1995
This copy may not be further
reproduced or distributed in any way
without specific authorization in each
instance, procured through the
Director of Libraries, Massachusetts
Institute of Technology.
PM-1 814"x4" PHOTOGRAPHIC MICROCOPY TARGET
(NBS 1010a ANSI/ISO #2 EQUIVALENT
rT “ . ls
WA23 Wd. Was
PRECISION™ RESOLUTION TARGETS
24:1ROOT AND STRUCTURE-PRESERVING TRANSFORMATIONS
by
JOSEPH E, EMONDS
M.A., University of Kansas
(1964)
SUBMITTED IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE
DEGREE OF DOCTOR OF a
PHILOSOPHY
at the
MASSACHUSETTS INSTITUTE OF
TECHNOLOGY
August, 1970
ery
Signature of Author... . eC. ih
Ma Foreign Lite: 8 and Linguistics
March 3, 1970
certified by eae cece eee de.
Thesis Supervisor
Accepted by... Nada. tyin% te eee .
Archives
Chairman, Departmental Comittee
on Graduate StudentsNOTICE
This copy may not be further
reproduced or distributed in
any way without specific
authorization in each instance,
procured through the
Director of Libraries
Massachusetts Institute of Technologya4
ROOT AND STRUCTURE-PRESERVING TRANSFORMATIONS
by
Joseph E. Bnonds
Submitted to the Department of Foreign Literatures
and Linguistics on March 3, 1970, in partial ful!iliment
of the requirements for the degree of Doctor of Philosophy.
ABSTRACT
In this study, I have been motivated by essentially the same problem
as was Ross in his doctoral dissertation (Ross, 1967): that of imposing
constraints on grammatical transformations. ‘Like Ross, I start from the
premise that various conditions and restrictions on when varicus trane-
formations may apply are not ad hoc specifications which are art of the
individual rules, but are rather reflections of some deeper grammatical
principles which define the formal framework in which transformational
rules operate. ‘The constraints proposed in Ross's dissertation (for example
the co-ordinate structure constraint or the complex NP constraint, but the
others as well) essentially specify that constituents cannot be moved out,
of certain structural configurations, even though these configurations
Othervise satisfy the structural deseription of transformation which
would move these constituents. .
This study ie an attempt to define the conditions when the structural
changes of transformations can actually apply, just in the way that Ross
tried to define the conditions when the structural descriptions of trans-
formations are actually applicable (more correctly, "satisfied"), That is,
ny constraints essentially specify that constituents cannot be moved into
certain structural configurations. OO
From a study of English transformations, I have arrived at the
following tentative hypothesis, subject of course to testing and refine-
ment and perhaps rejection on the basis of other languages. Constituents
are to be divided into "phrase nodes" (NP, 8, VP, PP, AP) and “non-phrase
nodes,” Various vay in which the class of phrase nodes my be formally
characterizable are mentioned in appropriate places in the text, but in
any case, they can be listed as a set of substantive universale in lin-
guistic theory. A phrase node X in a tree T can be moved, copied, or
inserted into a new position in T, according to the structural change of
4 transformation whose structural description T satisfies, only if at
east one of two conditions is satisfied: (1) In its new position in T,
X is immediately dominated by the highest § or by any 8 in turn inmediate-
dy dominated by the highest 8. (A transformation having such an effect
is a "root transformation.") (11) The new position of X is a position In
which a phrase structure rule, motivated independently of the transforma-
tion in question, can generate the category X. (A transformation having
such an effect is a "structure-preserving transformation.)dat
Non-phrase nodes may also be moved, copled, or inserted by either a
root or a structure-preserving transformation, but they may also be moved
by @ third type of transformation: a single, specified non-rhrase node
may be moved over a single specified adjacent node (but not over more
than one such node or over @ variable). This is a "minor mcvement rule."
(Further restrictions are placed on this type of rule in Chapter V.)
It is interesting to note at the outset that such important trans-
formations as wh fronting, dative movement, there insertion, adjective-
over-noun movement, complex NP shift, and (verbal) affix movement®
(among many others, depending on the analysis used) are counter-examples
to this claim in their usual formations. Such transformations will be
discussed in the text in the appropriate place.
* ACTix movement becomes a counter-example when one realizes that
verbal affixes mst move over pre-verbal adverbs as well as verbs.iv
ACKNOWLEDGEMENTS
Many thanks are owed to all my teachers and fellow students at
M.I.T, and elsewhere who discussed this material with me, but most
of all I have benefited from the close readings and careful criticisms
of Noam Chomsky, Morris Halle, John Ross, and Michael Brame. I also
owe mich to extended discussions with Ray Jackendoff, John Bovers,
Frederick Newmeyer, Ken Hale, Adrian Aknajian, and many others.
I would like to thank Katrina Streiff for preparing the final copy,
and to apologize at the same time (though words without deeds mean
little) for utilizing the unfair divisions of lebor of our academic
world which inevitably heap the boring work of typing on women who
have better things to do.TABLE OF CONTENTS
CHAPTER I: ROOT TRANSFORMATIONS L
I.1 Preliminaries i
1.2 The Phrase Structure Rule for Expanding S 4
3.3 Subject-Auxiliary Inversion 5
I.4 Verb Placement in German 10
1.5 Tag Questions 12
1.6 Adverb Preposing 14
1.6.1 Negated Constituent Preposing 4
1.6.2 Directional Adverb Preposing 15
1.7 Parenthetical Clauses 17
1.7.1 Another Use of Suject-Simple Verb
Inversion 17
1.7.2 Quote Preposing 19
1.7.3 Non-Factive Parentheticals al
1.8 Some Other Preposing Rules 2h
1.8.1 Topicalization ak
1.8.2 VP Preposing ek
1.9 Right and Left Dislocation 25
1.10 Preposing around Be 27
Conclusion to Chapter T 32
CHAPTER II: STRUCTURE-PRESERVING NP MOVEMENT
TRANSFORMATIONS 33
II.1 The Passive Construction 3h
II.1.1 The Active-Passive Relation in
English 3h
II.1.2 The Structure-Preserving Property 37
1.1.3 The Passive Rules in the Structure-
Preserving Framework 39
II.1.4 The Optionality of the By Phrase
in the Passive 43
II.2 There Insertion 47
12.2.1 There Insertion as Structure-Preserving 7
II.2.2 The Status of the Auxiliary Be 55
II.3 Dative Movement Rules ~ 62
Ir.4 It Replacement 61
II.5 Conjunct Movement 2
11.6 ‘he Possessive Transformation B
II.6.1 The Phrase Structure Rule for
Expanding NP ‘ 13
11.6.2 A Comparison of NP Preposing and
the Possessive Transformation 78
II.7 Obligatorily Reflexive Verbs 8
11.8 There Replacement 88
Conclusion to Chapter IT 89vi
CHAPTER III; THE STATUS OF EXTRAPOSITION OL
TII.1 Arguments that Sentences and Infinitives
Are not NP's 93
TII.1.1 Two Exceptional Classes of Gerunds 93
III.1.2 The Deep Structure of Sentence and
Infinitive Complements 95
III.1.3 The Position of Object Complements 97
III.1.4 Sentences and Infinitives in Other
Non-Subject NP Positions 99
TII.1.5 Sentences and Infinitives in
Subject Position 103
III.1.6 Other Arguments Confirming the
Non-NP Status of Infinitives and
Sentences 210
III.2 An Alternative to Rosenbaum's Explanations 119
IIL.2.1 Rosenbaum's Noun Phrase Complements 119
III.2,2 Rosenbaum's Verb Phrase Complements 125
Conclusion to Chapter IIT 133
CHAPTER IV: EXTENDING THE STRUCTURE-PRESERVING
CONSTRAINT 13h
IV.1 Adjective Phrase Movement Rules 23h
IV.1.1 Adjective Movement 23h
IV.1.2 Adverbial Adjective Phrases ay
Iv.2 Structure-Preserving Rules Which Move 8 Nodes 163
IV.2.1 Cleft Sentences 163
IV.2.2 Extraposition from Noun Phrase aye
IV.2.3 Complement Extraposition aps
IV.3 The Node PP and its Distribution 180
Iv.4 Wh Fronting as a Structure-Preserving Rule 190
IV.5 Complex NP Shift 204,
CHAPTER V: MOVEMENTS OF NON PHRASE NODES 206
V.1 Structure-Preserving Movements of Pre-Terminal
Symbols 208
V.1.1 NEG Transportation 208
V.1.2 Affix Movement ann
Y.1.3 AUK Movement 218
V.1.4 Some Considerations on Clitic Placement
in French and Spanish 226
V.2 Minor Movements of Pre-Terminal Symbols 234
V.2.1 DEG Postposing 234
V.2.2 DET Incorporation ' 235
V.2.3 Particle Movement 236
V.2.4 Article Movement 237
V.2.5 Verb Clitic Inversion in Spanish 238,
V.2.6 Quantifier Postposition 238
V.2.7 Further Restrictions on Minor Novement
Rules ae
Bibliography 2hbyv
CHAPTER I: ROOT TRANSFORMATIONS
This study is an attempt to restrict the notion of "gramatical
transformation” in a generative grammar. In particular, I will claim
that those synchronic, syntactic rules that have until now been
included in the rather general class of transformations are in fact
members of one of three much more restricted classes of grammatical
rules. These three kinds of rules will be called “root transformations,"
"structure-preserving transformations," and "minor movement rules."
The definitions of these classes and the arguments for their
existence will require discussion of many syntactic processes. An
adequate formulation will entail making revisions and innovations in
the accepted forms of @ transformational generative grammar, so it is
well to begin by reviewing those grammatical notions that provide a
relatively fixed basis of the discussion.
I.1 Preliminaries I assume familiarity with many of the notions of
transformational grammar; in particular, with those underlined in this
section. A generative grammar consists of semantic, phonological, and
syntactic components. Part of the syntactic component is a set of
phrase structure rules. A phrase structure rule consists of « synbol
A on its left-hand side and a sequence of symbols B. . . C on its
right-hand side. .
‘These rules define a set of formal objects called phrase-markers,
which can be diagrammatically represented as trees. In exposition,
I will ignore this distinction, referring to phrase-markers as trees.
‘A partial tree can be constructed (graphically) from a phrase structurerule R by writing its left-hand symbol A over its right-hand symbols
B. ..C, and connecting A to each of B,. . ., C by lines. A is then
said to be expanded as B. . . C by R. A full tree is then constructed
by expanding in turn each of B, . . . , C according to the phrase
structure rules, and subsequently expanding the results of this until
pre-terminal symbols are reached. Pre-terminal symbols are defined
as those which do not appear on the left-hand side of phrase structure
rules. To complete the tree, a terminal element, chosen from the
lexicon, is placed under each pre-terminal element. ‘The terminal
elements are lexical entries of semantic, phonological, and syntactic
features; we will use capitals to represent all synbols in phrase
structure rules and all features, but we will abbreviate entire lexical
entries by using, as closely as possible, English orthography (underlined).
‘Te lexicon and the phrase structure rules together comprise the base.
‘The trees defined or generated by the base are called deep structures.
They are subject to a linearly ordered series of mappings called
transformations, each of which maps input trees which satisfy its
structural description into output trees according to the instructions
in the transformation's structural change. The trees which result from
applying in order to the deep structures all the transformaticns whose
structural descriptions are satisfied are the sutface structures, These
are the input to the phonological component. It has previously been
held that the deep structures were the input to the semantic component,
but some recent work, especially that of Chomsky (forthcoming) and
Jackendoff (1969), indicates that thie input may consist of peirs of a
deep and a surface structure (DS,SS), where SS is the surface structure
‘that results from applying all the transformations to DS.‘The branch points in the trees are nodes, whose labels are given
by the phrase structure rules. Node A (immediately) dominates node B
if iv appears (immediately) over B in the path from B to the highest
node in the tree, called the root of the tree. A string of lexical
entries (morphemes) which is dominated by a node labeled A (more briefly,
dominated by A) when A dominates no other lexical entries, is a
constituent of type A; i.e., it is an A.
If the reader is not familiar with the terms underlined adove, he
should consult discussions of them in Chomsky (1965) and Chomsky (1957).
Although the aim of this paper is to define three restricted classes
of transformational rules, possible membership in these classes will be
seen to depend on the phrase structure rules and not on the notation
used to write the transformations. Since the empirical content of my
claims will rest on the precision and adequacy of the phrase structure
rules, mich of the discussion will center on them. .
If my claims prove to be correct, it may be that the transformational
notation devised by Chomsky in earlier work should be revised in order
to reflect the restrictions on transformations imposed by these claims.
In fact, different notations may be appropriate for the three classes
of rules. ‘These questions have not been approached in this study;
suffice it to say that, at least for expository purposes, the
transformational notation of Chomsky (1961) has been adequate.
I assume familiarity not only with the notion "phrase structure,
rule," but also with the particular phrase structure rules that have
been proposed for English, as, for example, those given in the second :
and third chapters of Chomsky (1965). But none of these will in the
end be adopted without some revision, All the rules that will be usedwill be introduced explicitly in the first discussion that directly
concerns them.
A few other preliminary notions may be useful. The nodes N, V, A,
(w= noun, V = verb,
M, and P will be referred to as Lexical nodes.
A = adjective, M= modal verb, and P = preposition.) The nodes NP, VP,
AP, S, and PP will be celled phrase nodes. A relation holds between
lexical nodes and phrase nodes: each phrase node can inmediately
dominate at most one lexical node. A lexical node dominated by a phrase
node is said to be the head of that phrase node. Furthermore, NP's
mey have only N heads, and similarly for VP and V, AM and A, S and M,
and PP and P. (But I do not claim that a phrase node must have a head. )
The class of phrase nodes may be characterizable as the only nodes under
which an unlimited amount of recursion i8ypermitted in surface structure,
but I have not investigated this possibility in detail.
‘The above restrictions do not rule out the possibility of phrase
structure rules which expand phrase nodes as a sequence of non-lexical
nodes. Nor do they exclude lexical (or other) nodes from dominating
phrase nodes; however, @ restriction on recursion under non-phrase
nodes would limit this latter phenomenon.
In attempting to define and delimit the classes of root and
structure-preserving transformations, I will confine myself to studying
the grammatical processes of English, unless another language is explicitly
mentioned. The permanent linguistic value of the notions to be intrpduced,
of course, cannot be established until they are tested and refined through
a study of other languages.
I.2 The Phrase Structure Rule for Expanding § The root of e tree,
as mentioned earlier, is the highest node. In English, and probablyuniversally, the root must be 8. We need now to write the phrase
structure rule for expanding 8, since this rule plays an important
part in defining root transformations. In writing this rule, I exclude q
from the constituent "verb phrase" (VP) certain elements which do not
occur’ in English in the clauses traditionally termed "non-finite";
i.e., in infinitives, participles, and gerunds.* These elements include
present and past tense verb endings (TENSE) and modal auxiliaries.
Thus, I take the underlined phrases in (1) to be VP's.
Q) John continued counting his money.
To have been chosen for this post is a great honor.
He helped by being cheerful at the right moment.
Formal justification for using the symbol VP in this way will be given
in Chapter V. The resulting S expansion rule is (2): |
(2) S —» COMP - NP - TENSE - (M) - VP ‘
COMP is the morpheme that, and perhaps than and as. M is will, can, |
may, ete. Subject pronouns occur only in NP's immediately doninated
by 8. TENSE in (2) can be either the morpheme s/f ( the present tense
ending) when it (TENSE) occurs with the syntactic feature -PAST, or the
morpheme e4 (the past tense ending) when it occurs with +PAST.
1.3 Subject-Auxiliary Inversion Rather than sét down immediately @
definition of root transformation, I will approach this central idea
gradually, through the rules which it covers. In this way, needed
modifications of accepted terms can be made more plausible, and apparent
*POBS-ING and FOR-TO complements in the terminology of Rosenbaum (1967).counter-examples can be eliminated before hand.
We will consider first the subject-auxiliary inversion which occurs
in direct questions, except when the subject itself is questioned. By
auxiliary in this study I will always mean Just those verbal elements
which undergo this rule: modals, have (in certain uses), be, and do.
Whether or not thie term, used here for convenience, corresponds to some
syntactic feature or node AUX will be discussed in later sections and
chapters. This inversion is also used in certain exclamations, wishes,
and sentences with preposed negative constituents.
(3) Wasn't that brave of him!
Isn't it cold out!
May you always be as thoughtful as she was!
Never in my life have I spoken to him.
‘The mechanics of this transformation are well-known; if the subject
of the highest 8 is preceded by a questioned (wh) or a negated constituent,
the order of the subject and the following auxiliary, which includes
TENSE affix and a possible n't are reversed. According to Katz end Postal
(1964), there is an underlying preposed questioned constituent whether
even in yes-no questions, which causes auxiliary inversion like other
wh constituents. This whether appears explicitly in embedded questions.
(4) John wonders whether Mary is coming.
‘The question of whether they will support us is still undecided.
Iwill show in Chapter V that the first post-subject auxiliary ,
(including do, have, and be) is always in the M position when subject-
auxiliary inversion applies. I will assume that the rule essentially
changes (5a) into (5b), where X ie a node that dominates wh or Nid.(5) (a) (b) 8
coe be
The important characteristic of this inversion rule for our
purposes is that it applies only in the highest 8; it doesn't apply
in relative clauses and indirect questions:
(6) We talked about how we would escape.
She doesn't know why they weren't co-operative.
The papers which you burned were important.
“Highest 8," hovever, does not describe exactly the contexts for
subject-auxiliary inversion, since this rule also applies in conjoined
sentences immediately dominated by the highest S:
(7) She didn't do the dishes, and why should she?
I know it was expensive, but never in my life have I been
so thrilled.
When is he coming, and where is he from? ‘
Come in right now, or do I have to use force?
(The conditions under which such mixed sentence types can be conjoined
are not known with precision, but this is of no consequence here.)
Sometimes the conjoined sentence exhibiting inversion 1s only an
abbreviated (tag) form of a full sentence. Thus, "neither-auxiliary-NP"
is derived from "NP-auxiliary-not either" by preposing the negated
constituent not either, which in turn causes the inversion, Like full
sentences containing inversion, such tage cannot be embedded:
(8) *Bill didn't come to the party because neither did Mary.
*John thought that Bill hadn't come, and that neither had Mary.
‘Mery doesn't know why Susan is leaving, and we don't know
why is she either.
The affirmative tag "NP-auxiliary-too" can also be paraphrased bypreposing too and changing it to so, yielding, after inversion,
auxiliary-NP," Again, the inverted form can't be embedded.
(9) *I am in great danger, and the knowledge that so are you
isn't very comforting.
*I'm worried, and I'm sure so are you.
Inversion in negative imperatives also takes place only in a root
8 or in a conjunct S immediately dorinated by an 8:
(ao) Don't you be so impolite!
Don't you talk like that, or I'll send you home!
Don't anybody move!
Another class of 8's which are not strictly speaking the highest
8's and in which subject-auxiliary inversion applies are the "reported
sentences" or "quotes" of direct discourse.
(11) John said, "What should she eat?"
The man asked, "Can my son have a key?"
She exclaimed, "Ien't it cold out!"
As we proceed, we will see that the class of 8's in which subject-
auxiliary inversion regularly occurs in English 1s also the cless of 8's
in which a great many other rules also apply. The purpose of this chapter
is to enumerate and study this class of rules, which I will call "root
transformations." For this, we need two definitions.
Definition From this potnt on, a root will mean either the highest
8 ina tree, an 8 inmediately dominated by the highest 8, or the reported
8 in direct discourse.
I do not rule out the possibility that the last condition in the
preceding definition is superfluous, since the surface structure of
direct discourse may well be as in (12).(a2) Ed
——
a a
4S
wo OWE eo OMP OUP
| lia dhe
John V what should she V
|
said eat
In the cases so far covered, a root S inmediately dominated by
another S is set off by commas. We will be able to extend this to
other cases as we proceed.*
* Something should be said here about the inversions that occur
after the S-introductory words than and as. The auxiliary inversion
that occurs in "sentence relatives” introduced by as is to be expected,
since these 8's are probably root 8's.
John mst do his own laundry, as must every student here.
Twas looking for faults in his presentation, as was my friend.
Comparative clauses introduced by than and as are not root 8's. They
therefore should not exhibit inversion according to the rules of Modern
English proposed in this paper. If the following sentences are completely
regular and to be accounted for by the subject-auxiliary inversion rule,
there is a deficiency in my analysis.
We saw the same man as did John
She hasn't bought as many souvenirs as has her husband.
She spoke more convincingly than did Harry.
‘Bill seems smarter in math than does Harry in science.
‘The Chinese are asready to fight as are the Japanese to talk.
*I hope you found the play more interesting than did we.
xOur friends can't afford to buy records as often as can you.
John likes Beethoven more than do I.
¥iie'11 cut more cake than will Mary be able to eat.
However, the fact that pronouns cannot invert in the above sertences
indicates that subject-auxiliary inversion does not operate the same
way in as and than clauses as it does in questions, exclanations, etc.
Furthermore, even the acceptable sentences with inversion are somewhat
less acceptable than their counterparts without (inversion:
We saw the same man as John did.
She hasn't bought as many souvenirs as her husband has.
She spoke more convincingly than Harry did.
Bill seems smarter in math than Harry does in science. +
‘The Chinese are as ready to fight as the Japanese are to talk.
Another difference between inversion in as and than clauses and normal
subject-auxiliary inversion is that the former is optional and the latter
is obligatory.
For these reasons, I am willing to regard the as and than clauses
(other than sentence relative as clauses) that exhibit iavers‘on as either
not fully grammatical in English or as derived by means of sone secondary
inversion rule which is not a root transformation but rather a "mnor
movement rule" of the type defined and described in Chapter V.10
Definition A root transformation is one in which any constituents
moved, inserted, or copied are inmediately dominated by a roct in the
derived structure.
Of cow'se, to simply enumerate a class of grammatical rules which
have @ certain property, as I will do in the remainder of this chapter,
4s without empirical interest, if nothing further can be said about the
rules which do not have the property. However, I will try to show in
subsequent chapters that non-root transformations are subject to severe
restrictions, so that the class of possible grammatical transformations
which can attach nodes to non-roots is greatly narroved. In the
remainder of this chapter, I want to precisely delineate that class
of transformations in English (root transformations) which witl not be
subject to the constraints offiFnsformations proposed later in this
study.
Our first example of a root transformation is of course subject-
auxiliary inversion. On the other hand, 1t should be clear that a great
many Well-knoyn transformations are not root transformations. For
example, the passive transformition, affix movement, and wh fronting,
fas these have been formated in transformational literature, are not
root transformations, since they can apply in embedded sentences freely.
I.4 Verb Placement in German German is in interesting contrast to
English with regard to the placement of the verb. In German, the
* Ina theory which distinguishes between "cyclic" and "last cyclic"
transformations, it is clear that every root transformation is last-
cyclic, although the converse is not necessarily true. There are problems
connected with these notions which I'will not treat here. It vould be
extremely interesting if the class of root transformations and the class
of last-cyclic rules were co-extensive.
\n
finite verb, i.e., the verb which carries the tense ending, is generally
in second position in @ root.
(3) Gestern ist er nach Hause gekonmen,
Welches Buch kinnen die Studenten nehmen?
Mit einem Messer wird er den Kuchen schneiden.
Hans stahl ein Buch.
Mich hat er geschlagen.
In embedded (non-root) sentences the finite verb is last in the verb
phrase. (In this section,
ast" means "last except for sentence
complements .")*
(ab) Thm tat es leid, dass er gestern nach Hause gekommen war.
Er weiss nicht, welche Biicher die Studenten genohmen haben.
Er wird mehr Kuchen mit dem Messer schneiden, als ich essen kann.
Weil Hans einen Bleistift gestohlen hat, wird er bestraft.
Dass er ein Auto stehlen wiirde, erstaunte seine Eltern und
seinen Lehrer.
If the German verb is lest in its clause in deep structure, as
argued in Bierwisch (1963), the rule which moves it to second position
in root S's can be formulated as a root transformation. That is, the
V moved can be attached to the highest S. The alternative position would
‘be to assume that the German verb is generated in second position by
* German has two indirect discourse constructions. In one of these,
the reported sentence is like other non-root S's in that it is introduced
by dass, its verb is in final position, certain fronting transformations
cannot occur in it, etc.
Er sagte dass er krank sei.
"Er sagte dass gestern er nach Hause gekonmen sei. .
"Er sagte dass mich sie geschlagen habe.
In the other type of indirect discourse, the reported sentence, like the
reported sentence in direct discourse, is a root 8. The verb is second,
certain fronting transformations can apply; dass does not introduce the
clause, and the clause is set off by a comma,
Er sagte, er sei krank.
Er sagte, gestern sei er nach Hause gekommen.
Er sagte, mich habe sie geschlagen.a
1
the phrase structure rules and is moved, in non-root 8's only, to final |
position in the VP. Cf. Ross (1967).
The constraints I will place on movement transformations in
embedded sentences in subsequent chapters are in fact inconsistent
with this letter assumption, as they will in principle exclude such a
movement rule. (At the same time, they predict that a rule moving the
verb from final position to second position could only apply in root
S's.) Thus, if the constraints on transformations which I will propose
in this study are correct, the German verb must be last in its clause
in deep structure.
I.5 Tag Questions A tag question is a declarative sentence, followed
by a repetition of the first auxiliary (do if the preceding declarative
has no auxiliary) and a pronominalized form of the subject. ‘Thus, some
typical tags in teg questions are shouldn't he? hasn't he? dia they?
Were you? etc. Two rules are probably involved in forming teg questions:
a tag formation rule copies an entire declarative sentence with probably
the addition of whether (deleted, as usual, in roots in surface structure)*,
and the subsequent deletion of what follows the first auxiliary is due to
an independently motivated "VP deletion" rule.
(15) Mary had come, hadn't she?
Mary won't buy this dress, will she?
Bill dates someone, doesn't he?
You were dissatisfied, were you?
T assume that the surface structure that results from the tag
formation rulesis as in (16). The fact that the right-hand § in (16)
* I am !gnoring the problems associated with positioning NEG
correctly in these constructions.13
is a root means that subject-auxiliery inversion may apply. Also, the
comma that sets off the tag from the preceding declarative is consistent
with the remark about commas and root 8's made in section 1.3.
(16)
‘The tag formation rule itself is a root transformation, since tag
questions are never non-roct S's; the highest S immediately dominates
the tag S.
(27) *Bill wanted to know whether Mary had come, hadn't she?
¥BLl1 asked if he could date someone, could he?
*The question of who Mary dates, doesn't she, doesn't bother him.
*The idea that Bill knew whether Mary had come, hadn't she,
, is preposterous.
Sometimes the tag S is a copy of the complement to I imagine,
I suppose, I guess, etc., as in (18).
(18) I imagine he is dating my wife, isn't he?
I guess he likes foreign beers, doesn't he?
However, the definition of a root transformation Only demands that the
constituents copied by tag formation be inmediately dominated by a root 8
in derived structure. There is no reason to believe that the tags in’
(18) are not so attached, given the examples of (19).
(19) The idea that I imagine he is dating my wife (4sn't he)
bothers him.
‘The foreign beers that I guess he likes (*doesn't he) are
expensive.wy
1.6 Adverb Preposing Several kinds of adverbs may be preposed in @
sentence. A basic division can be made between those which, in preposed
position, are followed by an optional (sometimes obligatory) coma or
preath pause, and those which are not. This division is further justified
by noting that the former class never causes inversion in the main clause,
while the latter class in almost every instance does.
Possible deep structure sources for the former class of adverbs,
and the question of whether they can freely appear in embedded sentences
will be discussed in cection IV1.2; I will say no more about them here.
‘The latter class, which we can call inversion adverbs, results from root
transformations, and hence is subject matter for this chapter.
1.6.1 Negated Constituent Preposing Preposed negative constituents,
which are sometimes NP's but are usually adverbs, cause subject~auxiliary
inversion and are never separated from the main clause by @ conma.
(20) Under no conditions may they leave the area.
Never have I had to borrow money.
At none of the beaches are the lifeguards alert.
Only on weekends did I see those students.
Tt was expensive, but seldom has John been so pleased.
Few movies have we enjoyed so thoroughly.
Nothing did I see that I liked.
In not many years will Christmas fall on Sunday.
(cf. In not many years, Christmas will fall on Sunday.)
Examples generally indicate that this preposing is limited to 8's
which are roots.15
(21)* -*If under no conditions may they leave the area, hov can they
pay their debt?
2 have worked so hard that never have I had to borrow money.
*he proof that at none of the beaches are the lifeguards alert
is that there have been many fatalities.
*The students that only on weekends did I see are living in the
country now.
‘We wouldn't do it again, even though seldom has John been so
pleased.
*The employees are happy that in not many years will Christmas
fall on Sunday.
1.6.2 Directional Adverb Preposing If the verb of a sentence is in the
simple past or present tense (no auxiliaries being allowed), a preposition-
al phrase indicating spatial direction may be preposed. This includes
adverbs of direction which, like other adverbs of time and space, will be
considered intransitive prepositions in this paper. (For example, away,
around, down, etc.) This construction seems limited to exclanatory
statements, Also, the simple present seems to paraphrase the present
progressive of sentences with normal word order.
* Some of the examples of (21) do not sound totally unacceptable.
However, the fact that auxiliary inversion clearly occurs after wh
constituents only in roots indicates that such examples may not be
strictly grammatical.
Moreover, the general heuristic subscribed ,to here for classes of
sentences of doubtful grammaticality (acceptability judgments in such
classes being erratic) is that they are ungrammatical, provided that they
are not semantically difficult or of undue length or embedding. The
reason for this is that it would be hard to explain even slight unaccept-
ability for sentences that are relatively short and simple, semc-.tically
clear, and perfectly grammatical. But it is to be expected that intelligent
language users vould possess strategies of interpretation to render sen-
tences that are relatively short and simple, semantically clear, and
slightly ungremmatical perfectly understandable and nearly perfectly
acceptable. According to this heuristic then, I teke the sentences of
(21) to be ungrammatical,16
(22) In came John!
Down the street rolled the baby carriage!
Up trotted the dog!
Round and round spins the fateful wheel!
Here he comes! (Cf. the synonymous "he is coming here.
Away they rant
Substituting verb forms containing auxiliaries into (22) produces
ungrammatical. examples.
The sentences of (22) are not obtained by simply exchanging the
first and last constituents of the corresponding declaratives, since
pronoun subjects are not inverted. Rather, two processes seem
involved; the first is the preposing of the adverbial PP, which is
perhaps part of some other rule, and the second is the movement of the
simple verb into second position. This is similar to the movement of
the auxiliary into second position after the preposing of wh and negated
constituents. (Cf. section I.3.) I will call this subject-stmple verb
inversion (because no auxiliaries can be involved); at the point of this
inversion in a derivation, the pronoun subject is apparently already
attached to the verb form as a prefix, so that the verb is already in
second position after the preposing of the directional adverb.
Both of these rules, directional adverb preposing and subject~
simple verb inversion, take place in the highest 8; 1.e., they are
root transformations.
(23) *I noticed that in came John. .
*It seems that away they ran.
*The fact that down the street it rolled amazed her.
*I was surprised when up trotted the dog.
Simple verb inversion cannot be combined with auxiliary inversion
because (i) in this case, if an auxiliary is present, the rule does not1
apply, whereas in the other case, the rule only applies to auxiliaries,
and (11) pronominal subjects do not count as taking up the second
position in simple verb inversion, but in auxiliary inversion they do.
Actually, it is hard to combine even the preposing of the directional
adverb with any other rule, because it depends on the sentence's verb
form not containing any auxiliaries, and no other adverb preposing rule
has this proviso.
(2h) *In John was coming!
Down the street the baby carriage was rolled!
*Hlere he does come!
*Round and round the wheel has spun!
‘Away they didn't runt
It appears therefore that directional adverb preposing and subject-
simple-verb inversion are distinct rules, and that moreover they are
respectively separate from other adverb preposing rules and subject-
auxiliary faversion. If this is true, two adverb preposing mules are
root transformations, negated constituent preposing and directional
adverb preposing, and two subject inversion rules are root transformations,
subject-auxiliary inversion and subject-simple verb inversion. Although
@irectional adverb preposing never occurs without subject-sinple verb
inversion, the latter has another use, as described in the following
section.
1.7 Parenthetical Clauses
T.7.1 Another Use of Subject-Simple Verb Inversion A main verb in’
English without auxiliaries inverts over the (non-pronoun) subject
(1) 4f a directional adverb is preposed, as discussed above, and (11)
if part of a direct quote which in deep structure is a complement to18
‘this verb is preposed. The identical conditions on absence of subject
pronouns and auxiliaries leave little doubt but that a single rule is
involved. This inversion is optional in most dialects. (Nothing depends
on this in what follows.)
(25) In John came!
Down the street the baby carriage rolled!
Round and round the fateful wheel spins!
Into the parking lot the car lurched!
‘There John goes!
2Up the dog trotted!
In case (11), the inversion is clearly optional, since normal or inverted
word order is allowed in (26).
(26) "John may come," said Mary, "but he will not be welcome."
"John may come," Mary said, "But he will not be welcome."
"On weekends," exclaimed Bill, "I really feel lonely."
"On weekends," Bill exclaimed, "I really feel lonely."
Of course, if the conditions for subject-simple verb inversion are not
satisfied, normal word order mst be used.
(27) *"John may come," said they, "But he will not be welcome."
*"0n weekends," did Bill exclaim, "I really feel lonely."
‘The examples of (23) show that one use of subject-simple verb
inversion can not occur in embedded sentences. Its use with preposed
quotes cannot be embedded either: ‘
(28) *"The game 1s over," I think exclaimed Bill.
*I know that, "let's get out of here," shouted Joha.
Thus, subject-simple verb inversion 1s a root transformation in all
its uses.19
1.7.2 Quote Preposing When part or all of a direct quote is placed
at the beginning of a sentence so that it precedes the clause in which
it wae originally embedded as in the examples of (26), this latter
clause is then called a parenthetical, or, more accurately, a direct
quote parenthetical. ‘These parenthetical clauses are clearly the highest
$ in deep structure (the entire quote being a complement to the verb in
the parenthetical), and, as the possibility of subject-simple verb
inversion in (26) shows, they are the highest 8 in surface structure
also. (This is not circular because we demonstrated in connection with
directional adverb preposing that this inversion was a root transformation. )
Therefore, for examples like (29), the deep structure is (30). As
was pointed out in section I.3, the circled 8 in (30), at least when the
rule of subject-auxillary inversion is reached, 1s a root.
(29) "John will buy any dress," Sue said, “which he thinks Mary likes
"John," said Sue, "will buy any dress which he thinks Mary likes.”
(30)
‘ion he thinks =.
May likes"
For explicitness, I assume that direct quote preposing works as follows:
when part of the circled (root) 8 in (30) 1s preposed, both this part
and the part left behind are dominated by 8, as in (31). (Circled 8's
>20
in (31) are roots.)
(31)
r 7
John will she Vv @®
buy any \
dress sald 8
which he wilh
thinks Mary buy any
Likes dress which
he thinks
Mary likes
The non-branching 8's in (31) are pruned (absent) in surface structure.
The comma or breath pause which separates the parenthetical from
the quote material is predictable from the presence of root (circled)
8's in (31), Since some of these roots can be pruned, it wuld appear
that the comma is present before pruniag. ,
The moat important observation to be made is that this preposing
of part or all of a quote is a root transformation. For the preposed
material must be moved to the front of the entire sentence, which implies
that it 1s attached to the root 8. (No one would claim that the preposed
quote is attached to the subject NP.)
(32) *I think, "The game is over," Bill exclaimed.
*I started, "I like beef," to say, "put not beef liver."
“The game is over," I think Bill exclaimed.
"I like beef," I started to say, "but not beef liver."
Incidentally, the examples in (33) indicate that a condition on
this rule is that the quote be sentence final in the deep structure;
this has no bearing on matters here.a.
(33) John said, "I know I'm guilty, but I shouldn't be punished,”
Just to attract attention.
know I'm guilty," John said, “but I shouldn't be punished."
know I'm guilty," John said, "but I shouldn't be punished,
Just to attract attention.
*"
It will be obvious when the constraints on non-root transformations
are defined in subsequent chapters that direct quote prevosing violates
such constraints. My general claim then would predict that direct quote
preposing should always attach material only to roots. This is borne
out, as shown above, and moreover my analysis is consistent with the
appearance of verb inversion in, and commas around, the parenthetical
clause itself.
I.7-3 Non-factive Parentheticals There is another kind of parenthetical
besides the direct quote parenthetical, which almost certainly is related
to it. Some interesting properties of this construction have been studied
by Rardin (1968).
The construction results from placing part or all of certain non-
factive complements at the beginning of a sentence so that this precedes
the clause in which it was originally embedded. Some typical non-factive
parentheticals are it seems to me, it is true, we can assume, one would
think, it is believed, I don't doubt, etc. Any of these, for example,
can be inserted in (34) at the blank.
(3h) John will buy any dress, ) which he thinks Mary likes.
In many cases, the positions in which a non-factive parenthetical can
occur ere just those where a direct quote parenthetical occurs. (Cf. (29)).
If and how these two types of parentheticals differ in their distribution22
cannot be determined by a few more random examples, so I will leave the
pursuit of this matter to the reader's curiosity. But the sinilarity
is striking enough to lead us to believe that the syntactic analyses for
the two cases must be parallel.
Among other things, it is easy to show that non-factive preposing
4s like direct quote preposing in that it too is a root transformation.
Parallel to the non-grammatical examples of (32) are those of (35):
(35) *It amazed me that a criminal, they claimed, had no rights.
*John replied that we had forgotten, it seemed to him, our
origins.
*Te fact tha. the econamy, we were led to believe, was sound
caused us to invest in stocks.
If non-factive preposing were not a root transformation, the examples of
(35) would be grammatical paraphrases of those in (36).
(36) It amazed me that they claimed a criminal had no rights.
John replied that it seemed to him we had forgotten our origins.
The fact that we were led to believe the economy was sound
caused us to invest in stocks.
My analysis of non-factive parentheticals would be roughly the same
as that given for direct quote parentheticals. The material preposed
as well as that left after the parenthetical clause are dominated by B's.
‘At Least the preposed material is dominated by @ root 8 (since this 8
ie immediately dominated by the highest 8), Thus, the surface structure
of (34), with the 8 which is pruned out circled, is as in (31).23
om
8 VP
John will L
buy any dress
which he thinks
Mary likes
‘he fact that the parenthetical clause and the preposed mater‘al are
both under root S's is attested to by the inversion which is sametimes
possible in one or both clauses. (I make no attempt to predict such
inversions; this analysis predicts only that they are possible. )
(38) Will John buy any dress, do you suppose, whic’ he thinks:
Mary likes?
John 4s quite a bore, don't you think?
Will he come in, I wonder, right during the dinner?
A final group of sentences should be mentioned.
(39) Would her father die at night, Mary wondered, or would he
survive until daybreak?
(40) Mary wondered if her father would die at night or if he would
survive until daybreak.
Tt seems unlikely that (39) is derived from (40). If dt were, we could
not explain why the 6 complement to wonder in (40), which is not a root,
suddenly becones a root in (39). I.e., why 1s there inversion in the
second clause in (39) but not in (40)?) Rather, 1t 1s more plausible
that (39) is derived from (41); the construction exhibited tn (42) 18
analogous to the indirect discourse without dass construction in German
in that the complement 8 of wonder 1s @ root 8.
(42) Mary wondered: would her father die at night or would he
survive until daybreak?2k
1.8 Some Other Preposing Rules
3.8.1 Topicalization Consider the following sentences:
(2) These steps I used to sweep with a broom,
Each part John examined carefully.
Our daugiters we are proud of,
Poetry we try not to memorize.
These examples result from a transformation called “topicalization," = ——
which moves noun phrases to the front of the sentene that is, it
attaches them to the highest 8, This transformation ie a root trans-
formation, since examples like (42) cannot be embedded.
(43) "Have I shown you the broom (that) these steps z used to
are proud of.
*Are you aware (of the fact) that poetry we try not to memorize?
XDo you think Socialist theory many Czechs wourd deny?
“That this house he left to a friend was generous of him.
1.8.2 VP Preposing Consider the following sentences:
(4b) Mary once predicted that John would
could not.
John intends to make a table, and make one he will.
We thought someone would fail the exam, and fail 1G plenty
of people have.
These sentences are derived from the structures underlying those in (45)
by preposing a VP to the front of an 8. Iwill show in Chapter Vv (ast
have mentioned previously) that the first auxiliary have, be, or do is
outside the VP in M position when rules like thie apply to VP; hence,
‘the rule is not overly complicated,25
(45) Mary once predicted that John would pass an exam eventually,
and he now has passed one.
John hoped that Mary would find his hat, but she could not
find it.
John intends to make a table, and he will make one.
We thought someone would fail the exam, and plenty of people
have failed it.
The rule in question, VP preposing, cannot apply in non-root S's.
~
(46) Mary once predicted that John, who now has passed an exam,
would pass one eventually.
“Mary once predicted that John, who pass(ed) an exam now has,
would pass one eventually.
John hoped that Mary would find his hat, but I wonder how she ever
could find it.
‘John hoped that Mary would find his hat, but I wonder how find
it she ever could.
*John intends to make a table, and his wife thinks that make one
he could.
*John intends to make a table, and we're afraid thet make one he
will.
*John intends to make a table, and I'll get the materials so that
make one he can.
‘We were looking for someone who failed the exam, and now we've
found someone who fail(ed) it (he) has.
‘Thus, VP preposing is also a root transformation.
1.9 Right and Left Dislocation In this section, we consider rules
which remove NP's from their ordinary position in sentences, set them off
‘by commas, and replace them with pronouns. Oné such rule, “left dis-
location", moves an NP to the beginning of the sentence.
(47) This room, it really depresses me.
John's sister, she won't do anything rash.
Bill, you ought to see a doctor. (May be a different rule)
‘These clams, I buy them right at the shore.
This movie, I told you you wouldn't like it much.
Jane, she visits this park every weekend.26
This rule apparently attaches NP's to the highest 8; if this
condition is not fulfilled, the impression 1s that of broken speech.
(48) *I told you that this movie, you wouldn't like it much.
Bil] hopes that John's sister, she won't do anything rash.
‘They put so much furniture in here that this room, 1¢
really depresses me.
¥The fact that these clams, I buy them right at the shore
means that they are sure to be fresh.
I'm afraid that Bill, you ought to see a doctor.
¥He doesn't like the park that Jane, she visite it every
weekend.
I conclude that left dislocation is a root transformation.
Corresponding to left dislocation there is a rule of right
dislocation as exemplified in (49).
(49) It really bothers me, John's big cigar.
She won't do anything rash, John's sister.
You ought to see a doctor, Bill.
I buy them right at the shore, these clams.
I told you you wouldn't like it much, this movie.
Jane visits it every weekend, this park.
The crucial examples for determining whether or not right dislocation
is a root transformation are trees which have an embedded sentence which
4s not rightmost in a constituent (NP or VP) inmediately dominated by a
root. An S which is rightmost in a subject NP or in a VP immediately
dominated by a root does not qualify because the right "dislocated" NP
could be attached to the root, as in (50). (bn the other hand, if (50)
4s unacceptable, it may mean that right dislocated NP's only appear on
the right of the highest VP; thus, we can conclude nothing from (50).)aT
(50) 8
ao
<— I~, ran
N i 8/6 { ap
a girls that chee a be very conscientious
visit
it every
weekend
Crucial examples are therefore those in (51).
(51) ‘*John has sold the garage that you store it in, that old car,
to Mary.
*I predicted that her attempt to do something daring, Job's
sister, would end in disaster.
The fact that the girl Bill bought it for, the camera, is
vioiting him doesn't seem to improve his disposition.
#Hle doesn't realize that the girl he bought it for, the
camera, doesn't like him.
*I didn't say that it bothered me, riding in the back seat,
on the trip out.
‘John gave the boy she used to go out with, his girl friend,
@ dollar.
Since these crucial exumples are ungrammatical, I conclude that right
Aislocation is a root transformation. But I leave undecided the question
of whether it can attach NP's between the subject NP and the VP, as in
(50).
3.10 Preposing around Be In this section, I will discuss some
constructions that are not so clearly root transformations, as are those
previously discussed. These constructions throw into relief rather tthe
possibility of falsification of the hypothesie which I will propose in
Subsequent chapters, and ways in which other syntactic or eenantic
Processes may interrelate with the constraints on transformations I
will propose. The rules which produce these constructions are not
~28
structure-preserving, according to the definition of that term to follow
in Chapter IT; nonetheless, they sometimes (not always) can apply in
Ron-root S's in normal English speech. In order to retain the hypothesis
Twill make, that non-structure-preserving major constituent movement rules
must be root transformations (the "structure-preserving constraint"), I
will be forced to assert that the use of these rules in embedded sentence\_
18 ungrammatical in tne strict sense, and that the structure-preserving
constraint 18 being broken for purposes of emphasis, clear comunication,
ete. Since I am not in a position to be able to characterize the
Conditions under which ungrammatieal sentences can be used, my theory,
im the only sense that I can make it precise, does not always coincide
with Judgments of acceptability. However, it seems likely to me that
‘he way to correct it 1s to study the conditions under hich the structure
Preserving constraint can be broken, and not to abandon the constraint
itself.
Consider first the rule by which predicate adjective phrases whose
heads are compared by means of more, less, most, least, or_as are permted
with the subject NP. (Whether or not the rule operates in sentences which
have linking verbs other than be is of no interest here.) Call this rule
comparative substitution."
(se) More important has been the establishment of jegal services,
Just as surprising was his love for clothes.
Most embarrassing of all was losing my keys. .
No less corrupt was the ward boss.
Equally difficult would be a solution to Russell's paradox.
‘The substitution of an AP for an NP is not a structure-preserving rule,
Sccording to the definition of this to be given in Chapter IL. ‘Therefore,
Comparative substitution should be a root transformation, end29
constructions like those in (52) should not occur in embedded sentences.
And in some embedded sentences, they are indeed unacceptable.
(53) *Bil1 wonders why more important has been the establishment
of legal services.
*A love for clothes that just as surprising was also got him
into debt.
hat equally difficult would be a solution to Russell's
paradox is not at all clear.
*The deputies could extort with impunity, as long as no less
corrupt was the ward boss.
In other instances, however, the sentences of (52) sound acceptable
when embedded, at least to some speakers.
(54) TWe convinced the authorities that more important would be the
establishment of legal services.
?Your admission that just as surprising was his love for clothes
indicates a lack of understanding.
2 am sure that most embarrassing of all was losing your keys.
John Bowers has pointed out to me (personal communication) that the
embedded sentences which violate the structure-preserving constraint in
(54) a1 seem to be complements to verbs, nouns, or adjectives which
report attitudes or statements of their subjects; 1.e., these verbs,
nouns, and adjectives are functioning to introduce indirect discourse.
If part of the meaning of a sentence 1s to be found in the surface order
of elements (for a discussion of thie possibility see Chomsky (1968) and
‘
Jackendoff (1969)), then the simplest way to report this meaning in
indirect discourse would be to break the structure-preserving constraint
and reproduce this order directly. .
By mentioning the possibility of an explanation along these lines,
I do not mean to deny that the structure-preserving constreint, as it
now stands, makes some doubtful predictions about the sentences in (51).30
But as I said above, the proper path for further study would seem to be
investigation of the conditions under which the constraint can be broken.
In the great majority of cases, as we will continue to see, the constraint
makes Just the right predictions; for example, if the constraint were
abandoned, what would explain the ungrammaticality of the sentences in
(53)?
Another rule which preposes constituents around be we might term
"participle preposing." In section II.2 I will give evidence that what
follows the progressive or passive uses of be is a single sister con-
stituent to be @ VP). The sentences of (55), which have limited use in
English, seem to corroborate this contention.
(55) Speaking at today's lunch will be our local congressman.
Taking tickets at the door vas a person I had previously
roomed with.
Examined today and found in good health was our nation's
first executive.
‘Taking turns, as usual, were his two sisters.
Inasmuch as participle preposing is used as a rule of English, it
seems to exhibit root transformation behavior quite clearly.
(56) *Bill wonders why speaking at today's lunch will be our local
congressman.
*he person who taking tickets at the door was had roomed
with me at Yale.
Since examined today and found in gpod health was our nation's
chief executive, we can all breathe more easily.
*he fact that taking tickets at the door was my old Yale
roommate made it easy to get in.
*Bi11 said that taking turns, as usual, were his two sisters.
YBL1L announced that speaking at today's lunch would be our
local congressman.
‘A third transformation which permtes constituents with subject
NP's over be (and a few other verbs, in this case) is @ rule which we31
can call "PP substitution."
(57) In each hallway is (hangs, has long stood) a large poster
of Lenin.
Among the guests were (sat) John and his family.
On the porch is a large wicker couch.
Upstairs is (stands, lies) all the wine we bought in Europe.
Here will be (will stand) the memorial to the war dead.
Since PP substitution is not a structure-preserving rule (according to
the definition of Chapter II), it should be, according to the structure-
preserving constraint, a root transformation. And, us was the case with
comparative substitution, there are embedded 8's in which it may not
operate.
(58) *I have no idea how often among the gueste were (sat) John and
his family.
*The posters that in each hallway are (have long stood) subtly
influence the children.
«That here will stand a memorial to the war dead upsets the
pacifists.
*I won't be satisfied until upstairs is all the wine we bought
in Europe.
‘Now that on the porch is a large wicker couch, ve can all
relax.
However, there are also cases of acceptable embedded sentences which
exhibit PP substitution (at least for some speakers).
(59) 2The belief that in each hallway is (hangs, has long stood)
a large poster of Lenin 1s erroneous.
ne convinced Bill thet among the gheste were (sat) John and
his family.
‘Bi11 1s happy thet on the porch is a large wicker couch.
?I've noticed that upstairs is (lies) all the wine you bought
in Europe.
2BL11 was just explaining that here w:
war memorial.
41] be (will stand) a
I believe that what was said earlier vis-A-vie comparative
substitution 1s also appropriate here. It should be remarked that the3
contexts in (53) and (58) are similar, as are those in (5l) and (59).
It seems to me that the structure-preserving constraint which I will
state in the next chapter (that rules like PP substitution can apply
only in root S's) should be retained in its simple form, and efforts
should be made to isolate special properties of the rules end lexical
items involved in (54) and (59) to see what allows this constraint to
be broken.
Conclusion to Chapter I A large class of transformational rules of
highly specific character, the class of root transformations, has been
isolated. These rules attach constituents only to root S's. (One
could define the traditional grammatical clause "independent clause"
as a clause in which root transformations apply.) In Chapter III, one
further important root transformation will be added to our list, and
still ancther will be discussed in Chapter IV. In Chapter ITI, we will
also see that further general restrictions can be placed on the class
of root transformations.
I will now try to snow how the class of noti-root transformations
of English can be constrained in an empirically interesting way.CHAPTER IZ: STRUCTURE-PRESERVING NP MOVEMENT TRANSFORMATIONS:
In Chapter I, I defined a root transformation, and 1 enumerated and
Aiscussed the root transformations of English. In this chapter, I will
define, in the course of discussing the English passive construction,
a structure-preserving transformation, The bulk of the chapter will —~
then be given over to showing that almost all the (non-root) transforma~
tions of English which move NP's should be formulated so as to be
structure-preserving. For many transformations, this means only a trivial
formal modification in the generally accepted formations of these rules,
while for others, it will involve shoring inadequecies in accepted for-
mulations which can be removed by assuming that they have the structure-
preserving property.
In following chapters, I will discuss transformations vwateh move
other nodes besides NP’s, and show that they too have the structure-
preserving property. Again, {n some cases, accepted formulations need
major revisions. But in each case of a major revision, evidence that
this revision is necessary will be presented.
Ultimately, the cleim of this paper is that every non-root
transformation is structure-preserving, unlers jt is in the highly
restricted class of minor movement rules yet to be defined. As might
be expected, presently accepted notions of grammatical structure mst in
many cases be modified in order to make our claim hold in general.
‘Although one or two of these modifications will weaken our initial claim
to some degree, the narrowing of the notion "possible transformational
rule" that emerges from this study is considerable. |34
II.l The Passive Construction
TI.1.1 the Active-]
sive Relation in English Transformational
grammarians generally agree that English contains a "passive" rule
relating pairs of sentences like the following:
(1) (a) Russia defeated Germany.
(b) Germany was defeated by Russia.
Furthermore, there is agreement that the trees corresponding to (1a)
and (1b) after the application of the passive rule are (2) and (3),
respectively. (There is disagreement over the status of the passive
auxiliary be, which will be retuined to later; in (3) it is simply
represented in ad hoc fashion.)
(2) (3)
tee we
he b OX | Tt
Germany la AUK
* ol A
defeat Germany ve en P ONP
defeat
»}
Russia
There is ample justification for assigning the structure
Op G ry ) (we X) ) to an “agent phrase" like by Russia in (3). For
example, it behaves like a PP in relative clauses and questions:
(4) By whom has this book been read?
Who has this book been read by? ‘
‘The people this book has been read by are not typical.
Also, this by, like other prepositions, cannot have a for-to (infinitive)
object:35
(5) The tension would be lessened by for Europe to disarm,
‘The bankers would be angered by to suggest devaluation.
‘The relation in (1), the active-passive relation, is sometimes
thought to hold also between pairs of noun phrases such as those in (6)
and (7).
(6) Russia's defeat of Germany
(7) Germany's defeat by Russia
There are, however, differences between "passive noun phrases" like (7)
and passive sentences. If and exactly how passive noun phrases and
passive sentences are granmatically related is uot of direct interest
here, since I will only be demonstrating that the rule or rules which
derive passive sentences from active ones have a certain property. I
return to the derivation of passive noun phrases in section 11,6.
There is some disagreement over what the common deep structures of
actives and passives should look like. In particular, it 18 not certain
whether the agent phrase (Russia in the preceding examples) should
originate in subject position or in the by phrase. The force of what
follows does not depend on the resolution of this question. However,
for purposes of exposition, I will assume that agent phrases are su jects
in deep structure, in line with most transformational work. Under this
assumption, (3) 1 a transform of (2).*
* The alternative would be to assume with Fillmore (1968) that the
by phrase 1s non-empty in deep structure and that any subject of an active
which can alternatively appear in a by phrase in @ passive is derived from
that source by an "agent preposing" rule. In this view, the (a) and the
(b) sentence below are derived from the deep structure (c),
‘a) John received the letter.
(b) The letter was received by John.36
A second question concerning the proper formulation of the passive
rule(s) is whether, in fact, one rule which moves two NP's or two rules
which move one NP apiece are involved. Again, the results of this
section are neutral with respect to this question. For purposes of
(footnote cont.)
(c) 8
oe
a v NP
BP
| m“~N
receive e ow
rete by aha
Such an analysis would have certain advantages. First, the fact that
‘the subject NP of an 8 is an obligatory node means that it must be
filled at same point in a transformational derivation. Thus, either
NP preposing or agent preposing must apply to (c), yielding (a) or (b).
(If we use the agent postposing rule adopted in the text, we mst impose
an ad hoc obligatory condition on NP preposing, so that agent postposing
does not apply alone, leaving (c) as an acceptable surface structure.)
By contrast, NP's do not have an obligatory subject NP as do B's, and the
possessive transformation which plays a role similar to NP preposing need
not apply inside NP structures analogous to (c); "The reception of the
letter by John."
Second, an argument for agent preposing (and against agent post-
posing) might be constructed from the following evidence:
fa) The { fale } slipped into the closet.
b:
The (RITE, ary) flew across the room.
(e) The closet was slipped into by the thief.
*The closet was slipped into by the soap.
(f) The room was flown across by the bird.
“The room was flown across by the dictionary.
These sentences show that if agent postposing 1s a rule, 1t will sometimes
have to take into account the deep structure intérpretation of the NP to
be postposed. For while the sentences of (d) are ambiguous when they have
an animate subject, the grammatical sentences of (e) and (f) are not,
showing that only an agent NP can be postposed over an intransitive yerb.
Such @ condition on a transformational rule would seem to be undesirable.
On the other hend, if agent preposing is a rule, the agent inter-
pretation of a subject could be associated with a deep structure by phrase
position, and the non-agent interpretation could be associated with the
subject NP position. While this would seem to be a promising chanuel for
investigation, I have not been able to formulate a comprehensive analysis
of all the consequences of using an agent prepcsing rule at this time,
80 Iwill not go into the matter further.37
exposition here, I will assume that one rule moves the deep structure
subject NP into the by phrase, and that a separate rule moves a deep
structure object into the subject position. Following the terminology
of Chomsky (1970), I will call these rules “agent postposing" and "NP
preposing" respectively.
1.1.2 The Structure-Preserving Property Agent postposing and NP
preposing are both good examples of the structure-preserving transforma-
tions which are to be defined and studied in this chapter. Both move
NP constituents into positions where NP's are permitted by independently
motivated phrase structure rules. Agent postposing moves the subject NP
into the NP position provided for in the rule: PP P - NP. NP preposing
moves an object NP into the position provided for by the rule:
8 NP - TENSE - (M) - VP. This notion of moving a constituent labeled
X into a position where a node X is already provided for by the phrase
structure rules is the central idea in the definition of a structure-
preserving transformation.
To make this notion precise, I drop the requirement that all nodes
dominate terminal elements. Rather, I require only that a node dominate
terminal elements at some point in a transformational derivation. Thus,
trees may contain "empty" nodes like NP, 8, P, etc., which dominate
nothing. (These empty nodes are ignored by semantic interpretive rules
and selection restrictions.)
Definition A structure-preserving movement rule is a transformation
such that (1) the structural description specifies the location in trees
of two nodes By and By bearing the same label X, and (11) the structural
change moves By and all the material dominated by it into the position38
of B,, deleting B,.
The deletion of B, is subject to a recoverability condition, as
in Chomsky (1965), so By is either empty or else what it dominates is
specified in the rule that deletes it. Similarly, I assume that an empty
node labeled X remains in the original By position, and that the movement
rule can specify material to be inserted into the B, position.
Structure-preserving insertion and deletion rules can be defined
enalogously; an insertion specifies the location of a node in trees and
replaces it with material explicitly mentioned in the rule, vhereas a
deletion rule specifies the location of a non-empty node in trees and
removes the material it dominates, leaving an empty node. By this
definition, a structure-preserving movement rule transforms
We (gX)-¥-(;_)-% into W-Gj_) - ¥- GX)
(of course, the movement may be from left to right or right to left.)*
We may wish to extend the term "structure-preserving" later to any
transformation which is @ combination of structure-preserving transfor-
mations as we have just defined them. Of course, only certain specific
kinds of combinations should be highly-valued grammatical rules. For
example, one could in effect substitute a fully specified NP which lacked,
however, a relative clause 8 for an NP which was empty excep: for a
relative clause 8 by substituting a sequence DAT - NUM - N - PP for a
similar empty sequence. I will not introduce such extensions, however,
unless they arise from the discussion of individual rules.
* I don't mean to imply that any structure-preserving rule is a
possible rule. Other constraints are certainly operating in the 7
Several such constraints are proposed and discussed in Ross (1967.39
An obligatory node X may be empty in deep structure, just as an
optional node may be. However, an obligatory node differs from an
optional one by definition in that it mist be chosen, Since require
(tn the second paragraph of this section) that any node dominate terminal,
elements at some point in a transformational derivation, it follows that
obligatory nodes are just those which mst be Present and non-empty at ™
Some point in a transformational derivation. (Optional nodes need not
de since they need not be chosen in deep structures
37.1.3 The Passive Rules in the Structure-Preserving Framevork ‘The
——= OSS ucture~Preserving Framework
subject) moves an NP into the by phrase object position, as shova by the
arrow in (8). By formulating this rule as structure-preserving (i.e.,
by assuming that by and an empty object NP are present in the deep
structure of the passive), we can account for the derived PP structure
which results from this rule without postulating any ad hoe symbols
such as PASSIVE in deep structure.
(8)
ob
| —™
acrlat Germany iP
by
The optional phrase structure choice of a PP with the preposition by
Provides an empty node for receiving @ postposed agent NP, This formal
mechanism (essentially that utilized by Chomsky (1965) to account forho
‘the PP structure of the by phrase) explains why the output of agent
postposing results in @ prepositional phrase similar to other prepositional
phrases in its position under VP, in its internal structure, and in its
transformational behavior.
The structure-preserving framework also provides a principled
reason for why agent postposing 1s obligatory, once agentive by with
an empty object NP is chosen in deep structure. (There are no sentences
like "*Russia defeated Germany by.") The agentive by is like a number of
other prepositions such as with, for, toward, at, etc. (as opposed to in,
around, near, locative by, etc.) in that it has an obligatory object NP.
Since an obligatory node mist be non-empty at some stage of a transfor-
national derivation (not necessarily in deep structure), an object NP
is always generated with agentive by in deep structure, and this NP mst
be filled either in deep structure or by a transformation.*
Very often in this paper, I will be able to use the structure-
preserving framework in this way to eliminate the need for stating that
@ given rule is optional or obligatory by means of ad hoc conditions on
the rule itself. Perhaps the structure-preserving framework could give
rise, in the long run, to some general claim about when rules are optional
and when they are obligatory, with special conditions always being due to
the phrase structure involved (sub-categorization especially might play
a role) and other more general considerations. For the moment, I will
limit myself to showing, at several points in the paper, how the structure-
preserving constraint explains obligatory or blocking conditions on
* Whether this NP is ever filled in deep structure is a separate
question from our concern here.aL
several rules and eliminates the necessity of placing conditions of
applicability on the rule itself.
One of the reasons that an attractive general claim aboxt when
rules are optional and obligatory cannot be made is the fact that the
next rule to be discussed, NP preposing, is obligatory. Since the deep
subject NP of a passive sentence is non-empty in deep structure (at least
if the by phrase is present in surface structure), there is 10 reason,
in the framework developed so far, why NP preposing should be obligatory.
Yet it does seem that the requirement that this rule apply whenever
agent postposing removes a deep subject is related to the fact that
‘the subject ‘NP of an § is an obligatory node. This problem was discussed
in more detail in the last footnote of section II.1.1. In any case, NP
preposing in sentences is probably not the same rule that preposes deep
object NP's inside noun phrases, as in ‘Germany's defeat by Russia.”
Arguments to this effect are given in section II.6.
The important fact about NP preposing, for our purposes here, is
that it 1s structure-preserving. It moves object NP's into the empty
NP position provided for by the phrase structure rule that expands 8.
(I treat the passive auxiliary in totally ad hoc fashion in this section;
it 18 discussed in more detail in later sections.)
(9) 8
en defeat | Germanyhe
It may be well to dwell here at the outset on the importance just
of showing that transformations have the structure-preserving property,
even if no other problems could be resolved by this observation. ‘The
reason we want to limit transformations to being of certain types (root
or structure-preserving) is that the expressive power of transformations
ie too great. Many hypothetical rules which are non-structure-preserving
are Just as easily expressed in the generally accepted algebra of
transformations as those which are.
For example, let A be a hypothetical language identical to English
except that the agent phrases in passive constructions are placed
immediately’ before the passive verb or noun rather than among the other
complement PP's. Thus, the grammar of A yields the following grammaticality
judgments, where # signifies ungremmatical in A, and * represents un-
grammatical in English.
(10) Some countries couldn't defeat Germany with convensional arms.
#Some countries couldn't with conventional arms defeat Germany.
countries couldn't Germany defeat with conventional arms.
#Some
(*4#) #Germany couldn't be defeated with conventional arms by some
countries.
(*##) Germany couldn't be by some countries defeated with conventional
arms.
This country's second defeat of Germany in this ceatury.
8 country's second Germany defeat in this century.
iis country's second in this, century defeat of Germany.
ny's second defeat by this country in this ceatury.
Germany! 8 second by this country defeat in this ceatury.
(*##) #the second defeat of Germany in this century by this country.
(*##) ‘The second by this country defeat of Germany in this century.
In presently accepted transformational theory, the grammars of A and of
English differ in an accidental way -- in the formation of the agent
postposing rule -- and each language is equally highly valued. But yet43
English seems a more regular language in a significant way: the postposed
egent PP in English appears in a typical PP position, whereas the agent
PP of language A does not appear in a typical PP position. (Recall that
language A has the same phrase structure rules as English.) Thus, it
would seem that the evaluation measure should be revised to favor English
over language A. Adoption of the structure-preserving hypothesis achieves
this, since A could not be described without an ad hoc addition of a PP
to the phrase structure rules for VP and NP expansion. More generally,
adoption of the structure-preserving hypothesis excludes on principle
any language whose transformations, like those of language A, do not
"obey" its phrase structure rules in embedded sentence:
Any number of hypothetical alternatives to English can be devised
whose passive rules, like those of language A, do not have the structure-
preserving property. In all cases, there are fairly clear intuitions
that these lenguages are irregular in a sense that English is not. It is
this sense that the claim that non-root transformations are structure-
preserving makes precise.
1.1.4 The Optionality of the By Phrase in the Passive The agent by
phrase is optional in English passive constructions:
(12) Germany was defeated (by Russia).
Germany's defeat (by Russia).
‘Transformationalists have accounted for this by postulating an optional
rule which deletes by - NP. In the framework of this paper, an alternative
to this ad hoc device is to attribute the optional presence of the by
phrase in passive constructions to the optionality of PP under the nodes
VP or NP. In this view, the deep structure of a passive lacking an agent
“wy
phrase in surface structure has an empty NP in the subject position.
Agent postposing cannot apply to such deep structures since there is no
empty node to move the subject NP onto. Thus, the deep structures of
(22) ana (13) outa be (14) and (15), respectivery.
(12) Germany was defeated.
(13) Germany's defeat.
(uh) 8 (a5) 8
[| -—_ — I~
we ‘TENSE ve DET N PP
Lom |
ea ov we NP defeat & WP
cleat oS cole
In this way, by phrase deletion can be eliminated from the grammar.
Eliminating by phrase deletion also eliminates the syntactic
problem of determining what element is actually deleted in this rule.
That is, in (15a), the understood agents are, respectively, everyone,
someone, and his father.
(15a) John wants to be left alone in his room.
A hitchhiker here will probably be picked up.
He was never physically harmed by his father, but he was
often threatened.
One might ask what the difference is in such a framework between
the differently interpreted "the corn grew" and "the corn was grown."
One possible answer is that the deep structure grammatical relation of
the corn to grow is the subject relation ir the first case and the object
relation in the second. Another possible answer is that the deep structure
of the second example contains the deep structure of the first, and also
an abstract "causative" verb which 1s replaced by grow in surface structure.45
Thus, the deep structure of "the corn was grown" would be as in (15b),
while that of "the corn grew" would be as in (15c).
(15) a
NP TENSE W
a7 ~~
lL v 8
————_
"cause" NP 7
the v
corn |
grow
(15¢)
IP TENSE y
tHe lL v
corn |
grow
This analysis of agentless passives will also account for a verb
Like be born (in the sense of be given birth to, not that of be carried),
which cannot appear in the active and which never occurs with an agent
by phrase.
(16a) *This child was born by my sister on Memorial Day.
‘*My sister bore this child on Memorial Day.
The use of the progressive with be born shows that it 1s a verb rather
than an adjective, since the progressive is not used with adjectives
formed from past participles:
(16d) Unwanted children are being born every minute.
Unwanted children are very often being still-born.
We need only stipulate that bear (in the sense of give birth to)46
may not take a deep structure subject; this has nothing to do with the
fact that an (empty) subject NP appears before this verb in deep structure,
once we admit the notion of empty node. Thus, the deep structure of (16c)
is (17).
(16e) This child was born on Memorial Day.
(27)
ed Vv ow
velr wale co Borsa
child Day
In connection with bear (give birth to), which can not have a deep
structure subject, a few verbs which mst should be mentioned: precede,
follow, gun:
“The dinner was preceded.
*The speech was followed.
*Bome of the chairs are owned.
‘The normal case is that a verb may or may not take a deep structure
subject.
This concludes the introduction of the concept of a structure-
preserving transformation. The main purpose of this section has been
to illustrate how the two NP movement transformations, NP preposing and
agent postposing, which together account for the passive construction in
English, satisfy what I have defined as the structure-preserving property.
‘In the remainder of this chapter, other non-root transformations which
move NP's will be shown to have this property. In Chapter IV, a number
of structure-preserving transformations which move other phrase nodes
(AP, 8, and PP) will be studied, and in Chapter V rules which moveuy
Ron-phrase nodes are discussed in sone detail. All in all, almost every
movenent transformation which has found scceptance in transformational
literature as relevant to the description of English 18 discussed
Somewhere in this paper.
T1.2 There Insertiou
a teertion
11.2.1 There Insertion as Structure-Preserving Transformationalists
Generally agree that the folloving pairs of Sentences are transformation-
ally related by a rule usually called there insertion.
(18) A small dog 1s in that room,
There 18 a enall dog in that room,
Several prizes are distributed on Saturday,
There are several prizes distributed on Saturday,
Rone children have been playing in the yard.
There huve been some children Playing in the yara.
Anew house 16 being built next door,
‘There 16 @ new house being bullt next door.
Few students are entirely without means of support,
‘There are fev students entirely withone means of support,
A hatless stranger appeared,
There appeared a hatless stranger.
A solution to this problem nay not exist.
‘There may not exist a solution to this problem.
This rule moves the subject NP to a Position after the first be in
the following verb phrase, provided no Ron-auxiliary verb intervenes,
and provided thet the subject NP hich 18 moved has en indefinite’ —».,
determiner, as in (18). The vacated Subject NP position is filled with
the morpheme there, which other transformations treat as an NP.+
——_—
* The operation of this rule when the main verb 18 occur, appear,
sxist, etc., 18 discussed in section 11.7,
Morris Halle (personal Communication) has suggested to me that there
4
ee48
We can assume that two Processes are involved; the first, which we
can more appropriately cali. indefinite subject movement, moves the subject
\
NP out of the subject Position, and the second, there insertion proper,
fille this empty NP with the morpheme there. Probably the feature of
Our Particular interest in this section is to show that the NP
movement. rule, indefinite subject movement, is structure-rreserving,
Three arguments to this effect follow,
First, we note that be has an obligatory complement, Just like the
Prepositions for, with, agentive by, ete, mentioned in the Preceding
Section,
(20) Some of the workmen vere,
qOnly B411 and John vill be,
“Some people who like beer are,
Vacuum cleaner making that noise,
“are) my friends outside, i4g
in Chapter III, (There is a sense of be meaning exist which sometimes
occurs without @ complement, but euch cases are marginal.)
In the discussion of the passive we saw how an empty node in deep
structure could fulfill an obligatory sub-categorization condition (on
by) provided that this node be filled at some stage of the transformational
derivation. Thus, by had an empty object NP in deep structure, which had
to be filled by a transformation. With reference to be, this means that
an empty deep structure predicate nominal NP which is filled by indefinite
subject movement should fulfill the condition that be have a complement.
If indefinite subject movement is not formated as a (structure-
preserving) substitution of subject NP for the Predicate nominal NP, a
very awkward set of ad hoc conditions results: indefinite subject movenent
As obligatory instead of optional just in case nothing follovs be and
be can be missing a deep structure complement just in case it has an
indefinite subject. And in fact, just these conditions hold:
(21) *Lote of people who don't like beer are.
There are lots of people who don't like beer.
*One person with a hat on was.
‘There was one person with a hat on.
Thus, we see that the structure-preserving formulation removes a
cumbersome ad hoc restriction from the indefinite subject movement rule.
A second argument for the structure-preserving nature of indefinite
subject movement can be made by examining the use of be as a modal. (M)
either of expectation or necessity or of futurity (4n connection with
going or "gonna"), These uses, exemplified in (22), are modal uses
because they do not occur ineide VP's (1.e., after other modals or in
infinitives, gerunds, etc.; I Justify analyzing M as a sister to VP in
Chapter V).(22) A demonstration is ging to be held at six.
*A demonstration may be going to be held at six.
*They expect the demonstration to be going to be held at six.
Three senators are to be here for the conference.
*The senators' being to be here surprises me.
You are to read one book every evening.
*They insisted on your being to read one book every evening.
If indefinite subject movement is not structure-preserving, a simple
statement of it without ad hoc conditions concerning modale would place
the subject NP after the first occurrence of be, as in (23).
(23) There is a new house being built next door.
*There is being a new house built next door.
There were only three students being obnoxious.
“There were being only three students obnoxious.
But when the first be 1s an M, such a statement of the rule yields the
wrong results:
(24) ‘There is a demonstration going to be held at six.
*Mhere are three senators to be here for the conference.
Rather, the indefinite subject movement moves the subject NP to a
position after the first be under the VP:
(25) There 18 going to be a demonstration held at six.
There are to be three senators here for the conference.
‘The structure-preserving constraint not only accounts for this but,
explains it, in the sense that another formation of indefinite subject
movement (for English) which would place the subject NP between M and V
when M is be would be impossible, since the phrase structure rules do
not provide an empty NP in that position. Thus, the structure-51
Preserving constraint removes a second ad hoc condition on indefinite
subject movement.
As a preliminary to a third argument that indefinite subject movement
is a structure-preserving rule, I digress to examine how predicate
nominals and predicate adjectives are generated by the phrase structure
rules. I will refer to both predicate adjectives and predicate nominals
as predicate attributes.
Semantically, a predicate attribute modifies the subject NP, whereas
a direct object NP or an adverbial AP complete the meaning of the verb
without standing ina direct relation with the subject NP. Examples of
predicate attributes are underlined in (26).
(26) ‘The boy was @ good swimmer,
Bome teenagers are very submissive. ,
Three of the managers became vice-presidents.
John became violently 111.
One boy seemed too polite.
He seems the perfect choice.
The girl looked happy.
He remained the only Latin teacher in the school.
A variety of syntactic facts correlate with the difference in
grammatical relations between predicate attributes and verb complements.
(2) In many languages with declined adjectives, the head of a predicate
attribute AP (but not an adverbial AP) agrees with the subject NP in
number, case, and grammatical gender, (11) A predicate attribute NP
agrees with the subject NP in number and, in many languages with case
systems, in case. (iii) In English, any AP immediately dominated by
VP or 8 is marked with an ly suffix, the exceptions (besides a few
lexical items like fast and hard) being Just the predicate attribute
AP's, (iv) The NP preposing (passive) rule and the rule which moves