT
U5, 1*6
1*5.
INDEF BEFORE QUANT DEL 1*6, 1*7
1*6.
ANY - NO
1*7.
QUANT MOVE
25
1*1, 1*2
VT, 1*8
52
853
RULES - k
w.
ALL THE
k9.
ALL THREE
50.
BE INSERT
51, 55, 57,
51.
PREVERB PART PLACE
55
52.
PREVERB ADV
53.
CONJ SPREAD
55, 56
5*.
WH COPY
55, 56
55.
AUX ATTRACT
56, 58, 59
56.
WH DEL
57.
AFFIX SHIFT
58, 59
58.
DO SUPPORT
59
59.
NEG CONTRACTION
60.
TO DEL
61.
TO BE DEL
62.
PREP DEL
^9
61
85U
RULE - 5
GERUND [+FACT]
Order No.
1
Rule:
S.I.
X
[the fact of
MT>
NP
i
S.C. (l)
(2)
(3)
(U)
NP
[
S
[# NP
AUX
[TNS (M) (PERF) (PROG)] X
1
Chomsky adjoin POSS as last right daughter of 2.
If 3 • +PAST and 5 = 0, attach PERF as left sister of 6.
Replace 3 and k by -ing.
[-EQUI NP DEL] + [+EQUI NP DEL].
(PROG)
Ordering Arguments:
(1)
Must precede FACT DEL, because the fact of provides the
environment for the rule.
(2)
Must precede EQUI NP DEL to account for
(a)
(3)
I regretted leaving.
Procedes TO REPLACE AUX, because otherwise we would derive
(b) *I regretted to leave.
855
RULE - 6
Order No.
GERUND [-FACT]
1
Rule:
S.I.
\mNPfX
x ) v
C[+GERp
12
S.D.
(1)
COND:
Obligatory
Example:
S
[XVAUX
iTIY [TNS]
i
X
3
Replace U by ing
John avoided leaving.
Tree:
S^
MOD
PROP
V
[+GER]
NEUT
\
NP
NP
MOD
PROP
AUX
/
K\
TE
(B
0
Ordering Arguments:
(1)
Precedes EQUI NP DEL to account for
(a)
(2)
John avoided leaving.
Precedes TO REPLACE AUX to avoid deriving
(b) *John avoided to leave.
(3)
Precedes PREP-SPREAD in order to guarantee that the distinction
between "real" prepositions and "case-marker" prepositions (see CASE
PLACE) can be maintained, since real prepositions demand gerundivization but prepositions introduced transformationally do not.
856
RULE - 7
PREP-SPREAD
Order No.
2
Rule;
S.I.
X
/vU+Ci PREP « ]
X
2
h
W'—'
1
S.C.
Example:
3
*
[PREP
5
Attach 3 to 5, erase 2-3
John puzzled over the problem.
Tree:
MOD
I
AGT
AUX
I
PAST
puzzle
[+NEUT PREP^ove?
/
PREP
NP
John
the problem
Ordering Arguments:
(l)
Must precede OBJ, since that rule moves a marked preposition down
under the verb.
857
RULE - 8
FACT DEL
Order No. 3
Rule:
S.I.
i
NP
X
[
i
the X fact of
i
12
S.D.
NP
i
i
[S] X
I
3
Erase 2
COND:
(1) Optional
Example:
John regretted (the fact of) Bill's having left.
Tree:
NP
D
NOM
/the\^ N
PREP
\^ fact
of)
NP
S
0
Ordering Arguments:
(1) Precedes EQUI NP DEL to account for
(a)
(b)
(c)
(d)
I regretted the fact of my leaving/having left.
I regretted the fact of leaving/having left.
I regretted leaving/having left.
*I regretted my leaving/having left.
particularly the ungrammaticality of (l.d). The only way we can account for
these examples is for FACT DEL to precede EQUI NP DEL, because otherwise
FACT DEL would be obligatory just in case EQUI NP DEL has applied.
(2) Must precede FOR INSERT because of [+FACT] [+EMOT] words like "tragedy",
as in:
(e)
It would be a tragedy for him to leave.
858
RULE - 9
Order No. k
FOR INSERT
Rule:
1.1.
X
J+EMOT]
(gg\
(PREP)
j
[#
|
2
NP
X
U
5
1
3
for
as left sister of k.
[+PREP]
S.C.
Attach
COND:
Obligatory
Example:
S
"P[
It is desirable for him to do it.
Tree:
S
MOD
PROP
V
[+EMOT]
C
/ \
PREP NP
I
^S
for \
[+PREP]
NP
MOD
PROP
Ordering Arguments:
(1) We have no strong arguments that this rule must precede EQUI NP DEL,
but only if it does will we have the same derivation for
(a)
(b)
(2)
Must precede PREP DEL to account for
(c)
(d)
(3)
John prefers (for Mary) to go.
It scared him (for Mary) to Jump off the roof.
John hoped tbt for Mary to go.
It scared him Hi to Jump off the roof.
Must precede ACCUSE MARK to get accusative in
(e)
For him to come early surprised me.
(U) Must precede ONE DEL, because the latter rule depends on for or
POSS in its environment.
859
RULE - 10
Order Ho.
EQUI NP DEL
5
Rule:
S.I.
NP
X
S
[
[NP X]
DAT
[X NP] X
12
S.C.
Erase 2
COND:
(1)
(2)
(3)
Example:
AGT
[X NP]
h
3
2=3
If 2 j 3, or if 3 = 0, then 2 = U
Optional with "transparent" nouns , if the noun was not
deleted (cf. John resented the fact of his leaving)
John hoped to go. [AGT identity]
Tree:
MOD
hope
PREP
NP
PREP
I
S
NP\ MOD
John
PROP
V
I
Ordering Arguments:
(1) Must precede RAIS OBJ to prevent REFLEX from applying to an identical
subject raised to object Just in case a verb like expect has both EQUI NP
DEL and RAIS OBJ, to avoid ambiguous derivation of:
(a)
We believe that
(b)
I expect myself to go.
(a)
is derived from:
I expect [I SJC go] of myself,
with EQUI NP DEL from DAT, which undergoes REFLEX.
Note that:
(c)
I expect him to go.
is ambiguous between:
(d)
I expect [he SJC go]
860
RULE - 11
(e)
(2)
I expect [he SJC go] of him
Must precede SOME-ANY to account for:
(f)John couldn't persuade anyone to come.
That is, if SOME-ANY has applied, there is no longer the required
identity between
•John couldn't persuade anyone [someone TNS come].
Transparent nouns are here opposed to picture nouns.
861
RULE - 12
Order No. 6
RAIS TO OBJ
Rule:
s
[# MOD
V
[ [t
X
S.C.
(1)
(2)
Attach 5 as right sister of 2
Erase 3 and original 5
COND:
(l)
2 contains the feature [+RAIS OBJ], and does not contain
the features [-STAT REDUCT] or [-FUT REDUCT].
Obligatory
(2)
pROp[
W S
S.I.
PREP
NP
X
Mary believes herself to be attractive.
S
[-PAST]
PROP
/
V
I
/NP\
\ )
NEUT
y\
believe XMary/ PREP
f+RAIS
+RAIS OBJ
[+STAT REDUCTJ
DAT
PREP
NP
be attractive
Ordering Arguments:
(1) This rule is governed, hence must precede the general rule of OBJ
PLACE, cf.
(a)
(b)
I believed B. to be a fool (applied).
I believed that B. is a fool (not applied).
[where object of "believe" is underlined]
(2) Must precede REFLEX, since the latter rule works in simplexes , and an
identical subj. raised to obj. must undergo REFLEX; e.g.
(c)
He considered/believed/etc. himself to be free.
(3) Must precede THAT INSERT because the latter rule depends on the
presence of a whole clause (real S). I.e., all sentences deforming rules
must precede THAT INSERT.
(U)
Must precede ACCUSE MARK to get accusative in, e.g.,
(d)
F. expected him to go.
862
RULE - 13
(5)
Must precede PASS SUBJ PLACE to account for
(e)
(6)
He was believed to be a fool.
Must precede TO REPLACE AUX to account for:
(f)
(g)
I expect B. to leave (applied).
I expect that B. will leave (not applied).
(7) Must precede SJC DEL, because the latter rule is triggered (among
others) by absence of subject.
863
RULE - Ik
M-OBJ (a)
Order No. 7
Rule;
S.I.
X
[Nj
+Cj -» OBJ]
_ [PREP NP]
C.
l
3
S.C.
(l)
(2)
X
k
_ [PREP NP]
v.
J
5
6
7
X
8
Attach 7 as right sister of 2
Delete 6-7
COND:
2 through 7 are a constituent
Example:
John aimed the gun at Mary.
Tree:
aim
[+INS -} OBJ]
at Mary
PREP
[+INS]
NP
[+INS]
the gun
PROP
aim
at Mary
the gun
Ordering Arguments:
(1)
See U-OBJ (h).
861*
(by) John
RULE - 15
M-OBJ (b)
Order No. 7
Rule:
S.I.
X
»••[PREP NP] X
NEUT
«
[+C
J
COND:
Example:
[PREP
NP] X
L
-^ OBJ, NEUT]
[+Prep]
12
S.C. (1)
(2)
(3)
Cn
3
k
5
6
7
8
Attach 7 as right sister of 2;
Attach [+Prep] (from 2) to 3;
Delete 6-7.
2 through 7 are a constituent
He loaded the wagon with hay.
Tree:
load
on the wagon
[+LOC -» OBJ, NEUT ]
[+with]
PROP
AGT
load
Ordering Arguments
(1)
See U-OBJ (b)
865
RULE - 16
Order No. 7
U-OBJ (a)
Rule:
S.I.
X jj
1
S.C.
COND:
c
[PREP
NP]
[+Prep]
2
(l)
(2)
(3)
(1)
(2)
Example;
h
3
X
C,
X
5
6
7
Chomsky-adjoin 3 as right sister of 2;
Attach k as right sister of 2;
Erase 3-U.
2 through 6 are a constituent
If 5 is null and 6 = LOC, the rule does not apply.
He insisted on an answer.
PROP
Tree:
AGT
insist
PREP
[+pnj
•22.1
V
NP
^^
an answer
Ordering Arguments:
(1)
See U-OBJ (b).
866
PREP
[+on]
an answer
he
RULE - 17
Order No. 7
*
X
(I ]c tFREP NP] X C, X
1
S.C.
COND:
Example:
2
3
5 6
7
Attach k as right sister of 2;
Erase 3-b.
(1)
(2)
(l)
(2)
2 through 6 are a constituent
If 5 is null and 6 is LOC, the rule does not apply.
He lost his mind.
Tree:
PROP
V
NEUT
lose
U
PREP
DAT
"^NP
A
he
lose
his mind
he
his mind
Ordering Arguments:
(1) The four varieties of objectivalization are disjunctive with
respect to each other: if one applies, the other three are excluded,
simply because the four structure indices are mutually exclusive: two
different features govern the two M-OBJ rules, and the two U-OBJ rules
are distinguished by the presence vs. absence of a marked preposition
on the first actant that follows the head. In respect to ordering
arguments, the four may therefore be treated as a single rule OBJ.
(2) OBJ must precede all SUBJ rules because it provides a condition
for the first of the SUBJ rules, namely the removal of the case node
over the NP that is permitted to be raised to subject, or to become
passive subject.
867
RULE - 18
SINGLE-ACTANT-of
Order No. 8
Rule:
J.I. X |J}
S.I.
12
S.C.
k
3
5
Attach [+of] to 3 and delete features other than [+PREP] on 3.
COND:
Example;
2-k is a constituent
The shooting of the hunters...
NOM
NOM
Tree:
ACT
N
i
shooting
i
PREP
[+by]
shooting
PREP
[•of]
the hunters
the hunters
Ordering Arguments:
(l) The rule must precede all SUBJ rules, since the SUBJ rules can
move an actant to the left of the head item and leave behind a single
actant which could then—but should not—be affected by this rule.
868
RULE - 20
Order Ho. 10
RAIS OBJ TO SUBJ
Rule:
S.I.
S
X
[#
MOD
12
pRQp[
3
X
U
•*[
S[X
V
5
NP
X
6
S.C.
(1)
(2)
Attach 6 as right sister of 2
Erase original 6
COND:
(1)
(2)
U contains the feature [+RAIS OBJ TO SUBJ]
Optional
Example:
John is difficult for Mary to please.
Tree:
difficult
[+RAIS OBJ TO SUBJ]
NP
I
Mary
MOD
PROP
I
TNS
Ordering Arguments:
(1) Precedes ACT SUBJ PLACE, because the latter rule is oblig., and RAIS
OBJ TO SUBJ is optional. Verbs like easy, difficult, etc. allow both, as do
(a)
(b)
(2)
The book is easy for John to read.
For John to read the book is easy.
Precedes ACCUSE MARK, because we want to derive
(c)
He is difficult for Mary to please,
and not
(d) *Him is difficult for Mary to please.
869
RULE - 19
Order No. 9
SOME-ANY
Rule:
S.I.
X
[+AFFECT]
X
1
2
3
f-SPEC I
x
L-INDET]
L.-1JJ
U
5
s.c.
Change [-INDET] to [+INDET] in U.
COND:
(1)
(2)
(3)
(U)
Example:
2 commands h
If 2 = [+N]; or [+V]; or [+PREP], then U does not command 2
Complex NP constraint holds
Obligatory
John dislikes anyone meddling in his affairs.
Tree:
MOD
NEG
AUX
V
/
/
/
dislike
[+AFFECT]
/someone meddles in John's affairs
Ordering Arguments:
(1) Must precede SUBJ PLACE to define context for NEG; i.e., to get
no-one left from NEG leave someone, since SOME-ANY applies to the right.
(2) Must precede SOME-ANY (REL) because a converted any can then
trigger SOME-ANY REL below it.
(3) Must precede S INITIAL ADV PLACE, since the latter rule moves ADV
to the left of NEG, and SOME-ANY only works to the right of NEG.
(U) Must precede ANY-NO, because SOME-ANY provides the environment for
the latter rule.
870
RULE - 21
Order No. 11
RAIS TO SUBJ
Rule:
S.I.
X
S[
'
i
1
M0D
u
2
PROP[
X
i
NPr
s
I
U
[# NP
X
*
5
6
7
s.c.
(1)
(2)
Attach 6 as right sister of 2
Erase original 6
COND:
(1)
(2)
U contains the feature [+RAIS SUBJ]
Optional
Example:
John is likely to have arrived.
Tree:
NEUT
likely
PREP
[+RAIS SUBJ]
NP
I
S
/NP\ MOD
/ I ] i\
\John/ TNS PERF
\V
arrive
Ordering Arguments;
(l) Precedes ACT SUBJ PLACE, because the latter rule is oblig., and verbs
like appear, likely» etc. take either
(a)
(b)
That he left is likely.
He is likely to have left.
871
RULE - 22
PASS-SUBJ-BY-PLACE
Order No. 12
Rule;
S.I.
S.C.
COND:
Example:
X \y\
NP
X
PREP
NP
X
1
3
^
5
6
7
2
[+Prep,
(1)
(2)
(3)
byj replaces features on 5
2-6 is a constituent
2 has the feature [+PASS]
If 2 • N, then 5-6 immediately dominated by AGT or INS
The destruction of the city by the enemy...
Tree:
NOM
destruction
[+PASS]
the enemy
[+Prep]
[+by_]
872
RULE - 23
PASS-SUBJ
Order No. 12
Rule:
MOD
DET
l[-Dem]
S.I.
X
S.C.
(l)
(2)
(3)
(k)
COND:
Example:
(1)
(2)
(3)
V
N
NP
X
PREP
[+bv_]
NP
Attach U as left sister of 2;
If 3 • I, attach the feature [+Genitive] to U;
If 3 • V, attach be + en as right daughters of 2;
Erase original k.
3-6 is a constituent
If 3 = N, the rule is optional
If 3 - V, the rule is obligatory
The saint's canonization (by someone)...
Tree:
canonization
NOM
NOM
the saint
by someone
the saint
[+GENITIVE]
canonization
by
someone
Ordering Arguments:
(l) Must precede ACT-SUBJ because it is a governed rule whose application removes a set of possible candidates from the domain of ACT-SUBJ.
873
RULE - 2k
M-ACT-SUBJ
Order No. 13
Rule:
8.1.
S.C.
COND:
Example;
X (DET
I)
X
1
3
U
\MOD
2
(1)
(2)
(1)
(2)
r
4
[PREP
NP]
6
5
8
If 3 is V, attach 6 as left sister of 2; delete 5-6.
If 3 is N, attach 6 to 2; add [+GENITIVE] to 6; delete 5-6.
3-7 is a constituent;
3 has a feature of the form [+C. -> SUBJ]
The. pool filled with water.
Tree:
MOD
*
V
I
fill
[+LOC -* SUBJ]
[in]
with water
87U
RULE - 25
Order No. 13
M0D
PROp£V
DET
[N
NOM
[PREP
NP]
X ]
IPROP I
| NOM I
8
S.C.
COND:
Example;
(1)
(2)
(1)
(2)
(3)
If 3 is V, attach 7 as left sister of 2; delete 5-6-7.
If 3 is N, attach 7 to 2 with the feature [+Genitive]
added to it; delete 5-6-7.
Obligatory if 3 = V, or if 3 = N and 5 - DAT
8 • LOC, or is null
5 t LOC
The package arrived at the airport,
Tree:
Ordering Arguments:
(l) Since REFLEX (ignoring crossover conditions) is most easily stated
as working on LEFT-RIGHT surface order, and our underlying structure
has active subjects generally right-most, this rule preceded REFLEX.
875
RULE - 26
Order No. lU
REFLEX
Rule:
S.I.
X up I pgip I ART X J X N X J X «p [ ppyp
1
2
3
S.C.
Attach
COND:
(1)
(2)
(3)
(10
(5)
(6)
Example:
U
[+AXCH]
5 6 7
ART
X ] X N X ] X
+DEF
-DEM
-GENITIVEJ
89
10
11 12 13 lfc 15
to 13 and 10
2 is immediately dominated by lowest S or HP that dominates 9
6 = head of its NP
13 = head of its NP
5,6,7 « 12,13a1*
If 3 • [+DEF,-GENERIC], then 3 = 10 and U = 11
If 3 i 10, then 11 = 0, and 10 = [-1,-11]
If 3 • [+1] or [+11] optional, otherwise obligatory
The boy saw himself.
Tree:
Ordering Arguments:
(1) Must precede NOUN REDUCT TO ONE, so that the latter can be obligatory
for items marked [+REFLEX].
(2)
Must precede ONE DEL to get
(a)
(3)
Patting oneself on the back is ungracious.
Must precede YOU DEL to get
(b)
Help yourself!
(U)
Must precede ART DEL with proper nouns,
(5)
Must precede PR0N CONJ to get
(c)
John and Mary shot themselves.
876
RULE - 27
Order No. 15
PARTITIVE POSTPOSE
Rule:
S.I.
X
D[
X
PARTl
X
Attach 3 as right sister of h
Erase original 3
S.C.
(1)
(2)
COND:
Obligatory
Example:
NOM[Nl
Three (boys) of the boys left.
Tree:
Ordering Arguments:
(1)
Must precede PARTITIVE REDUCE because otherwise we get
(a)
#
three of the ones boys
from the above.
(2)
Must precede QUANT MOVE because otherwise we derive
(b)
each of them
from each of the boys, instead of the correct each of the boys or the
boys each.
(3)
Must precede ACCUSE MARK to derive
(c)
each of them
rather than *each of they.
(h) As far as we know, this rule is not ordered with respect to any of
the preceding rules. That is, this rule does not seem to have to follow
any rule.
877
RULE - 28
Order No. l6
PARTITIVE REDUCE
Rule:
S.I.
X „p[X
N
X [ of
2
Np[X
[+PART]
X] ] X ] X
5
r
Ec PL
6 PL
£ ACCUSE
.r REFLEX)
/* ACCUSE
.^REFLEX
_>
i_
N
S.C.
(1)
(2)
Attach [+PRO] to 3
Delete all features of 2 not specified in 3
COND:
(1)
(2)
2=5 (except for NUMBER, CASE, REFLEX)
Obligatory
Example:
John met many (ones) of the boys.
Tree:
NP
PROP
NP
NOM
NP.
/ \
PREP
NP
/\
D
N
John
met
many
(boys)
of
I
the
I
boys
Ordering Arguments:
(1)
Must precede QUANT MOVE because otherwise we get
(a)
*boy of them each
instead of the boys each.
(2)
Must precede ACCUSE MARK for the same reason as PARTITIVE POSTPOSE
must precede the former rule.
IT^-S
This
rule slightly changed from earlier version.
878
RULE - 29
OF-INSERT
Order No. 17
Rule:
S.I.
S.C.
COND:
Example:
X
N
NP
X
1
2
3
1+
Attach
PREP
[+of ] as left sister of 3
2 and 3 are Immediately dominated by NOM
The amazement of the child at the performance...
Tree:
amazement
the performance
amazement
the child
PREP
1
the performance
Ordering Arguments:
(1) The rule must follow OBJ, since its distinctive environment is
established by that rule.
(2) It must precede ACCUSE-MARK because the inserted preposition of
provides a condition for applying the rule.
879
RULE - 30
Order No. 18
ACCUSE MARK
Rule:
S.I.
(
PREP;
12
„„[ ART
X
3
5
k
S.C.
Attach [+ACCUSE] as feature to h
COND:
(1)
2 and 3 must be a constituent
(2)
Obligatory
Example:
She gave gj» the apple
? .
(_the apple to him)
Tree:
Ordering Arguments:
(l) Must precede TO REPLACE AUX because the accusative provides one of
the environments for the latter rule; i.e. the subject of the nominal is
marked [+ACCUSE], thus permitting TO REPLACE AUX to take place.
•^-This rule slightly changed from earlier version.
880
RULE - 31
Order No. 19
TO REPLACE AUX
Rule:
S.I.
X
NP
[ S[ (for HP)
Q^fl (PERF) (PROG)
X
i
1
2
S.C.
(1)
(2)
COND:
Obligatory
Example:
3
U
Replace 2 by to
If 2 = [+PASTTT and 3 does not contain PERF, then attach
PERF as right sister of 2
John intends to win.
Tree:
John
TE
I
[-PAST]
intend
Ordering Arguments:
(1) Must precede THAT INSERT (NOM) because THAT insertion depends on
presence of AUX.
(2)
Must precede SJC DEL because there must be an AUX to replace.
(3)
Must precede ONE DEL so that the derivation of e.g.,
(a)
(b)
NP is easy for NP[+ACCUSE] to please.
NP is easy
to please.
is the same. That is, so that TO REPLACE AUX should be triggered by the
same condition in both examples above, but if ONE DEL precedes TO REPLACE
AUX, then in example
(a) the rule would be triggered by the presence of
the for (the feature [+ACCUSE]), while in example (3.b) the rule would be
triggered by the absence of the subject.
881
RULE - 32
Order No. 20
SJC DEL
Rule:
S.I.
X
1
SJC
2
X
3
S.C.
Erase 2
COND:
Obligatory
Example;
John demands that Mary be dismissed.
Tree:
PROP
John
7
MOD
PROP
be dismissed
Ordering Arguments:
(l) We have no strong arguments why this rule must precede THAT INSERT
(NOM). That is, this is simply the earliest place the rule can have in
the ordering.
882
RULE - 33
Order No. 21
THAT INSERT (NOM)
Rule:
S.I.
X •*[
S
[
#
J
12
NP
Ainr[TNS X
AUX
»-
3
S.C.
Insert that as right sister of 2
COND:
Obligatory
Example:
That he came at all astonished us.
Tree:
Ordering Arguments:
(l) Precedes EXTRA, because THAT clauses can be extraposed, while
GERUNDIVES cannot.
883
RULE - 3*»
Order No. 22
ONE DEL (GENERIC)
Rule:
S.I.
NPr: SD[#
for NP [one ]
AUX [to X
12
S.C.
Erase 2
COND:
Optional
Example:
3
To collect butterflies is amusing.
Tree:
collect butterflies
Ordering Arguments:
(1) Must precede NOUN REDUCT TO ONE because the abstract one can be
pronominal!zed to him, but only when the antecedent has not been deleted;
cf.,
C»hinf)
(a) laughing at people and expecting them to like 1 onej ...
(b) One shouldn't laught at people and expect them to like him.
88U
RULE - 35
Order No. 23
EXTRA (from OBJECT)
Rule:
PR0P
S.I.
X
S.C.
(1)
(2)
Attach U as last daughter of 2
Replace U by it
[-PRO]
COND:
(1)
Obligatory
(2)
U does not dominate *inr[ing X]
Example:
[
v
[+EXTRA]
^[S]
X
John likes it that Mary is faithful.
Tree:
Ordering Arguments:
(l)
Must precede EXTRA (from SUBJECT), because of e.g.,
(a)
It clinches it that she is a fool that she put the
eggs in the bottom of the basket.
Sentences like this one have the extraposed object inside the extraposed
subject. [Note that the example is not acceptable to all of our informants,
For dialects not accepting the sentence cited above, EXTRA (from OBJECT)
and EXTRA (from SUBJECT) appear to be disjunctively ordered.]
885
RULE - 36
Order No. 2k
EXTRA (from SUBJECT)
Rule:
S.I.
s
X
[ ^[S]
12
3
MOD
h
PROP]
5
X
6
S.C.
(1)
(2)
Attach 3 as right sister of 5
Replace 3 by it
[-PRO]
COND:
(l)
(2)
Obligatory if 5 dominates V[+EXTRA], optional otherwise
3 does not dominate AUxTiiig X]
Example:
It appears that John is a fool.
Tree:
Ordering Arguments
(l) Must precede THAT DEL (NOM) , since all
(THAT clauses) disallow THAT DEL (NOM).
886
sentences in subject position
RULE - 37
0rder No
THAT DEL (NOM)
"
Rule:
S.I.
X
PR0P
[
V
NP[ Sf
that
x
[-fact]
i
i
1
S.C.
Erase 2
COND:
Optional
Example:
2
3
Fred thought he saw John.
Tree:
NP
MOD
Fred
TNS
[+PAST]
PROP
V
NP
I
I
think
[-FACT]
S^
f^haft he sav John
/
Ordering Arguments:
(l)
As far as we know, this could be the last rule.
887
25
RULE - 38
Order No. 26
SOME-ANY (REL)
Rule:
[-SPEC ~|
X ,jp[
S.C.
Change [-INDET] to [+INDET] in 2
COND:
(1)
(2)
Example:
D[
f+AFFECT]]
[+AFFECT]
[[+INDET]JX]
S.I.
x
NOM[NOM gX
(-INDETJ
X]]] X
1 = lowest S dominating 2
Obligatory
Anybody who ever swears at me better watch his step.
Tree:
PROP
D
I
better watch his step
any
[+INDET]
PROP
NOM
I
some
pSPEC "I
N
UlNDETJ
body
ever swears at me
t
[+INDET]
Ordering Arguments:
(1)
Must precede NEG ATTRACT because of
(a)
No-one who ever swears at me is likely to leave
this room.
888
RULE - 39
Order No. 27
NOUN FEATURE TO ART
Rule:
S.I.
X
[ ART
xrol
NP
X
N
X ]
X
l|
"c^COUNT
(PfflJMAN)
(JMASC)
(A PLURAL)
S.C.
Attach 5 as features to 2
COND:
(1)
(2)
Example:
U - head of its NP
Obligatory
When I saw the man, he was eating.
Tree:
The whole set of features under ART will become he_ in the second lexical
look-up.
Ordering Arguments:
(1) This rule does not seem to be ordered with respect to the preceding
rules; i.e., this is simply the latest place at which it fits.
(2)
This rule must precede WH ATTACH, because WH ATTACH deletes head nouns.
(3) This rule must precede NOUN REDUCT TO ONE because the latter rule wipes
out the environment on which this rule operates.
889
RULE - UO
Order No. 28
WH ATTACH ("NOM-S" Analysis)
Rule:
S.I.
S.C.
COND:
X
NOM
g[
#
X
3
h
[ X
ART + NOM]
X
# ]
8
9
X
10
(2)
•WH
+REL
L+PROJ
Replace f^pj in 6 by [+DEF]
(3)
(1»)
Erase 3 and 7
Replace 9 by "Half-Fall"
(1)
(2)
2-7
r+qpFr-T
6 dominates _^C
(3)
If there is a [+WH] anywhere within the S immediately
dominating 7 which is also [-REL], the S.I. is not met
Obligatory
(1)
Replace [-WH] in 6 by:
CwH J
(U)
The picture which I took is out of focus.
\
Ordering Arguments:
(l)
This rule must precede WH FRONT for obvious reasons,
890
RULE - kl
Order No. 28
WH ATTACH ("NP-s" Analysis)
Rule:
S.I.
X
NP g[ # X
12
S.C.
COND:
Example:
Np[
3 U 5
D N]
X #] X
67
89 10
+WH
+REL
+PR0.
(1)
Replace [-WH] in 6 by
(2)
(3)
(U)
If 6 dominates [-DEF], replace it by [+DEF]
Erase 3 and 7
Replace 9 by "Half-Fall"
(1)
(2)
(3)
(k)
2=5, and h f X + that
Obligatory
6 dominates [-WH]
If there is a [+WH] anywhere within the S immediately
dominating 7, which is also [-REL], the S.I. is not met
The picture which I took is out of focus.
Tree:
Ordering Arguments:
(l)
This rule must precede WH FRONT for obvious reasons.
891
RULE - U2
WH ATTACH
Order No. 28
("ART-S" Analysis)
Rule:
8.1.
S.C.
X
D[X
1
2
(1)
(2)
(3)
(U)
COND:
(1)
(2)
(3)
(U)
Example:
ART g[# N
3
Np[ART
U 5 6 7
8
N] X #]
N
X
9 10 11 12 13 lU
+WH
+REL
+PR0
If 8 dominates [-DEF], replace it by [+DEF]
Erase 5 and 9
Replace 11 by "Half-Fall"
Replace [-WH] in 8 by
3 + 13 = 8 + 9
If there is a [+WH] anywhere within •+, which is also [-REL],
the S.I. for the rule is not met.
8 dominates [-WH]
Obligatory
The picture which I took is out of focus.
Tree:
Ordering Arguments:
(l)
This rule must precede WH FRONT for obvious reasons.
892
RULE - 1»3
WH FRONT
Order No. 29
Rule:
S.I.
X
s[
12
S.C.
COND:
Example:
(1)
X jpl X
X
jjpt ART ]
X
k
5
6
8
3
(2)
Chomsky adjoin 6 as left daughter of 2, OR Chomsky adjoin
5 + 6 as left daughters of 2 (in accord with Pied Piping
convention)
Erase original (5), 6
(1)
7 dominates
(2)
(3)
General constraints on movements hold
Obligatory
+WH
+REL
+PRO
+DEF
The picture which I took is out of focus.
Tree:
the
Ordering Arguments:
(1) This rule precedes THAT DEL because deletion of that is possible
only if the that precedes another NP, a condition to be found only after
WH FRONT.
(2)
This rule is not ordered with respect to REL^THAT, but should precede it,
(3)
This rule precedes REL REDUCT for obvious reasons.
893
RULE - kk
CLAUSE POSITIONING
("ART-S" Analysis only)
Order No. 30
Rule:
S.I.
K
jjp[ X
D
1
[ S ]
2
X ]
X
3
S.C.
(l) Attach 2 as last daughter of 1
(2) Erase original 2
COND:
(l)
Example:
(2)
+WH
+REL
+PR0
2 does not dominate an S which dominates
(3)
(1+)
1 is the highest NP dominating 2
Obligatory
2 dominates
+WH
+REL
+PR0
John bought the car which I want.
Tree:
Ordering Arguments:
(l) This rule is not strictly ordered with respect to any rules in this
block except REL REDUCT and ADJ PREPOSE , which it must precede.
89^
RULE - U5
REL=»THAT
Order No. 31
Rule:
S.I.
X
ART X
+WH~
+REL
+PRO
+F
12
3
S.C.
Attach [+THAT] as feature to 2
COND:
(1)
(2)
Example:
1 t X + PREP
Optional
The picture that I took is out of focus.
Tree: (NOM-S)
[+THAT]
Ordering Arguments:
(1)
This rule precedes THAT DEL for obvious reasons.
(2)
This rule is not ordered with respect to REL REDUCT and ADJ PREPOSE
895
RULE - U6
Order No. 32
POSSESSIVE FORMATION
Rule:
S.I.
[
m
NP
1
S.C.
COND:
Example:
[X ART X N X] [NP BE NP
] ]
S
[+Dative]
23U5678
(1)
(2)
Attach 9 to 3
Erase 7, 8, 9
(3)
Add [+Genitive] to 3
(l)
(2)
9
3 does not dominate NP
7 dominates [+THAT]
John's house...
Tree:
.
Hn 1
Add (J+Genitive]
Ordering Arguments:
(1)
Must follow REL -» that, for condition (2).
(2)
Must precede Genitive Postposing to obtain:
896
that house of John's.
RULE - U7
GENITIVE POSTPOSING
Order No. 33
Rule:
B.I.
[
AKr
1
S.C.
COND:
(1)
(2)
(1)
(2)
Example:
NP
]
[+Genitive]
X
N
X
2
3
k
5
Attach 2 as right sister of k
Delete 2
1/ [+Def]
(Note:
This is understood strictly: if 1
dominates anything in addition to
[+Def] the rule does not apply.)
3 does not directly dominate NP
That house of John's..
Tree:
Ordering Arguments:
(l)
Follows Possessive Formation to obtain:
897
that book of John's.
RULE - U8
THAT DEL (REL)
Order No. 3U
Rule:
S.I.
X
ART
NP
+WH ]
X
+REL|
1
2
S.C.
Erase 2
COND:
Optional
Example;
3
1*
The picture I took is out of focus.
Tree:
Ordering Arguments:
(1)
This rule is not ordered with respect to REL REDUCT or ADJ PREPOSE.
898
RULE - U9
ELSE
Order No. 35
Rule:
S.I.
X
[+ATTACH]
12
[+ATTACH]
3
U
X
5
Attach else as right sister of h
Erase 3
S.C.
(l)
(2)
COND:
Obligatory
Example;
other
"somebody else"(from "*some other body" )
Tree:
NP
some
other
[+ATTACH]
NP
NOM
D
N
some
body
)dy
[+ATTACH]
NOM
N
else
body
Ordering Arguments:
(1) This rule must precede ART ATTACH because other (which is the source
for else) stands between ART and NOUN; e.g.,
(a)
some other person ^ someone else
899
RULE - 50
Order No. 36
ART ATTACH
Rule:
S.I.
X
D
[ [+ATTACH] ]
12
S.C.
(1)
(2)
(3)
[ [+ATTACH] ]
X
h
3
Insert "4" as right and left sister of 3.
Insert 2 as left sister of 3.
Erase original 2.
(i.e., l-0-§ + 2+3 + §-U)
COHD:
Example:
Obligatory
where "§" is an ad hoc symbol for word-formation
"everything", "anyone"
Tree:
NP
NOM
I
I
thing
N
every
[•ATTACH]
+N
•ATTACH
-HUMAN
Ordering Arguments:
(1) This rule must precede ATTACH BLOCK, since nouns marked [+ATTACH]
cannot stand alone, yet they are not constrained in the P.S. rules to
co-occur only with ART's marked [+ATTACH].
(2)
This rule precedes ADJ PREPOSE
(a)
somebody nice
i.e., the fact that attached ART + NOUN structures have the ADJ following
them.
900
RULE - 51
Order No. 37
ATTACHMENT BLOCK
Rule:
S.I.
X
D
NQM[
N
[+ATTACH] ]
X
l
I
1
S.C.
Erase 1
COND:
Obligatory
Example:
*each body-
Tree:
NP
NOM
each
[-ATTACH]
N
body
[+ATTACH]
Ordering Arguments;
(l) This rule must be last in this block of rules, because it blocks
ungrammatical strings like
(a) • three bodies
(b)* each body
which would otherwise be formed on the analogy of somebody, .anybody. etc.
901
RULE - 52
Order No. 38
REL REDUCE A
Rule:
S.I.
X
NOM,
[-PRO]
i
cLART
s
1
i
12
S.C.
(1)
(2)
COND:
Optional
Example:
TNS
[+REL]
be
X ]
—
X
1
3
U
5
Erase 3
Attach U as last daughter of 1
The boy here is tall.
Tree:
PRUNED
Ordering Arguments:
(l) This rule must precede ADJ PREPOSE to derive pre-nominal adjectives
like
(a)
the tall girl
without needing two rules which reduce relative clauses.
902
RULE - 53
Order No. 38
REL REDUCE B
Rule:
S.I.
X
NOM
[-PRO]
I
TNS
(NEG)
V
X ]
1
23U56
X
I
78
(1) Erase 3
(2)
(3)
Attach ing. to 5, erasing [+/-PAST], OR
If 5 dominates [+PAST], attach -ing have En as daughter
of 5, and erase [+PAST]
Attach U-7 as last daughters of 1
(U)
COND:
ART (ADV)
[+REL]
1
1
S.C.
[
S
Optional
Example:
People owning large houses pay large taxes.
Tree:
pay large taxes
[+DEFl
+GENJ
N
>^
people
Ordering Arguments:
(1)
This rule must precede ADJ PREPOSE to derive e.g.,
(a)
the sleeping child...
without having two rules which reduce Rel. Clauses.
903
PRUNED
RULE - 51*
Order No. 39
ADJ PREPOSE
Rule:
S.I.
NOM
[-PRO]
X
r[+ADj|
MOD
[(AW)]
] X
]
)fiM
+p
TNS H
AST]r
y
N
[-PRO]
S.C.
(1)
(2)
COND:
Obligatory
Example:
NOM-S: attach 3 as first daughter of 2
ART-S, NP-S: attach 3 as left sister of 2
the tall boy, the sleeping child, the well-fed horse,
the burned carpet
Tree:
Ordering Arguments:
(1) There are no strong arguments for ordering this rule before GENPOSTPOSE.
However, we wanted to keep the Rel.Clause rules together.
(2) This rule must precede NOUN REDUCT TO ONE, since the latter rule
deletes identical modifiers.
90U
RULE - 55
Order No. UO
NOUN REDUCT TO ONE
Rule:
S.I.
X
NP[X NOM[X
N
X] X]X ^[X
NQM[X
N
X] X]X
8
<* PLURAL
(+ACCUSE)
(+REFLEX)
^PLURAL
(+ACCUSE)
tr REFLEX
i
30
S.C.
(1)
(2)
Attach [+PRO] to 9
Erase all features in 8 not specified in 9
COND:
(1)
(2)
2 = 7, 3 = 8, 5 = 10
If H-= [+], obligatory; if ft = [-], optional
Example:
1
11
Bob has a red pencil and I have a blue one (one • pencil),
Tree:
Bob has a red pencil
a
blue
pencil
-PLURAL
-REFLEX
+PR0
(The set of features becomes one in the second
lexi c al look-up.)
Ordering Arguments:
(1) This rule must precede NOUN=^0, because only the reduced noun one
(from this rule) can be deleted; e.g.,
(a)
(2)
*three boys of the boys =Mthree ones of the boys ^
three of the boys
Must precede PROPER NOUN THE DEL, in order to derive pronouns as in
(b)
When John came in, he_ was tired.
905
RULE - 56
He starts out as a definite article on the second occurrence of John,
and must be prevented from being deleted. This is accomplished by
ordering this rule before PROPER NOUN THE DEL, since the former rule
deletes (among other things) the feature [-COMMON], so that the
structure index of PROPER NOUN THE DEL will no longer be met.
906
RULE - 57
Order No. Ul
NOUN^0
Rule:
S.I.
X
jjpf
DET[
<*PL
X [+NDEL] & COUNT
+PRO
-REFLEX
X ]
X
N
S.C.
(1)
(2)
COND:
(1) If-- [-].«*/«[•].«* 3- [jgg[]
(2)
Example:
Attach [+PRO] to 3
Erase U
then 5 = 0
Obligatory
After he had eaten, Fred went to the movies.
Tree:
NP
NP
I
I
NOM
D
I
ART
I
-ACC
+DEF
-I
-II
+III
-DEM
+MASC
-PL
+NDELJ
ART
-ACC1
. N
+N
+COUNT
-PL
+PRO
=»
+PRO
" I
(the)
m
\
(one)
he
'
Ordering Arguments:
(1) Must precede ALL THE because the latter deletes of_ in the string
all of the NOUN, and NOUN=>0 creates the environment for ALL THE to
apply.
(2)
Must precede QUANT MOVE, in order to avoid
(a)
*ones the boys each left
907
RULE - 58
PROPER NOUN THE DEL
Order No. 1+2
Rule:
S.I.
X
1
„[ ^[gg]
2
S.C.
Erase 2
COND:
(1)
(2)
Example:
N[-COMMON]
]
3
X
1+
U does not contain (#) S + X
Obligatory
John ate the meat.
Tree:
NP
PROP
D
NOM
ART
|
N
1
7+DEF~j\
L-DEMJJ
1
1
ate the meat
John
[-COMMON]
Ordering Arguments:
(l) There are no strong arguments that this rule must precede any of the
following rules; i.e., this is simply the earliest place at which it can
be ordered.
908
RULE - 59
Order No. U3
S-INITIAL ADV PLACE
Rule:
S.I.
#
S.C.
(l)
(2)
COND:
Optional
Example:
NP
MOD [ X
ADV
AUX ]
X
Attach 2 as right sister of 1
Erase original 2
Hardly ever does John forget his lunch.
Tree:
MOD
NEG
/ADvN
PROP
AUX
I / I / I
hardlylevel/ [-PAST]
forget
his lunch
[the NEG (hardly) will be preposed by the next rule (NEG ATTRACT)]
Ordering Arguments:
(1)
Must precede NEG ATTRACT to account for
(a)
(b)
(2)
Hardly ever does John forget his lunch.
Sometimes John forgets his lunch.
Must precede AUX ATTRACT for the same reason (same examples).
909
RULE - 60
Order No. kh
NEG ATTRACT
Rule Part a:
S.I.
X
[+INDET] (QUANT)
1
X
NEG
X
1
S.C.
(1)
(2)
Attach k as left sister of 2
Erase h
COND:
(1)
(2)
(3)
If h = ADV, then 2 + [-HARDLY] + X
1 f X [+INDET] X
Obligatory
Example:
No-one showed up.
Tree:
NP
D
NOM
I
any
[+INDET]
PROP
MOD
(NEG)
showed up
AUX
I
[+PAST]
N
one
j
Part b:
S.I.
X NEG
12
S.C.
(1)
(2)
Attach 2 as left sister of U
Erase 2
COND:
(1)
(2)
(3)
3 1 X [+INDET] X
5 i X QUANT
Obligatory
Example:
X
3
[+INDET]
k
X
5
No-one showed up.
Tree:
\
(NEG)
NP
PROP
NOM
/-A,
(NEG)
any
I
I
one
N
[+INDET]
910
showed up
RULE - 6l
Ordering Arguments:
(1) Must precede INDEF-BEFORE-QUANT DEL, so that the [+INDET] ART
which triggers NEG ATTRACT is still present.
(2) Must precede ANY-NO, because the latter rule depends on a NEG as
first daughter of the D(eterminer) node.
911
RULE - 62
Order No. U5
INDEF-BEFORE-QUANT DEL
Rule:
T-DEF]
S.I.
1
2
s.c.
Erase 2
COND:
Obligatory
Example:
QUANT
X
3
It
Many people left.
Tree:
Ordering Arguments:
(1)
Must precede QUANT MOVE, because otherwise
(a)
a each of the boy
will become
(b)
*a the boy each
and we will need a separate rule deleting the indef. ART preceding the
definite ART.
(2)
Must precede ANY-NO, otherwise we would derive
(c)
(from
NEG
*no many people
ART^INDETI
(d)
QUANT^8"3^ Pe0Ple^
not many people
912
instead
°f
RULE - 63
ANY-NO
Order No. k6
Rule:
S.I.
[-DEF I
NEG
[+INDETJ
[•COMPLETE]
12
3
X
x
k
S.C.
(1)
(2)
Add 2 as feature to 3
Erase original 2
COND:
(l)
Optional if 3 dominates ever, and 1 ^ #
Obligatory otherwise
Example:
No-one showed up.
Tree:
Ordering Arguments:
(1) No arguments why this rule must precede BE-INSERT; i.e., ANY-NO is
the last rule of this block.
913
RULE - 6U
Order No. hf
QUANT MOVE1
Rule:
S.I.
X
QUMT[+SHIFT]
12
S.C.
COND:
Example:
of
Np[D
3
N]
X
TNS
X
U5678
(1)
Attach 2 as left sister of 7
(2)
(3)
Erase 2 and 3
Erase [+ACCUSE] in U
Optional
The boys each have left.
Tree:
Ordering Arguments:
(l) There are no strong arguments why this rule must precede ALL THE,
because both rules are optional. Hence
^ (QUANT MOVE) =»the boys all left
(a)
all of the boys
"^(ALL THE)
=*.
all the boys left
(2) The rule is not ordered with respect to ALL THREE, since the latter
rule has
[+INTEGER] in its S.I. , while QUANT MOVE has QUANT[+SHIFT].
(3)
Must precede NBR AGREE to derive
(b)
(c)
They each have left.
Each of them has left.
(U) Must precede PRE-Vb ADV MOVE because the QUANT, after it was moved
into the ADV slot (the one preceding TNS),can be moved like an Adv; e.g.,
(d)
The boys have all left.
9lU
RULE - 65
Although our rules do not actually move the QUANT like an adverb (into
post-verbal position), we feel that in principle this is how things
should work.
(5)
We account for examples like
(e)
(f)
Each of them had a piece of pie.
They each had a piece of pie.
by the ad hoc device of deleting the feature [+ACCUSE], This is not
offered as a solution, but only as a way of avoiding an ordering paradox.
We know very little about ACCUSE MARK, and do not account for e.g.,
(g)
(h)
Who did you come with?
The man who John came with...
This rule slightly changed from an earlier version.
915
RULE - 66
Order No. U8
ALL THE
Rule:
[bpthj
of
^t+DEF]
X
Ordering Arguments:
(l) Must precede ALL THREE, because it sets up the environment for the
latter rule.
916
RULE - 67
Order No. U9
ALL THREE
Rule:
r+DEF]
S.I.
X
-DEM
all
ART
1
2
X
3
S.C.
Erase 3
COND:
Optional
Example:
+I
QUANTt[•INTEGER]
All (the) three boys left.
Tree:
PROP
|+DEFJ\ three
L-DEMJ/
^S
I
boys
' [+INTEGER]
Ordering Arguments:
(1) We have no arguments that this rule must precede PRE-VERBAL PRT
PLACE; i.e., this rule is the last of the preceding block.
917
RULE - 68
Order No. 50
BE INSERT
Rule:
S.I,
X
S.C.
Insert be as first daughter of 2
PROP * ^+ADJ^
X
[+V]
COND:
Example:
Obligatory
John is not happy.
Tree:
NP
1
John
PROP
MOD
/\
NEG
AUX
/
1i
/I
/ v
j
/ happy
TNS
+v
/
r
l
/
|_+ADJJ
1
[-PAST]
( [•VI
Ordering Arguments:
(1) Precedes AUX ATTRACT because the be_ inserted by this rule is one of
the AUX's attracted by the latter rule; cf. ,
(a)
(2)
Precedes PRE-Vb PRT because the latter rule also attracts the be_, cf.,
(b)
(3)
John is not happy.
Precedes AFFIX SHIFT to account for e.g.,
(c)
(U)
Is he happy?
Mary isn't pretty.
Precedes NEG CONTRN to account for e.g.
(d)
Mary isn't pretty.
918
RULE - 69
Order No. 51
PRE-VERBAL PRT PLACE
Rule:
S.I.
X
!
I
S.C.
(1)
(2)
COND:
Obligatory
Example:
(TNS
V
)
(M
\
{
(TNS)HAVEV X^
NEC (ADV)
1
1
L
Attach 2 as right sister of 3
Erase original 2
John didn't often visit his mother.
Tree:
Ordering Arguments:
(1)
Must precede AUX ATTRACT because NEG can attract with AUX, cf.,
(a)
(b)
Doesn't he like it?
Hasn't he left?
919
RULE - 70
Order No. 52
PRE-VERBAL ADV PLACE
Rule:
r
S.I.
X
M
")
TNS )HAVE>
ADV
(BE
I
12
X
i
3
S.C.
(1)
(2)
COND:
Optional
Example:
J
h
Attach 2 as right sister of 3
Erase original 2
John has often seen the sea.
Tree:
NP
John
MOD
\
/ADVA
AUX
PROP
seen the sea
Ordering Arguments:
(l)
There are no strong arguments for placing this rule here.
920
RULE - 71
Order No. 53
CONJ SPREAD SCHEMA"1
Rule:
CONJ
A,... A
[•X] Z^
COND:
Z\
•ft
• • •
CON,
•NJ\\
[+X]
•x]
J\
CONJ\\.
v
[+x]
^
Obligatory
Example:
CONJ
[+and]
John sang Bill danced
CONJ
[+and]
John sang
CONJ
[+and]
Bill danced
Ordering Arguments:
(1) Must precede AUX ATTRACT because it is one of the reasons why the
AUX is attracted.
(2)
Must precede EITHER INSERT (CONJ p.110)
(3) Must precede WH DEL (if there is such a rule), since the latter rule
deletes the feature [+WH].
Note:
CONJ
f+OR"]
isLs one of the CONJUNCTIONS spread by this schema.
L+WHJ
CONJ is used for alternative (Yes-No) questions.
This rule slightly changed from earlier version.
921
This
RULE - 72
Order No. 51*
WH COPY
Rule:
I
S.I.
CONJ[+WH]
1
2
3
s.c.
(1)
(2)
(3)
COND:
Obligatory
Example:
#
X
#
It
5
6
CONJf+WHl
7
8
f X #
9
10
Attach 3 and 8 as right sisters of U and 9 respectively
Erase 3 and 8 from complex symbols 2 and 7 respectively
Insert CONT as left sister of 6
Does he snore (or doesn't he snore)?
Tree:
#
CONJ
#
#
he snores
CONJ
+orl
+WHT
§
—L he NEG snores
/[+WH])
>^
'
Ordering Arguments:
(1)
Must precede WH DEL (if that rule is needed) for obvious reasons.
(2) Must precede AUX ATTRACT, because [+WH] is one of the reasons why
AUX is attracted.
(3) (This rule may not be needed if CONJ SPREAD SCHEMA is changed as
indicated on the preceding page.)
922
RULE - 73
Order No. 55
AUX ATTRACT
Rule:
S.I.
(S CONJ)« § [ADVI
[X
feSl] X] X TNS (
U
) (IBB) (ADV) X #
8
5
S.C.
(1)
(2)
Attach 5,6,7 as right sister of 3
Erase original 5,6,7
COND:
(1)
If 6 = 0, 9 = [+Jm\l
(2)
(3)
Obligatory
Last cyclic
Example:
|HAVE|
9 10
Seldom has he mentioned her.
Does he snore (or doesn't he snore)?
Tree:
ADV
seldom
[+NEG]
MOD
he
TNS
PROP
En)
mention
her
Ordering Arguments:
(1) Must precede DO SUPPORT for obvious reasons.
last-cyclic, and not post-cyclic.
Therefore, the rule is
(2) Must precede WH DEL, since the WH provides the environment for AUX
ATTRACT.
(3) Must precede NEG CONTRCN, because the latter rule operates on the
environment created by AUX ATTRACT.
923
RULE - Ik
Order No. 56
WH DEL (THIS RULE MAY NOT BE NEEDED)
Ride:
B.I.
#
[+WH]
TNS
12
S.C.
Erase 2
COND:
(1)
(2)
Example:
X
3
Last-cyclic
Obligatory
Are you coming (or aren't you coming)?
Tree:
[+WH]
TNS
|
I
V
[-PAST]
you
0
be-Ing
NP
PROP
or aren't you coming
t
V
I
come
Ordering Arguments:
(1) (We can do without this rule if we change the CONJ SPREAD SCHEMA as
indicated above.)
92U
RULE - 75
AFFIX SHIFT
Order No. 57
Rule:
S.I.
3
S.C.
(1)
(2)
Chomsky adjoin 2 to the right of 3
Erase original 2
COND:
(1)
(2)
Obligatory
Last-cyclic1
Example:
John loved Mary.
Tree:
NP
I
John
Ordering Arguments:
(1)
Must precede DO SUPPORT in order to get
(a)
(b)
John didn't go home.
Did you see the man?
since affixes can only be shifted across elements marked [+VERBAL].
(2)
Must precede NEC CONTRCTN, because NEG contracts only in the
environmentft+TNS f).
U+SJC]]
This rule must be last-cyclic, applying to all levels of the tree.
This is because all embedding rules which deform AUX require deep structure
AUX's for input and introduce new stems and affixes in their output; hence,
the embedded AUX must not have undergone AFFix SHIFT on its own cycle.
925
RULE - 76
DO SUPPORT
Order No. 58
Rule:
12
3
S.C.
Attach do as left sister of 2
/TERF\
COND: (1) 2 is not dominated by I PROG /
(: j
(2)
Example:
Obligatory
Does he snore (or doesn't he snore)?
Tree:
TNS
I
[-PAST]
NP
I
he
PROP
I
I
snore
r'
/
TNS
do^
NEG
I
[-PAST]
V
I
he
PROP
I
V
I
snore
Ordering Arguments:
(l)
NP
Must precede NEG CONTRCTN for obvious reasons.
926
RULE - 77
NEC CONTRCN
Order No. 59
Rule;
S.I.
X CTNS]
NEG
(SJCJ
1
ffl
2
3
k
5
S.C.
Add [+CNTR] as feature to 3
COND:
Obligatory if U • NP; optional otherwise
Example:
John hasn't seen the doctor yet.
Tree:
PERF
PROP
A
have TNS
I
John
[-PAST]
seen the doctor yet
NEC
/P
[+CNTR]
Ordering Arguments:
(l) There are no strong arguments why this rule must precede any of the
following rules; i.e., this is simply the earliest point at which this
rule may be ordered.
927
RULE - 78
Order No. 60
S.I.
XV
[+T0 DEL]
NP
s
[ °[
AUX.
.
[to X]
1
s.c.
Erase 2
COND:
Obligatory-
Example:
2
3
Fred saw John beat his wife.
Tree:
Fred
See [+PAST]
[+T0 DEL]
AUX
&
PROP
beat his wife
<^
Ordering Arguments:
(l) There are no strong arguments for placing this rule here, since the
rule is not ordered with respect to any rule other than TO REPLACE AUX,
which it must follow and TO BE DEL, which it must precede. Because this
rule is governed, it is not surprising that ordering it is not crucial.
928
RULE - 79
Order No. 6l
TO BE DEL
Rule:
S.I.
X
V
[+TOBEDEL]
Fred
NPr
r
PROP.
[ S [ X to
[ be X
think
[+TOBEDEL]
TNS
I
[+PAST]
NP
MOD
MARY
AUX
to
PROP
V
ADJ
be) impulsive
0^
Ordering Arguments:
(l) There are no strong arguments for placing the block. TO DEL - TO
BE DEL here. This is not surprising, because both rules are governed, and
are ordered only with respect to TO REPLACE AUX, which they must follow.
929
RULE - 80
Order No. 62
PREP DEL
Rule:
S.I.
X
1
S.C.
PREP {PREpf X
2
3
k
Erase 2
COND: (1)
(2)
Example:
3 is not dominated by ART
Obligatory
John hoped that Mary would be here.
Tree:
John
Ordering Arguments
None.
July 1969
930
SAMPLE LEXICON
Contents
Page
I.
BIBLIOGRAPHY
1
II.
FIRST LEXICAL LOOKUP
1
A.
Discussion
1. Introduction
2. Order of Insertion of Lexical Items
3. Form of Lexical Entries
k. Feature Specification
5. Redundancy Rules
6. Problems with Cases
1
1
2
"
9
13
26
B.
Sample First Lexicon
33
III.
SECOND LEXICAL LOOKUP
78
A.
Discussion
78
B.
Sample Second Lexicon
83
931
LEX - 1
I.
BIBLIOGRAPHY
Chapin, P.G. (1967) On the Syntax of Word-Derivation in English
Chomsky,N. (1965) Aspects of the Theory of Syntax
Fillmore.C.J. (1967a) "The Case for Case"
(1967b) "The Grammar of Hitting and Breaking"
Friedman,J. and T.H. Bredt (1968) Lexical Insertion in Transformational Grammar
Gruber, J.S. (1965) Studies in Lexical Relations
(I967h) "Look and See"
(1967c) Functions of the Lexicon in Formal Descriptive
Grammars
Lakoff, G.P. (1963b)"Toward Generative Semantics"
(1965) On the Nature of Syntactic Irregularity
McCawley,J.D. (1966) "Concerning the Base Component of a Transformational Grammar"
Matthews,G.H. (1968) "Le Cas Echeant"
Weinreich,U. (1966) "Explorations in Semantic Theory"
II.
FIRST LEXICAL LOOKUP
A.
Discussion
1.
Introduction
There are many ways in which the present lexicon is provisional and exploratory. Late changes in a number of rules
(particularly Nominalization) have prevented testing it for
internal consistency; the decision not to include selectional
restrictions systematically has left crucial areas unexplored;
doubts about the number of cases in the proposition have made it
difficult to resolve a number of questions; and the fact that the
ordering of the rules had not been definitely fixed at the time the
lexicon was compiled has meant that the redundancy rules are incomplete. Moreover, new problems arose during the compiling of
the lexicon which it has not been possible to investigate fully
in relation to the grammar as a whole.
However, this preoccupation with the problems does not
mean that no progress has been made in specifying lexical entries.
Small-scale computer tests of lexical insertion using interim grammars
"AFESP Case Grammars I and II" were run in March and May 1968
respectively at Stanford University, employing J.Friedman's
system (Friedman and Bredt, 1968) and the results were encouraging
enough to suggest that the form of the lexical entries is at least
932
LEX - 2
coherent. Since the time when the tests were run, the scope of
the grammar has increased considerably with a consequent increase
in the complexity of the lexical entries, but it is assumed that
essentially the same format will continue to work.
2.
Order of Insertion of Lexical Items
In the earliest kinds of transformational grammars
lexical items were introduced by the terminal rewriting rules
of the phrase structure component. Chomsky (1965) suggested two
alternative ways of inserting lexical items so as to take into
account strict subcategorization and selectional restrictions.
In the first of these, the base component includes rewriting rules
which introduce complex symbols (sets of specified syntactic
features) so that the output of the base component is a "preterminal string" consisting of complex symbols and grammatical
formatives. The lexicon consists of an unordered list of lexical
entries, each of which is a phonological matrix for a lexical
formative accompanied by a complex symbol containing a collection
of specified syntactic features. A terminal string is formed by
inserting for each complex symbol in the preterminal string a
lexical formative whose complex symbol is not distinct from that
of the given complex symbol. (Two symbols are not distinct if
there is no feature which is positively specified in one symbol
and negatively specified in the other.) However, the use of
rewriting rules to introduce complex symbols into the preterminal
string of a tree has the effect of changing the base component
from a phrase structure grammar to a kind of transformational
grammar. Consequently, Chomsky proposed an alternative method
of inserting lexical entries. For this purpose, the base component is divided into a categorial component and a lexicon.
The categorial component is context-free phrase structure grammar
whose output is a string of dummy symbols, '&", (to mark the
position of the lexical categories ) and grammatical formatives.
The lexical items are then inserted by a substitution transformation where the complex symbol in the lexical entry is the structure
index for the transformation, and the lexical item is appropriate
for substitution if the tree meets the conditions of the structure
index specified by the complex symbol.
It is the second of these alternatives that we have
adopted, primarily for the practical reason that it permits
greater latitude and flexibility in making and changing decisions
about the lexicon while leaving the categorial component fixed.
However, for a variety of reasons, both practical and theoretical,
we have incorporated a feature of Friedman's system whereby
verbs are inserted before nouns. Chomsky (1965: H^-115) argued
against the insertion of verbs before nouns on the grounds that
933
LEX - 3
the complex symbols for the nouns would require such features as
[ PREand
+[+ABSTRACT]-SUBJECT, PRE- +[+ANIMATE]-OBJECT]
[ POST- +[+ABSTRACT]-SUBJECT, POST- +[+AHIMATE]-OBJECT]
for the subject and object respectively of a verb such as frighten.
Chomsky pointed out that these specifications were excessively
redundant since "the feature [PRE- +[+ANIMATE]-OBJECT] is irrelevant to choice of Subject Noun, and the feature [POST- +[+ABSTRACTSUBJECT] is irrelevant to choice of Object Noun". Chomsky
maintained that there was "no alternative to selecting Verbs in
terms of Nouns ... rather than conversely." However, it turns
out that the insertion of verbs first need not lead to such unwieldy specifications.
This is because of what Friedman has called "side effects."^ '
Side effects are effects on other nodes in a tree after an item
had been inserted. Thus, if verbs are inserted first, the selectional
features in the complex symbol for the verb must be specified for
the relevant category nodes in the tree. Friedman and Bredt give
the example of admire, which is positively specified for animate
subject, thus requiring the corresponding NP to be so specified.
+ N
+ V
+ ANIMATE
- ABSTRACT
... N
+ TRANS
...
V
+ ANIMSUBJ
adm: .re
[Friedman and Bredt, 1968: 30]
Side effects thus achieve the same ends as were gained by Chomsky
in making verbs selectionally dependent on nouns, so that, in
many ways, Chomsky's system and Friedman's can be considered
notational variants of each other.
(1)
This notion is similar to that of "transfer features" as
proposed by Weinreich (1966) to account for certain semantic
questions of disambiguation, selectional deviance, etc.
93U
LEX - k
We have provisionally adopted Friedman's approach
because the notion of side effects seemed sufficiently promising
to bear further exploration, particularly in terms of a deep
case grammar. Moreover, the insertion of verbs first makes for
much more economical testing in a computerized program, because
random selection of nouns will lead to a large number of "impossible"
strings in which no verb can be inserted. Nor is this a purely
practical issue, since in a very real sense verbs are selectionally
dominant. It must be admitted, however, that the theoretical
implications of side effects need investigating more fully than
we have been able to do thus far. Part of the difficulty is that
we have not investigated selections! restrictions in any depth
but even at this early exploratory stage it is clear that there
are problems which we do not yet know how to handle. For
example, as Friedman and Bredt point out, negatively specified
selectional restrictions cause problems since a verb marked for
[-ANIMATE SUBJECT] would be acceptable for insertion in a
tree such as
[Friedman + Bredt, 1968:32]
though this is presumably wrong. Consequently, Friedman and Bredt
conclude that contextual features containing selectional restrictions should be positively specified. We have adopted this
principle but it causes problems for which we have at present
no solution. The difficulty is not with the animacy of subjects
since we are assuming that agents and datives are always [+ANIMATE],
though as we shall see below this is not altogether correct.
The trouble arises with a selectional restriction which
applies to an optional contextual feature. For example, the verb
answer must take an agent and may or may not take a dative or
neutral case:
i
^
11)
,
^
(a)
(b)
(c)
(AGT
)
Nobody answered /DAT\
Nobody answered Jonn (jjEUT)
Nobody answered the question
If we now wish to place a sele.ctional restriction on the verb to
allow only an abstract objec&v this must be positively specified
(l)
We are assuming that the verb in he answered the door is a
different verb. The example is perhaps unfortunate because of
sentences such as he answered the letter, which raises other
questions, but it is the point being illustrated which is important not the particular example.
935
LEX - 5
[+ [+ABSTRACT] OBJECT]. However, we have now assigned (by side
effects) a feature to a constituent which may not be present.
If we followed Chomsky, the restriction could be negatively
specified, [- [+CONCRETE] OBJECT], and this would leave the
presence of the object optional, but as we have seen negatively
specified features cannot have side effects. This may be an
important argument against the use of side effects, but we are
still hopeful that the principle may be saved. What we need
is some kind of device that will indicate that if optional feature
[F.] is present then it is positively specified for feature [?^].
We could call this device "implicational specification" and
employ a notation such as [&F [+F.] ] which would mean [+F ]
if and only if [+F ], otherwise [-F.]. We could not use an^alpha
convention because, for reasons stated below, optional contextual
features are left unspecified. However, we have not attempted
to incorporate such a device into our specification of features
because it is not absolutely clear that something of this nature
will be required.
There are further problems in the ordering of lexical
insertion which we have not resolved. In Friedman's algorithm
embedded sentences are considered in lowest to highest, right
to left orderW Lexical items are inserted for each lexical
category node in the order specified in the lexicon (e.g. V N PREP
ART, which would mean that verbs were to be inserted first,
followed by nouns, prepositions, and articles). In each category,
the order is left to right in the tree. This is the type of
algorithm that was used in the test grammars I and II. However,
since then certain problems have arisen. One of them is that
the order of insertion of the category nodes has not yet been
fixed, although we are assuming at the moment that verbs will
be inserted before nouns. A more important point is that the use
of side effects to specify selectional restrictions will in
some cases require that subtrees be considered in highest to
lowest order. For example, verbs such as persuade and force
require that the verb in the lower embedded sentence be
[ STATIVE]:
(2)
(a) He persuaded them to be good
(b) *He persuaded them to be delighted
(1)This is what the text says; in the accompanying diagram (p. 25)
the order is shown as left to right. As the choice is presumably
arbitrary and of no substantive significance the discrepancy
is unimportant.
936
LEX - 6
(c) She forced him to eat it
(d) *She forced him to know it
If such restrictions are to be specified by side effects,
the verb in the higher sentence will have to be inserted first,
which means top to bottom processing. On the other hand, these
particular selectional restrictions are too weak to base such
a decision on, since the feature [+/-STATIVE] itself is not
particularly transparent and there are many putative [+STATIVE]
verbs which can occur after persuade and force.
(g)
(h)
He persuaded them to like it
He forced them to respect him
It seems likely that the constraints imposed by these verbs are
directly related to the Imperative, and only indirectly to the
feature [STATIVE]. W
3.
Form of Lexical Entries
The form of the lexical entries follows, in principle, the
lines of the Stanford University Computational Linguistics Project
(Friedman and Bredt, 1968). Each vocabulary word has associated
with it a complex symbol containing four types of features:
category features, contextual features, inherent features and rule
features. A category feature denotes a lexical category such as
noun or verb. In the present format each complex symbol contains
only one positive specification for a category feature and this
means that there is no disjunctive ordering of related lexical
entries. Thus, each vocabulary word which belongs to more than
one lexical category, e.g. torment, empty, has associated with it
a separate complex symbol for each lexical category. Derivational
processes have also been ignored in the present lexicon. Although,
in principle, we would like to have a single complex entry for
items such as produce, productive, production, product, etc.
and though we have tentatively explored some possibilities in
this direction, there are so many complex problems that nothing
has reached a formalizable state. (See NOM for further discussion.)
(1) This whole question needs further investigation along the lines
suggested by Gruber (1965), who posits causative agents, passive
agents and non-agents. Some such classification is relevant
to the feature [STATIVE], as can be seen in the following
examples:
This report deals with export subsidies/*is dealing with
John deals with your requests usually/is dealing with...today
That matter does not concern me/*is not concerning me.
I concern myself with such matters/I am concerning myself
This problem is also related to that of Genericness (cf. Chapin,
1967). See below for the relationship between agency and stativity.
937
LEX - 7
A number of the contextual features are represented by a
"case-frame" (Fillmore, 1967a:35) in which the cases that can
occur with a lexical item are shown. For example, Fillmore
suggests the case-frame [
DAT (INS) ACT] as a suitable one for
the verb kill in, say, the farmer killed the chicken (with an axe),
where the parentheses round the instrumental case show that the
instrument may be omitted. However, this case-frame will not
account for the sentence the poison killed the chicken, since there
is no agent, which is obligatory in the above frame. This situation
can be covered by a second entry for kill with the case-frame
[
DAT INS] where the instrument is now obligatory and the
agent omitted. Fillmore suggests an ingenious notation for
combining these two entries by means of linked parentheses, which
indicate that at least one of the two elements thus specified
must be chosen, [
DAT (INSOAGT)] to account for the sentences
Mother is cooking the potatoes, the potatoes are cooking and
Mother is cooking. However, cook may also optionally take a
locative and an instrument, Mother is cooking on the stove,
Mother is cooking with gas and these optional cases cannot be
included in the case-frame with linked parentheses, given our
decision that the order of cases is fixed, with LOC and INS
both intervening between NEUT and AGT. Furthermore, we have
(for reasons given below) chosen to specify obligatory contextual
features positively, impossible contextual features negatively,
and omit optional contextual features. Thus our case-frames for
kill are:
[
- NEUT + DAT - LOC + AGT]
[
,(1)
- NEUT + DAT - LOC + INS - AGT]
(The other contextual features are listed in the complex symbol
immediately following the case-frame.)
However, this means that the number of entries is multiplied
as an artifact of the system of notation. Although there are a
number of ways in which this multiplication of entries could be
avoided we have not adopted one because the choice at this stage
would be arbitrary and would have the effect of concealing the
problem rather than solving it. On the other hand, there are
also polysemous items which need separate entries for distinct
readings in any lexicon not simply as a consequence of the
notation used. For example, sick in John is being sick must
(l)
Whether the second entry should be specified [-AGT] or left
unspecified for AGT is a question which appears to be an
artifact of the representation. (Unspecified uses fewer
features but predicts an unrealized ambiguity.)
938
LEX - 8
2
be kept distinct from sick in John is sick.
to a difference in the case-frames:
This corresponds
SICK1
+[
-NEUT -DAT -LOC -INS +AGT]
SICK2
+[
-NEUT +DAT -LOC -INS -AGT]
There is, thus, an important difference between the two entries
for kill, which are a consequence only of the lack of disjunctions
of features and sets of features in the present system of notation,
and the two entries for sick, which are semantically distinct,
though related. We have accordingly chosen to indicate multiple
entries of the kill type by superscript lower-case letters (e.g.
KILLa) and polysemous items of the sick type by superscript
numerals (e.g. SICK ). As might be expected, it is not always
easy to decide whether two entries are substantively different
or not. For example, we have chosen to represent sick in he is sick
of arguing about linguistics as SICK-3 although it might also belong
with SICK^. This is a traditional problem for lexicographers and
no attempt has been made to deal with it systematically in the
present lexicon. However, the problem forced itself on our attention
because of the semantic nature of deep-case relationships (e.g. the
relationship between the presence of AGENT and stativity — see below)
and the use of such semantically based syntactic features as
[+/- FACT] and [+/- EMOT]. This is one of the ways in which the
nature of the present sample lexicon has changed as a consequence
of new rule features introduced into the grammar. Moreover, it
has become increasingly obvious that the kind of features employed
in the present grammar need to be defined much more precisely than
they have been so far. One of the benefits of even a small sample
lexicon such as the present one is that it draws attention to
difficulties in feature specification which might otherwise be
overlooked.
Inherent features denote qualities such as animate, human
and abstract. Rule features refer to the transformations which can
apply to the lexical item, e.g. EXTRA (position), TO-DEL(etion).
(See NOM)
The number of inherent features will ultimately depend on
where the dividing line between syntax and semantics is drawn.
Since selectional rules are not included in the present grammar the
number of inherent features needed is quite small and no attempt
has been made to incorporate many of the features suggested in
recent treatments of semantic theory (e.g., Lakoff, 1963; Weinreich,
1966). There is thus in this formulation no essential difference
between inherent features and rule features.
939
LEX - 9
h.
Feature Specification
When Chomsky (1965:81-83) first proposed the use of
features for the specification of lexical entries similar to
the form of phonological entries in a distinctive feature
matrix, he only allowed three values for a feature, namely,
positive, negative or unspecified. However, it is probable
that at least four and possibly five values are necessary.
This is partly because different kinds of features may require
different values to be specified. For example, contextual
features and rule features differ in this respect.
For contextual features, positive specification [+F^]
means that such an element must occur in the proposition to allow
insertion of the lexical item and negative specification [-F;]
means that the lexical item cannot be inserted in the presence
of such an element. Similarly, for rule features positive or
negative specification will indicate whether a given governed
rule must or cannot apply. However, there is an important
difference between the two kinds of features when the feature
may be either positively or negatively specified for a single
lexical item. In the case of contextual features such a feature
is genuinely optional since its presence or absence does not
affect the insertion of the lexical item. Thus, in the present
lexicon optional contextual features are left unspecified since
the lexical item can be inserted whether the element is present
or not. For example, the verb cook, as mentioned above has
two entries, one for the transitive verb in Mary is cooking
the meat (on the stove) (with gas) and the other for the intransitive verb in the meat is cooking (on the stove). The
case-frames for these two entries are
cooka
[
- DAT
cookb
[
• NEUT - DAT - INS - AGT]
+AGT]
The first case-frame shows that cooka must take an agent, cannot
take a dative, and may or may not take neutral case, a locative
or an instrument. The second case-frame shows that cook must
take neutral case, cannot take a dative, instrument or agent, and
may or may not take a locative, For contextual features, therefore,
absence of specification means that the element may or may not be
present.
The situation is rather different with respect to rule
features. Let us consider the following examples:
(3)
(a)
(b)
I saw him leave.
Mary helped him (to) do it.
9^0
LEX - 10
(c)
(d)
The government wanted him to accept.
He avoided (looking ")at her.
(*to lookj
We can see that with respect to the rule for TO-deletion (see
NOM) there are not three possibilities but four. In (3a)
the rule must apply, in (3b) the rule may or may not apply, in
(3c) the rule does not apply, and in (3d) the rule is irrelevant since the structural description for the rule is not met.
Items which never meet the structural description of the rule
can be left unmarked but items where the rule is optional cannot
be left unmarked for that feature because the rule will be specified as obligatory and will require the governing item to be
positively specified. Consequently, in such cases we have
"obligatory specification" [*F.], which means that the value
of the feature is left unspecified in the feature index of the
complex symbol but must be specified either positively or negatively before the complex symbol is inserted in a tree. Thus,
for example, the complex symbols for the verbs see, help, want and
avoid will contain the following specifications for the rule
feature TO-DEL(etion): **'
see
help
want
[ + TO-DEL] [ * TO-DEL] [ - TO-DEL]
avoid
[
]
However, since the optionality of governed rules is handled by
"obligatory specification" and there are no transformations which
required a feature to be negatively specified, it is possible for
negatively specified rule features to be left unmarked in the
lexical entry. This is equivalent to a redundancy rule:
[uF^ =^ [-F^]
where FA = rule feature
In this respect, rule features and inherent features are
treated differently.
It is possible that a five-valued system might be necessary for inherent features. For example, [+HUMAN] nouns must be
specified for gender in order to allow correct pronominalization;
thus, boy, man and brother are [+MASC] and girl, woman and sister
are [-MASC]. However, nouns such as neighbor, teacher, doctor and
cousin can be specified either positively or negatively for the
(1)This is probably more mechanism than we need in many cases.
However, our analysis has not yet reached the degree of
subtlety where we can attempt to distinguish between major and
minor rules. See Lakoff (I9b5) for a careful analysis of the
possibilities.
9^1
LEX - 11
feature [MASC], though it is not clear whether this is optional
or obligatory specification. In any case, it is different
from the situation with the [-HUMAN] higher mammals, e.g. horse,
monkey and dog, which may be (but need not be) specified for
gender. These in turn are possibly different from other forms
of life which are seldom, if ever, specified for gender, e.g.
fruitfly, worm and jellyfish. If five values are necessary we
could adopt the following convention:
(1) + positive specification
(2) - negative specification
(3) * obligatory specification
(U) +/- optional specification
(5)
absence of specification would mean that the feature
was irrelevant
This would provide (partial) entries of the following kinds:
boy
girl
neighbor
mare
horse
fruitfly
+ N
+ N
+ N
+ N
+ N
+HUMAN
+HUMAN
+HUMAN
-HUMAN
-HUMAN
-HUMAN
+ MASC
- MASC
*MASC
-MASC
+ H
"
+/-MASC
-
_i
_
_i
However, it is far from obvious that this is the right way to
handle these relationships. In the first place, a sentence
such as I haven't met the teacher yet feels intuitively unspecified
for gender, although whenever an anaphoric pronoun is used it
must be either he_ or she and not it_. Secondly, he_ often occurs
as an unmarked form with indefinites, e.g. everyone did his best,
which does not imply that everyone is [+MASC]; everyone did his or
her best sounds extremely pedantic and everyone did their best
is often stigmatized as substandard, but the three sentences
seem to be variants. Thirdly, there is the problem of it^ as an
anaphoric pronoun for [+ANIMATE] [-HUMAN] nouns. As we have seen
above many of these (perhaps all of them) can be specified for
gender but they need not be. Perhaps we need a feature [+/-GENDER]
such that [+GENDER] requires specification for the feature [MASC],
whereas [-GENDER] nouns would not require such specification and
and be anaphorically replaced by it. This, however, will not help
with nouns such as neighbor. Alternatively, we might have a
feature [+/-FEMININE] in addition to the feature [+/-MASC] so
that it_ would replace a noun which was negatively specified for
both features. However, it seems ad hoc and counter-intuitive
9U2
LEX - 12
to make nouns such as neighbor and teacher hermaphroditic with
a positive specification for both features. In the absence of
convincing evidence as to the correct choice we have decided
to treat inherent features like rule features and have eliminated
specification (h) above. This means that items such as horse
must either be classed with neighbor or with fruitfly and
the latter choice seems preferable. Finally, it is possible
that selection of gender for items such as neighbor is fundamentally semantic (as McCawley (iy66) has argued) and thus some of
the above discussion may relate to a pseudo-problem, but within
the scope of the present grammar we have no alternative to a
syntactic solution.
As was stated above, optional contextual features are
left unspecified whereas optional rule features and optional
inherent features have "obligatory specification", indicating
that the feature must be positively or negatively specified
before the lexical item is inserted into a tree. This means that
the entry for a lexical item will show the rule features and inherent features which are relevant to that item but will show
only those contextual features which are positively or negatively
specified, indicating that their presence or absence is obligatory.
To know which contextual features are optionally allowed one must
know the set of possible contextual features and consequently
which features have been omitted from the feature index. For
example, verbs and nouns which can take a neutral case may
take a sentential complement, either dominated directly by
neutral case or dominated by the fact (see NOM), unless such
features are negatively specified. Accordingly, destroy,
which does not allow a sentential complement of either kind,
must be marked [-FACT] and [-S]; regret, which allows only
factive sentential complements, must be marked [-S]; and expect,
which allows only non-factive sentential complements, must be
marked [-FACT]. This may appear confusing at first sight since
factive verbs are identified by the specification [-S] and nonfactive verbs by the specification [-FACT]. The absence of
both negative specifications in a verb which takes a neutral
case would mean that the verb takes both factive and nonfactive sentential complements, but in the present lexicon such
verbs have two entries.^'
Deep structure articles, pronouns and prepositions which
will later be given their appropriate phonological representation
in the Second Lexical Lookup are listed in the first lexicon under
identifying labels in lower case letters between quotation marks,
e.g. "the", "much/many". These labels are identificatory only
(l)
The multiplication of entries
here since there is clearly a
in He remembered telling her,
in He remembered to tell her,
9U3
is not altogether unmotivated
difference between remember
which is factive, and remember
which is non-factive.
LEX - 13
since such items have no phonological representation until
the Second Lexical Lookup.
5.
Redundancy Rules
Redundancy rules help to reduce the number of feature
specifications in a complex symbol whenever predictable features
can be added by a general rule. The usual form of such rules is
outlined in the GENERAL INTRODUCTION-FORMAL ORIENTATION (see
under "Lexical Rules"). In addition we allow complex symbols on
the left in redundancy rules and such complex symbols may include a
feature with "obligatory specification" ([*F1]). For example,
r+N
L*
I
HUMAN!
=>
[+ANIMATE]
since any noun thai is speci fied for the feature HUMAN must be
ANIMATE.U' It is important to note that this rule is equivalent
to the three rules
[-•H
1
-/
[+ANIMATE]
HUMAN]
=>
[+ANIMATE]
+ N
* HUMANj
^
[+ANIMATE]
[+
HUMAN]
r 1
[_-
since the redundancy rules apply before the insertion of a lexical
item in a tree and thus there may be items where the value "*"
has not yet been expanded. Examples for the above feature are
man [+HUMAN], horse [-HUMAN], and champion [*HUMAN].
(l)
It is important to note, as Friedman and Bredt point out
(1968:10), that rules of the kind used by Chomsky (1965:82),
e.g. [+ANIMATE]/[+/-HUMAN], are not redundancy rules but
generative rules, since the feature HUMAN is certainly not
optional for all animates (if any).
9UU
LEX - lU
Fillmore (l96Ta:3^) suggests redundancy rules of the following
kind: (1)
p
r+N —ii
1+
AGTJ
r+N i
L+
DATJ
=*>
[+ ANIMATE]
$
[+ ANIMATE]
However, in the present grammar it would be impossible to interpret
such rules since syntactic cases are not assigned to nouns and even
if they were this would not have happened by the time the redundancy rules apply. On the other hand, it would be useful to
capture the generalization that agents and datives are usually
[+ ANIMATE] and also that locatives and instruments are usually
[- ANIMATE]. It is unfortunately not true that this is always
the case:
F+ ACT
y ANIMATE]
{h) (a) The wind opened the door.
T+ DAT
L- ANIMATEj
(b) John robbed a bank.
[+ LOC
(c)
(d)
1
1+ INS
. J
[+ ANIMATE(?U
1+ ANIMATE(?)j
He hit me in the face
fac with his fist
LOC
ANIMATE]
There are thieves*in the crowd
[:
It is clear that the problem is not simply one of the character
of the cases but also involves the little explored nature of inherent
features such as [ANIMATE]. In the above examples, it may be that
natural forces such as wind which are the principal class of
[-ANIMATE] nouns that can appear as agents are in fact a subclass
of [+ ANIMATE] nouns. Similarly, many [-ANIMATE] nouns such as
bank which have human associations can often take the genitive and
otherwise behave in some sense like [+ANIMATE] nouns. On the other
hand, fist and face, though parts of an animate being, share few
selectional restrictions with nouns such as man, horse and fruitfly,
and thus they are [-ANIMATE], Perhaps the feature we need should not
Xl)
This is the form of the rules in the pre-publication version.
In the published version Fillmore gives a different formulation
which is closer to our rules given below.
9^5
LEX - 15
be labelled ANIMATE but something like AUTONOMOUS. This might
exclude collectives such as crowd. In any event, the question
of redundancy rules on cases is complicated by the fact that we
are working with such ill-defined features.
It was also thought at one time that locatives and instruments might be predictably [-ABSTRACT]. However, this turns out
to be wrong:
C LOC
1
|± ABSTRACTJ
(5) (a) He found the idea in one of Chomsky's footnotes.
1+ INS
1+ ABSTRACTJ
(b) He destroyed my argument with several counterexamples .
There is, nevertheless, an interesting constraint on verbs such
as find which can take a [+ABSTRACT] locative only with a [+ABSTRACTJ object:
(c)
f+ NEUT
[+ LOC
"1
U ABSTRACTJ
1=: ABSTRACTJ
He found the pencil in a drawer.
1+ NEUT
1+ LOC
L+ ABSTRACT]
U ABSTRACT]
(d) *He found the idea in a drawer.
[+ NEUT
(+ LOC
Lr ABSTRACT)
|+ ABSTRACT!
(e) *He found the pencil in a footnote.
Since many lexical items in locative position can be either
[+ABSTRACT] or [-ABSTRACT], the concreteness of the object will
determine the concreteness of the locative:
.
f+ LOC
L-
(f)
ABSTRACT]
He found the pencil in a book.
(g)
f+ LOC
1+ ABSTRACT!
He found the idea in a book.
Any redundancy rule that would capture this relationship would
presumably also require neutral case to precede locative case in
the insertion of lexical items. At the present stage of uncertainty
as regards the ordering of lexical insertion this is conceivable
and it seems reasonable that the order should not be completely
arbitrary, but it is too early to know what consequences this
would have.
9U6
LEX - 16
An alternative proposal has been put forward by Matthews
(1968) where dative case "refers to a person or thing which is
affected in some way by the action of the verb", whereas in the
neutral case (absolutive case, in Matthews's terminology) the
referent is acted upon by the action of the verb but "not affected
by this action". Thus Matthews contrasts
(AGT)
(DAT)
(INS)
(6) (a) The workman broke the window with a hammer.
. ,
(AGT)
(NEUT)
(DAT)
(b) The doctor broke the bad news to the child's
(INS)
parents with a telegram.
Matthews argues that the bad news is not in the dative case
because it is not affected by the action of the verb, that is,
it "is the same before and after it is broken to the child's
parents". Although this captures a distinction between neutral
and dative that is not handled in the UESP grammar, the examples
are not convincing. In the first place, it seems unlikely that we
are dealing with the same verb in he broke the window and he broke
the news since the former can occur freely with physical objects
of a certain degree of regidity (Fillmore, 1967b:25), but the latter
is extremely restricted even with abstract objects:
/the tabled
(7) (a) He broke jthe stick /
I his leg ?
(^a cup
)
(b)
/ the news
"*
He broke \ the story
[
) *the idea
>
j *a proposal \
( *his thoughts^
Secondly, there is no apparent parallelism between the examples
of the dative case in
(o)
(a)
He broke the window.
(b)
He broke the news to the child's parents,
nor can they be switched.
(c) *He broke the child's parents.
(d) *He broke something to the window.
9^7
LEX - 17
Thirdly, Matthews makes use of pseudo-cleft constructions to distinguish
between neutral case and dative case:
(9)
(a)
What the workman did to the window was break it.
(b)
What the doctor did with the bad news was break
it to the child's parents.
However, pseudo-cleft constructions with appended prepositional
phrases can be so freely generated that it is dangerous to base
decisions of this kind on them:
/to ~)
/break it
~")
(c) What John did \aboutS the window was ) oil the hinges
r
(_with )
(^replace the glassy
(d)
/keep quiet
\
What John did/about) the bad news was\tell his father
/
fwith >
J
about it
;>
^
^
)suppress it
^
(telephone his mother;
Accordingly, we have not adopted Matthew's use of dative though
semantically, at least, the distinction between an object which
is affected by the action of the verb and one which is not is
clearly important. However, we do not feel that it can best be
captured in the present kind of case-grammar by a contrast between
neutral and dative.
datives.
There are a few residual problems with apparently [-ANIMATE]
For example, in the following sentences
(10) (a)
(b)
(c)
John gave the house a coat of paint.
He attributed his success to good looks.
This evidence lends credence to his argument.
House, good looks and argument all seem possible datives. However,
in (10a) house is clearly not a dative in the same sense as Peter
in
(d)
(e)
(f)
John gave Peter his old car.
John gave his old car to Peter.
John gave Peter his old car and then took it back
again.
9U8
LEX - 18
since there are no equivalent examples with house to (I0e,f)
(g) *I gave a coat of paint to the house,
(h) *I gave the house a coat of paint and then took
it back again.
In fact, house must be a locative with obligatory objectivalization:
I gave a coat of paint on the house ^ I gave the house a
coat of paint.
Supporting evidence that this is the right analysis comes from
the sentences
(i) I put a coat of paint on the house,
(j) *I put the house a coat of paint,
(k) I gave the house a coat of paint and then took
it off again.
This will, of course, be a different lexical item from give in
(lOd-f). The other two cases are harder to account for without
adding to the number of cases or making some apparently ad hoc
changes to the rules. In the present analysis they remain as datives.
In Test Grammar II the following rules were proposed:
[+ N
[+ ANIMATE]
[+ N
I- ANIMATE]
p INS NOUN*!
(- LOC NOUNJ
h
~'
\- AGT NOUN]
L DAT NOUNJ
_v
where INS NOUN, LOC NOUN, AGT NOUN, DAT NOUN were abbreviations of
contextual features which amounted to "able to appear as head noun
in an INS (LOC, AGT, DAT respectively) case frame." Although, as
we have seen, these rules are not completely accurate we have
decided to retain them until they can be replaced by rules which
better capture the generality which lies behind them.
(l)The usual restriction with [-ANIMATE] nouns is exactly the
opposite, I brought the water to the table/*! brought the table
the water, though there are some problems as to whether table
is a dative or a directional locative.
9»»9
LEX - 19
Another suggestion which Fillmore has made regarding the
intrinsic content of cases is his claim that only verbs which have
an agent in the sentence are non-stative:
The transformation which accounts for the true
imperatives' can apply only to a sentence containing
an A[gent]; and the occurrence of B[enefactive]
expressions, progressive aspect, etc., are themselves
dependent on the presence of A[gent]. No special
features indicating 'stativity' need to be added to
verbs, because only those verbs which occur in Propositions] containing an A[gent] will show up in those
sentences anyway.
[Fillmore, 1967a:k2]
This is an important claim since it would, if correct,
support the view that deep cases reflect semantic relations in an
economical and non-ad hoc manner. However, the statement as it
stands is clearly inadequate and is contradicted by one of the
examples given by Fillmore two pages before it, namely the
potatoes are cooking, where there is no agent in the sentence to
account for the progressive, unless Fillmore means that there
is a deleted agent in this sentence. It turns out that there
are two main groups of possible exceptions to Fillmore's claim:
(11)
Verbs in the progressive without an animate subject
(a) The string is breaking.
(b) The potatoes are cooking.
(c) This material is losing its sheen.
(d) The train is arriving.
(e) The water is filling the barrel.
(f) The garden is swarming with bees.
(12)
Verbs in the progressive with an animate subject which
is putatively in the dative case
(a) He is dying.
(b) John is dreaming.
(c) I am hoping to hear from them very soon.
(d) I'm regretting it already.
(e) She is expecting that there will be a big crowd.
(f) They are hating it.
Although such examples show that Fillmore's claim cannot be
accepted as it stands, it does not prove that there is no correlation
between stativity and lack of agent in the sentence since stativity
is not merely a matter of tolerance for the progressive aspect. We
will consider the examples in (ll) first since the absence of an
animate subject otherwise coincides with the criteria for stativity:
950
LEX - 20
(13)
Imperative
(a) *Break, string!
(b) *Cook, potatoes!
(c) *Lose your sheen, material!
(d) •Arrive, train!
(lU)
Do-something
(a) "What the string did was break.
(b) *What the potatoes did was cook.
(c) *What the material did was lose its sheen.
(d) *What the train did was arrive.
(15)
Do-so
(a) *The string broke and the rope did so, too.
(b) *The potatoes cooked and the meat did so, too.
(c) "This material lost its sheen and that material
did so, too.
(d) *The train arrived and the bus did so, too.
(16)
Suasion
(a) *I persuaded the string to break.
(b) *I forced the potatoes to cook.
(c) *I made the material lose its sheen.
(d) *I ordered the train to arrive.
(17)
Agentive
(a) *The
(b) *The
(c) "The
(d) *The
(18)
In-order-to
(a) *The string broke in order to open the parcel.
(b) *The potatoes cooked in order to feed the people.
(c) "The material lost its sheen in order to be less
ostentatious.
(d) *The train arrived in order to disgorge its passengers,
adverbials
string willingly broke.
potatoes cooked carefully.
material deliberately lost its sheen.
train carefully arrived.
It is clear that by the above criteria the verbs in (ll)
are non-stative in spite of the fact that they can take the progressive aspect.
Nevertheless, there remains the problem of why the stative
verbs in (11), if they are stative, can take the progressive aspect.
The confusion arises because BE+ING has more than one use:
(19)
(a)
(b)
Look, the young bird is actually flying. (Now)
John is flying to London next week. (Future)
951
LEX - 21
(c)
(d)
John is flying to Europe or Africa these days.
(Habitual within a limited period.)
John is always flying off somewhere. (Uttered
as a complaint)
If we look at the examples in (11) we find that it is not simply
a matter of BE+ING:
(20)
(a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
(i)
(j)
(k)
(l)
*The string is breaking tomorrow.
*The string is breaking these days.
The string is always breaking.
*The potatoes are cooking tomorrow.
*The potatoes are cooking these days.
*The potatoes are always cooking.
*This (piece of) material is losing
its sheen tomorrow,
*This (piece of) material is losing its sheen
these days,
*This (piece of) material is always losing its sheen,
The train is arriving tomorrow,
*The train is arriving these days,
*The train is always arriving.
However, note also
(m) ?This (kind of) material is losing its sheen these
days,
(n) This (kind of) material is always losing its sheen,
(o) The train is arriving late these days.
(p) The train is always arriving late.
There is, apparently, some relationship between the classes of
verbs and the uses of BE+ING. In connection with such problems
Vendler (1967:97-121) has some interesting observations to make.
As well as distinguishing between "activity" verbs and "state"
verbs, Vendler has two additional categories, "achievement" verbs
and "accomplishment" verbs. Vendler's "activity" verbs, e.g. run,
walk, swim, push, etc., are unambiguously in the category of nonstative verbs, and his "state" verbs, e.g. know, believe, like,
hate, etc., correspond closely to stative verbs. It is the other
two categories which are especially interesting. Vendler gives
as examples of "accomplishment" verbs
paint a picture, build a
house, draw a circle, give a class, play a game of chess, etc.,
in all of which the perfective use of the verbs requires the
completion of a finite task. In other words, if John begins to
(l)The fact that Vendler gives examples of verb phrases rather
than verbs is an indication that we are dealing with a fairly
complex situation.
952
LEX - 22
draw a circle but stops before the task is completed we cannot say
John drew a circle, while with an "activity" verb such as run there
is no such requirement. As examples of "achievement" verbs Vendler
gives recognize, realize, identify, find, win the race, reach the
summit, etc. At first sight, it is not obvious that "achievement"
verbs differ significantly from "accomplishment" verbs but the basis
of the distinction is that "achievement" verbs take place at an instant
of time, whereas "accomplishment" verbs take place over a period of
time. Vendler's example is that if it takes you an hour to write a
letter you can say at any time during that hour I am writing a letter,
but if it takes you three hours to reach the summit you cannot say
at any moment during that period I am reaching the top. Since it
might be argued that the latter remark is possible, it might be
safer to say that it would at least be inappropriate as a reply to
the question What are you doing?
However, perhaps more important that the distinction between
"achievement" and "accomplishment" verbs is the difference of both
of them from "activity" and "state" verbs. Vendler argues that
"activity" and "state" verbs do not require unique or definite
periods of time. By this, Vendler apparently means that "activity"
and "state" verbs do not place definite limits on the duration of
the action or state. For example, he is swimming in the sea and he
knows the answer do not imply a specific termination of the "activity"
of swimming or the "state" of knowing. On the other hand, he is writing
a book and he is winning the race require that the terminal point has
not been reached; that is, that the book is not yet finished nor the
race over.
If we look back at the examples in (ll) we find that (lla)
and (lie) are similar to Vendler's "achievement" verbs and that (lib),
(lid) and (lie) are similar to Vendler's "accomplishment" verbs. If
this identification is correct it might help to explain why such
verbs allow BE+ING when it is used to indicate a process of indefinite
duration. In (lla) the process must end when the string breaks and
in (lie) when the train arrives; in (lib), (lid) and (lie) the time
will come when the potatoes are cooked, the barrel is filled and the
material has lost its sheen. At such a point the process will stop
and it will no longer be appropriate to use BE+ING.
It is possible that there is some better classification of such
verbs than "achievement" and "accomplishment" but Vendler's distinction at least supports the view that the occurrence of BE+ING with
the verbs in (ll) is not in itself sufficient grounds for excluding
them from the category of stative verbs, in view of the overwhelming
evidence from the other criteria that they are in fact stative, and
we accordingly treat them as such. However, this also means that the
occurrence of BE+ING is not always predictable on the basis of stativity.
This seems a small price to pay compared with the advantage of predicting stativity on the basis of deep case relationships.
953
LEX - 23
The verbs in (ll), with one exception, are therefore
considered to be [+STATIVE] although sucn a feature will not
be marked in the lexicon since it is totally predictable.
The recalcitrant example is (llf). For convenience, we
repeat the example:
(ll)
(f)
The garden is swarming with bees.
This is clearly closely related to
(21)
Bees are swarming in the garden.
However, there are disagreements as to whether (llf) and (21)
are paraphrases. Those who argue that they are not synonymous
point to the difference in (22):
(22)
(a) The garden is swarming with people,
(b) *?People are swarming in the garden.
Those who reject (22b) claim that swarm in (21) is used in a
technical or literal sense, which is inappropriate for people,
whereas in (llf) and (22a) it is used in a metaphorical sense.
A similar distinction can be seen in
(23)
(a)
(b)
The cat was crawling with lice,
Lice were crawling on the cat.
However, if we consider bees, people and lice as agents in
these sentences regardless of whether they occur as surface
subjects or not the verbs are predictably [-STATIVE], which
is what we want. The difference between (llf) and (21) can
then be seen as a difference in topic focus, either involving
a slight change in meaning or setting up two different verbs,
though the latter view seems unnecessary.
We will now consider the examples in (12), namely, the
sentences with an animate subject which is putatively in the
dative case although the verb is in the progressive. These
examples caused considerable trouble at first because the
criteria for dative subjects are the same as those for stative
verbs; that is, we wish to say that an animate subject is in the
dative case if the verb does not require active voluntary
participation on the part of the subject. Thus see and hear
take dative subjects in contrast to look and listen, which
have agentive subjects. However, the presence of the progressive
in the sentences of (12) raised doubts about the validity
of the criteria involved and it was not at first clear whether
the notion of passive, involuntary participation outweighed
the use of the progressive, or vice versa. The discussion of
95U
LEX - 2k
the use of BE+ING in (ll), however, shows that there may be
an explanation for the apparent discrepancy between the two
sets of criteria, though the situation is considerably more
complicated than with the inanimate subjects.
In the first place, it is not always clear to what
extent mental states or activities are under voluntary control.
For example,
{2k)
(a) I forget his name.
(b) *I am forgetting his name.
(c) *He persuaded me to forget what had happened.
(d) He told me to forget what had happened.
(e) Forget it!
(f) I tried to forget it.
(g) ?He deliberately forgot to tell her.
In (2Ua-c) it is clear that the sense of forget is something
that is not under voluntary control, whereas in (2Ud-g) it
somehow is. In (2Ud-f) forget is roughly equivalent to ignore
and in (2Ug), if the sentence is acceptable, it is closer to
neglect. Thus, in the sentences of (2U) it is not so much
the basic meaning of forget which predicts the degree of voluntary control, it is rather the use of the verb which predicts
the meaning.
In the second place, Vendler distinguishes "achievements that start activities from achievements that initiate
a state" (1967:112). His illustration of the latter is when
someone who is trying to find the solution to a mathematical
problem suddenly shouts out Now I know it 8 Another example
of know used in an achievement sense might be
(25)
He told me to know the answer by tomorrow.
It is clear that know in (25) is roughly equivalent to learn
and it is interesting that in other languages this distinction
may be expressed by an aspectual difference rather than by lexical suppletion as is usually the case in English. In Spanish,
for example, lo sabia ayer means 'I knew it yesterday', whereas
lo supe ayer means *I found out about it yesterday'.
Accordingly, although it has not been possible to work
out the full implications of the decision, we are assuming
that stativity is predictable from the absence of an agent in
the sentence and that there are convincing explanations for
955
LEX - 25
the apparent exceptions. One of these explanations is that
verbs which normally take dative subjects may occasionally
be found with agentive subjects with a corresponding effect
on the semantic interpretation of the verb. Thus, know is
listed in the lexicon as taking a dative subject although in
(25) it takes an agent. This is similar to the manner in which
the count/non-count distinction may be overridden in a sentence.
For example, although butter is marked [-COUNT] it appears in
the following examples as a count noun:
(26)
(a)
(b)
This is a very fine butter.
Some butters are more expensive than others.
The fact that not everyone will accept the sentences in (26)
is not important. The point is that if they are acceptable
they must be interpreted in a count sense. Similarly, (25)
may not be acceptable to everyone but if it is acceptable it
requires an agentive subject for know. Consequently, we consider
the examples in (12) to have dative subjects.
Thus stative verbs such as know, believe, understand
have no agent in the case-frame and take the dative as subject.
Other verbs such as annoy, amuse, scare, frighten need two
entries, one with an agent where the action of the verb is done
"deliberately", the other without an agent where the action of
the verb "happens" without the deliberate intention of an agent.
The first is non-stative and the second stative:
(AGT)
(27) (a) John (deliberately) frightened Mary (by bursting
a balloon behind her back).
(AGT)
(b) John was frightening Mary (by bursting balloons
behind her back).
(INS)
(c) John (accidentally) frightened Mary (by opening
the door suddenly).
(INS)
(d) The noise frightened Mary.
(INS)
(e) *John was (accidentally) frightening Mary.
(INS)
(1)
(f) *The noise was frightening Mary.
(l)This sentence is, of course, perfectly grammatical in the sense
"Mary was growing more and more frightened because of the noise"
but it is ungrammatical if taken as parallel to "was frightening"
in (27b). This is another example of the complex relationship
between BE-ING and stativity.
956
LEX - 26
6.
Problems with Cases
As mentioned above there are six cases in the present
grammar but it is clear that more will be needed since there
are many sentences which cannot be generated by the grammar in
its present form. The number of cases that may ultimately be
required is uncertain for two reasons. The first is the
doubt as to the status of the category adverb and the relationships of such a putative category both inside and outside of
the proposition. The second reason for uncertainty is that the
addition of one case may have implications for the adoption or
exclusion of another. In the light of such doubts the following
discussion is purely exploratory.
Among the possible additional cases that have been suggested
are BEN(efactive), COM(itative), DEG(ree), MAN(ner), MEANS,
REF(erential), RES(ultative), SOU(rce) and TIME. For example,
(BEN)
(28) (a) I built a house for father
(COM)
(b) He brought a friend with him.
(DEO)
(c) He liked it extremely.
(MAN)
(d) The chancellor spoke threateningly.
(e)
(f)
(g)
(h)
(i)
He drained the water from the tank with a hose
(MEANS)
by sucking on it like a straw.
(REF)
She wouldn't tell us anything about the accident.
(RES)
He broke the chocolate bar into three pieces.
(SOU)
My mother taught me Russian from a book.
(TIME) "
The concert lasted for three hours.
For each of these cases, however, there is considerable uncertainty as to its scope and definition. As we have seen above,
there are problems even with dative, agent, instrument and locative, which are far from intuitively simple categories, but the
problems are multiplied with most of the cases illustrated in
(28). For instance, the Benefactive in (28a) can have at least
three different interpretations:
I built a house for father.
(i) He'll get the rent from it each
month, (for the benefit of)
(ii) His lumbago has been bothering him.
(in place of)
957
LEX - 27
(iii) We'll move him in on the first
of the month, (intended/
reserved for)
Moreover, (i) and (iii) might well be datives since the same
interpretations would apply to I built father a house. This
would simply mean that build had an idiosyncratic dative preposition since the sentence I built a house for myself could
not have this interpretation but only those of (i) and (iii).
In the absence of a clearer notion of Benefactive either inside
or outside the proposition we have chosen to exclude it from
the propositional frame and treat examples (i) and (iii) above
as datives.
A similar argument regarding the reflexive applies to
(28b) where the ungrammaticality of *he brought it with himself
shows that the Comitative is also outside of the proposition.
Moreover, there seem to be no verbs which would either obligatorily require or exclude such a case as a contextual feature.\
and thus no justification for including it within the frame.
The situation is quite different with regard to Manner
and Degree. Although there are no verbs which require such
cases ^2' there are many verbs which exclude them:
(29)
/John killed him
-\
\He died
/
) I heard a noise
(
LHe keeps it in his drawer^y
/'•completely
\ "utterly
\ *slightly
( "moderately
(30)
/He knows the answer
"^
\ She resides in Sacramento/
\ John is intelligent
7
I The room is empty
J
/"carefully^
\ "easily
I
\ *slowly
r
("freshly J
(1)
The situation is complicated by the fact that with him
in he brought it with him is pleonastic since the sense of
with is contained in bring.
(2)
This is an overstatement because of examples such as The
guards treated the prisoners badly/*The guards treated the
prisoners, but it is not clear how many verbs are like treat
in this respect.
958
LEX - 28
It seems likely that Manner and Degree should be included in the
propositional frame but lack of an analysis of adverbs outside
of the proposition has so far prevented us from incorporating
them.
Of the other cases illustrated in (28) Time certainly and
possibly Resultative are closely related to adverbials, while
Means is often difficult to distinguish from instrumental or a
third possibility which might be called Method. Although we
have investigated some of the possibilities we have not found
convincing arguments for the exclusion or inclusion of these
cases and we shall not discuss them further here. The remaining
two cases illustrated in (28) raise interesting problems.
The inclusion of Source as a case would affect the character
of the locative case. For example, a verb such as drain may
objectivalize the locative case or subjectivalize the neutral case:
(ACT)
(NEUT)
(LOC)
(31) (a) He drained the water from the tank.
(b) He drained the tank.
(c) The water drained from the tank.
However, there is also the possibility of an additional prepositional phrase which might be considered a second locative:
(d)
(e)
He drained the water from the tank into the
barrel.
The water drained from the tank into the barrel.
However, it is not possible to objectivalize this second locative:
(f) *He drained the barrel from the tank.
One solution would be to consider from the tank as Source and
into the barrel as the sole locative. One disadvantage of this
is that it loses the parallel with
(g)
The water in the tank drained into the barrel,
which seems much closer to (31e) than
(h)
The water from the tank drained into the barrel.
On the other hand, we could allow two locatives with [+DIRECTIONAL] verbs, one [+T0], the other [-TO]. This would help with all
transitive verbs that are "motional" in Gruber's sense:
(32)
(a)
(b)
He brought his old car from England to the United
States.
The Martians have sent a rocket from their planet
to the earth.
959
LEX - 29
It would also help with the distinction between locative and
dative in
(c)
He sent a letter from New York to London and
it got there in two days.
(d) He sent a letter from New York to his brother
(in London) and it got there in two days.
(e) He sent his brother a letter from New York.
(f) *He sent London a letter from New York.
The last sentence would be unstarred if London is an abbreviation
for our branch in London or some other entity with human associations , but then it could properly be treated as a dative. From
such examples it is not clear that there are grounds for setting
up a case such as Source. There are, however, examples of a
quite different sort, to which we now turn.
It is tempting to look to deep cases for the expression
of converse relations. For example, if John bought a car from
Peter implies Peter sold a car to John and vice versa, and similarly if John borrowed ten dollars from Bill implies Bill lent
ten dollars to John and vice versa, one way to express these
paraphrase relations would be if John, Peter and Bill were in
the same case in each of the pairs of sentences:
(ACT)
(NEUT) (DAT)
(33) (a) Peter sold a car to John.
(DAT)
(NEUT)
(AGT)
(b) John bought a car from Peter.
(AGT)
(NEUT)
(DAT)
(c) Bill lent ten dollars to John.
.
(d)
(DAT)
(NEUT)
(AGT)
John borrowed ten dollars from Bill.
In the first place, it is important to note that the verbs buy
and borrow are not [+STATIVE] and this would contradict the claim
that only verbs with agentive subjects are [-STATIVE]. Moreover,
if it were not for the converse relations there would seem no
good syntactic reason for considering the subjects of sentences
(33b) and (33d) as other than agents. In addition, the number
of lexical items which have strict converse relations of this
kind is fairly small and hardly Justifies the inclusion of such
a principle in the grammar. On the other hand, if we do not
adopt an analysis of this kind we are left without a suitable case
for from Peter in (33b) and from Bill in (33d). This could be
Source, if such a case were admitted into the proposition. At
the moment, we are rejecting the analysis which shows the converse
relations and we are also not yet clear enough about the nature
of the possible case Source to include it.
960
LEX - 30
Sentence (28f), which illustrates a possible Referential
case, also raises other interesting questions. For convenience,
we repeat the example:
(REF)
(28) (f) She wouldn't tell us anything about the accident.
In the first place, either the indefinite noun or the prepositional
phrase can be omitted but not both:
(31*)
(a) She wouldn't tell us anything.
(b) She wouldn't tell us about the accident.
(c) *She wouldn't tell us. (Only possible as a response
to a question.)
However, the indefinite and the prepositional phrase could also
appear alone:
(d)
Anything about the accident would interest them,
or with the other indefinites
(e)
(f)
Nothing about the accident appeared in the paper.
Something about the accident is bound to leak out.
and the prepositional phrase cannot appear with the verb if
the object is a definite pronoun:
(g) *She wouldn't tell it to us about the accident.
Thus only example (3Ub) suggests that the prepositional phrase
is a case on the verb; the other examples make it appear to
be a case on the indefinite noun. However, we have no clear
notion of the specific constraints that the dummy noun might
have, though it seems that there are some. To take another set
of examples:
(35)
(a) I read something by Chomsky.
(b) ?I ate something by Chomsky.
(c) Something by Chomsky was on the table.
(d) *Something to Chomsky arrived yesterday. (??)
(e) "Something into the city arrived yesterday.
(f) Something from Chomsky arrived yesterday.
(g) Something on the table must have caught her eye.
These could perhaps be analyzed as reduced relative
but then the problem is simply pushed one step back to that
the occurrence of prepositional phrases in the predicate of
copular sentence, which we have regarded elsewhere as cases
predicate noun, bringing the problem back full circle.
961
clauses,
of
a
on a
LEX - 31
The problems with the configuration of cases in the
proposition would be greatly reduced if we adopted a proposal
put forward by Matthews (1968). Matthews suggested that instead
of having the phrase structure component assign cases to the
several noun phrases introduced in the expansion of the proposition,
the proposition could be expanded into a verb followed by any
number of phrases (up to a certain number). Then the cases
would be assigned to the phrases by the particular choice of a
verb from the lexicon, which would be marked for the number of
phrases that may appear with it and the cases to be assigned
to each of them. This suggestion is attractive in many ways
and should be further explored but at the moment we retain a
Fillmore-type base with the cases assigned by the phrase
structure component.
A special problem has arisen with the extension of
case grammar to the structure of the NP, having to do with apparent
"inherent cases" of certain nouns.
In testing an early sample lexicon on Friedman's system,
it was noted at various stages that finding nouns of given types
(e.g. animate nouns, instrument nouns) that could take a wide
range of cases was sometimes difficult. The easiest cases are
deverbal abstract nouns such as shooting, destruction, etc.,
which generally take the same cases as the corresponding verb,
although instrument seems to be generally excluded from noun
complements. Other abstract nouns with a variety of cases are
also fairly easy to find (the sort for which Lakoff et al posit
an underlying verb, e.g. idea, novel, portrait). Animate nouns
which take agent are much harder to find than those which take
other cases; and this would seem readily explainable by the fact
that most animate nouns which take cases at all are themselves
"agent nouns". Thus owner, donor, and guide seem to be agent
nouns derived from corresponding verbs (Latin in one case), and
accordingly take all the cases of the corresponding verbs
except agent. Similarly for the non-deverbal king, father,
ambassador, nurse. The few found so far that allow agent seem
themselves to be datives, i.e. captive, victim, employee,
delegate. But emissary seems to allow both a dative and an agent.
The same problem appears from a different angle when we
note that author takes book as an object, while book takes author
as an agent. And triples such as employment, employer, employee;
lease, lessor, lessee clearly have a deverbal member, an agent
member, and a dative member. The problem thus boils down to
the fact that many of what we have thus far regarded as head nouns
are really case-marked themselves. One approach to a solution,
would be to have headless NP's with rules for inserting one
of the case-marked nouns into head position. This would obviously
require some pretty complex mechanisms. Perhaps something more
moderate could be worked out with redundancy rules. In any case,
the area needs and deserves much study. In our present grammar,
962
LEX - 32
no representation of these apparent "inherent cases" is given;
it is simply taken (incorrectly) as an accidental fact that
author takes an object, book an agent, etc.
It is clear from the foregoing that many problems remain
to be solved in specifying lexical entries and the sample
lexicon which follows makes no claims to do more than illustrate
some of the information which a more developed lexicon ought to
include, but as the other areas of the grammar are more fully
explored we hope to expand the lexicon and make it more representative than we have been able to do so far.
July 1969
963
LEX - 33
B.
Sample First Lexicon
Redundancy Rules
+ V
s + NEUT^
\ + DAT /
1 + LOC \
[-ESS]
) + INS \
K + AGT )
r+v i
[+NEUT -DAT -LOC -INS -AGT]
L+ ESSJ
[:3
r*i
[•EXTRA]
-I
[+ANIMATE]
L* HUMANJ
[+ N
ABSTRACT!
=>
PANIMATE
[-HUMAN
r+N
=>
l-INS NOUN]
U-LOC NOUN]
=>
-AGT NOUN
-DAT NOUN
i
[+ ANIMATEJ
[_ ANIMATEJ
|+ ART]
|_+ DEFJ
r:
+ ART
+ DEM
[-ATTACH]
=>
961.
+N-DEL
-WH
]
LEX - 3k
+ ART
- DEF
+ DEM
I+ATTACH]
[-N-DELJ
1+ ART]
L-
[-GEN]
DEFJ
r+
ART]
|_+ DEMJ
[-GEN]
-PRO
-INDET
-NEG
-PL
"+ ART "I
965
LEX - 35
"a/sm"
ADVOCATE
+
*
•
+
+
*
*
ART
DEF
DEM
SPEC
ATTACH
COUNT
V
ADJ
+NEUT -LOC -INS +AGT]
[
FACT
INDIC
WH-S
PASS
GER
ADMITJ
V
ADJ
ACCUSE
+ V
- ADJ
+NEUT +DAT -LOC -INS +AGT]
+ [
FACT
IMPEF
WH-S
PASS
GER
DAT -> OBJ
PREP NEUT of
V
ADJ
+NEUT -LOC -INS +AGT]
[
FACT
IMPER
WH-S
PASS
STAT-REDUCT
RAISE-TO-OBJ
V
ADJ
[
S
PASS
+NEUT -LOC -INS +AGT]
AFTER
+ PREP
+ TEMPORAL
- AFFECT
AIM
ACKNOWLEDGE'
V
ADJ
[
S
PASS
ADMIT/
+
+
*
ACKNOWLEDGE-1
+
+
*
+NEUT -LOC -INS +AGT]
[
FACT
IMPER
WH-S
PASS
STAT-REDUCT
RAISE-TO-OBJ
+
+
*
*
+NEUT -LOC -INS +AGT]
966
V
ADJ
-NEUT -DAT +LOC +AGT]
[
PASS
INS—5 OBJ
LEX - 36
ALL
AMUSED
+
+
+
+
QUANT
DIST
N-DEL
ATTACH
[[+DEF]
[f+SPEC]"
INTEGER
SHIFT
AMUSEMENT
N
-NEUT -LOC -AGT]
[
COMMON
COUNT
ABSTRACT
PREP INS at
ALWAYS
+
+
.
-
V
ADJ
[
-NEUT +DAT -LOC -AGT]
PREP INS at
ADV
TEMPORAL
DEF
SPEC
"and"
AMBASSADOR
+ CONJ
+ AND
N
[
-NEUT -DAT -INS -AGT]
COMMON
COUNT
HUMAN
ANGRY
+
+
+
+
+
AMUSEa
+
+
*
*
V
ADJ
[
PASS
EMOT
V
ADJ
-NEUT +DAT -LOC +INS -AGT]
[
DAT-^SUBJ
PREP INS at
-NEUT +DAT -LOC +INS -AGT]
ANNOUNCE
+
+
*
AMUSE1
V
ADJ
(
S
PASS
+NEUT -LOC -INS +AGT]
-NEUT +DAT -LOC +AGT]
ANNOUNCEMENT
N
-LOC -INS]
[
COMMON
COUNT
967
ABSTRACT
LEX - 37
ANNOY*
ANSWER
+ V
- ADJ
N
-LOC -INS]
[
FACT
IMPER
WH-S
COMMON
COUNT
ABSTRACT
[[+ ABSTRACT] OBJ]
PREP NEUT about
-NEUT +DAT -LOC +INS -AGT]
• PASS
* EMOT
ANNOYb
-NEUT +DAT -LOC +AGT]
ANTICIPATE
V
ADJ
[
ANNOYANCE
N
[
+NEUT -DAT -LOC
-INS +AGT]
FACT
IMPER
WH-S
PASS
GER
STAT-REDUCT
RAISE-TO-OBJ
[[+ABSTRACT] OBJ]
-NEUT -LOC -AGT]
COMMON
COUNT
ABSTRACT
PREP INS at
ANNOYED
+
+
+
+
ANXIOUS1
V
ADJ
[
-NEUT +DAT -LOC -AGT]
PREP INS at
+
+
+
+
+
ANSWER
+ V
ADJ
[
-LOC -INS +AGT]
FACT
IMPER
WH-S
PASS
[[+ABSTRACT] OBJ]
V
ADJ
[
+DAT -]
FACT
INDIC
WH-S
EMOT
EQUI-NP-DEL
ANXIOUS^
+
+
+
+
968
V
ADJ
[
+DAT -LOC -INS -AGT]
S
PREP NEUT about
LEX - 38
hNY
ARM
QUANT
DIST
ATTACH
N-DEL
[[+DEF]
[f+SPEC]"
INTEGER
SHIFT
+
+
+
+
-
N
[
-NEUT -LOC -INS -AGT]
COMMON
COUNT
ABSTRACT
ANIMATE
ARRIVE
APPEAR
+
+
*
+ V
- ADJ
+
[
- FACT
- S
V
ADJ
+NEUT -LOC -INS -AGT]
[
FACT
IMPER
WH-S
RAISE-SUBJ
+NEUT -DAT -INS -AGT]
ASIC
V
ADJ
APPRECIATE
+
+
*
+
V
ADJ
+NEUT +DAT -LOC -INS -AGT]
[_
S
PASS
[[+ABSTRACT] OBJ]
[
+NEUT -LOC -INS +AGT]
FACT
INDIC
PASS
EQUI-NP-DEL
DAT —>OBJ
PREP DAT of
AS?
APPRECIATION
V
ADJ
N
-LOC -INS -AGT]
I
S
COMMON
COUNT
ABSTRACT
[
+NEUT -LOC -INS +AGT]
FACT
S
PASS
DAT ->- OBJ
PREP NEUT for
969
LEX - 39
ASSUME
AVAILABLE
V
ADJ
+NEUT -DAT -LOC -INS +AGT]
+ [
FACT
IMPER
WH-S
PASS
STAT-REDUCT
RAISE-TO-OBJ
+
+
+
-
N
+NEUT -INS -AGT]
AVOID
V
ADJ
[
ASSUMPTION
[
V
ADJ
[
FACT
S
+NEUT -DAT -LOC
-INS +AGT]
FACT
IMPER
WH-S
AGT-IDENT
PASS
GER
EQUI-NP-DEL
AFFECT
-DAT -LOC -INS]
FACT
IMPER
WH-S
COMMON
COUNT
ABSTRACT
[[•ABSTRACT] OBJ]
AWARE
AT
+ V
+ ADJ
+ PREP
+ LOC
- DIR
+ [
+NEUT +DAT -LOC
-INS -AGT]
- s
AUTHOR
AWARENESS
N
[__
-DAT -LOC -INS -AGT]
N
FACT
S
COMMON
COUNT
HUMAN
-LOC -INS -AGT]
[_
S
COMMON
COUNT
ABSTRACT
970
LEX - kO
'be'
BEGINNING
+ V
- ADJ
+ [
N
-DAT -LOC -INS]
[
FACT
S
COMMON
COUNT
ABSTRACT
+ESS]
- FACT
- S
BEARER
BELIEF
N
[
-INS -AGT]
FACT
S
COMMON
COUNT
HUMAN
MASC
-LOC -INS -AGT]
FACT
IMPER
WH-S
COMMON
COUNT
ABSTRACT
PREP NEUT in
BEFORE
+ PREP
+ TEMPORAL
+ AFFECT
BELIEVE
+ V
- ADJ
+ [
BEGIN1
+ V
- ADJ
+NEUT -DAT -LOC -INS -AGT]
+ [
FACT
IMPER
WH-S
GER
RAISE-TO-SUBJ
•
+
*
FACT
IMPER
WH-S
PASS
STAT-REDUCT
RAISE-TO-OBJ
BETWEEN
+ PREP
+ LOC
- DIR
BEGIN*
+ V
- ADJ
• [
FACT
S
PASS
+NEUT -DAT -INS +AGT]
971
+NEUT +DAT -LOC
-INS -AGT]
LEX - Ul
BIG
+
+
+
-
BOY
V
ADJ
[
FACT
S
N
+NEUT -DAT -LOC -INS -AGT]
-NEUT -LOC -INS -ACT]
[
COMMON
COUNT
HUMAN
MASC
BODY
BREAK
N
PRO
ATTACH
HUMAN
MASC
COUNT
PLURAL
+ V
- ADJ
+ [
FACT
S
PASS
BOOK
BRIDGE
N
[
+NEUT -DAT -LOC]
+NEUT -DAT -LOC -INS +AGT]
FACT
S
COMMON
COUNT
ABSTRACT
ANIMATE
OBJ-DEL
PREP NEUT fabout
Ion
N
-NEUT -DAT -INS -AGT]
[
COMMON
COUNT
ABSTRACT
ANIMATE
PREP LOC over
"but"
+ CONJ
+ BUT
BOTH
QUANT
DIST
N-DEL
ATTACH
[[+DEF]
[[+SPEC]"
INTEGER
SHIFT
BUTTER
N
COMMON
COUNT
ABSTRACT
ANIMATE
972
LEX - 1*2
CAN
CAT
+ MODAL
+
+
-NEUT -LOC -INS -AGT]
+ COMMON
+ COUNT
HUMAN
MASC
CANONIZATION
-NEUT -LOC -INS]
COMMON
COUNT
ABSTRACT
CERTAIN
V
ADJ
[
CANONIZE
+ V
- ADJ
+ [
PASS
-NEUT +DAT -LOC -INS +AGT]
*
+NEUT -DAT -LOC -INS
-AGT]
FACT
IMPER
WH-S
RAISE-TO-SUBJ
CERTAIN
CAREFUL
+ V
+ ADJ
• [
V
ADJ
[
-DAT -LOC -INS +AGT]
FACT
IMPER
EQUI-NP-DEL
•
+NEUT +DAT -LOC -INS
-AGT]
FACT
IMPER
WH-S
GER
CAREFUL
CHAIR
+
+
+
+
V
ADJ
-DAT -LOC -INS +AGT]
[
S
PREP NEUT about
N
COMMON
COUNT
ABSTRACT
ANIMATE
973
LEX - U3
CHILD
CHAMPION
N
N
[_
[_
-DAT -INS -AGT]
-NEUT -LOC -INS -AGT]
COMMON
COUNT
HUMAN
MASC
FACT
S
COMMON
COUNT
HUMAN
MASC
COME
+ V
- ADJ
+ [
CHANCE
+
+
*
*
+
+
+
+
+
N
-LOC -INS -AGT]
[
FACT
IMPER
WH-S
EMOT
GER
EQUI-NP-DEL
COMMON
COUNT
ABSTRACT
PREP NEUT of
-NEUT -DAT -LOC
-INS +AGT]
COMMAND1
+
+
*
+
V
ADJ
[
+NEUT -LOC -INS +AGT]
FACT
INDIC
WH-S
PASS
EQUI-NP-DEL
CHIEF
N
[
-NEUT -LOC -INS -AGT]
COMMAND
COMMON
COUNT
HUMAN
MASC
+NEUT -DAT -LOC
-INS +AGT]
- FACT
- S
* PASS
CHIEF
+ CHIEF
9lh
LEX - kk
CONFIDENT
CONTINUE
V
ADJ
[
V
ADJ
[
+NEUT +DAT -LOC -INS -AGT]
FACT
IMPER
WH-S
GER
FACT
IMPER
WH-S
AGT-INDENT
PASS
GER
EQUI-NP-DEL
CONSIDERl
V
ADJ
-NEUT +DAT -LOC -INS -AGT]
+ [
FACT
IMPER
WH-S
PASS
STAT-REDUCT
TO-BE-DEL
COOKa
+
+
-
V
ADJ
[
+NEUT -DAT -INS -AGT]
FACT
- S
+ [[-ABSTRACT] OBJ]
RAISE-TO-OBJ
COOKb
CONSIDER
V
ADJ
[
+NEUT -DAT -LOC
-INS +AGT]
V
ADJ
+NEUT -DAT -LOC -INS +AGT]
+
FACT
WH-S
PASS
GER
+NEUT -DAT +AGT]
[
FACT
S
PASS
OBJ-DEL
[[-ABSTRACT] OBJ]
CONTAIN
COVER
V
ADJ
+NEUT -DAT +LOC -INS -AGT]
+ [
FACT
S
LOC- SUBJ
975
+
+
*
+
V
ADJ
-NEUT -DAT +LOC +INS]
[
FACT
S
PASS
[[-ABSTRACT] OBJ]
LEX - 1+5
CROSS
DEDUCE
+ V
- ADJ
-NEUT -DAT +LOC -INS +AGT]
+ [
* PASS
+ LOC-•OBJ
[across]
V
ADJ
+
[
+NEUT +DAT -LOC
-INS -AGT]
FACT
IMPER
PASS
STAT-REDUCT
RAISE-TO-OBJ
DANGEROUS
V
ADJ
DEMAND
[
+NEUT +DAT -LOC -INS -AGT]
FACT
IMPER
WH-S
EMOT
AFFECT
V
ADJ
+NEUT -LOC -INS +AGT]
+ [
FACT
INDIC
WH-S
AGT-IDENT
PASS
EQUI-NP-DEL
PREP DAT of
DAUGHTER
N
[
-NEUT +DAT -LOC -INS -AGT]
COMMON
COUNT
HUMAN
MASC
DEMAND
N
DEAD
+ V
+ ADJ
+ [
-NEUT +DAT -LOC -INS -AGT]
-LOC -INS]
[
FACT
INDIC
WH-S
AGT IDENT
EQUI-NP-DEL
COMMON
COUNT
ABSTRACT
DEATH
DEMOLISH
N
[__ -NEUT -LOC -INS -AGT]
COMMON
COUNT
ABSTRACT
V
ADJ
+ [
FACT
S
PASS
976
+NEUT -DAT -LOC +AGT]
LEX - U6
DIFFICULT"1
DEPLORABLE
+ V
+ ADJ
+ [
V
ADJ
[
-NEUT -LOC +INS -AGT]
EMOT
+NEUT -DAT -LOC
-INS -AGT]
FACT
IMPER
WH-S
EMOT
RAISE-OBJ-TO-SUBJ
+ AFFECT
DEPLORE
+ V
- ADJ
[
DIFFICULT'
+NEUT +DAT -LOC -INS -AGT]
+ V
+ ADJ
S
PASS
EMOT
+ [
-DAT -LOC -INS +AGT]
- S
+ AFFECT
+ PREP NEUT about
DESTROY
V
ADJ
+ [
FACT
S
PASS
DISCOVER
+NEUT -DAT -LOC +AGT]
+ V
- ADJ
+ [
FACT
IMPER
PASS
STAT-REDUCT
RAISE-TO-OBJ
DESTRUCTION
-DAT -LOC]
FACT
S
COMMON
COUNT
ABSTRACT
DOG
N
[
COMMON
COUNT
HUMAN
MASC
DIE
+ V
- ADJ
+ [
+NEUT -DAT -LOC
-INS +AGT]
-NEUT +DAT -LOC -INS -AGT]
977
-NEUT -LOC -INS -AGT]
LEX - 1*7
DOUBT
DREAM
V
ADJ
+ V
- ADJ
+ [
+
+
*
+
[
+NEUT +DAT -LOC -INS -AGT]
FACT
IMPER
WH-S
PASS
AFFECT
*
+
+NEUT +DAT -LOC -INS
-AGT]
FACT
IMPER
WH-S
PASS
PREP NEUT about
DOUBTFUL
EACH
V
ADJ
[
+NEUT +DAT -LOC -INS -AGT]
FACT
IMPER
WH-S
GER
PREP NEUT about
V
ADJ
[
FACT
QUANT
DIST
ATTACH
N-DEL
[
[+PL]]
[
[-COUNT]]
- [T+DEF]
- [[+SPEC]
- INTEGER
+ SHIFT
DRAIN£
+
+
-
+
+
*
-
]
]
EAGER
-DAT +LOC +AGT]
V
ADJ
- S
* PASS
* LOC—»OBJ
[from]
+ [[-ABSTRACT] OBJ]
+ [[-ANIMATE] OBJ]
+DAT -LOC -INS -AGT]
[
FACT
INDIC
WH-S
EMOT
EQUI-NP-DEL
DRAIN
EAGERNESS
+
+
-
V
ADJ
[
FACT
S
-DAT +LOC -INS -AGT]
978
+
+
+
+
+
+
+
N
[
-DAT -LOC -INS]
FACT
INDIC
WH-S
EMOT
EQUI-NP-DEL
COMMON
COUNT
ABSTRACT
PREP NEUT for
LEX - U8
EARLY
EITHER
+ ADV
+ TEMPORAL
QUANT
DIST
ATTACH
N-DEL
[
[+PL]]
[
[-COUNT]]
[T+DEF]
]
[[+SPEC] _ * ]
INTEGER
SHIFT
EASINESS
-DAT -LOC -INS -ACT]
FACT
S
COMMON
COUNT
ABSTRACT
ELAPSE
+ V
- ADJ
+ [
EASY
V
ADJ
[
+NEUT -DAT -LOC -INS -AGT]
FACT
INDIC
WH-S
EMOT
RAISE-OBJ-TO-SUBJ
- FACT
- S
ELECT
V
ADJ
[
EAT
+
+
-
V
ADJ
[
FACT
+NEUT -DAT -LOC
-INS -AGT]
+NEUT +DAT -LOC
-INS +AGT]
FACT
IMPER
WH-S
PASS
STAT-REDUCT
TO-BE-DEL
RAISE-TO-OBJ
+NEUT -DAT -LOC +AGT]
- S
* PASS
* OBJ-DEL
+ [[-ABSTRACT] OBJ]
ELECTION
+ N
+ [_ -NEUT -LOC -INS]
+ COMMON
+ COUNT
+ ABSTRACT
979
LEX - U9
EMISSARY
EMPLOYMENT
N
L_
N
-NEUT -LOC -INS]
-NEUT -LOC -INS]
[
COMMON
COUNT
ABSTRACT
COMMON
COUNT
HUMAN
MASC
EMPTY
EMPHASIZE
+
+
•
V
ADJ
[
S
PASS
+
+
-
+NEUT -LOC -INS +AGT]
-DAT +LOC +AGT]
- S
* PASS
* LOC—>OBJ
[from]
+ [[-ABSTRACT] OBJ]
+ [[-ANIMATE] OBJ]
EMPLOY
+ V
- ADJ
+ [
PASS
V
ADJ
[
FACT
-NEUT +DAT -LOC -INS +AGT]
EMPTY1
+ V
- ADJ
+ [
EMPLOYEE
-NEUT -DAT +LOC -INS
-AGT]
N
[,
-NEUT -DAT -LOC -INS]
COMMON
COUNT
HUMAN
MASC
ENJOY
+ V
- ADJ
+ [
*
+
EMPLOYER
N
[__
-NEUT -LOC -INS -ACT]
COMMON
COUNT
HUMAN
MASC
980
FACT
IMPER
WH-S
PASS
GER
+NEUT +DAT -LOC -INS
-AGT]
LEX - 50
ENTER
EXPECT*1
+ V
- ADJ
-NEUT -DAT -INS +AGT]
+ [
* PASS
+ L0C- • OBJ
[in]
+ V
- ADJ
+ [
•
+
+
+NEUT -LOC -INS +AGT]
FACT
INDIC
WH-S
PASS
EQUI-NP-DEL
PREP DAT of
ENTRANCE
+ N
EXPLAIN*
-NEUT -DAT -INS -ACT]
• [
+ COMMON
+ COUNT
+ V
- ADJ
+ [
- ABSTRACT
- ANIMATE
+NEUT -LOC -INS +AGT]
- FACT
- IMPER
* PASS
EVERY
+
+
+
-
QUANT
DIST
ATTACH
N-DEL
[
[+PL]]
[
[-COUNT]]
[l+DEF]
]
[[+SPEC]
]
INTEGER
SHIFT
EXPLAIN"
+
•
-
V
ADJ
[
+NEUT -LOC +INS -ACT]
FACT
IMPER
• PASS
FACT
1
EXPECT
V
ADJ
+NEUT -DAT -LOC -INS +AGT]
+ [
FACT
IMPER
WH-S
PASS
FUT-REDUCT
RAISE-TO-OBJ
981
N
-DAT -LOC -INS -ACT]
[
FACT
IMPER
WH-S
COMMON
COUNT
ABSTRACT
LEX - 51
FEEL3
FAMILIAR
V
ADJ
[
+NEUT -LOC -INS -AGT]
FACT
S
PREP NEUT with
V
ADJ
+ [
FACT
S
PASS
FATHER
N
[
+NEUT -DAT -LOC +AGT]
"few/little"
-NEUT +DAT -LOC -INS -AGT]
COMMON
COUNT
HUMAN
MASC
+
+
#
QUANT
DIST
ATTACH
N-DEL
[_
[+PL]]
[-COUNT]]
[
[l+DEF]
[[-DEF]
FEELX
+ V
- ADJ
+NEUT +DAT -LOC -INS -AGT]
+ [
FACT
IMPER
WH-S
PASS
STAT-REDUCT
RAISE-TO-OBJ
FIDO
+
+
N
COMMON
HUMAN
MASC
FIFTH
+ ORD
FEEL
V
ADJ
[
FILL""
+NEUT +DAT -LOC -INS -AGT]
FACT
IMPER
WH-S
PASS
TO-DEL
RAISE-TO-OBJ
+ V
- ADJ
+ [
-DAT +LOC +AGT]
- FACT
- S
* PASS
+ LOC—»-OBJ,NEUT
[into]
[with]
+ [[-ABSTRACT] OBJ]
+ [[-ANIMATE] OBJ]
982
LEX - 52
FILLb
FONDNESS
• V
- ADJ
+ [
-DAT +LOC • -INS -AGT]
- FACT
- S
* LOC—K>BJtNEUT
[into]
[with]
+ LOC-*SUBJ
+
+
+
+
+
+
N
[
-LOC -11
FACT
IMPER
WH-S
GER
COMMON
COUNT
ABSTRACT
PREP NEUT for
FINISH
V
ADJ
+NEUT -DAT -LOC -INS +AGT]
[
FACT
IMPER
WH-S
AGT-IDENT
PASS
GER
FIRST
+ ORD
FORCE
+
+
*
+
+
V
ADJ
[
+NEUT + DAT -LOC
+ AGT]
FACT
INDIC
WH-S
PASS
EQUI-NP-DEL
DAT—>OBJ
FIVE
1
FORGET
QUANT
DIST
ATTACH
N-DEL
[
[+PL]]
[-COUNT]
[_
+ INTEGER
V
ADJ
+ [
FACT
IMPER
AGT-IDENT
PASS
+ EQUI-NP-DEL
+ AFFECT
+
*
FOND
V
ADJ
[
FACT
IMPER
WH-S
+ GER
+NEUT +DAT -LOC
-INS • -AGT]
+NEUT +DAT -LOC -INS -AGT]
983
LEX - 53
GENEROUS
FORGET
V
ADJ
V
ADJ
[_
+NEUT +DAT -LOC -INS -AGT]
S
• PASS
+ AFFECT
+ PREP NEUT about
•LOC -INS +AGT]
[
FACT
S
PREP NEUT with
GIFT
FOUR
N
+
+
+
+
-LOC -INS +AGT]
[
FACT
S
COMMON
COUNT
ABSTRACT
PREP AGT from
QUANT
DIST
ATTACH
N-DEL
[
[+PL]]
[
[-COUNT]]
INTEGER
GIRL
FOURTH
N
+ ORD
[__
-NEUT -LOC -INS -AGT]
COMMON
COUNT
HUMAN
MASC
FROM
+ PREP
+ LOC
+ DIR
GIVE
V
ADJ
[
FULL
-NEUT -DAT +LOC -INS -AGT]
98U
+NEUT +DAT -LOC
-INS +AGT]
FACT
S
PASS
DAT—» OBJ
LEX - 5U
GO
HAPPEN'
+
+
-
-NEUT -DAT -INS +AGT]
V
ADJ
[
FACT
S
+NEUT -INS -AGT]
GRASP1
+
+
*
+
V
ADJ
[
+NEUT -DAT -LOC +AGT]
FACT
S
PASS
[[-ABSTRACT] OBJ]
HARD
+ ADV
+ MANNER
HARD
+ V
+ ADJ
+ [
GRASP
+
+
*
+
V
ADJ
[
+NEUT +DAT -LOC -INS -AGT]
S
PASS
[[+ABSTRACT] OBJ]
FACT
IMPER
WH-S
EMOT
RAISE-OBJ-TO-SUBJ
HAT
GUILTY
+
+
+
+
+
*
+NEUT -DAT -LOC
-INS -AGT]
V
ADJ
[
+NEUT +DAT -LOC -INS -AGT]
FACT
IMPER
WH-S
GER
N
COMMON
COUNT
ABSTRACT
ANIMATE
HATEJ
HAPPEN1
+
+
*
V
ADJ
[
+NEUT -DAT -LOC -INS -AGT]
FACT
IMPER
WH-S
RAISE-TO-SUBJ
985
+ V
- ADJ
+ [
*
+
*
+
+
+NEUT +DAT -LOC
-INS -AGT]
FACT
IMPER
WH-S
PASS
EMOT
GER
EQUI-NP-DEL
AFFECT
LEX - 55
HATE£
HELP
+NEUT +DAT -LOC -INS -ACT]
V
ADJ
+ [
+DAT -LOC +AGT]
FACT
IMPER
WH-S
PASS
TO-DEL
EQUI-NP-DEL
HAVE
+
+
-
V
ADJ
[
FACT
S
HERE
+NEUT +DAT -LOC -INS -AGT]
+ ADJ
+ LOC
- FAR
HEAD
HIT1
+
+
+
+
-
+
+
*
+
N
[
-NEUT -LOC -INS -AGT]
COMMON
COUNT
ABSTRACT
ANIMATE
V
ADJ
[
-NEUT -DAT +LOC +AGT]
PASS
[[-ABSTRACT] OBJ]
HITb
HEAR
+
+
+NEUT +DAT -LOC -INS -AGT]
V
ADJ
[
-NEUT -DAT +LOC
+INS • -AGT]
* PASS
+ [[-ABSTRACT] OBJ]
986
LEX - 56
IGNORANCE
HOPE
N
V
ADJ
[
+NEUT +DAT -LOC -INS -ACT]
FACT
IMPER
WH-S
PASS
EQUI-NP-DEL
PREP NEUT for
-LOC -INS -AGT]
[_
S
COMMON
COUNT
ABSTRACT
AFFECT
IGNORE
+ V
- ADJ
+ [
HOPE
N
[
+NEUT -DAT -LOC
-INS +AGT]
-LOC -INS -AGT]
- S
* PASS
+ AFFECT
FACT
IMPER
WH-S
GER
COMMON
COUNT
ABSTRACT
IMAGINE
V
ADJ
[
HORSE
FACT
IMPER
WH-S
PASS
GER
STAT-REDUCT
RAISE-TO-OBJ
N
[
+NEUT +DAT -LOC
-INS -AGT]
-NEUT -LOC -INS -AGT]
COMMON
COUNT
HUMAN
IMPORTANT
ART
DEF
DEM
GEN
COUNT
I
II
III
PL
+ V
+ ADJ
+ [
EMOT
987
-NEUT -LOC +INS -AGT]
LEX - 57
IMPRISON
INQUIRE
+ V
- ADJ
+ [
PASS
+ V
- ADJ
+NEUT -LOC -INS +AGT]
+ [
FACT
INDIC
IMPER
PASS
PREP DAT of
-NEUT +DAT +AGT]
IMPRISONMENT
N
-NEUT]
[
COMMON
COUNT
ABSTRACT
INSIST-1
+ V
- ADJ
• [
+NEUT -DAT -LOC
-INS +AGT]
FACT
INDIC
WH-S
PASS
GER
PREP NEUT on
IN
+ PREP
+ LOC
- DIR
INNER
INSIST
+ CHIEF
+ V
- ADJ
+ [
INFORM
+ V
- ADJ
+NEUT +DAT -LOC -INS +AGT]
+ [
FACT
IMPER
WH-S
PASS
DAT—*OBJ
PREP NEUT of
988
*
+NEUT -DAT -LOC
-INS +AGT]
FACT
IMPER
WH-S
PASS
INSISTENCE
N
-DAT -LOC -INS]
[
FACT
WH-S
GER
COMMON
COUNT
ABSTRACT
PREP NEUT on
LEX - 58
INSULT
+ V
- ADJ
+ [
PASS
INTEREST
N
-LOC -INS -AGT]
[
FACT
IMPER
WH-S
GER
COMMON
COUNT
ABSTRACT
PREP NEUT in
-NEUT +DAT -LOC -INS +AGT]
INSULT
N
-NEUT -LOC -INS +AGT]
t
COMMON
COUNT
ABSTRACT
PREP AGT from
INTERESTED
+ V
+ ADJ
+ [
INTEND
+ V
- ADJ
+NEUT +DAT -LOC -INS -AGT]
+ [
FACT
INDIC
WH-S
PASS
EMOT
GER
EQUI-NP-DEL
INTENTION
+
+
FACT
IMPER
GER
PREP NEUT in
INTIMATE
+
+
*
V
ADJ
+NEUT -LOC -INS +AGT]
[
FACT
IMPER
PASS
+ N
-LOC -INS -AGT]
FACT
INDIC
WH-S
EMOT
EQUI-NP-DEL
COMMON
COUNT
ABSTRACT
• [
*
+
+
+
+
INTO
+ PREP
+ LOC
+ DIR
989
+NEUT +DAT -LOC
-INS -AGT]
LEX - 59
KEY
JOHN
+
+
+
N
N
COMMON
HUMAN
MASC
-NEUT -DAT -INS -AGT]
[
COMMON
COUNT
ABSTRACT
ANIMATE
KEEN
V
ADJ
KILL*
+NEUT +DAT -LOC -INS -AGT]
[
FACT
INDIC
WH-S
GER
EQUI-NP-DEL
PREP NEUT on
+
+
*
V
ADJ
[
FACT
S
PASS
-NEUT +DAT -LOC +AGT]
KILL'
KEEP1
+
+
*
V
ADJ
[
FACT
S
PASS
+ V
- ADJ
+ [
+NEUT -DAT -LOC -INS +AGT]
- FACT
- S
* PASS
KEEP
KILLING
+ V
- ADJ
+
+
+
+
+ [
-NEUT +DAT -LOC
+INS -AGT]
+NEUT -DAT +LOC -INS +AGT]
- FACT
- S
* PASS
N
[
-NEUT -LOC]
COMMON
COUNT
ABSTRACT
KING
KENNEL
N
[
N
COMMON
COUNT
ABSTRACT
ANIMATE
COMMON
COUNT
HUMAN
MASC
990
-DAT -LOC -INS -AGT]
LEX - 60
LAST
LEG
+ ORD
+
•
+
+
-
LATE
N
[
-NEUT -LOC -INS -AGT]
COMMON
COUNT
ABSTRACT
ANIMATE
+ ADV
+ TEMPORAL
LET
LAUGH
V
ADJ
[
-DAT -LOC -INS +AGT]
+NEUT +DAT -LOC
-INS +AGT]
FACT
INDIC
WH-S
TO-DEL
EQUI-NP-DEL
DAT—> OBJ
LEARN
LETHAL
+ v
•
•
*
•
+ V
+ ADJ
+ [
ADJ
[
-DAT -LOC -INS +AGT]
FACT
IMPER
AGT-IDENT
PASS
EQUI-NP-DEL
1
LIKE
+ V
- ADJ
+ [
LEARN2
+
+
*
-NEUT -LOC +INS -AGT]
V
ADJ
[
+NEUT +DAT -LOC -INS -AGT]
FACT
IMPER
PASS
991
*
+
*
+
+NEUT +DAT -LOC
-INS -AGT]
FACT
IMPER
WH-S
PASS
EMOT
GER
EQUI-NP-DEL
LEX - 61
LIKE'
MAIN
+
+
*
+
+ CHIEF
V
ADJ
+NEUT +DAT -LOC -INS -AGT]
[
S
PASS
EXTRA
LIKELY
+
+
+
*
V
ADJ
[
+NEUT -DAT -LOC -INS -AGT]
FACT
IMPER
WH-S
RAISE-TO-SUBJ
LISTEN
+
+
-
V
ADJ
[
FACT
+NEUT -DAT -LOC +AGT]
MAKE
+
+
*
V
ADJ
[
FACT
S
PASS
+NEUT -DAT +AGT
2
MAKE
+
+
*
+
+
+
V
ADJ
+NEUT +DAT -LOC
[
FACT
INDIC
WH-S
PASS
TO-DEL
EQUI- NP-DEL
DAT—>OBJ
s
PASS
PREP NEUT to
MARE
N
-NEUT -LOC -INS -AGT]
[
COMMON
COUNT
HUMAN
MASC
LOOK
+
+
*
+
V
ADJ
+NEUT -DAT -LOC +AGT]
[
FACT
S
PASS
PREP NEUT at
MARK*
+
+
*
+
LOWER
+ CHIEF
992
V
ADJ
-NEUT -DAT +LOC +AGT]
[
PASS
LOC- -OBJ
[on]
LEX - 62
MARK1
MOVE
+
+
»
+
+
+
*
V
ADJ
[
-NEUT -DAT +LOC +INS -AGT]
PASS
LOC—»OBJ
[on]
V
ADJ
[
FACT
S
PASS
+NEUT -DAT]
MARY
"much/many"
+
+
-
+
+
*
*
»
*
N
COMMON
HUMAN
MASC
MAY
QUANT
DIST
ATTACH
N-DEL
[
[+PL]]
[
[-COUNT]]
[T+DEF]
]
[f-DEF]
]
+ MODAL
MURDER
+ V
- ADJ
+ [
PASS
MILK
+
+
-
N
COMMON
COUNT
ABSTRACT
ANIMATE
-NEUT +DAT -LOC +AGT]
MURDER
MOTHER
N
[
-NEUT +DAT -LOC -INS -AGT]
COMMON
COUNT
HUMAN
MASC
+
+
+
+
+
N
-NEUT -LOC]
[
COMMON
COUNT
ABSTRACT
MUST
+ MODAL
993
LEX - 63
NEAR
OLD
+ PREP
+ LOC
- DIR
+ CHIEF
ON
NEXT
+ ORD
+ PREP
+ LOC
- DIR
NOISY*
ONE
+ V
+ ADJ
+ [
QUANT
DIST
ATTACH
N-DEL
[
[+PL]]
[-COUNT]]
[
- INTEGER
-NEUT -DAT -LOC -INS +AGT]
NOISY1
+ V
+ ADJ
+ [
-NEUT -DAT +LOC -INS -AGT]
ONE
+ ADV
+ TEMPORAL
- FAR
+
+
+
+
*
+
-
OFTEN
ONE
NOW
N
PRO
ATTACH
HUMAN
MASC
COUNT
PLURAL
ADV
FREQ
OLD
+
+
•
-
V
ADJ
[
FACT
S
N
PRO
ATTACH
HUMAN
MASC
COUNT
PLURAL
+NEUT -DAT -LOC -INS -AGT]
99»»
LEX - 6k
ONTO
OUT OF
+ PREP
+ LOC
+ DIP.
+ PREP
+ LOC
+ DIR
OPEN
OUTER
+ CHIEF
_ +NEUT -DAT -LOC]
FACT
S
PASS
[[-ABSTRACT] OBJ]
[[-ANIMATE] OBJ]
OWN
+ V
- ADJ
+ [
*
+
or
+ CONJ
+ OR
+NEUT +DAT -LOC
-INS -ACT]
FACT
S
PASS
[[-ABSTRACT] OBJ]
PASS1
ORDER
+
+
-
+ V
- ADJ
+NEUT -LOC -INS +AGT]
+ [
FACT
INDIC
WH-S
DAT-IDENT
PASS
EQUI-NP-DEL
V
ADJ
[
+NEUT -LOC +AGT]
FACT
- S
* PASS
• DAT-4 OBJ
+ [[-ABSTRACT] OBJ]
PASS2
+
+
*
*
V
ADJ
[
-NEUT -DAT -INS +AGT]
PASS
LOC—>OBJ
[by]
+ [[-ABSTRACT] OBJ]
ORDER
N
-LOC -INS]
[
FACT
INDIC
WH-S
DAT-IDENT
EQUI-NP-DEL
COMMON
COUNT
ABSTRACT
995
LEX - 65
PASS+
+
+
PLACE
V
ADJ
[
+NEUT -DAT -LOC -INS]
FACT
S
[[+ABSTRACT] OBJ]
N
PRO
ATTACH
HUMAN
COUNT
PLURAL
PAW
POOR
+
•
+
+
-
+ V
+ ADJ
N
-NEUT -LOC -INS -AGT]
[
COMMON
COUNT
ABSTRACT
ANIMATE
+ [
-NEUT +DAT -LOC
-INS -AGT]
POOR
PERF
HAVE
+ CHIEF
EN
PORTRAIT
+ N
+ [
PERPETRATOR
N
[__
+
+
*
-DAT -LOC -INS -AGT]
COMMON
COUNT
HUMAN
MASC
-NEUT +DAT -LOC
-INS +AGT]
COMMON
COUNT
ABSTRACT
ANIMATE
OBJ-DEL
PREFER
PERSUADE
+ V
- ADJ
+ [
V
ADJ
[
+NEUT +DAT -LOC +AGT]
FACT
INDIC
WH-S
DAT-IDENT
PASS
EQUI-NP-DEL
+NEUT +DAT -LOC
-INS -AGT]
FACT
INDIC
WH-S
PASS
EMOT
GER
EQUI-NP-DEL
996
LEX - 66
PREFERABLE
PROBABLE
V
ADJ
[
+NEUT -LOC -INS -ACT]
+NEUT -DAT -LOC
-INS -AGT]
FACT
IMPER
WH-S
PREFERENCE
PROG
N
-LOC -INS -AGT]
[
FACT
INDIC
WH-S
GER
COMMON
COUNT
ABSTRACT
PREP NEUT for
BE
ING
PROMOTE
+ V
- ADJ
+ [
-NEUT +DAT -LOC
-INS +AGT]
« PASS
PRIDE
+ N
+ [_
+
+
+
PROOF
-LOC -INS -AGT]
S
COMMON
COUNT
ABSTRACT
PREP NEUT in
N
-DAT -LOC -INS]
[
FACT
IMPER
WH-S
COMMON
COUNT
ABSTRACT
ANIMATE
PRINCIPAL
+ CHIEF
PROUD
+ V
+ ADJ
+ [
- S
997
+NEUT +DAT -LOC
-INS -AGT]
LEX - 67
PROVE6
REBUKE
N
V
ADJ
[
-NEUT -LOC -INS +AGT]
[
COMMON
COUNT
ABSTRACT
PREP AGT from
+NEUT -LOC +AGT]
FACT
IMPER
WH-S
PASS
STAT-REDUCT
TO-BE-DEL
RAISE-TO-OBJ
RECEIVE
V
ADJ
PROVE1
[
+NEUT +DAT -INS -AGT]
FACT
S
PASS
+ V
- ADJ
+NEUT -LOC +INS -AGT]
+ [
FACT
IMPER
WH-S
PASS
STAT-REDUCT
TO-BE-DEL
RAISE-TO-OBJ
RECEIVE
+ V
- ADJ
+ [
-NEUT +DAT -LOC
-INS +AGT]
* PASS
RAIN
+ V
- ADJ
+ [
REFUSAL
-NEUT -DAT -LOC -INS -AGT
-ESS]
REBUKE
+ V
- ADJ
+ [
-NEUT +DAT -LOC -INS +AGT]
* PASS
998
+
+
+
+
+
+
+
+
N
[
-DAT
FACT
IMPER
WH-S
AGT-IDENT
EQUI-NP-DEL
COMMON
COUNT
ABSTRACT
AFFECT
LEX - 68
REFUSE
REMEMBER
+
+
+
*
+
+
+ V
- ADJ
• [
V
ADJ
+NEUT -DAT -LOC -INS +AGT]
[
FACT
IMPER
WH-S
AGT-IDENT
PASS
EQUI-NP-DEL
AFFECT
*
•
+NEUT +DAT -LOC
-INS -ACT]
FACT
IMPER
PASS
EQUI-NP-DEL
REPUTE
+ V
+ [
REGRET
+
+
*
+
V
ADJ
+NEUT +DAT -LOC -INS -AGT]
[
S
PASS
AFFECT
+
+
+
S
IMPER
WH-S
PASS
RAISE-TO-SUBJ
RELY
REQUIRE
+
+
*
+
+
•
•
*
+
V
ADJ
[
+NEUT -DAT -LOC -INS +AGT]
S
PASS
PREP NEUT (up)on
+NEUT -DAT -LOC
-INS +AGT]
V
ADJ
+NEUT -LOC -INS +AGT
[
FACT
INDIC
WH-S
PASS
EQUI- NP-DEL
PREP DAT of
REMEMBER
+
+
*
V
ADJ
[
S
PASS
RESENT
+NEUT +DAT -LOC -INS -AGT]
+ V
- ADJ
• [
- S
* PASS
+ AFFECT
999
+NEUT +DAT -LOC
-INS -AGT]
LEX - 69
RESENTMENT
SCAREa
+
+
+
+
+
+
+
+
*
N
-LOC -INS -AGT]
[
S
COMMON
COUNT
ABSTRACT
AFFECT
PREP NEUT at
-NEUT +DAT -LOC +AGT]
SCARE1
+ V
- ADJ
+ [
RESIDE
+ V
- ADJ
+ [
V
ADJ
[
PASS
-NEUT +DAT -LOC
+INS -AGT]
* PASS
* EMOT
-NEUT -DAT +LOC -INS +AGT]
SECOND
RUMOR
+ ORD
V
+NEUT -DAT -LOC -INS +AGT]
s
SEE
IMPER
WH-S
+ PASS
+ V
- ADJ
+ [
RUN
+ V
- ADJ
+ [
-NEUT -DAT -INS +AGT]
SAY
*
+
•
+NEUT +DAT -LOC
-INS -AGT]
FACT
IMPER
PASS
TO-DEL
RAISE-TO-OBJ
SEEM
V
ADJ
+NEUT -LOC -INS +AGT]
[
FACT
IMPER
WH-S
PASS
STAT-REDUCT
RAISE-TO-OBJ
1000
V
ADJ
+NEUT -LOC -INS -AGT]
+ [
FACT
IMPER
WH-S
RAISE-TO-SUBJ
LEX - 70
SH0Wv
SEND
V
ADJ
V
ADJ
+NEUT -INS +AGT]
[
FACT
S
PASS
DAT—* OBJ
+NEUT -LOC +INS -ACT]
FACT
IMPER
PASS
STAT-REDUCT
RAISE-TO-OBJ
DAT—• OBJ
+ [
SEVERAL
+
+
+
-
SIC K1
QUANT
DIST
ATTACH
N-DEL
[
[-COUNT]]
[
[+PL]]
[T^SPEC]
]
INTEGER
SHIFT
+ V
+ ADJ
+ [
-NEUT -DAT -LOC
-INS +AGT]
SICK2
+ V
+ ADJ
+ [
SHALL
-NEUT +DAT -LOC
-INS -AGT]
+ MODAL
SHOW*1
SICK"
+
+
*
+
*
*
+ V
+ ADJ
+ [
V
ADJ
[
+NEUT -LOC -INS +AGT]
FACT
IMPER
PASS
STAT-REDUCT
RAISE-TO-OBJ
DAT—>-OBJ
1001
+
+
+NEUT +DAT -LOC
•INS • -AGT]
FACT
IMPER
WH-S
GER
AFFECT
LEX - 71
SIGNIFICANT
SON
+
+
+
*
+ N
+ [
V
ADJ
[
EMOT
-NEUT -LOC +INS -AGT]
+
+
+
+
-NEUT +DAT -LOC
-INS -AGT]
COMMON
COUNT
HUMAN
MASC
"SJC"
+ MODAL
SOON
+ ADV
+ TEMPORAL
SLEEPY.
+ V
+ ADJ
+ [
-NEUT +DAT -LOC -INS -AGT]
STALLION
N
[__
SMEAR
+ V
- ADJ
+NEUT -DAT +LOC +AGT]
+ [
FACT
S
PASS
LOC ->OBJ,NEUT
[on]
[with]
[[-ABSTRACT] OBJ]
[[-ANIMATE] OBJ]
+
+
*
*
STATUE
+ N
+ [
+
+
*
"some"
ART
DEF
DEM
WH
SPEC
COUNT
-NEUT -LOC -INS -AGT]
COMMON
COUNT
HUMAN
MASC
-NEUT +DAT -LOC
-INS +AGT]
COMMON
COUNT
ABSTRACT
ANIMATE
OBJ-DEL
STICK
N
COMMON
COUNT
ABSTRACT
ANIMATE
1002
LEX - 72
STOP
+
+
*
+
+
SUGGEST
V
ADJ
[
V
ADJ
+NEUT -LOC +INS -AGT]
+ [
FACT
IMPER
WH-S
PASS
+NEUT -DAT -LOC]
FACT
IMPER
WH-S
PASS
GER
AFFECT
SUGGESTION
STORY
N
+NEUT -DAT -LOC -INS +AGT]
[
FACT
IMPER
GER
COMMON
COUNT
ABSTRACT
OBJ-DEL
PREP NEUT about
+
+
*
+
+
+
+
N
[
-LOC -INS]
FACT
INDIC
IMPER
GER
COMMON
COUNT
ABSTRACT
PREP NEUT for
SUPPOSE
SUFFICE
V
ADJ
+ V
- ADJ
[
+ [
-NEUT -DAT -LOC +INS -AGT]
EMOT
SUGGEST•
+NEUT -LOC -INS +AGT]
1003
+NEUT +DAT -LOC
-INS -AGT]
FACT
IMPER
WH-S
PASS
STAT-REDUCT
RAISE-TO-OBJ
LEX - 73
TELL
SURE
V
ADJ
+NEUT -DAT -LOC -INS -ACT]
[
FACT
IMPER
WH-S
RAISE-TO-SUBJ
+
+
V
ADJ
f
- S
* PASS
+ PREP NEUT about
SURE'
"that"
+ V
+ ADJ
+ [
+NEUT +DAT -LOC -INS -ACT]
- FACT
- IMPER
- WH-S
« GER
+
+
+
+
*
*
+
ART
DEF
DEM
WH
FAR
N-DEL
COUNT
I
II
III
TABLE
N
COMMON
COUNT
ABSTRACT
ANIMATE
"the"
+ ART
+ DEF
- DEM
- GEN
* COUNT
- I
- II
+ III
TELL
+ V
- ADJ
+ [
+NEUT +DAT -LOC -INS +AGT]
FACT
WH-S
PASS
EQUI-NP-DEL
100U
+NEUT +DAT -LOC
-INS +AGT]
LEX - Ik
THEN
THINK-
+ ADV
+ TEMPORAL
+ FAR
+NEUT -DAT -LOC
-INS +AGT]
*
+
THERE
+ ADV
+ LOC
+ FAR
FACT
IMPER
WH-S
PASS
GER
THIRD
THING
+ ORD
N
PRO
ATTACH
HUMAN
COUNT
PLURAL
"this"
THI NK1
V
ADJ
+NEUT +DAT -LOC -INS -AGT]
[
FACT
IMPER
WH-S
PASS
+ STAT-REDUCT
» TO-BE-DEL
• RAISE-TO-OBJ
THINK2
+
+
+
V
ADJ
[
-DAT -LOC -INS +AGT]
S
PREP NEUT about
1005
+
+
+
ft
ft
+
ART
DEF
DEM
WH
FAR
N-DEL
COUNT
I
II
III
THREE
+
+
+
+
QUANT
DIST
ATTACH
N-DEL
[
[+PL]]
[
[-COUN
INTEGER
LEX - 75
TRY
TIME
+
+
+
+
*
N
PRO
V
ADJ
[
ATTACH
HUMAN
COUNT
PLURAL
+NEUT -DAT -LOC
-INS +AGT]
FACT
INDIC
WH-S
AGT-IDENT
PASS
GER
EQUI-NP-DEL
TIREDJ
+ V
+ ADJ
+ [
-NEUT +DAT -LOC -INS -AGT]
TWO
TIRED
+ V
+ ADJ
+ [
+NEUT +DAT -LOC -INS -AGT]
- FACT
- IMPER
- WH-S
+ GER
QUANT
DIST
ATTACH
N-DEL
[
[+PL]]
[-COUNT]]
[~
+ INTEGER
UNLIKELY
+ V
+ ADJ
+ [
"TNS"
* PAST
•
+
TO
+NEUT -DAT -LOC
-INS -AGT]
FACT
IMPER
WH-S
RAISE-TO-SUBJ
AFFECT
+ PREP
+ LOC
+ DIR
UNDERSTAND
TRAGIC
V
ADJ
[
-NEUT -LOC +INS -AGT]
EMOT
AFFECT
1006
V
ADJ
+DAT -LOC -INS -AGT]
+ [
FACT
IMPER
PASS
STAT-REDUCT
RAISE-TO-OBJ
LEX - 76
UPPER
WARN
+ CHIEF
+ V
- ADJ
• [
URGE
•
+
+
«
*
V
ADJ
[
+NEUT -LOC -INS +AGT]
FACT
INDIC
WH-S
DAT-IDENT
PASS
EQUI-NP-DEL
VERY
+ ADV
+ DEG
+NEUT +DAT -LOC
-INS +AGT]
- S
* PASS
+ PREP NEUT about
"we"
+
+
+
+
•
*
+
ART
DEF
DEM
GEN
COUNT
I
II
III
PL
WELL
WANT
+
•
•
•
•
V
ADJ
+NEUT +DAT -LOC -INS -AGT]
[
FACT
INDIC
WH-S
PASS
EMOT
EQUI-NP-DEL
+ ADV
+ MANNER
"what"
+
+
+
*
ART
DEF
DEM
WH
COUNT
WARN
+
+
*
*
V
ADJ
[
+NEUT -LOC -INS +AGT]
FACT
WH-S
AGT-IDENT
PASS
EQUI-NP-DEL
1007
"whether"
+ CONJ
+ OR
+ WH
LEX - 77
WILL
+ MODAL
"you"
+
+
+
+
ART
DEF
DEM
GEN
COUNT
I
II
- Ill
- PL
"you"
+
+
+
+
*
+
ART
DEF
DEM
GEN
COUNT
I
II
III
PL
1008
LEX - 78
III.
A.
THE SECOND LEXICAL LOOKUP
Discussion
The present grammar utilizes a second lexical insertion procedure which follows the last rule of the transformational component,
The function of the second insertion process is to attach phonological matrices to clusters of semantic-syntactic features that
have resulted from operations of the transformational component.
Such an operation is not unique to this grammar; the suggestion of
some such operation has been made informally many times before. In
particular, Fillmore proposed that pronouns were to be viewed as
feature clusters whose phonological realizations were not interestingly related and therefore ought to be inserted following the transformational operations (cf. Fillmore, 1966d).
Typical of the operations for which the second lexical insertion process is useful is the set of rules that produce the surface
pronouns in this grammar. The pronouns, as can be seen in the section on Pronominalization, are never inserted in their surface forms
in the first pass through the lexicon.
A non-anaphoric definite pronoun is derived from a full noun
phrase expanded by the PS-rules as (36).
(36)
DET
I
ART
NOM
I
N
To this tree the first lexical insertion procedure can attach the
definite article the and the PRO-noun one, with the following features as one possibility assigned by the first lexical lookup (but
with no phonological matrices):
1009
LEX - 79
NP
(37)
NOM
DET
1
1
ART
N
1
i
~+N
+ART "
+DEF
-DEM
-ATTACH
+PL
+1
-II
+HUMAN
*
+III .
At this point neither of the two constituents of the NP above
has accompanying phonological specifications. In addition, the
cluster of features that is dominated by N is identical to the
cluster of features that result from the N reduction rules that
form a part of the derivation of anaphoric pronominalization (cf.
PRO section).
Feature-copying rules (also in PRO section) copy the features
+PL, +HUMAN from the N onto the ART; the Deletion of Noun Node rule
(PRO section) deletes the N altogether, adding +PRO to the ART,
leaving the structure (38):
(38)
NP
DET
ART
+ART
+DEF
-DEM
+PL
+HUMAN
+1
-II
+III
L+PRO
There is still no phonological specification associated with this
complex symbol.
Finally the string of which this NP is a part emerges from
the transformational component, but the phonological rules cannot
yet apply because there are sentence constituents that are still
1010
LEX - 80
without phonological specifications. At this point the second
lexical lookup applies. In the case of the tree in (38) we will
be attached. If +ACCUS had been added by the objective case-marking
rule (see PRO), the form would be us_; addition of the feature +GENIT
would give our or ours, though, in fact, these genitive forms have
not been included in the sample second lexicon because of the problems in keeping the two feature specifications distinct (see discussion in GEN).
The second lexical lookup is utilized in the present grammar
to attach phonological matrices to already existing feature complexes. The operation as it is presently viewed does not permit
deletion of nodes or addition of nodes. For example, in a recent
paper J. Gruber (1967c) proposed insertion possibilities that would
allow a tree expanded as the following:
(39)
to be replaced by a single lexical item, namely, stallion. Such
an operation would account for the absence of such NP's as male
stallion, male steer, male tom-cat and male gander. The tree
above (39) differs in a rather profound way from the kind of tree
that Gruber's grammar would generate, but the principle is the
same. The present grammar disallows such power in the second lexical lookup.
A comparable operation would be the incorporation of Past
Tense in the case of irregular verbs in English. An affix movement
rule assigns the Past Tense Affix as the right daughter of a
Chomsky-adjoined V node like the tree below:
(UO)
V
V
Past
The node Past, under certain circumstances, would allow the
attachment of the Past Tense Affix [t], [d] or [id]. The present
constraint on the power of the second lexical lookup would not allow
the tree above to be changed as would have to be the case if the
lower V were an irregular verb; e.g. run, steal.... To allow the
tree above to be changed so that run + Past could be given the
phonological matrices of /ran/ would make the exclusion of stallion
1011
LEX - 81
ad hoc. It is difficult to see what possible limits there might
be if such attachment were permitted.
The question of whether the second lexical lookup should require non-distinctness or strict identity is a serious one. In
favor of the strict identity condition is the fact that many transformationally introduced features appear to be clear instances of
"marked" features, where the opposite value would never appear on
any item—e.g. +REFL, +ACCUS, +GENITIVE, and all the prepositional
features +OF, +WITH, etc. It would seem quite unnatural to have to
introduce -REFL etc. on all deep structure items of the category
on which the transformationally introduced feature could potentially
appear. On the other hand, where the same phonological form corresponds to several syntactic feature matrices which have a distinctive subset of features in common, it seems wasteful to have to
provide multiple entries in the second lexicon. Such is the case,
for example, with we, which must include [+1, +PL] as well as the
other features common to nominative personal pronouns, but is indifferent to [til], [till].
Both of these generalizations can be captured if the requirement for second lexical lookup is the following:
(kl)
The phonological matrix P associated with complex
symbol L in the second lexicon is assigned to the
terminal complex symbol S in a given surface structure tree if the features of L are a subset of the
features of S.
That is, if L contains +F , S must contain +F,; if L contains
-Fo, S must contain -F2; but S may contain some features not mentioned in L. This inclusion condition appears to capture the
desirable properties of both strict identity and non-distinctness.
Finally, the kinds of items for which the present grammar
utilizes the second lexical lookup are the following A'
(l)
In the sample lexicon that follows the features marked with an
asterisk could have been omitted in accordance with {kO) but
they have been retained in the interests of readability.
1012
LEX - 82
1)
2)
3)
U)
5)
6)
Determiners;
Pronouns—both independent and relative;
Negative adverbials, particles, quantifiers and
determiners;
Prepositions;
Conjunctions;
Quantifiers resulting from conjunction reduction.
In the sample Second Lexicon which follows representative
entries for items (l-k) are given*
1013
LEX - 83
B.
Sample Entries for Second Lexicon
Pronouns and Determiners
HE
+
+
+
+
+
+
+
ART
PRO
DEF
DEM
+
+
+
+
+
+
+
+
GENERIC
WH
ATTACH
N-DEL
I
II
III
COUNT
PL
HUM
ART
PRO
DEF
DEM
GENERIC
WH
ATTACH
N-DEL
I
II
III
COUNT
PL
HUM
MASC
ME
HIM
+
+
+
+
+
+
+
+
ART
PRO
DEF
DEM
+
+
+
—
+
+
+
+
+
+
GENERIC
WH
ATTACH
N-DEL
I
II
III
COUNT
PL
HUM
ACCUS
101U
ART
PRO
DEF
DEM
GENERIC
WH
ATTACH
N-DEL
I
II
III
COUNT
PL
HUM
MASC
ACCUS
LEX - Bh
HIM
+
+
+
+
+
+
+
+
+
+
HER
ART
PRO
+
+
+
—
+
—
+
+
—
+
+
REFLEX
DEM
GENERIC
WH
ATTACH
N-DEL
I
II
III
COUNT
PL
HUM
MASC
ACCUS
SHE
+
+
+
—
+
+
+
+
-
ART
PRO
DEF
DEM
GENERIC
WH
ATTACH
N-DEL
I
II
III
COUNT
PL
HUM
MASC
ACCUS
HER
ART
PRO
DEF
DEM
+
+
+
+
+
+
+
+
+
—
+
GENERIC
WH
ATTACH
N-DEL
I
II
III
COUNT
PL
HUM
MASC
1015
ART
PRO
REFLEX
DEF
DEM
GENERIC
WH
ATTACH
N-DEL
I
II
III
COUNT
PL
HUM
MASC
ACCUS
LEX - 85
IT
WE
+
*
+
-
ACCUS
US
+
*
+
+
+
*
*
+
+
+
IT
+
+
+
+
—
+
+
+
»
—
+
ART
PRO
DEF
DEM
GENERIC
WH
ATTACH
N-DEL
I
II
III
PL
HUM
ART
PRO
REFLEX
DEF
DEM
GENERIC
WH
ATTACH
N-DEL
I
II
III
COUNT
PL
HUM
ACCUS
1016
ART
PRO
DEF
DEM
GENERIC
WH
ATTACH
N-DEL
I
II
III
PL
HUM
ACCUS
LEX - 86
YOU
THEM
ART
PRO
DEF
DEM
GENERIC
WH
ATTACH
N-DEL
I
II
III
COUNT
PL
HUM
ACCUS
ART
PRO
DEF
DEM
GENERIC
WH
ATTACH
N-DEL
I
II
III
PL
HUM
MASC
] 1
+ ACCUS
THEY
THEM
ART
PRO
DEF
DEM
GENERIC
WH
ATTACH
N-DEL
I
II
III
PL
HUM
MASC
+
+
+
+
—
—
—
+
+
+
+
+
] ]
1017
ART
PRO
REFLEX
DEF
DEM
GENERIC
WH
ATTACH
N-DEL
I
II
III
PL
ACCUS
LEX - 87
SELF
SOME (sm)
PRO
COUNT
PL
REFLEX
ATTACH
+
*
+
SELVES
+
+
+
+
PRO
PL
REFLEX
ATTACH
+
*
+
ART
PRO
DEF
DEM
GENERIC
SPEC
WH
ATTACH
N-DEL
- I
+
j\+
-
II
III
COUNT I
PL
/
INDET
SOME
- I
+
+
-
ART
PRO
DEF
DEM
GENERIC
SPEC
WH
ATTACH
N-DEL
II
III
COUNT
PL
INDET
{
+
+
*
+
+
—
+
m
1018
ART
PRO
DEF
DEM
GENERIC
SPEC
WH
ATTACH
N-DEL
I
II
III
COUNT I
PL
/
INDET
LEX - 88
NO
SOME
+
+
*
+
-
ART
PRO
DEF
DEM
GENERIC
SPEC
WH
ATTACH
N-DEL
ART
PRO
DEF
DEM
GENERIC
SPEC
WH
ATTACH
N-DEL
I
II
III
COUNT
PL
INDET
NEG
- I
+
*
*
II
III
COUNT
PL
ONE
ART
PRO
DEF
DEM
GENERIC
SPEC
WH
ATTACH
N-DEL
I
II
III
COUNT
PL
INDET
NO
ART
PRO
DEF
DEM
GENERIC
SPEC
WH
ATTACH
N-DEL
I
II
III
COUNT
PL
INDET
NEG
1019
LEX - 89
ANY
NONE
+
«
+
+
+
+
ART
PRO
DEF
DEM
GENERIC
SPEC
WH
ATTACH
N-DEL
ART
PRO
DEF
DEM
GENERIC
SPEC
WH
ATTACH
N-DEL
- I
- I
- II
III
COUNT
PL
INDET
NEG
- II
III
COUNT
PL
INDET
NEG
ANY
THE
+
+
+
-
+
+
+
ART
PRO
DEF
DEM
GENERIC
SPEC
WH
ATTACH
N-DEL
ART
PRO
DEF
DEM
GENERIC
WH
ATTACH
N-DEL
- I
- I
+
•
*
- II
III
COUNT
PL
INDET
NEG
II
III
COUNT
PL
+ [
1020
N]
LEX - 90
THIS
THAT
ART
PRO
DEF
DEM
GENERIC
WH
ATTACH
N-DEL
I
II
III
COUNT
PL
FAR
ART
PRO
DEF
DEM
GENERIC
WH
ATTACH
N-DEL
I
II
III
COUNT
PL
FAR
THESE
THOSE
ART
PRO
DEF
DEM
GENERIC
WH
ATTACH
N-DEL
I
II
III
PL
FAR
+
*
+
+
*
ART
PRO
DEF
DEM
GENERIC
WH
ATTACH
N-DEL
- I
+
+
+
1021
II
III
PL
FAR
LEX - 91
THAT
WHICH
ART
PRO
DEF
DEM
GENERIC
WH
ATTACH
N-DEL
I
II
III
COUNT
HUM
[
NP[—
ART
PRO
DEF
DEM
GENERIC
WH
ATTACH
N-DEL
I
II
III
COUNT
PL
HUM
] ]
THOSE
WHAT
ART
PRO
DEF
DEM
GENERIC
SPEC
WH
ATTACH
N-DEL
I
II
III
COUNT
PL
1022
LEX - 92
Relative Pronouns
WHO
+
+
+
+
+
+
+
+
«
+
WHICH
ART
PRO
DEF
DEM
ART
PRO
DEF
DEM
GENERIC
SPEC
WH
REL
ATTACH
N-DEL
I
II
III
COUNT
PL
HUM
GENERIC
SPEC
WH
REL
ATTACH
N-DEL
I
II
III
COUNT
PL
HUM
WHOM
+
+
+
+
+
+
-
THAT
ART
PRO
DEF
DEM
GENERIC
SPEC
WH
REL
ATTACH
N-DEL
I
II
III
COUNT
PL
HUM
ACCUS
ART
PRO
DEF
DEM
GENERIC
SPEC
WH
REL
ATTACH
N-DEL
I
II
III
COUNT
PL
HUM
THAT
1023
LEX - 93
Advertaials and Negatives
TOO
NEVER
+ ADV
+ TOO
- SPEC
+
+
+
+
ADV
TEMPORAL
DEF
SPEC
INDET
NEG
EITHER
+
+
+
-
ADV
TOO
SPEC
INDET
NEG
HARDLY
+ NEG
- COMPLETE
NEITHER
NOT
+
+
+
+
+ NEG
+ COMPLETE
ADV
TOO
SPEC
INDET
NEG
N'T
+ NEG
+ CNTR
SOMETIMES
+
+
*
-
ADV
TEMPORAL
DEF
SPEC
INDET
EVER
+
+
+
-
ADV
TEMPORAL
DEF
SPEC
INDET
NEG
102U
LEX - 9k
Prepositions
ABOUT
TO
+ PREP
+ NEUT
+ PREP NEUT about
+ PREP
+ NEUT
+ PREP NEUT to
AT
UPON
+ PREP
+ NEUT
+ PREP NEUT at
+ PREP
+ NEUT
+ PREP NEUT upon
FOR
WITH
+ PREP
+ NEUT
+ PREP NEUT for
+ PREP
+ NEUT
+ PREP NEUT with
IN
OF
+ PREP
+ NEUT
+ PREP NEUT in
+ PREP
+ DAT
+ PREP DAT of
OF
TO
+ PREP
+ NEUT
+ PREP NEUT of
+ PREP
+ DAT
+ PREP DAT to
ON
AT
+ PREP
+ NEUT
+ PREP NEUT on
+ PREP
+ INS
+ PREP INS at
1025
LEX - 95
WITH
+ PREP
+ INS
+ PREP INS with
BY
+ PREP
+ AGT
+ PREP AGT b£
FROM
+ PREP
+ AGT
+ PREP AGT from
OF
+ PREP
+ Of
1026
BIBLIOGRAPHY
Abbreviations:
CLHU:
Computational Laboratory of Harvard University,
Mathematical Linguistics and Automatic
Translation, Report No. NSF
, to the National
Science Foundation, Anthony G. Oettinger,
Principal Investigator, Cambridge.
IBM:
International Business Machines Corporation,
Thomas J. Watson Research Center, Yorktown
Heights, N.Y.
LRP:
Linguistics Research Project, Principal
Investigator: F.W. Householder, Jr., Indiana
University, Bloomington.
POLA:
Project on Linguistic Analysis, The Ohio State
University Research Foundation, Columbus.
TDAP:
Transformations and Discourse Analysis Papers,
Zellig Harris, Director, University of
Pennsylvania, Philadelphia.
The Structure of Language: The Structure of Language,
Jerry A. Fodor and Jerrold J. Katz (ed.),
Prentice Hall, Inc., (1964), Englewood Cliffs,
New Jersey.
1027
BIBLIO - 2
Alexander, D. and W. J. Kuntz (1964)
"Some Classes of Verbs in English", LRP.
Alexander, D. and G. H. Matthews (1964)
"Adjectives before 'that' Clauses in English", LRP.
Anderson, Tommy (1961)
"The English Adverb", unpublished paper, U.C.L.A.
(1965)
A Contrastive Analysis of Cebuano Vlsayan and English,
unpublished Ph.D. dissertation, U.C.L.A.
Annear, Sandra (1964a)
"The Ordering of Prenominal Modifiers in English", POLA
//8, pp. 95-120.
(1964b)
"English and Mandarin Chinese: The Comparative Construction",
unpublished paper, Ohio State University.
(1965)
"English and Mandarin Chinese: Definite and Indefinite
Determiners and Modifying Structures", POLA #11, pp. 1-55.
(1967)
"Relative Clauses and Conjunctions", Working Papers in
Linguistics, Report #1, Ohio State University.
(1968)
"Restrictive Relative Clauses" and "Constraints on Relative
Clause Formation", Chapters I and II of dissertation (no
title given), Ohio State University.
Annear, Sandra and Dale E. Elliot (1965)
"Derivational Morphology in a Generative Grammar", LSA
Winter Meeting.
(1967)
"Some Problems of Derivational Morphology", Working Papers
in Linguistics, Report #1, Ohio State University.
Bach, Emmon (1967a)
"Nouns and Nounphrases" (to appear in Proceedings of the
Texas Conference on Language Universals, ed. by E. Bach
and R. Harms).
(1967b)
"Have and be in English Syntax", Language, 43.462-485.
1028
BIBLI0-3
Baker, C. LeRoy (1966a)
"Existentials and Indefinites in English", unpublished
paper, University of Illinois.
Bellert, Irena (1966)
"On Certain Syntactical Properties of the English Connectives
and and but", TDAP #64.
Bever, Thomas G. and John Robert Ross (1965)
"Underlying Structures in Discourse", unpublished preliminary
draft, M.I.T.
Bierwisch, Manfred (1967a)
"On Certain Problems of Semantic Features", unpublished
paper, M.I.T.
(1967b)
"Some Semantic Universals of German Adjectivals", Foundations
of Language 3.1-36.
Blake, Frank R. (1930)
"A Semantic Analysis of Case", Language Monograph #7, 34-49.
Bolinger, D. L. (1960)
"Linguistic Science and Linguistic Engineering", Word 16,
374-391.
(1961)
"Syntactic Blends and Other Matters", Language 37, 366-81.
(1967a)
"Adjectives in English: Attribution and Predication",
Lingua. 18, 1-34.
(1967b)
"Entailment and the Meaning of Structure", unpublished paper,
Harvard.
(1967c)
"The Imperative in English", To Honor Roman Jakobson, 355-362,
Mouton and Co., The Hague.
Bourton, Lawrence F. (1968)
"Do-so Revisited", unpublished paper, University of Illinois.
1029
BIBLIO-4
Boyd, Julian and James Peter Thorne (1968)
"The Deep Grammar of Modal Verbs", unpublished paper,
University of California, Berkeley, and University of
Edinburgh (PEGS Paper #31).
Brame, Michael (1968)
"On the Nature of Relative Clauses", unpublished paper,
M.I.T.
Bremer, Mary G. (1966)
"On the Relationship between 'have' and 'there is' in
English", unpublished paper, University of Illinois.
Bridgeman, Loraine I., et al.
(1965a)
"Further Classes of Adjectives", LRP.
(1965b)
"More Classes of Verbs in English", LRP.
(1965c)
"Nouns before That-Clauses in English", LRP.
Browne, Wayles E. (1964)
"On Adjectival Comparisons and Reduplication in English",
unpublished paper, M.I.T.
Chafe, Wallace L. (1967)
"English Noun Inflection and Related Matters from a Generative Semantic Point of View", unpublished paper, University
of California, Berkeley (PEGS Paper #15).
(1968)
"English Questions", unpublished paper, University of
California, Berkeley (PEGS Paper #26).
Chaiyaratana, Chalao (1961)
A Comparative Study of English and Thai Syntax, Indiana
University, Bloomington.
Chapin, Paul G. (1967)
On the Syntax of Word-Derivation in English, unpublished
Ph.D. dissertation, M.I.T.; and Information Systems
Language Studies, #16, MITRE Corporation, Bedford.
Chatman, Seymour (1961)
"Preadjectivals in the English Nominal Phrase", TDAP #22,
1030
BIBLIO-5
(1962)
"The Classification of English Verbs by Object Types",
Proceedings of the 1961 International Conference on
Machine Translation of Language and Applied Language
Analysis. VOL. I, pp. 84-95, Her Majesty's Stationery
Office, London.
(1964)
"English Sentence Connectors", Studies in Languages and
Linguistics in Honor of Charles C. Fries, Albert Henry
Marckwardt (ed.), pp. 315-334, The English Language
Institute, Ann Arbor.
Chomsky, Noam (1955)
The Logical Structure of Linguistic Theory, unpublished
paper, M.I.T.
(1957)
Syntactic Structures (Janua Linguarum, #4), Mouton and Co.,
The Hague.
(1958)
"A Transformational Approach to Syntax", Proceedings of
the Third Texas Conference on Problems of Linguistic
Analysis in English. 1958, A. A. Hill (ed.), pp. 124-158,
University of Texas Press, Austin, 1962; and in The
Structure of Language. pp. 211-245.
(1961)
"On the Notion 'Rule of Grammar'", Structure of Language
and its Mathematical Aspects. Proceedings of Symposia in
Applied Mathematics (1960), Roman Jakobson (ed.), Vol. XII,
pp. 6-24, American Mathematical Society, Providence; and
The Structure of Language, pp. 119-137.
(1964a)
"The Logical Basis of Linguistic Theory", Proceedings of
the Ninth International Congress of Linguists, Horace G.
Lunt (ed.), pp. 914-978, Mouton & Co., The Hague.
_(1964b)
"Current Issues in Linguistic Theory", The Structure of
Language. pp. 50-118.
1031
BIBLIO-6
_(1964c)
Current Issues in Linguistic Theory, (Janua Linguarum, #38),
Mouton & Co., The Hague.
.(1965)
Aspects of the Theory of Syntax, M.I.T. Press, Cambridge.
_(1966a)
"Topics in the Theory of Generative Grammar", in Current
Trends in Linguistics, Thomas A. Sebeok (ed.) Vol. Ill,
pp. 1-60, Mouton & Co., The Hague.
(1966b)
Cartesian Linguistics, Harper and Row, New York and London.
(1968)
"Remarks on Nominalization", unpublished paper, forthcoming
in Readings in English Transformational Grammar, ed. Jacobs
and Rosenbaum.
Chomsky, Noam and Morris Halle (1968)
The Sound Pattern of English, Harper & Row, New York.
Cressey, William W. (1967)
"Relative Adverbs in Spanish: A Transformational Analysis",
University of Michigan, forthcoming in Language.
Dean, Janet (1967)
"Determiners and Relative Clauses", unpublished paper, M.I.T.
Doherty, Paul C. and Arthur Schwartz (1967)
"The Syntax of the Compared Adjective in English", Language
43.903-936.
Doran, R. W. (1967)
External Formats Accepted by Transformational Grammar Testing
System, Report AF-12, Stanford University Computer Science
Department, Computational Linguistics Project.
Dougherty, Ray C. (1967a)
"The deep structure of plurals, conjoined noun phrases,
plural reflexives, and reciprocal pronouns", unpublished
paper, M.I.T.
1032
BIBLIO-7
(1967b)
"Coordinate Conjunction", unpublished paper, M.I.T.
Ehrman, Madeline E. (1966)
The Meanings of the Modals in Present Day American English,
(Janua Linguarum, Series Practica XLV), Mouton & Co., The
Hague.
Elliot, Dale E. (1965)
"Interrogation in English and Mandarin Chinese", POLA #11,
pp. 57-117.
Fidelholtz, James (1964)
"Coordination in Sentences: Universals (i.e. English
Extrapolated) or the Case for the Schem(a)ing Linguist",
unpublished paper, M.I.T.
Fillmore, Charles J. (1962)
"Indirect Object Constructions in English and the Ordering
of Transformations", POLA #1; and Monographs on Linguistic
Analysis No. 1, Mouton and Co., The Hague (1965).
(1963)
"The Position of Embedding Transformations in a Grammar",
POLA #3, pp. 1-33; and Word, 19.208-231.
(1964a)
"Desentential Complement Verbs in English", POLA #7, pp.
88-105.
(1964b)
"'Transportation' Rules in English", LSA Summer Meeting.
(1965)
"Entailment Rules in a Semantic Theory", POLA #10, pp. 60-82.
(1966a)
"A Proposal Concerning English Prepositions," Georgetown
Monograph Series on Language and Linguistics, No. 19, 19-33.
(1966b)
"Toward a Modern Theory of Case", POLA #13, pp. 1-24.
(1966c)
"Deictic Categories in the Semantics of 'come'", in Foundations
of Language. 2.219-227.
1033
BIBLIO-8
(1966d)
"On the Syntax of Preverbs", unpublished paper, Ohio State
University.
(1967a)
"The Case for Case", to appear in Proceedings of the Texas
Conference on Language Universals, ed. by E. Bach and R.
Harms.
(1967b)
"The Grammar of Hitting and Breaking", Working Papers in
Linguistics, Report #1, Ohio State University.
(1968)
"Lexical Entries for Verbs", (second draft), unpublished
paper, Ohio State University.
Fraser, James B. (1963)
"The Linguistic Framework for a Sentence Recognition and
Analysis Routine: Transformational Structure", Working
Paper w6266, MITRE Corp., Bedford, Massachusetts.
(1964)
"On Particles in English", LSA Summer Meeting.
(1965)
An Examination of the Verb Particle Construction in English,
unpublished Ph.D. dissertation, M.I.T.
Friedman, Joyce (1968a)
A Computer System for Transformational Grammar, CS-84, AF-21.
Stanford University: Computer Science Dept.
(1968b)
Computer Experiments in Transformational Grammar, forthcoming,
Stanford University: Computer Science Dept.
Friedman, Joyce and Thomas H. Bredt (1968)
Lexical Insertion in Transformational Grammar, Stanford
University Computer Science Department; Computational
Linguistic Project.
Friedman, J. and R. Doran (1968)
A Formal Syntax for Transformational Grammar, AF-24, CS-95,
Stanford University: Computer Science Dept.
1031*
BIBLIO-9
Fudge, Erik (1965)
"Investigation of Verb Noun Cooccurrence", in Linguistic
Analysis of English; Final Report, LRP.
Garcia, Erica C. (1965)
"Auxiliaries in Generative Grammar", LSA Winter Meeting.
(1967)
"Auxiliaries and the Criterion of Simplicity", Language
43.853-870.
Givon, Talmy (1967a)
Some Noun-to-Noun Derivational Affixes, Systems Development
Corporation, Santa Monica, California.
(1967b)
Transformations of Ellipsis, Sense Development and Rules of
Lexical Derivation, Systems Development Corporation, Santa
Monica, California.
Gleason, H. A., Jr. (1965)
Linguistics and English Grammar, (Chapters 10-12), Henry
Holt and Co., New York.
Gleitman, Lila R. (1960)
"Conjunction with 'each other"', unpublished M.A. thesis,
University of Pennsylvania.
(1961a)
"Pronominals and Stress in English Conjunctions", Language
Learning. XI.157-170.
(1961b)
"Conjunction with and". TDAP #40.
(1961c)
"A Grammar for English Conjunction", unpublished paper,
University of Pennsylvania.
(1963)
"Coordinate Conjunction in English", unpublished paper,
Eastern Pennsylvania Psychiatric Institute, Philadelphia.
.(1965)
"Coordinating Conjunctions in English", Language, 41.260-293.
1035
BIBLIO-10
Goodman, Ralph (1964)
"A Look at Transformational Grammars", In Part IV of An
Introductory English Grammar, by N. Stageberg, Holt and
Rlnehart, New York.
Gross, Maurice (1967)
"On Grammatical Reference", unpublished paper, University
of Aix-Marseille, and Centre National de la Recherche
Scientifique (Revised version of paper read at the 3rd
International Congress for Logic, Methodology and Philosophy
of Science, Amsterdam, Aug. 25-September 2, 1967).
Gruber, Jeffrey C. (1965)
Studies in Lexical Relations, unpublished Ph.D. dissertation,
M.I.T.
(1967a)
"Disjunctive Ordering Among Lexical Insertion Rules",
unpublished paper, M.I.T.
(1967b)
"Correlations between the Syntactic Constructions of the
Child and of the Adult", presented for the Society for
Research in Child Development, March 31, 1967.
(1967c)
Functions of the Lexicon in Formal Descriptive Grammars,
Systems Development Corporation TM-377O/OOO/O0, Santa
Monica, California.
(1967d)
"Look and See", Language 43.937-947.
Gunter, Richard (1963)
"Elliptical Sentences in American English", Lingua, 12.137-150.
Hale, Austin (1964)
"Quantification and English Comparative", to appear in
Monographs on Linguistic Analysis, (ed.) Wm. Wang.
Hall, Barbara C. (1962a)
"All about Predeterminers", unpublished paper, M.I.T.
(1962b)
"A Preliminary Attempt at an Historical Approach to Modern
English Predeterminers", unpublished paper, M.I.T.
1036
BIBLIO-11
(1963a)
"Preartides in English: Their Contemporary Grammar and
Its Historical Development", unpublished paper, M.I.T.
(1963b)
"Remarks on 'some' and 'any' in Negation and Interrogative
Constructions with a Note on Negation in Russian",
unpublished paper, M.I.T.
(1964a)
"The Auxiliary in English Sentences with 'if'", unpublished
paper, LSA Summer Meeting.
(1964b)
"Adverbial Subordinate Clauses", MITRE Corp., Bedford,
Massachusetts.
(1965)
Subject and Object in Modern English, unpublished Ph.D.
dissertation, M.I.T.
Harris, Zellig S. (1957)
"Cooccurrence and Transformation in Linguistic Structure",
Language. 33.283-340; and The Structure of Language, pp.
155-210.
(1965)
"Transformational Theory", Language. 41.363-401.
Hartung, Wolfdietrich (1964)
"Die zusammengesetzten Satze des Deutschen", Studia
Grammatics IV, Akademie-Verlag, Berlin.
Hasegawa, Kinsuke (1965)
"English Imperatives", Festschrift for Professor Naka.jima,
pp. 20-28, Kenkyusha, Tokyo.
(1967)
"The Passive Construction in English", forthcoming in
Language.
Hofmann, T. K. (1962)
"The English Verb Auxiliary, #1", unpublished paper,
University of Illinois.
1037
BIBLIO-12
(1964)
"The English Verb Auxiliary, #2", unpublished paper, M.I.T.
(1965a)
"Some Notes on Dictionaries", unpublished paper, M.I.T.
(1965b)
"Auxiliary Topics: The English Verb Auxiliary, //4", unpublished paper, M.I.T.
(1966)
"Past Tense Replacement and the Modal System", CLHU, NSF 17.
Hornby, A. S., E. Gatenby, and H. Wakefield (1963)
The Advanced Learner's Dictionary of Current English, Oxford
University Press, London.
Householder, Fred W. (1962)
"Review of The Grammar of English Nominalizations (R. B.
Lees)", Word, 18.326-353. "
(1965)
"Introduction" in Linguistic Analysis of English:
Report, LRP.
Final
Huddleston, R. (1967)
"More on the English Comparative", Journal of Linguistics,
3.1.91-102.
Jackendoff, Ray S. (1968a)
"Quantifiers as Noun Phrases", Studies in Transformational
Grammar and Related Topics, Principal Investigator: S. J.
Keyser ASCRL-68-0032, Brandeis University.
(1968b)
"Possessives in English", Studies in Transformational Grammar
and Related Topics, Principal Investigator: S. J. Keyser
ASCRL-68-0032, Brandeis University.
(1968c)
"On Some Incorrect Notions about Quantifiers and Negation",
Studies in Transformational Grammar and Related Topics,
Principal Investigator: S. J. Keyser ASCRL-68-0032, Brandeis
University.
(1968d)
"An Interpretive Theory of Pronouns and Reflexives", unpublished paper, M.I.T. (PEGS Paper No. 27).
1038
BIBLIO-13
(1968e)
"An Interpretive Theory of Negation", unpublished paper,
M.I.T.
(1968f)
"Speculations on Presentences and Determiners", unpublished
paper, M.I.T.
Jacobs, R. and P. Rosenbaum (1968)
English Transformational Grammar, Blaisdell, Boston.
Jespersen, Otto (1914-29)
A Modern English Grammar on Historical Principles, George
Allen and Unwin, London.
(1933)
Essentials of English Grammar. University of Alabama Press (reprint)
Karttunen, Lauri (1967)
The Identity of Noun Phrases, Rand Corporation Publication
No. P-3756, Santa Monica, California.
(1968)
What Do Referential Indices Refer to? Rand Corporation
Publication No. P-3854, Santa Monica, California.
Kat*, Jerrold J. (1964)
"Semi-Sentences", The Structure of Language, pp. 400-416.
Katz, Jerrold J. and Paul M. Postal (1964a)
"Semantic Interpretation of Idioms and Sentences Containing
Them", M.I.T. Quarterly Progress Report No. 70. Cambridge.
Kay, Martin (1967)
From Semantics to Syntax. Rand Corporation Publication No.
P-3746 (expanded version of paper presented at 10th International Congress of Linguists) To appear in Recent Advances
in Linguistics, eds. Bierwisch and Heidolph, Mouton and Co.,
The Hague.
Keyser, Samuel Jay and Robert Kirk (1967)
"Machine Recognition of Transformational Grammars of English",
Brandeis University.
Kimball, John (1967)
"Identity Crisis in Pronominalization", unpublished paper, M.I.T.
1039
BIBLI0-14
Kiparsky, Paul (1963)
"A Note on the English Imperative", unpublished paper, M.I.T.
Kiparsky, Paul and Carol Kiparsky (1968)
"Fact", unpublished paper, M.I.T., to appear in Recent
Advances in Linguistics, ed. Bierwisch and Heidolph, Mouton
and Co., The Hague.
Klima, E. S. (1960)
"Verb Classes in English for a Transfer Grammar", unpublished
paper, M.I.T.
(1962a)
"Structure at the Lexical Level and its Implications for a
Transfer Grammar", Proceedings of the 1961 International
Conference on Machine Translation of Language and Applied
Language Analysis, Vol. I, pp. 98-109, Her Majesty's
Stationery Office, London.
(1962b)
"Correspondence at the Grammatical Level", M.I.T. Quarterly
Progress Report, No. 64, Cambridge.
(1964a)
"Current Developments in Generative Grammar", to appear in
Kybernetika I, Prague.
(1964b)
"Relatedness Between Grammatical Systems", Language 40.1-20.
(1964c)
"Negation in English", in The Structure of Language, pp.
246-323.
(1964d)
Studies in Diachronic Transformational Syntax, unpublished
doctoral dissertation, Harvard University.
Koutsoudas, Andreas (1968)
"On Wh-Words in English", Journal of Linguistics 4.267-73.
Kuroda, S-Y (1965a)
"A Note on English Relativization", unpublished paper, M.I.T.
10U0
BIBLIO-15
_(1965b)
Generative Grammatical Studies in the Japanese Language.
unpublished Ph.D. dissertation, M.I.T.
_(1966a)
"Attachment Transformations" and "Wa"; slightly expanded
unpublished versions of Chapters 1 and 2 respectively of
the author's Ph.D. dissertation, M.I.T., Generative Grammatical Studies in the Japanese Language (1965), University
of California, San Diego.
_(1966b)
"English Relativization and Certain Related Problems",
unpublished paper, University of California, San Diego.
_(1967a)
"Review of Fillmore, 'Indirect Object Constructions in
English and the Ordering of Transformations'", forthcoming
in Language.
(1967b)
"On English Manner Adverbials", unpublished paper, University of California, San Diego.
(1967c)
"On English Manner Adverbials", unpublished paper, University of California, San Diego (revised version of 1967b).
(1968)
"English Relativization and Certain related "roMcns,"
Language 44.244-266.
Lakoff, George P. (1963a)
"Cycles and Complex Symbols in English Syntax", unpublished
paper, Indiana University.
(1963b)
"Toward Generative Semantics", unpublished paper, Research
Laboratory of Electronics, M.I.T.
(1964)
"Some Constraints on Transformations", unpublished paper,
Indiana University.
(1965).
On the Nature of Syntactic Irregularity, CLHU, NSF 16.
IOUI
BIBLIO-16
(1966a)
"Stative Adjectives and Verbs in English", CLHU, NSF 17.
(1966b)
"A Note on Negation", CLHU, NSF 17.
(1966c)
"Deep and Surface Grammar", unpublished paper, Harvard
University.
(1968a)
"Instrumental Adverbs and the Concept of Deep Structure",
Foundations of Language, 4, 4-29.
(1968b)
"Pronouns and Reference", unpublished paper, Harvard
University.
(1968c)
"Counterparts, or the Problem of Reference in Transformational Grammar", LSA Summer Meeting.
Lakoff, George and John R. Ross (1966a)
"A Criterion for Verb Phrase Constituency", CLHB, NSF 17.
(1966b)
"On the Ordering of Transformational Rules" unpublished
paper, M.I.T.
(1967)
"Is Deep Structure Necessary?", unpublished paper, M.I.T.
Lakoff, George and Stanley Peters (1966)
"Phrasal Conjunction and Symmatric Predicates", CLHU,
NSF 17.
Langacker, Ronald W. (1966)
"On Pronominalization and the Chain of Command", forthcoming
in Modern Studies in English: Readings in Transformational
Grammar, David Reibel and Sanford A. Schane (ed.), Prentice
Hall, Englewood Cliffs, N.J.
(1968)
"Mirror Image Rules in Natural Languages", unpublished paper
(preliminary version), University of California, San Diego
(PEGS Paper #23).
10U2
BIBLIO-17
Langendoen, D. Terence (1966a)
"Some Problems Concerning the English Expletive 'it'",
POLA #13, pp. 104-134.
(1966b)
"The Syntax of the English Expletive 'it'", Georgetown
University Monograph Series on Language and Linguistics,
No. 19, 207-216.
(1967a)
"The Use of the Expletive 'it' in Construction with Expressions of Place and Time", forthcoming in Proceedings of
the Tenth International Congress of Linguists.
(1967b)
"On Selection, Projection, Meaning, and Semantic Content",
Working Papers in Linguistics, Report No. 1, Ohio State
University.
(1968)
"An Analysis of Symmetric Predicates, and of the Formation
and Deletion of Reciprocal Elements in English", unpublished
paper, Ohio State University (very preliminary version).
Lee Gregory (1966)
"Causatives and Indirect Object Sentences", presented at
the Second Chicago Linguistics Circle Meeting, Chicago.
(1967a)
"Some properties of English be sentences", unpublished
paper, Ohio State University.
(1967b)
"The English Preposition WITH", Working Papers in Linguistics,
Report No. 1, Ohio State University.
Lees, Robert B. (1958a)
"Families of Nominalization Transformations", LSA Winter
Meeting.
(1958b)
"Generation of Nominal Compounds in English by Means of
Grammatical Transformations", unpublished paper, M.I.T.
10U3
BIBLIO-18
_(1958c)
"Some Neglected Aspects of Parsing", in Readings in Applied
Linguistics, (ed.) H. Allen, pp. 146-155, Appleton-CenturyCrofts, New York.
_(1960a)
The Grammar of English Nominalizations, UAL Publication 12,
Indiana University, Bloomington; and Mouton and Co., The
Hague.
_(1960b)
"A Multiply Ambiguous Adjectival Construction in English",
Language 36.207-221.
_(1960c)
"Analysis of the So-called 'Cleft Sentence' in English",
unpublished paper, IBM.
_(1960d)
"On Some Contributions of Empirical Linguistics", unpublished
paper, IBM.
_(1960e)
"Review of Interrogative Structures of American English
(D. Bolinger)", Word 16.119-125.
_(1960f)
"On the Stress of English Compounds", unpublished paper, IBM.
_(1961a)
"On Reformulating Transformational Grammars", unpublished
paper, IBM.
_(1961b)
"Grammatical Analysis of the English Comparative", Word
17.171-185.
_(1961c)
"The Constituent Structure of Noun Phrases", American
Speech. XXXVI.149-168.
J1963)
"Analysis of the So-Called 'Cleft Sentence' in English,"
Zeitschrift fur Phonetik Sprachwissenschaft und Kommunikationsferschung.
10UU
BIBLIO-19
(1964a)
"A Transformational Grammar of English", in Two Transformational Grammars of English, annotated by Earl Rand, English
Teachers Retraining Project, Taiwan Provincial Normal
University, University of Texas, AID.
(1964b)
"On Passives and Imperatives in English", Gengo Kenkyu No. 46.
(1965)
"Two Views of Linguistic Research", Linguistics 11.21-29.
Lees, R. B. and E. S. Klima (1963)
"Rules for English Pronominalization", Language 39.17-28.
Lieberman, Philip (1966)
Intonation, Perception, and Language, M.I.T. Press, Cambridge.
Lightner, T. M. (1964)
"On the Syntax of the Intransitive Verb 'wait'", M.I.T.
Quarterly Progress Report No. 75, Cambridge.
Long, Ralph B. (1961)
The Sentence and Its Parts, University of Chicago Press,
Chicago.
(1967)
"The 'Conjunctions'", unpublished paper, University of
Puerto Rico, San Juan.
LRP (1964a)
"A List of Adverbs in English According to Manner, Degree,
etc.", unpublished paper, Indiana University.
LRP (1964b)
"A List of Adjectives that Cooccur with Prepositions",
unpublished paper, Indiana University.
Lu, John H. T. (1965)
"Contrastive Stress and Emphatic Stress", POLA #10.
Lyons, John (1967)
"A Note on Possessive, Existential and Locative Sentences",
Foundations of Language 3.390-397.
101*5
BIBLIO-20
Malone, Joseph L. (1967)
"A Transformational Reexamlnation of English Questions",
Language 43.686-702.
Marchand, Hans (no date)
The Categories and Types of Present-Day English Word-Formation, Alabama Linguistic and Philological Series #13,
University of Alabama Press, no date (1967?)
Matthews, G. H. (1963)
"Transformational Grammar: Review Article", Archivum
Linguisticum, Vol. XIII, Fasc. 2.
(1965a)
"Problems of Selection in Transformational Grammar",
Journal of Linguistics 1.35-47.
(1965b)
"Studies in Transformational Syntax", Linguistic Analysis
of English: Final Report, LRP.
(1968)
"Le Cas Echeant", Parts I and II, unpublished paper, M.I.T.
Matthews, Peter H. (1967)
"Review of: Aspects of the Theory of Syntax (Chomsky)",
in Journal of Linguistics, 3.1.119-152.
McCawley, James D. (1964)
"Quantitative and Qualitative Comparison in English", LSA
Winter Meeting.
(1966)
"Concerning the Base Component of a Transformational
Grammar", LSA Linguistic Institute, U.C.L.A., unpublished
paper (see 1968b).
(1967a)
"How to Find Semantic Universals in the Event that there
are Any", To appear in Proceedings of the Texas Conference
on Language Universals, ed. by E. Bach and R. Harms.
(1967b)
"Can You Count Pluses and Minuses before You Can Count?",
read at the third annual Midwest Regional Conference of
the Chicago Linguistic Society, May 6, 1967.
101+6
BIBLIO-21
(1967c)
"Where Do Noun Phrases Come From?", unpublished paper,
University of Chicago.
(1967d)
"Why Auxiliary Verbs are Verbs", presented May 2 at University
of Wisconsin.
(1968a)
"The Annotated Respective", unpublished paper, University of
Chicago.
(1968b)
"Concerning the Base Component of a Transformational Grammar,"
Foundations of Language 4.243-369.
McKay, John (1968a)
"Some Generative Rules for German Time Adverbials", Language,
44. 25-50.
(1968b)
The Free Adverbial in a Generative Grammar of German,
unpublished Ph.D. dissertation, U.C.L.A.
MITRE Corp. (1964)
English Preprocessor Manual, MITRE Corp., Language Processing Techniques Subdepartment of Information Sciences
Department, Bedford.
(1965)
English Preprocessor Manual (revised), MITRE Corp., Language
Processing Techniques Subdepartment of Information Sciences
Department, Information System Language Studies No. 7,
Bedford.
Moore, Terence H. (1966)
"A Proposal to Constrain Embedding", LSA Summer Meeting.
(1967)
The Topic-Comment Function; A Performance Constraint on
a Competence Model, unpublished Ph.D. dissertation, U.C.L.A.
Newmeyer, Frederick J. (1968)
"Durative Keep in English", unpublished paper, University
of Illinois.
10U7
BIBLIO-22
Nilsen, Don Lee Fred (1967)
English Adverblals, unpublished Ph.D. dissertation,
University of Michigan.
Otanes, Fe (1966)
A Contrastive Analysis of English and Tagalog Complement
Structures, unpublished Ph.D. dissertation, U.C.L.A.
Partee, Barbara Hall (1968)
"Negation, Conjunction, and Quantifiers: Syntax vs.
Semantics", unpublished paper presented at the Conference
on Mathematical Linguistics, Budapest-Balatonszabadi,
Hungary, September, 1968.
Perlmutter, David M. (1967)
"The Two Verbs 'Begin'", unpublished paper, Brandeis
University, Massachusetts (PEGS Paper #28).
(1968a)
"On the Article in English", prepublication version, to
appear in Recent Developments in Linguistics, ed. Bierwisch
and Heidolph, The Hague: Mouton and Co.
Peters, Stanley (1967)
"Co-ordinate Constructions in English",
Preliminary draft of Ph.D. dissertation, M.I.T.
Peters, Stanley and Emmon Bach (1968)
"Pseudo - Cleft Sentences",(preliminary version), unpublished
paper, University of Texas.
Peterson, Thomas H. (1966)
"A Transformational Analysis of some Derived Verbs and
Adjectives in English", an expanded version of a paper
presented at LSA Summer Meeting.
Polutzky, H. J. (1960)
"Cleft Sentences" unpublished paper, M.I.T.
Postal, Paul (1966a)
"The Method of Universal Grammar", presented at Conference
in Linguistic Method, LSA Linguistic Institute, U.C.L.A.
(1966b)
"On So-called 'Pronouns' in English", Georgetown Monograph
Series on Language and Linguistics, No. 19, 177-206.
10U8
BIBLIO-23
(1967a)
"Crazy Notes on Restrictive Relative Clauses and Other
Matters", unpublished paper, IBM.
(1967b)
"Linguistic Anarchy Notes: Series A: Horrors of Identity:
No. 2, Coreferentiality and Physical Objects", unpublished paper, IBM.
(1968)
Cross-Over Phenomena: A Study in the Grammar of Coreference,
preliminary version of unpublished monograph, IBM.
Poutsma, H. (1904-29)
A Grammar of Late Modern English, P. Noordhoff, Groningen.
Querido, A. [n.d.]
"Transformations de Pronominalisation", unpublished paper,
Universite de Montreal.
Robbins, Beverly (1962)
"The Transformational Status of the Definite Article in
English", TDAP #38.
(1963)
"Relative Clause Adjuncts of a Noun", TDAP #47.
Roberts, Paul (1964)
English Syntax, Alternate Edition, Harcourt, Brace & World,
New York.
Robinson, Jane (1966)
"A Dependency Grammar for Transformations", IBM.
(1967)
A Dependency-Based Transformational Grammar, IBM.
Robinson, Jane and Shirley Marks (1965
PARSE: A System for Automatic Syntactic Analysis of
English Text, 2 Volumes, RAND Corporation, Memorandum
RM-4654- RR, Santa Monica, California.
Rosenbaum, Peter S. (1965)
"A Principle Governing Deletion in English Sentential
Complementation", Research Paper RC-1519, IBM.
10U9
BIBLIO-24
(1967a)
The Grammar of English Predicate Complement Constructions.
Research Monograph No. 47, The M.I.T. Press, Cambridge
(previously available as an unpublished dissertation, M.I.T.,
1965).
(1967b)
"Phrase Structure Principles of English Complex Sentence
Formation", Journal of Linguistics, 3.1.103-118.
(1968)
English Grammar II, Section 1 of Scientific Report 2 on
Specification and Utilization of a Transformational Grammar,
IBM.
Rosenbaum, Peter S. and Fred Blair (1966)
Specification and Utilization of a Transformational Grammar;
Final Report, IBM.
Rosenbaum,Peter S. and D. Terence Langendoen (1964a)
"Summary Report: Seminar on English Complement Constructions",
unpublished paper, IBM-M.I.T. Summer Linguistics Working
Paper.
(1964b)
"Prepositional Phrase Adjuncts of 'wise' Class Adjectives",
unpublished paper, IBM - M.I.T. Summer Linguistics Working
Paper.
(1964c)
"On Sentential Adjuncts of Transitive and Intransitive Verbs",
unpublished paper, IBM - M.I.T. Summer Linguistics Working
Paper.
(1964d)
"Toward a Grammar of the English Copula", unpublished paper,
IBM - M.I.T. Summer Linguistics Working Paper.
Rosenbaum, Peter S. and Dorita Lochak (1966)
"The IBM Core Crammar of English" in Specification and
Utilization of a Transformational Grammar, by D. Lieberman,
Scientific Report No. 1, IBM.
Ross, John Robert (1963)
"Negation", unpublished paper, University of Pennsylvania.
1050
BIBLIO-25
(1964a)
A Partial Grammar of English Superlatives, unpublished
M.A. thesis, University of Pennsylvania.
(1964b)
"The Grammar of Measure Phrases in English", LSA Winter
Meeting.
(1966a)
"A Proposed Rule of Tree Pruning", CLHU, NSF 17.
(1966b)
"Relativization in Extraposed Clauses", CLHU, NSF 17.
(1966c)
"Adjectives as Noun Phrases", LSA Winter Meeting.
(1967a)
"Auxiliaries as Main Verbs", unpublished paper (preliminary
version), M.I.T.
(1967b)
"Gapping and the Order of Constituents", Unpublished
paper, M.I.T.
(1967c)
Constraints on Variables in Syntax, unpublished Ph.D.
dissertation, M.I.T.
(1968a)
"On the Cyclic Nature of English Pronominalization", To
Honour Roman Jakobson, Vol. Ill, 1669-82, Mouton & Co.,
The Hague.
(1968b)
"On Declarative Sentences", to appear in Jacobs and Rosenbaum
(eds.), Readings in English Transformational Grammar.
Rutherford, William E. (1968)
Modern English: A Textbook for Foreign Students.
Brace and World, Inc.
Harcourt,
Sadock, Jerrold M. (1967)
"A Note on Higher Sentences", unpublished paper, University
of Illinois.
1051
BIBLIO-26
Schachter, Paul (1961a)
A Contrastlve Analysis of English and Pangasinan, unpublished
Ph.D. dissertation, U.C.L.A.
(1961b)
"Some Problems in the Transformational Analysis of English
Verbs", unpublished paper, LSA Winter Meeting.
(1962)
"Review of The Grammar of English Nominalizations (R. B.
Lees)", UAL 28.134-146.
(1964)
"Kernel and Non-Kernel Sentences in Transformational Grammar",
Proceedings of the Ninth International Congress of Linguistics,
(ed.) H. Lunt, pp. 692-696, Mouton and Co., The Hague.
Schane, Sanford A. (1966)
A Schema for Sentence Co-ordination, Information Sciences
Dept., Information System Language Studies Number Ten, The
MITRE Corp., Beford.
Sgall, Petr (1967)
"Functional Sentence Perspective in a Generative Description," Prague Studies in Mathematical Linguistics 2,
Publishing House of Czechoslovak Academy of Sciences, Prague.
Shopen, Timothy (1967)
"Reference: or, How Anyone Knows What in the World Anyone
Else is Talking About", unpublished paper, U.C.L.A. (PEGS
Paper #35).
Sloat, Clarence (1968)
"Proper Nouns in English", University of Oregon, forthcoming
in Language.
Smith, Carlota S. (1961a)
"A Class of Complex Modifiers in English", Language 37.342365.
(1961b)
"Determiners", M.I.T. Quarterly Progress Report No. 63,
Cambridge.
1052
BIBLIO-27
(1964)
"Determiners and Relative Clauses in a Generative Grammar
of English," Language 40.37-52.
(1965)
"Ambiguous Sentences with 'and'", unpublished paper, M.I.T.
Stockwell, Robert P. (1960)
"The Place of Intonation in a Generative Grammar",
Language 36.360-367.
(1963)
"On Simultaneous Categories and the Noun in English",
unpublished notes for RAND Seminar.
Stockwell, Robert P., J. Donald Bowen, and John U. Martin (1965)
The Grammatical Structures of English and Spanish.
University of Chicago Press, Chicago.
Stockwell, Robert P. and Paul Schachter(1962)
"Rules for a Segment of English Syntax", unpublished paper,
U.C.L.A.
Thomas, Owen (1965)
Transformational Grammar and the Teacher of English,
Holt, Winston, and Rinehart, New York.
Thorne, J. P. (1966)
"English Imperative Sentences", Journal of Linguistics
2.69-78.
Traugott, Elizabeth C. and John Waterhouse (1968)
"Already and Yet: A Suppletive Set of Aspect-markers?",
unpublished paper, University of California, Berkeley,
(PEGS Paper #33).
U.C.L.A. English Syntax Project (1967)
"September Conference Papers on Phrase Structure, Lexicon,
Determiners, Relativization, Passive, Interrogative,
Imperative, Negation, Conjunction, Pronominalization,
Nominalization, Verb Complementation", Distributed through
PEGS (not otherwise available).
Vendler, Zeno (1963)
"The Transformational Grammar of English Adjectives", TDAP
#52.
1053
BIBLIO-28
(1964)
"Nominalization", TDAP //55.
(1967)
Linguistics In Philosophy. Cornell University Press, Ithaca,
New York.
(1968)
Adjectives and Nominalizations, Papers on Formal Linguistics
No. 5, Mouton and Co., The Hague.
Walker, D. E., et al. (1966)
Recent Developments in the MITRE Syntactic Analysis Procedure, Information Sciences Department, Information System
Language Studies Number Eleven, The MITRE Corp., Bedford.
Weinreich, Uriel (1966)
"Explorations in Semantic Theory", in Current Trends in
Linguistics, Thomas A. Sebeok (ed.), Vol. 3, pp. 395-447,
Mouton and Co., The Hague.
Wierzbicka, Anna (1967)
"Against 'Conjunction Reduction'", unpublished paper, M.I.T.
Wilkinson, Robert W. (1968)
"An Investigation into the Nature of English Noun Phrases",
unpublished paper, University of Illinois.
Wilson, Robert D. (1964)
An Algorithm of Derived Constituent Structure, unpublished
Ph.D. dissertation, U.C.L.A.
(1966)
"Review of Indirect Object Constructions in English and
the Ordering of Transformations (Charles J. Fillmore),"
UAL 32.405-409.
Winter, Werner (1965)
"Transformations without Kernels?", Language, 41.484-489.
Wolck, Wolfgang and P. H. Matthews (1965)
"A Preliminary Classification of Adverbs in English",
LRP.
Wolfe, Patricia M. (1967)
"The Operation of Pronominaiization within the NP, with
Particular Reference to English", unpublished paper, U.C.L.A.,
(PEGS Paper #52).
105**
BIBLIO-29
(1968)
"Definite and Indefinite Pronominalization in English",
LSA Summer Meeting.
Zvicky, Arnold M. (1968)
"Naturalness Arguments in Syntax", (Proceedings of the
4th Annual Meeting of the Chicago Linguistic Society).
Zvicky, A. M., et. al. (1965)
"The MITRE Syntactic Analysis Procedure for Transformational
Grammars", The Fall Joint Computer Conference, The MITRE
Corp., Bedford.
1055
Addendum
The addendum to the Bibliography contains several titles
inadvertantly omitted from the original list as well as several which
came to our attention quite late in the preparation of the grammar.
As a result we were not able to take full advantage of the latter in
the presentation of the problems to which they are relevant.
Baker, C. LeRoy (1966b)
Definiteness and Indefiniteness in English, unpublished M.A.
Thesis, University of Illinois.
Bendix, Edward H. (1966)
Componential Analysis of General Vocabulary: The Semantic
Structure of ai Set of Verbs in English, Hindi, and Japanese.
Mouton, The Hague. [and UAL Supplement]
Bowers, J. (1964)
"Generic Sentences in English", unpublished paper, M.I.T.
(1968)
"English Complex Sentence Formation", Journal of Linguistics,
4.di-8».
Carden, G. (1967a)
"The Deep Structure of English Quantifiers", unpublished
paper, I.B.M. Programming Center, Boston.
(1967b)
English Quantifiers, unpublished M.A. Thesis, Harvard.
(1968)
"English Quantifiers"(revised), CLHU, NSF 20.
Dean, Janet (1968)
"Nonspecific Noun Phrases in English", CLHU, NSF 20.
Emonds, Joseph (1967)
"The Place of the Phrase Structure Rules in a Generative
Grammar", I.B.M. Programming Center, Boston.
Jacobs, R. and P. Rosenbaum (1967a)
Grammar I. Ginn and Company, Boston.
(1967b)
Grammar II.
Ginn and Company, Boston.
1056
Addendum - 2
Katz, Jerrold J. and Paul M. Postal (l96Ub)
An Integrated Theory of Linguistic Descriptions.
Press, Cambridge.
M.I.T.
Lakoff, George P. (I968d)
"Repartee: Negation, Conjunction and Quantifiers",
unpublished paper, Harvard.
Perlmutter, David M. (1968b)
Deep and Surface Constraints in Syntax, unpublished Ph.D.
dissertation, M.I.T.
S^renson, Holgersteen (1959)
Word Classes in English.
Copenhagen.
Wagner, K. Heinz (1968)
"Verb Phrase Complementation:
Linguistics, U.88-92.
1057
A Criticism", Journal of
Security Classification
DOCUMENT CONTROL DATA -R&D
(Security classification ot title, body ot abstract and indexing annotation must be entered when the overall report is classified)
1. ORIGINATING ACTI VI TV (Corporate author)
2a. REPORT SECURITY CLASSIFICATION
UNCLASSIFIED
University of California
Los Angeles, California
26. GROUP
N/A
3
REPORT TITLE
INTEGRATION OF TRANSFORMATIONAL THEORIES ON ENGLISH SYNTAX
4. DESCRIPTIVE NOTES (Type ot report and Inclusive dates)
None
5. AUTHOR(S) (Firsi name, middle Initial, last name)
Robert P. Stockwell
Paul Schachter
Barbara Hall Partee
6
REPORT DATE
October 1968
BU. CONTRACT OR GRANT NO.
AF l9(628)-6007
b. PROJEC T NO.
la. TOTAL NO. OF PAGES
7b. NO. OF REFS
532
9a. ORIGINATOR'S REPORT NUMBER(S)
ESD-TR-68-419, Vol. II
9b. OTHER REPORT NOW (Any other number* that may be assigned
this report)
10. DISTRIBUTION STATEMENT
This document has been approved for public release and sale; its distribution is unlimited,
II. SUPPLEMENTARY NOTES
12. SPONSORING MILITARY ACTIVITY
Command Systems Division, Electronic Systems
Division, Air Force Systems Command, USAF,
L G Hanscom Field, Bedford, Mass. 01730
13. ABSTRAC T
Integration of Transformational Theories on English Syntax
This study attempts to bring together most of the information about the
transformational analysis of the grammar of English that was available up through
the summer of 1968, and to integrate it into a single coherent format. The
format chosen is that of C. Fillmore (the "Deep Case" hypothesis) combined with
the "Lexicalist" hypothesis of N. Chomsky. The areas of close investigation were
the determiner system; pronominalization; negation; conjunction; relativization;
complementation and nominal!zation; the systems of interrogative, passive,
imperative, and cleft sentences; the genitive; the lexicon; and the ordering of
rules for these areas of the grammar.
DD
FORM
1473
Security Classification
Security Classification
KEY
WORDS
English Syntax
Transformational Grammar
Case Grammar
Deep Case Hypothesis
Lexicalist Hypothesis
Determiners
Pronominalization
Negation
Conjunction
Relativization
Complementat i on
Nominalization
Interrogative
Imperative
Genitive
Cleft
Passive
Lexicon
Case Placement
Security Classification
Source Exif Data:
File Type : PDF File Type Extension : pdf MIME Type : application/pdf PDF Version : 1.4 Linearized : Yes Page Count : 536 Page Mode : UseNone XMP Toolkit : XMP Core 4.1.1 Metadata Date : 2010:12:15 18:28:56-05:00 Creator Tool : LuraDocument PDF Compressor Server 5.5.50.41 Create Date : 2010:12:15 18:28:56-05:00 Modify Date : 2010:12:15 18:28:56-05:00 Producer : LuraDocument PDF v2.41 Part : 1 Conformance : B Document ID : uuid:uuid:e714651a-4f73-bd93-c66d-4388226989a4 Version ID : 1 Creator : LuraDocument PDF Compressor Server 5.5.50.41EXIF Metadata provided by EXIF.tools