235 67 27MB
English Pages 208 Year 1981
Linguistische Arbeiten
96
Herausgegeben von Herbert E. Brekle, Hans Jürgen Heringer, Christian Rohrer, Heinz Vater und Otmar Werner
Susan Olsen
Problems of seem/scfaeman Constructions and their Implications for the Theory of Predicate Sentential Complementation
Max Niemeyer Verlag Tübingen 1981
ClP-Kurztitelaufnahme der Deutschen Bibliothek Olsen, Susan: Problems of seem/scheinen constructions and their implications for the theory of predicate sentential complementation / Susan Olsen. - Tübingen : Niemeyer, 1981. (Linguistische Arbeiten ; 96) ISBN 3-484-30096-5 NE:GT
ISBN 3-484-30096-5
ISSN 0344-6727
© Max Niemeyer Verlag Tübingen 1981 Alle Rechte vorbehalten. Ohne ausdrückliche Genehmigung des Verlages ist es auch nicht gestattet, dieses Buch oder Teile daraus auf photomechanischem Wege zu vervielfältigen. Printed in Germany. Druck: fotokop wilhelm weihert KG, Darmstadt.
CONTENTS
PREFACE
viii
0.
INTRODUCTION
1
1.
THE NATURE OF SEEM CONSTRUCTIONS
5
Analysis I: The Complements of SEEM Verbs as Sentential Subjects Rosenbaum 1967 1.1.1 1 . 1 . 1 . 1 Rosenbaum ' s System 1 . 1 . 1 . 2 Classification of seem, appear, happen and turn out 1 . 1 . 1 . 3 Analysis of seem, appear, happen and turn out 1 . 1 . 1 . 4 Criticism of the Analysis Kiparsky/Kiparsky 1970 1.1.2 1 . 1 . 2 . 1 Kiparskys 1 Analysis of the SEEM Verbs 1 .1 . 2 . 2 Criticism of the Kiparskys 1 Analysis 1.1.3 Ohring 1977 1 .1.3.1 ohring's Objection to the Kiparsky-Thesis 1 .1 . 3 . 2 Ohring 's Solution 1 .1 .3.3 Criticism of ohring's Analysis 1.1.4 Summary Analysis II: The SEEM Verbs as Psychological 1.2 Predicates 1 .2.1 Flip 1 . 2 . 1 . 1 Lakoff 197O 1 . 2 . 1 . 2 R.Lakoff 1968 1 . 2 . 1 . 3 Comments on Flip 1 .2.2 Psych Movement 1 . 2 . 2 . 1 Postal 197O, 1971 1 . 2 . 2 . 2 Postal 1974 1 . 2 . 2 . 3 Comments on Psych Movement 1.2.3 Comparison of Analysis I and Analysis II 1 .3 Analysis III: The Complements of SEEM Verbs as Sentential Objects 1 .3.1 Kajita 1967, Culicover 1976 and Ruwet 1976a 1.3.1.1 Syntactic Tests 1 .3.1 .2 Conclusion 1 .3.2 Emonds 1970 1.1
6 6 6 7 8
10 12 12 14 15 15 18 18 19 21 21 21 23 24 25 25 27 30 32 34 34 35 47 48
vi 1 .3.3
Bresnan 1 9 7 2
52
1 .3.3.1
Bresnan's Fixed Subject Constraint
52
1 .3.3.2
Bresnan 's Solution
55
1 .3.3.3
Shortcoming of Bresnan's Solution Summary
57
1.4
2.
58
ENGLISH SYSTEM OF PREDICATE SENTENTIAL COMPLEMENTATION
62
2.1
Derivation of SEEM Constructions within REST
62
2.2
67
2.2.1
The Distribution of Sentential Complements Extraposition vs. Intraposition
67
2.2.1.1
Extraposition
67
2 . 2 . 1 .2
Intraposition
69
2.2.2
Lexicalist Argument
71
2.2.2.1
Chomsky 197O
72
2.2.2.2
Jackendoff 1977
74
2.2.3
Sentential Subjects
78
2.2.3.1
Fullest Extension of the Intraposition Premise: Koster 1978
78
2.2.3.2
2.2.3.3
A Counterargument to Intraposition and Solution
its
81
2.2.3.2.1
Higgins 1973
81
2.2.3.2.2
Chomsky 1977
83
Derived Sentential Subjects within an Intraposition Framework
86
2.2.4
Conclusion
92
2.3
The Nominal Gap Hypothesis
93
2.3.1
The Insufficiency of Chomsky and Lasnik's 1977 and Jackendoff 's 1977 Analyses 93
2.3.2
Control
95
2.3.2.1
Interpretation of S Final Sentential Subjects
96
2.3.2.1.1
Derived Sentential Subjects
96
2.3.2.1.2
Underlying Sentential Subjects
2.3.2.2
100
Interpretation of S Final Sentential Objects
102
2.3.2.2.1
NA Complements
102
2.3.2.2.2
Oblique Complements
103
2.3.3
It Insertion
106
2.4
Explanatory Power of the NGH
110
2.5
Conclusion
114
vii 3. 3.1
3.1.1 3.1 .2 3.2 3.2.1 3.2.1.1 3.2.1.2 3.2.2 3.2.2.1 3.2.2.2 3.3 4. 4.1 4.1.1 4.1.2 4.1.3
4.1.3.1
THE NATURE OF SCHEINEN CONSTRUCTIONS
SCHEINEN + daß Clause Definition of Class Summary SCHEINEN + Infinitival
Definition of Class Bech 1955 Transformational Grammar Derivation of SCHEINEN + Infinitival NP-Movement V-Raising Summary GERMAN SYSTEM OF PREDICATE SENTENTIAL COMPLEMENTATION
The Nominal Gap Hypothesis Distribution of Sentential Subjects Distribution of Sentential Objects Lexical Redundancy and Control Assigning the Syntactic Functions 'Subject' and Object 1 to Sentential Complements Es_ Insertion PP Complements Scheinen and heißen Explicative and Implicative Constructions Bech 1955, 1957 Transformational Grammar Other Sentence Variants Topical iza t ion
117 117 117 122 125 125 125 132 134 134 138 145
147 147 147 149 151 151
4.3.3 4.3.4 4.4
Pseudo Clefting Summary
155 157 161 163 163 169 174 174 177 178 179 180
5.
SUMMARY
18fi
6-
BIBLIOGRAPHY
189
7.
GERMAN SUMMARY
194
4.1.3.2 4.1.3.3 4.1.4 4.2 4.2.1 4.2.2 4.3 4.3.1 4.3.2
Right and Left Dislocation Scheinen and heißen
V11J
PREFACE
The present study is a slightly revised version of my doctoral dissertation presented to the Philosophical Faculty of the University of Cologne during the summer semester of 1980. I am greatly indebted to Prof. Dr. H. Vater, my doctoral advisor, for his support, encouragement and critical comments during the four-year period that this work was coming into being as well as to P r o f . Dr. A. Wollmann, who also served on my dissertation committee, for his encouragement. My thanks also .go to J.Erickson, H.Heltay, J.Koster, J.Lenerz and C.Thiersch for valuable comments and discussion along the way. Special appreciation should be extended to T.Höhle without whose professional advice this work would not have been possible. Finally, I would like to express my gratitude to Prof. Dr, J. Untermann for his support during the period of my graduate studies. A copy of the German version of this thesis is available in the Universitäts- und Stadtbibliothek Köln.
Cologne, September 198O
Susan Olsen
.
IN'IRüDUC'ilOU
Tne verb seem has cropped up under a broad spectrum of topics in the linguistic literature since the beginning of transformational grammar. In these discussions seem has always been an important verb and yet frequently - in some way or another - slightly irregular. One has often been left with the impression that there is more to the nature of seem than meets the eye. The verb seem takes a finite sentential complement which looks very much like a subject complement. (1)(a)
It seems that money is the root of all evil.
But this putative subject complement can only occur sentence finally, hardly the behavior of 'regular' sentential subjects. ( 1 ) ( b ) »That money is the root of all The sentences in
evil
seems.
(2) demonstrate the regular distribution of
a sentential subject in English. (2)(a) (b)
It is likely that money is the root of all evil. That raonjiy is the root of all evil is
likely.
On the basis of such a superficial distribution should the sentential complement to seem be considered an object complement? If v/e consider further data, we find that seeia shares with likely an infinitive complement that was, for quite some time, thought to be derived by the process of subject-to-subject raising. (3)
r-ioney seems to be tne root of all
evil.
(4)
Money is likely to be the root of all evil.
This construction would tend to speak for the former analysis and not for the latter. On the other hand, one must note, the raising process is not well understood and only applies to a limited number of verbs. Seem itself is probably the prime example of a raising verb. The more usual type of complement is the equi construction.
infinitive
If we accept the raising argument illustrated in (3) and (4) that seem and likely exhibit parallel sentential complement structures, and therefore the complement to seem must also be a subject complement, then in order to rule out ( 1 b ) , the putative sentential subject to seem must be limited to sentence final position. Thus, having gained regularity in the raising data by this assumption, we are faced with what seems to be irregular syntactic behavior on the part of seem with a finite complement, since ordinary subject clauses have an optional distribution between subject and sentence final position. These insecurities that arise around an attempt to characterize seem constructions are not idiosyncratic to the English language. In German the daß complement to scheinen displays the same peculiar distributional property that we observed in English seem constructions. ( 5 ) ( a ) Es scheint, daß Geld die Wurzel allen Übels ist. (b) »Daß Geld die Wurzel allen Übels ist, scheint. Once again we have in (5a) a superficial construction that looks very much like an extraposed sentential subject. Stimmen is a good example of a verb that clearly takes a sentential subject. ( 6 ) ( a ) Es stimmt, daß Geld die Wurzel allen Übels ist. (b) Daß Geld die Wurzel allen Übels ist, stimmt. From (6) we see that also in German regular sentential subjects occur optionally in subject position. German scheinen also engages in an infinitive construction that cannot be derived by a rule of equi NP deletion. (7) Geld scheint die Wurzel allen Übels zu sein. But there are serious reasons to doubt that a raising rule can account for this structure. Furthermore scheinen lacks the likely counterpart which would show that another intransitive verb engages in this construction and thus point to the conclusion that the two sentential complements must be parallel in structure. Could it possibly be that constructions like (1a) and (5a) which share such striking superficial similarity with extraposed sentential subject constructions are not instances of sentential subjects? It is our goal in this dissertation to see if upon closer investigation of the data surrounding seem and scheinen constructions some of the question marks left by
previous analyses cannot be removed. In both languages other verbs as well share the behavior of seem and scheinen. We will term these (rather small) classes of verbs the SEEM and SCHEINEN verbs. In each case we will contrast the SEEM and SCHEINEN verbs with verbs and predicates that can unambiguously be shown to occur with sentential subjects. These classes will be termed the LIKELY and STIMMEN predicates and verbs and will be further defined in Chapter 1 for English and in Chapter 3 for German. If upon further empirical investigation we should turn up evidence of a sound syntactic nature that SEEM and SCHEINEN constructions have for the most part been misunderstood in the relevant linguistic literature, would the impact of this topic to linguistic theory cease with this observation? That is, are we merely engaging in a trivial - but perhaps interesting quarrel over whether the that and daß complements to SEEM and SCHEINEN sentences shouldn't be termed Objects' instead of 'subjects' with no further ramifications for the theory of grammar? We suggest that this is not the case. If the SEEM and SCHEINEN verbs should turn out to be transitive in nature and not intransitive as previously supposed in most of the linguistic literature dealing with predicate sentential complement systems within the framework of transformational grammar, this simple empirical observation could have a definite effect on the overall structure of such systems. That is, linguistic inquiry does not content itself with descriptive adequacy but sets as its goal explanatory adequacy. In such a case the grammar would carry the burden of offering an explanation of how the speakers of the language it describes know to differentiate on a syntactic plane between structures ( 1 b ) and (2b) and (5b) and ( 6 b ) inspite of the superficial similarities between the corresponding (a)-constructions. This is an especially interesting problem within a theoretical framework like the present state of transformational grammar which predicts that all interpretation takes place at the level of surface structure. The superficially similar sentences ( 1 a ) / ( 2 a ) and ( 5 a ) / ( 6 a ) would have to receive different interpretations. Is there a device present in the present theory of grammar that could be plausibly
extended to accomplish such a task? A solution to this problem will be tackled in Chapter 2 for the English predicate complement system and in Chapter 4 for German. This inquiry into the problems of SEEM and SCHEINEN constructions and their ramifications for a theory of predicate sentential complementation was intentionally set up as a contrastive study. The SEEM and SCHEINEN constructions can, of course, only be systematically examined within their own English and German predicate complement systems and the consequences of the results of this examination are only definable within the target system. But, by then being able to contrast at least two individual systems of language with regard to the topic at hand, similar conclusions that result from the study have less chance of being ad hoc or language-specific in nature. If studies of further language systems should reveal similar results on the question, we would then be justified in considering the consequences of this topic shown to be necessary in several individual grammars to be of equal consequence for universal grammar.
1.
THE NATURE OF SEEM CONSTRUCTIONS
We will begin our discussion of the syntactic nature of the SEEM constructions in section 1.1 with the first major work in transformational grammar that dealt with the predicate sentential complement system of English - Rosenbaum 1967 and show how the verb seem was originally classified within this system, what type of complement structures it was thought to display and how these structures were derived. Rosenbaum's initial suggestions were immediately takenup in the subsequent linguistic literature and subjected to more extensive study. As examples of works that built upon Rosenbaum's postulates concerning the SEEM constructions, we will consider the Kiparskys" 197O study of factive predicates in section 1 . 1 . 2 and ohring's 1977 constrastive study of the extraposition phenomenon in English and German in section 1.1.3. In sections 1.2 and 1.3 competing viewpoints to that represented by Rosenbaum on the nature of SEEM constructions will be introduced. Firstly, whereas Rosenbaum and the majority of linguists after him have held the sentential complements to the SEEM verbs to be sentential subjects, Lakoff 197O, Postal 1970 and 1971 and other students of generative semantics have prefered to view seem as a psychological predicate with an underlying transitive nature. Secondly, returning to the standard theory of TG, works like Kajita 1967, Culicover 1 9 7 6 , Ruwet 1976a, Emonds 1970 and 1976 and Bresnan 1 9 7 2 find syntactic reasons to disagree with Rosenbaum's original postulates about the nature of SEEM constructions and consider the complements to the SEEM verbs to be sentential objects. Thus, the question to be explored in this chapter is what is the syntactic nature of the complement structures to the SEEM verbs? That is, can evidence be brought forth from the syntactic structure of the language to bear unambiguously on a correct characterization of the SEEM constructions?
1.1
Analysis I: The Complements of SEEM Verbs as Sentential Subjects
1.1.1
Rosenbaum 1967
1.1.1.1
Rosenbaum's System
Rosenbaum 1967 is a study of the system of English predicate complementation, a term used by Rosenbaum to designate the sentential embeddings which function as either subjects or objects to the verb of the matrix sentence. His basic assumption is that the symbol S can be reintroduced into the phrase structure rules of the base component of the grammar as an immediate expansion of NP or VP. Thus, he speaks of noun phrase complementation in the former case and verb phrase complementation in the latter. This establishes a two-way classification. First, a complement may be either a noun phrase or a verb phrase complement depending on the node that introduces it into the phrase marker. Secondly, this complement may function as either subject or object to its matrix verb, depending on where it is introduced in the phrase marker. A subject complement would then be defined as an S which occurs under the immediate domination of the subject NP node, illustrated in Rosenbaum's terms in diagram ( 1 ) :
NP
POP
S
An object complement would appear in the configuration shown in ( 2 ) . An object complement is an S which is dominated by the VP node, whether directly (as in ( 2 a ) ) or mediated by an NP node (as in ( 2 b ) ) . (2)(a) S. (b) NP
A noun phrase complement, as can be seen from diagrams (1) and ( 2 b ) , may function as either the subject or the object to its matrix verb. A verb phrase complement is always an object complement. For Rosenbaum, the defining criteria of noun phrase complementation is whether the verb plus complement construction may undergo syntactic processes typical of a verb plus simple NP structures, namely passive and pseudo cleft operations. Verb phrase complementation does not exhibit either of these syntactic properties. Furthermore, a verb that takes a verb phrase complement does not usually allow a nominal object. Verb phrase complementation, moreover, entails for Rosenbaum that the subject of the embedded complement is identical to some NP in the matrix sentence (either the subject or an object N P ) , a circumstance which triggers his identity erasure transformation (in later literature called equi NP deletion). 1.1.1.2
Classification of seem, appear, happen and turn out
Rosenbaum classifies seem, appear, happen and turn out as intransitive verbs that take subject noun phrase complementation. (Cf. Rosenbaum's appendix of verb classifications ( 1 9 6 7 , 1 2 1 ) . ) That is, these verbs are subcategorized to appear in the following configuration under the symbol V. Their sentential complements occur under the S symbol embedded within the subject NP. The DET and N are present because Rosenbaum assumes that for noun phrase complementation, a pronoun it accompanies the sentential complement. (3) S. _?DP DET
N
"S
VP
(1967,
71)
V
Rosenbaum does not discuss any syntactic evidence for this classification of the complements of these intransitive verbs as subject complements. Apparently his decision was based on his intuitions; explicitly no reason is offered.
1.1.1.3
Analysis of seem, appear, happen and turn out
Rosenbaum establishes that for these intransitive verbs extraposition is obligatory; the subject complement may not occupy the sentence initial position. This is not the case for transitive verbs which take sentential subjects. Compare (4) and (5) to (6) : ( 4 ) ( a ) *That John is right turns out. (b) It turns out that John is right. ( 5 ) ( a ) *That John came early happened. (b) It happened that John came early. ( 6 ) ( a ) That you came early surprised me. (b) It surprised me that you came early. ( 1 9 6 7 , 7 1 f . & 79) Rosenbaum's rule of extraposition is thus made sensitive to this syntactic factor: it is obligatory when applying to the sentential complement of an intransitive verb and optional in other cases. The fact that extraposition is obligatory with seem, appear, happen and turn out enables him to explain two other peculiarities, the first having to do with his 'POSS-ing 1 complementizer (the gerund construction) and the second with the
'for-to'
complementizer (from which he derived all infinitive constructions) . Firstly, examples (7) and (8) show that seem, appear, happen and turn out do not allow the gerund construction, whereas their transitive counterparts do. (7) »John's playing the bugle seems. (8) John's playing the bugle annoys me.
( 1 9 6 7 , 79)
Rosenbaum attributes this to the conflicting set of facts that extraposition must apply to the complements of the SEEK verbs but that the gerund construction is never extraposable: (9) *It seems John's playing the bugle. (10) *It annoys me John's playing the bugle.
( 1 9 6 7 , 79)
Secondly, Rosenbaum (who does not yet have access to a raising transformation) explains the following construction in ( 1 1 ) in two steps. ( 1 1 ) John happened to find gold. Extraposition v/hich has already been stated to apply obligatorily to these verbs applies to an underlying structure like (12a) to yield ( 1 2 b ) . ( 1 2 b ) is then transformed into ( 1 2 c ) by pronoun replacement and subsequent for deletion ( f o r is deleted
immediately before the infinitival marker t o ) . ( 1 2 ) ( a ) t N p [ N i t ] [ c for John to find g o l d ] ] happened. (b)
t N p t N it]]
happened [
for John to f i n d gold]
(c)
[ N p John] happened [ g to find gold]
(1967, 7 8 f . )
It is important to note that sentences like the following are not exceptions to Rosenbaum's analysis that extraposition must apply to these verbs. ( 1 3 ) That John came early happened to annoy Bill. ( 1 9 6 7 , 72) To derive (13) Rosenbaum procedes from the underlying structure presented in ( 1 4 a ) . Extraposition must apply on the cycle processing the complement to happen yielding ( 1 4 b ) . Pronoun replacement then applies replacing the matrix subject it: by the sentential subject of the complement sentence producing (14c) which equals ( 1 3 ) . (14)(a) [ Ν ρ [ ς that John came early][ annoy B i l l ] ] happened (b)
[ N p it]
happened [ g [ N p that John came early][ v p annoy B . ]
(c)
[,,p that John came early] happened [ vp to annoy Bill]
To briefly recapitulate Rosenbaum's analysis, seem, appear, happen and turn out are subcategorized for subject clauses. The sentential subjects are instances of noun phrase complementation and accompanied by a sister node N dominating the pronoun it. Extraposition must apply to their subject complements irrespective of which of the two possible complementizers, that or for (the POSS-ing complementizer is impossible, c f . ( 7 ) ) , has been inserted into the complement by the complementizer placement transformation. Extraposition shifts the sentential complement to sentence final position leaving the pronoun it which accompanies the complement as a remnant of the original structure behind. If the complementizer that has been inserted, the derivation up to this point would yield ( 1 5 a ) . If the for complementizer was chosen, pronoun replacement is required to apply, after which the for is deleted. Thus, we arrive at ( 1 5 b ) . A gerund complement like ( 1 5 c ) will not be generated by the grammar, because special rule markings will prevent complementizer placement from inserting the POSS-ing complementizer into these complements.
10
( 1 5 ) (a) It seems that times are getting harder. (b) Times seem to be getting harder. (c) »Times getting harder seems. 1.1.1.4
Criticism of the Analysis
In section 1 . 1 . 1 . 1 Rosenbaum's defining criteria for noun phrase complementation were mentioned. If a V + S structure can undergo processes typical of V + NP structures, such as passivization and pseudo cleft formation, the complement S is considered to be dominated by an NP. The following are such examples of (object) noun phrase complementation taken from Rosenbaum ( 1 9 6 7 , 10 & 1 4 ) . ( 1 6 ) ( a ) That the world is not flat was demonstrated by Columbus, (b) What Columbus demonstrated was that the world is not f l a t . Note that neither of these processes apply to the SEEM verbs. ( 1 7 ( a ) *It was seemed that John is a nice guy. (b) *What seemed was that John was a nice guy. Passive would not be expected to apply because these verbs are intransitive and therefore have no object NP which can be preposed to become the derived subject of the passive. But, crucially, seem does not exhibit pseudo cleft variants. Pseudo clefting is possible with the transitive counterparts to seem. (18) What surprised me was that he was able to get a job. (19) What annoys her is that John is so damn bull-headed. It is not at all clear at this point why the SEEM verbs are classified by Rosenbaum as taking noun phrase complementation. They fail to passivize and form pseudo cleft variants, which are the tests for noun phrase complementation. If the SEEM verbs do not comply with his tests for noun phrase complementation, one wonders why Rosenbaum classifies them as such. It might be interesting to see if they fit into another of Rosenbaum's classes. Let us experiment with the idea that they might be instances of object noun phrase complementation and assume an expletive _it to be the subject. (20) shows that they don't meet the criteria for noun phrase complementation either. ( 2 0 ) ( a ) It seems that John is a nice guy. (b) *That John is a nice guy is seemed (by i t ) . (c) *What it seems is that John is a nice guy. (d) «What does it seem?
11
If the complements to the SEEM verbs are not instances of noun phrase complementation, could they be classified as verb phrase complements under the same assumption that it is an expletive j.t that occurs as subject? The characteristics of verbs taking verb phrase complements in Rosenbaum's system are (i) they are intransitive, (ii) they trigger the identity erasure transformation (equi) by having a subject or object NP which is identical to the subject NP of the complement and (iii) they never take the that complementizer ( c f . Rosenbaum 1967, 2 7 ) . The SEEM verbs seem to conform to ( i ) ; perhaps they are intransitive. However, they definitely do not comply with properties (ii) and ( i i i ) . They exhibit raising and not equi structures and they accept the that complementizer as readily as the for-to complementizer. It appears that there are compelling reasons to doubt that the SEEM verbs are appropriately classified in Rosenbaum's system. At least it would be accurate to observe that Rosenbaum's classification of the SEEM verbs leaves many questions about the nature of these verbs unresolved. Up to this point we have taken a critical look at primarily the classification of the complement structures of seem, appear, happen and turn out as noun phrase complements. The reason for this is that Rosenbaum devotes a considerable amount of discussion to defining the nature of noun phrase and verb phrase complementation and yet it is immediately apparent that the complement structures to the SEEM verbs display a great deal of irregularity with regard to the defining characteristics of their classification. Let us now scrutinize another aspect of Rosenbaum 1 s analysis of the SEEM class of verbs. As already mentioned in section 1 . 1 . 1 . 3 , the process of extraposition plays a fundamental role in the derivation of SEEM sentences. Extraposition is obligatory; a complement to SEEM may never occupy its underlying position in surface structure. Furthermore, a gerund complement cannot occur with the SEEM verbs because it is not extraposable. In contrast, extraposition applies optionally to the counterparts to SEEM, i.e. to the transitive verbs which take subject noun phrase complementation such as annoy. In other words, the subject clauses of these transitive verbs may optionally occur
12
in subject position in the surface structure. Rosenbaum's extraposition transformation is thus conditioned to apply obligatorily to intransitive verbs
(like seem) and optionally
to all others. But this obligatoriness condition on extraposition in the syntactic environment of 'intransitivity' can be called into question. Rosenbaum has only looked at verbs in a very limited sense; there are intransitive verbs which he has overlooked and other intransitive predicates which do allow their sentential subject to remain in subject position in surface structure. (21)
That Björn Borg will be playing in the Cologne Cup this November is exciting.
(22)
That the number of political skyjackings is steadily increasing is alarming.
(23)
That Elvis died so young is tragic.
(24)
Whether or not a man is successful matters a lot
to
most women. (25)
That he should lie about such a trivial matter doesn't make sense.
(26)
That he has no love for his country is odd.
(27)
For him to have eaten a dozen pieces of cake is crazy.
1.1.2
Kiparsky/Kiparsky 1970
1.1.2.1
Kiparskys' Analysis of the SEEM Verbs
In their study of factivity the Kiparsky's touch upon the verbs seem, appear, happen and turn out simply because these verbs are prime examples of nonfactive predicates.
They assume
that the complements to the SEEM verbs are sentential subjects. Their analysis of the SEEM verbs proceedes along a slightly different line from Rosenbaum's, however, in accordance with their topic
of interest.
1 A factive predicate is a predicate that presupposes the truth of its complement sentence; a nonfactive predicate is one which does not commit the speaker in any way to the truth of the complement.
13
The Kiparskys 1 main thesis in 'Fact 1 is that an important semantic distinction in predicates, that of factivity vs. nonfactivity, influences their syntactic behavior. One important example of how factive and nonfactive predicates d i f f e r with respect to a syntactic process in the language is extraposition. Extraposition - according to the Kiparskys of the sentential subject out of subject position is optional for factive predicates and obligatory for nonfactives. ( 2 8 ) ( a ) That there are porcupines in our basement makes sense to me. (b) It makes sense to me that there are porcupines in our basement. ( 2 9 ) ( a ) *That there are porcupines in our basement seems to me. (b) It seems to me that there are porcupines in our basement. (Kiparsky & Kiparsky 1970, 1 4 5 ) A little later on in their paper, under the topic 'Additional Notes on Factivity 1 they speculate as to why it is that the sentential complement to factive verbs may occupy the initial position in the sentence, but why the complements to nonfactive verbs must be extraposed. They believe that there is a correspondence between initial position and factivity. To demonstrate this they choose the predicate report which is ambiguous between a factive and nonfactive reading. ( 3 0 ) ( a ) The UPI reported that Smith had arrived. (b) It was reported by the UPI that Smith had arrived. The sentences in (3O) are normally read nonfactively. It does not have to be the case that Smith arrived. Sentence ( 3 1 ) , on the other hand, tends to receive a factive reading. ( 3 1 ) That Smith had arrived was reported by the UPI. (197O, 1 6 7 ) If it is true that a complement in sentence initial position tends to be understood factively, then this observation would offer an explanation as to why complements of nonfactive verbs are extraposed. The Kiparskys 1 analysis of seem, appear, happen and turn out resembles that of Rosenbaum's. The SEEM verbs are classified as taking sentential subjects and extraposition must apply in the derivation of the sentences to shift the subject clause out of subject position. However, the Kiparskys have enlarged the class of intransitive predicates to include likely, sure,
14
possible, true, false, chance and so forth. The reason for the obligatory application of extraposition to this class of verbs - and here their analysis differs from Rosenbaum's is that these verbs are nonfactive and their complements may not remain in sentence initial position which is a position with an a f f i n i t y for factive complements. 1.1.2.2
Criticism of the Kiparskys 1 Analysis
The trouble with the Kiparskys 1 analysis is that it isn't true that the complements of nonfactive verbs must be extraposed. Seem, appear, happen and turn out require the extraposition of their complement, but the following examples construed with the nonfactive predicates true, likely, clear and certain are acceptable. (32) That the US policy is shifting on whether or not to deal with the P . L . O . is true, (33) That the price of coffee might recede slightly in November is likely. ( 3 4 ) That Ohio State won't make it to the Rosebowl this year is clear. (35) That he will come is certain. The grammatical status of parallel sentences with the SEEM verbs is of an entirely d i f f e r e n t nature from that of sentences (32) - ( 3 5 ) . They are entirely out of the question. ( 3 6 ) *That the US policy is shifting on whether or not to deal with the P . L . O . happens. (37) *That the price of coffee might recede slightly in November turns out. (38) »That Ohio State won't make it to the Rosebowl this year seems. ( 3 9 ) »That he will come appears. When the Kiparskys claim that extraposition is optional with factive predicates and obligatory with nonfactive ones, their statement is simply false, cf. ( 3 2 ) - ( 3 5 ) . It really only applies to the SEEM verbs. However, let us abstract away from the SEEM verbs and just consider the other nonfactives such as likely. We could then modify the Kiparskys' statement and say that the nonextraposition of subject clauses is more acceptable with factive predicates than with (our new group of) nonfactive predicates. That is, the sentence initial positioning
15
of complements to nonfactive predicates is more restricted than with factives. To summarize the Kiparskys 1 analysis we must grant that seem, appear, happen and turn out are nonfactive verbs, but there is a noticable distinction in their syntactic behavior with regard to extraposition from that of their putative classmates, likely, sure and certain (which we will group together in contrast to the SEEM verbs and term the LIKELY predicates). The nature of this distinction has not been resolved by the Kiparskys 1 analysis. 1.1.3
Öhring 1977
Brigitte Öhring, in a thorough study of the extraposition phenomena in English and German, accepts - besides the extraposition transformation itself - some of the basics of Rosenbaum's classification of verbs. She implicitly assumes, for example, that the verbs relevant to us here - seem, appear, happen and turn out - are to be classified with other verbs taking sentential subject complementation. 1.1.3.1
Öhring's Objection to the Kiparsky-Thesis
The Kiparskys 1 thesis that the obligatory vs. optional application of extraposition to subject clauses depends on the semantic nature of the predicate involved (when the predicate is nonfactive, extraposition must apply) would be an extremely important feature of the analysis of sentential subject complementation if it could be upheld, for it would explain under what conditions underlying subject clauses would have to be displaced from their underlying position by the time they reach surface structure. Öhring sees immediately that the Kiparskys 1 thesis does not hold up to closer scrutiny. There are nonfactive predicates which yield perfectly good sentences when extraposition has not applied to their subject complements. (40)(a) (b)
That there are porcupines in our basement That there are porcupines in our basement (Öhring 1977, ( C f . also examples ( 3 2 ) - ( 3 5 ) in section 1 . 1 . 2 . 2 ) The most notable property, Öhring believes, of the
is true, is false. 43) extraposi-
16
tion phenomena is the markedness of the nonextraposed positioning of a subject complement as opposed to the nonmarked character of the extraposed positioning. She notes that the extraposition of a sentential complement is always possible (never restricted) and more frequent than the nonextraposition of the complement which is subject to certain restrictions. However, against the background of this state of a f f a i r s , there is a general tendency to be acknowledged which she terms the communicative intention of the speaker: there is a tendency for the speaker to place communicatively less important material toward the front of the sentence and communicatively 2 more important material toward the end of the sentence. In accordance with this tendency, it would be most natural to read an extraposed clause, i.e. a clause placed at the end of the sentence, as containing the communicatively important material of the sentence. Conversely, in the case of a nonextraposed clause, the matrix predicate by virtue of being closer to sentence final position would receive the communicative emphasis. This should contribute towards explaning the fact that the nonextraposed form of sentential subjects is a less frequent occurrence than the extraposed form. Nonextraposition would be subject to the restriction that the subject clause contain previously knowninformation and that the matrix predicate carry the communicative weight of the statement. At this juncture the relevance of the Kiparskys 1 factive/nonfactive distinction in predicates can be brought into the discussion and a correlation can be drawn between the principle of communicative importance and the fact that nonextraposition is more common with factives than with nonfactives. The nature of factive verbs is such that they presuppose the truth of their complement sentence. The complement sentence, as a presupposition, is an assumption on the part of the speaker and a factive verb is used by the speaker to make some kind of statement about that assumption. This very characteristic may 2 Her notion of communicative importance is very similar to the idea that a sentence possesses a topic-comment structure, topic being already known information or the theme of the discourse and comment being the new information brought into the discourse, what is being said about the topic. Comment usually follows topic in a sentence.
17
enable factive verbs to build the communicative center of the sentence, allowing the communicatively less weighted complement to precede the predicate. Here the two factors, the factive/nonfactive distinction and the speakers communicative intention dovetail. The following example (of ours) illustrates a factive predicate carrying the communicative import of the sentence as its complement (whose truth is presupposed by the verb) precedes it. ( 4 1 ) That he can't control his emotions any better than that is absolutely mortifying. Sentence ( 4 2 ) , in which the matrix verb is nonfactive, is not as acceptable as ( 4 1 ) . (42)
That he can't control his emotions any better than that is possible.
The communicative weight of ( 4 2 ) / is not centered in the nonfactive predicate possible, which in fact seems to have little semantic content. Rather, the subject complement carries the import of the utterance. But it does not occur toward the end of the sentence. As Ohring's principle predicts, the result is a more marked construction. Öhring does not rule out the possibility that nonfactive predicates may carry the communicative intention of the speaker. This can especially occur when the matrix predicate is emphasized. ( 4 3 ) That it snows in May does happen. ( 1 9 7 7 , 46) ( 4 4 ) That he will come is possible but not probable. (1977, 47) In ( 4 3 ) the matrix predicate is emphasized by means of the morpheme do, in ( 4 4 ) by the contrasting of two predicates. When this occurs, the nonextraposition of the complement is more acceptable than sentence ( 4 2 ) . In sum, Öhring attempts to correct the Kiparskys' claim that extraposition is obligatory with nonfactive predicates by attributing the unusualness of a nonextraposed clause plus nonfactive matrix verb to a combination of other factors. First, a nonextraposed position is a more marked position than an extraposed position in general. Secondly, the semantic structure of a sentence tends to fall into a pattern of communicatively more important material following communicatively less weighted information.
In accordance with this tendency, a sub-
ject clause can occupy nonextraposed
(sentence initial) posi-
18
tion if the matrix predicate is communicatively weighted, whereas, if the subject clause is extraposed, it usually comprises the communicatively more important content of the sentence. 1.1.3.2
Ohring's Solution
The Kiparskys claim that factivity influences the type of complementizer that may occur together with certain predicates. Whereas both factive and nonfactive verbs take that complements, only factives take gerunds and infinitives are typical of nonfactive predicates. One type of infinitival construction typical of nonfactives is the result of what has come to be known as raising. (45)
He is likely to accomplish even more.
(46)
There seems to have been a snowstorm. (Kiparsky/Kiparsky 1970, 144) But only a selective number of nonfactive verbs allow raising. Ohring believes that there is a correlation between those verbs that require extraposition and those that allow raising. She asserts that it is this feature, i.e. the ability of a predicate to undergo raising, which requires the application of extraposition if raising h a s n ' t applied. ( 4 7 ) ( a ) *That John is a liar seems. (b)
John seems to be a liar.
(c) It seems that John is a liar. ( 1 9 7 7 , 5O) Ohring formalizes this feature as [+raising] and presumably has it accompanying the appropriate verb. She labels the raising rule optional and orders it before the extraposition rule. If raising applies, the environment for extraposition has been destroyed. If raising does not apply, extraposition must. These facts account for a paradigm like that shown in ( 4 7 ) 1.1.3.3
Criticism of Ohring's Analysis
Even if one could establish such a correlation between verbs that allow raising and verbs that require extraposition, it isn't quite clear how this correlation could be used as an explanation for the fact that the underlying subject complement to a few verbs cannot occupy subject position in surface structure. It is, of course, an interesting observation, but the
19
analysis is dealing with ad hoc rule markings. An analysis of the SEEM verbs would be much more convincing if the syntactic behavior they display could be seen to follow as regular behavior from some natural assumption underlying the analysis of SEEM in the grammar. Furthermore, Öhring's observation isn't correct that precisely that set of verbs which allows raising also requires extraposition. Rather, the set of verbs that allows raising to apply to their sentential complements is broader than the set of verbs that requires extraposition to apply, cf. ( 4 8 ) and ( 4 9 ) . The nonextraposed position of the subject complement is, for example, possible for the predicates in ( 4 8 a ) , whereas they allow raising, as (48b) demonstrates. The same is true for likely in ( 4 9 ) . ( 4 8 ) ( a ) That he will finish in just under 16 seconds is (b) (49)(a) (b)
1.1.4
certain/sure. He is certain/sure to finish in just under 16 seconds. That the price of coffee might recede slightly in November is likely, The price of coffee is likely to recede slightly in November. Summary
Each of the three works considered in this section has assumed that the verbs seem, appear, happen and turn out take subject sentential complementation. A consequence of an analysis that assumes the complement structures of the SEEM verbs to be underlying subjects is the obligatory application of the extraposition transformation to the putative underlying subject clauses at some point in the derivation, shifting them from underlying subject position to extraposed position, or in other words, to the end of their matrix sentence, before they reach surface structure. The obligatory application of extraposition, which is normally an optional rule when applying to subject clauses, is necessary in these cases because the surface positioning of the complements to the SEEM verbs in subject position yields an ungrammatical structure. (50) »That times are getting harder fseems Iappears |happens Iturns out.
20
Rosenbaum attribute the necessary application of extraposition to complements of the SEEM verbs to the syntactic criterion 'intransitivity' and thus distinguishes between verbs like seem and those like annoy, the latter being transitive and thus allowing extraposition to apply optionally. But the syntactic criterion 'intransitivity 1 is clearly not sufficient to describe when the subject complement may not occupy
its
underlying position in surface structure. For this is only true of seem, appear, happen and turn out, but there are many more intransitive predicates - e.g. is obvious, is likely, is alarming, is true, makes sense, matters etc. - and these allow their complements to occur in either subject or extraposed position. The Kiparskys' attempt to attribute the obligatory application of extraposition to the semantic criterion 'nonfactivity' is also not correct. Apart from seem, appear, happen and turn out, nonfactive predicates do allow their subject complements to remain in subject position. The initial positioning of the complements to nonfactive verbs is sometimes more unusual than with factive verbs, and may be meshed with other factors (as Öhring believes), but the initial positioning is possible; this positioning does not yield ungrammatical structures. Finally, Öhring's attempt to limit the Kiparskys' thesis further by stating that it is not the feature [+nonfactive] , but within the group of nonfactives the feature [+raising] that determines which predicates require extraposition is also not correct. More nonfactives than just seem, appear, happen and turn out allow raising, for example certain, sure and likely, and these verbs do not require extraposition. The attempts to explain when the application of extraposition is obligatory, whether it be attributed to the intransitivity o^1 nonfactivity of the matrix predicate or to the ability of this predicate to exhibit raised structures, have not been able to adequately account for the grammatical facts. The facts indicate that seem, appear, happen and turn out are irregular in each of these three systems in a way in which the other predicates of their class are not and this irregularity all centers around the obligatoriness condition on the extraposition rule. Obligatory extraposition is a very curious property to be forced on a group of verbs. The claim that is being made is
21
that these clauses are underlying subjects but can never in comparison to all other subject clauses - occupy subject position on the surface. 1.2
Analysis II:
SEEM Verbs as Psychological Predicates
A possible analysis of seem and appear has been discussed by a few linguists respresenting the direction of generative grammar termed generative semantics. The analysis to be discussed in this section has been termed f l i p by George Lakoff and referred to as psych movement by Paul Postal. The f l i p / psych movement analysis shares certain properties with Analysis III
(see section 1 . 3 ) : the complement structures of seem and
appear are set in object position in underlying structure. Nevertheless, the flip/psych movement analysis retains at the same time the basic feature of Analysis I, namely the complement clause reaches subject position at some juncture in the derivation and then undergoes obligatory extraposition. Inspite of the similarity in part to both Analysis I and Analysis III, Analysis II d i f f e r s from both these analyses in its motivation. The flip/psych movement analysis will not be adopted here, but, because seem and appear have been mentioned in this connection in some of the important linguistic literature, it will be briefly explicated - in part to distinguish it from Analysis I and Analysis III and in part to explain why it has not been found suitable to yield the theoretical basis for this study of the complement types displayed by the verbs seem, appear, happen and turn out. 1.2.1
Flip
1.2.1.1
Lakoff 1970
George Lakoff postulates a rule f l i p which interchanges ( f l i p s ) subject and object among adjectival and verbal pairs of sentences . (51)(a) (b) (52)(a)
What he did amused me. I was amused at what he did. I enjoy movies.
(b) Movies are enjoyable to me. (197O, 126) The postulation of this rule enables Lakoff to assume that one
22
set of subject-verb-object relations is basic to the pair and that the other sentence is derivable from this basic relation by the rule f l i p , thus capturing the obvious relationship between the two sentences. The nominalization, Lakoff assumes, reveals the basic subject-verb relations. ( 5 1 ) ( c ) My amusement at what he did. ( 5 2 ) ( c ) My enjoyment of movies. Therefore, in sentence pair ( 5 1 ) the adjectival variant is basic ( = ( 5 1 b ) ) and if flip applies to this basic ordering the verbal variant is derived. The verbal variant is basic to the sentence pair ( 5 2 ) ( = ( 5 2 a ) ) and the adjectival sentence has been derived by means of flip. Lakoff does not formalize the rule flip. However, in a later argument he shows his conception of a flip derivation of a sentence containing a complement clause ( c f . Lakoff (1970, 128-133)). (53) It is necessary for me to know that. In the underlying representation the sentential complement occurs in object position and the I_ of the for me phrase is the subject. (54) NP
VP
N I I
necessary
1 ifc
for I to know that After equi applies to the complement, f l i p operates shifting the object complement clause to subject position of the matrix sentence and placing the subject 1^ in derived object position. (55) FLIP NP I N
for me to know necesthat sary Then extraposition applies. (56) EXTRAPOSITION
I ι
23
For deletion, tense spelling, prep spelling and be addition yield the surface structure.
(57)
it
to know that This derivation claims that the complement clause is an underlying object and that it occupies subject position within the course of the derivation due to the operation of flip. Finally, the rule of extraposition may place it in sentence final position yielding the surface structure. 1.2.1.2
R.Lakoff 1968
Robin Lakoff sees a correlation between seem and appear sentences and certain flip constructions such as ( 5 8 ) . (58) It is surprising to me that John gave up his earldom. (59) It seems to me that John is a fool. ( 1 9 6 8 , 40) tinder the flip analysis the underlying structure of ( 5 9 ) would be ( 6 O ) , to which f l i p applies yielding ( 6 1 ) . (60) (61)
FLIP
John be a fool S.
that J. is fool
seems
( 1 9 6 8 , 41
(1968, 42)
Then extraposition must shift the complement out of subject position, the result being ( 5 9 ) . Lakoff justifies her postulation of the complements of seem and appear as object clauses in deep structure - thus allowing the parallel to f l i p verbs to be drawn - by citing facts concerning not-transportation. Not-transportation applies generally to object complements but it applies to the complements of seem and appear as well, indicating that they are object clauses.
24
1.2.1.3
Comments on Flip
There are two points that should be commented on here. First, R.Lakoff actually considers syntactic reasons (not-transportation) for assuming the complements to seem and appear to be objects. The fashion in which she chooses to derive the sentences claims that at some level of intermediate structure (i.e. after f l i p has applied) these complements become subjects, for they occupy subject position. In fact, the crux of the flip analysis seems to be that the complement starts out as an underlying object and becomes the derived subject. Her f l i p analysis, however, has obscured her original assumptions that these clauses are object clauses. The f l i p analysis claims that they are derived subjects. Secondly, the parallel of seem to surprise is quite vulnerable. In the f l i p data there are paradigms of related sentences where both options are present in the grammar - the option representing the underlying subject-verb-object order and the option derived by flipping the underlying subject with the underlying object ( c f . ( 6 2 ) and ( 6 3 ) with the f l i p predicate surprise). Sentence ( 6 2 ) represents the basic underlying order, as the nominalization my surprise testifies. Flip has applied in the derivation of ( 6 3 ) , with ( 6 3 b ) having undergone additional extraposition of the complement. (62)
I am surprised that John gave up his earldom.
( 6 3 ) ( a ) That John gave up his earldom is surprising to me. (b) It is surprising to me that John gave up his earldom. It was sentence (63b) that made R.Lakoff think that seem and appear sentences should also be handled by flip. But it is conspicuous in the case of seem and appear that there are considerable gaps in the paradigm predicted by the f l i p analysis. Seem and appear are only possible in the (63b)-variant of the flip paradigm. ( 6 4 ) ( a ) *I seem that John is a fool. (b) »That John is a fool seems to me. (c) It seems to me that John is a fool. 3 Lakoff has overlooked the fact that not-transportation applies to the complement of likely as well which is a subject clause. ( i ) ( a ) It is likely that John isn't here, (b) I t ' s not likely that John is here.
25
The fact that ( 6 4 a ) is an impossible structure does not o f f e r positive evidence for the assumption implicit in the f l i p analysis that the logical subject-verb relation is !_ seem. Whereas this assumption is justifiable in the case of the other f l i p predicates I am surprised/amused/irritated/ flabbergasted etc., I seem is clearly discrepant. G.Lakoff consulted the nominalization of the verb to support his intuitions. In the case of geern the nominalization permits only the impersonal form (65)(a)
the semblance of an angel
(b)
a semblance of anger
and is inherently incompatible with a subjective genitive ( 6 6 ) ( a ) *my semblance of anger (b)
*her semblance of an angel
whereas the nominalization my surprise at ... is perfectly OK. Thus, seem and appear do not seem to fit paradigm at all.
well into the f l i p
If they are considered to be f l i p verbs, their
syntactic behavior is highly irregular. Crucially
( 6 4 a ) is not
possible, which is supposed to represent the underlying structure. Furthermore, the nominalization semblance does not offer evidence that the animate being of the to phrase of seem sentences is the logical subject. Indeed, this nominalization substantiates the intuition that it is an impersonal expression. Hence, R . L a k o f f ' s assumed likeness of the it construction with it
seems to me that
surprises me that is a little shaky. Her
analysis is very brief and sketchy and does not convincingly o f f e r any revelations as to the nature of the verb seem. 1.2.2
Psych Movement
1.2.2.1
Postal 197O and 1971
Postal (1970, 6O-71 and 1 9 7 1 , 39-54) discusses a rule similar in effect to f l i p . He calls it psych movement because it applies to verbs and adjectives that typically designate a psychological state, activity, event etc.
of some animate being mentioned
in the accompanying experiencer phrase. Some examples of psych movement verbs are amuse, bore, disgust, excite, f r i g h t e n , horrify,
irritate and so forth - Postal believes that the num-
ber of verbal and adjectival elements that make up the class of psych verbs is quite high, i.e.
several hundred. Psych
26
movement verbs occur in paradigms like the following. ( 6 7 ) ( a ) Max disgusted me. (b) Max was disgusting to me. (c) I was disgusted by Max. (197O, 6 4 f . ) The psych movement analysis entails that the sentences of this paradigm share a common set of logical verb-nominal relations. Psych movement has applied in sentences ( 6 7 a ) and ( 6 7 b ) altering the underlying order, for the me of these sentences is the logical subject. Psych movement has not applied to (67c) ; in this sentence the underlying order of the verb-nominal relation is intact. The importance of Postal's psych movement for this topic is that he views seem and appear as psych movement verbs. Semantically seem and appear are similar to psych movement verbs in that they also have to do with the psychological state of affairs of the NP mentioned in the to phrase. They furthermore share the three syntactic factors common to verbs having undergone psych movement: they take an experiencer phrase in object position, the adverb personally can occur with this experiencer (Postal believes that it is otherwise restricted to accompany4 ing the subject only ) and, finally the experiencer cannot be a reflexive. The following is how Postal conceives the derivation of seem and appear constructions. He explicitly assumes the complements to seem and appear to be objects. The underlying structure for the seem constructions in ( 6 8 ) is ( 6 9 ) . 4 Postal (i)(a)
( 1 9 7 1 , 2 5 8 ) gives examples to illustrate this, Harry seems to me personally to be the best man for the job. (b) »Harry personally seems to me to be the best man for the job. 5 For instance (i) *I seem insane to myself. ( 1 9 7 1 , 258) Cases having undergone psych movement cannot take a reflexive pronoun, where those that haven't may. ( i i ) (a) I am amused with myself. (b) *I am amusing to myself, ( i i i ) ( a ) I was irritated at myself. (b) *I was irritating to myself. ( 1 9 7 1 , 47) Postal's main argument in 'Cross-Over 1 is that movement transformations such as passive, tough movement, about movement and psych movement that reorder NPs, cannot cross an NP over a coreferential NP within the same immediate clause. The violation of this constraint would be the source of the ungrammaticality of (i) as well as (iib & i i i b ) , because psych-movement would interchange 1^ with !_. Reflexivization is possible in (iia) and (ilia) because these sentences represent the underlying order, no movement rule has applied, crossing the coreferent NPs. If
27
(68)(a) (b)
It seemed to Max that Paulette was nervous. Paulette seemed to Max to be nervous. ( 1 9 7 O , 67;
(69)
NP
was
I Paulette
V
I
nervous
(1970, 68)
The (b)-sentence has undergone raising and then psych movement applies exchanging the derived object Paulette with Max. The (a)-sentence has not undergone raising so that psych movement exchanges the object complement S with the subject Max. After this step, extraposition must apply to place the complement at the end of the sentence. 1.2.2.2
Postal 1974
Postal's 1974 treatment of raising phenemena ignores - at least upon first glance - the psych movement approach to seem and appear constructions and offers a contradictory analysis of seem and appear to that of Postal 197O and 1 9 7 1 . Postal discerns two types of raising phenomena: raising with Α-verbs in which the raised NP attains superordinate subject position, c f . ( 7 0 ) , and raising with B-verbs where the raised NP is lifted into superordinate object position, cf. ( 7 1 ) . 6 (70) Harry seems to be a wierdo. ( 1 9 7 4 , 29) ( 7 1 ) I believe that to be incorrect. ( 1 9 7 4 , 30) He formulates together with Perlmutter a supposed universal principle that explains when an NP is raised to superordinate subject position and when to superordinate object position. This principle - the functional succession principle - is meant to pertain to not just the operation of raising, but to all promotional rules. 5 cont. the rule of psych movement is rejected, there would be no means to handle cases ( i ) , (lib) and ( i i i b ) in a parallel fashion to the other cross-over phenomena. 6 The claim that raising occurs into superordinate subject position (as with seem, likely etc.) is not a controversial claim. Postal sets out in On Raising to prove that parallel to this type of raising the subject of the complement sen-
28
(72)
The Functional Succession Principle When a promotional rule extracts one NP, NPa from a containing NP, NPc, in the output NPa takes on the grammatical function manifested by NPc in the input.
(1974,287) The FSP predicts that an NP lifted from a subject clause becomes the superordinate subject and an NP lifted from an object clause becomes the superordinate object. Since the raised NP of the complement to seem reaches superordinate subject position, it follows from the FSP that Postal considers the complement a subject. The same holds for the verbs appear, happen and turn out as well as for likely etc. These verbs are classified in Chapter 9 as Α-verbs and adjectives - cf. the list on page 191. In fact, Postal uses the verb seem as his prime example of an Α-verb R(aising)-trigger throughout the text of
On Raising : in his original arguments
in section 2.2 in justification of a raising analysis for Averb constructions (as opposed to the inappropriateness of an equi analysis, for example, to handle the same d a t a ) , Postal uses primarily the verb seem to exemplify Α-type raising sentences. In addition, Postal lists 17 arguments in Chapter 12 that serve as tests for determining when a particular verb is an Α-verb. Each of these arguments is demonstrated with the verb seem. These observations would lead the reader to believe that Postal has rejected his 1970/1971 viewpoint that seem and appear belong to the psych movement predicates. For if they were psych verbs that take underlying object complements as shown in ( 6 9 ) , they would blatantly violate the FSP, because they raise into subject position. However, Postal makes a confession in footnote 2 of page 29O. His true view on the underlying structures of seem and appear is really in accordance with the psych movement analysis proposed in Postal 1970 and 1971 and at variance with the classification throughout this work, i.e. 6 cont. tence may be lifted into superordinate object position (as with believe, find e t c . ) . This latter claim has been attacked, most notably by Chomsky ( 1 9 7 1 b , 1 9 7 3 ) . Postal devotes Chapters 3 and 4 to showing how the underlying subject of the complement sentence with B-verbs is treated on the matrix cycle as a derived object by a series of syntactic rules, thus justifying a raising process into superordinate object position.
29
that seem and appear are Α-verb R-triggers. But because the psych movement phenomena are not pertinent to the discussion at hand and because in end e f f e c t , seem and appear are much like the Α-verb R-triggers in that the raised NP does end up in superordinate subject position, it is convenient for him not to take explicit issue with the traditional analysis of the complements to seem and appear as subject clauses. He therefore continues to use them as prime examples of A-construction raising phenomena because of the ease with which they exemplify the array of types of raised NP. Under Postal's true conviction of the derivation of seem sentences, seem, as a psych predicate, takes an underlying object complement; the experiencer is found in subject position. The subject of the object complement undergoes raising to the superordinate object position (notice there is no violation of his functional succession principle involved in the derivation). Psych movement is required to apply. It interchanges the new derived object with the experiencer NP. Underlying structure: SEEM me [LOVE Melvin Lucy] Ο
Raising:
SEEM me Melvin
Psych movement:
SEEM Melvin me
D
Ο
[LOVE Lucy] [LOVE Lucy]
Subject formation must then apply to place the verbs in second position. In this view, seem and appear are really B-verb Rtriggers like believe, with the added stipulation that psych movement must apply after raising. As already mentioned, Postal does not actually incorporate this view into the classification of R-triggers and the explanation of the raising phenomenon in the text of this work. Rather, he skims over the surface of the question, classifying seem and appear as Α-verbs because the raised NP does eventually end up in superordinate subject position (although in reality he believes that this position is due to the obligatory application of psych movement after raising has applied to an underlying B-verb construction). Postal, in On Raising therefore has not reneged on his original psych movement analysis of seem and appear. 7 Postal assumes verb initial underlying order because this enables him to state A and Β type raising in one rule, see Postal ( 1 9 7 4 , 5-28) for discussion.
30
1.2.2.3
Comments on Psych-Movement
As mentioned at the outset of this discussion, Postal considers the scope of the psych movement phenomena quite encompassing. Further sentence types to which psych movement is assumed to apply are: ( 7 2 ) ( a ) It came to me that Schwarz is insane. (b) It hit me that you are wrong. (c) It is clear to me that hippopotamuses can't fly very far. ( d ) It is obvious to me that hippopotamuses c a n ' t fly very f a r . (e) It is evident to me that hippopotamuses can't fly very f a r . (f) It is plain to me that hippos can't fly very f a r . (g) It is important to me that hippos c a n ' t fly very far. (h) It occurs to me that John is a fool. (i) It strikes me that you are right. (j) It is vital to me that g i r a f f e s eat figs. (k) It concerns me that you are growing so slowly. (1971, 45-46) Vie will not o f f e r arguments to refute Postal's psych movement analysis for this large class of supposed psych movement constructions. Our concern here is merely a relevant and revealing discussion of the SEEM verbs. The discussion will take the form of as straightforward and precise an analysis of these verbs as possible. Our goal is not an attempt at a justification, substantiation or refutation of any particular direction within the framework of the present theory of transformational grammar. For the purposes of this study, however, a claim such as Postal's that the embedded clauses occurring with the predicates in (72a-k) are underlying object clauses is not acceptable for syntactic reasons. Such syntactic reasons ( e . g . pronominalization patterns, parenthetical constructions, pseudo cleft variants, wh-questioning etc., which will be discussed in detail in section 1 . 3 ) offer compelling evidence that these complements are underlying subject not object clauses. Postal's psych movement analysis applies an identical underlying structure to the verbs seem and appear and their complement structures as to the (likewise supposed psychological)
31
predicates clear, obvious, evident, plain, vital and so forth plus their sentential complements. Syntactic arguments will be given in the next section that seem d i f f e r s in a fundamental way from clear, obvious etc. and, thus, cannot share an identical deep structure with them. Furthermore, Lakoff and Postal's analyses set seem and appear apart from happen and turn out. Happen and turn out cannot be considered psychological predicates because no experiencer phrase is possible with these two verbs. ( 7 3 ) ( a ) It seems to me that John is a nice guy. (b) It happens (*to me) that John is a nice guy. (c) It turns out (*to everyone) that John is a nice guy. Happen and turn out, when accompanied by a complement clause, are similar in a straightforward syntactic way to seem and appear. This objection will become clearer in section 1.3. In summary, the flip/psych movement analyses do not capture the similarity of seem and appear to happen and turn out, nor do they capture the difference of these verbs to predicates like clear, evident, important etc. A further question pertaining to the analysis of seem and appear as f l i p verbs or psych predicates arises. Implicit in this analysis, if the sentential complement is assumed to bear the logical object relationship to the verb, is that there is some term that bears the logical subject relation to the verb. Postal calls this term the experiencer. But a large number of seem and appear sentences occur without this term. The question is obvious: is there, in these cases, a term that is postulated as occurring in subject position of the underlying representation, and, if so, how is this term subsequently deleted in a non-ad hoc fashion? Furthermore, both seem and appear occur in construction with a clausal complement that has a sense which is imcompatible with an experiencer. (74)
She left the party early. It seems that she got a telephone call that her mother had taken sick. (75) Max is in the hospital. It appears that he has had an automobile accident. In these sentences seem and appear have the function of narrating an event which the speaker himself has not experienced, but has only been told. This sense may be translated into German by a modal verb.
32
(76)
Sie soll einen Anruf bekommen haben.
(77)
Er soll einen Autounfall gebaut haben.
Q
There is absolutely no reason for postulating an experiencer phrase in cases such as these. 1.2.3
Comparison of Analysis I and Analysis II
It has been noted that Analysis I and Analysis II share certain features and are, for this reason, similar in essence. Let us briefly compare the derivation of seem with a that clause under Analysis I and Analysis (78)
II.
Analysis I, underlying structure (a)
[„„[.it] [„ο that John is a nice g u y ] ] MF w
seems to me
EXTRAPOSITION (b) (79)
it
seems to me [ Ο that John is a nice guy]
Analysis II, underlying structure (a)
I stern [ i'i c [N i t ] [ _D that John is a nice g u y ] ]
PSYCH MOVEMENT
(b)
U j p t r , i t H g that John is a nice g u y ] ] seem to me
EXTRAPOSITION (c)
it
seems to me [ Ο that John is a nice g u y ]
Apparent from the comparison of the two analyses is that Analysis II involves an additional assumption. It selects an underlying representation that is more abstract, that is,
further
removed from the actual surface structure, than Analysis I. To this underlying structure Analysis II applies obligatorily the psych movement ( f l i p ) rule. Once this initial stage is over, Analyses I and II proceed similarly, cf.
(78a & b) with
(79b & c ) . Lakoff and Postal's reasons for conceiving the underlying representations of sentences the way they do are not straightforward syntactic reasons. Indeed, a distinct level of syntactic deep structure, v/here lexical insertion has take place before the transformational process begins to apply, does not 8 It was Hilary Heltay who brought this sense of seem and appear to my attention.
33
exist for them. Within the theory of grammar which they espouse, the level of deep structure is not distinct from the level of semantic representation and, thus, of necessity somewhat more abstract and more remote from surface structure than a purely syntactic deep structure. Lakoff and Postal's considerations in assuming the flip/psych movement analysis to handle certain data of English have centered around the goal of capturing in the underlying representation of such sentences the logical meaning of a sentence. This difference in Analysis I and Analysis II is what was referred to in the introduction of this section as the difference in the motivation of these two analyses. In sections 1 . 2 . 1 . 3 and 1 . 2 . 2 . 3 we have taken issue with f l i p and psych movement as they apply to seem and appear. In summary the kernel of our objection to Analysis II is as follows. First, Analysis II makes the same mistake as Analysis I in classifying seem and appear together with predicates like obvious, evident, matters etc. and is, therefore, not in a position to account for fundamental syntactic differences between these two groups of predicates. (These syntactic differences have just been hinted at up until now, but will be discussed more thoroughly in the next section.) Furthermore, Analysis II offers no more enlightenment into the syntactic nature of seem and appear than Analysis I: the underlying assumptions of both analyses force highly irregular syntactic behavior on these verbs. Perhaps Analysis II is even less revealing than Analysis I. Analysis I requires the obligatory application of extraposition to the complements of the SEEM verbs, thus loosing the generalization that subject clauses may optionally occupy subject position of a sentence. But Analysis II is guilty of this and more. It claims that seem and appear are f l i p or psych movement predicates but that the flip or psych movement rule is obligatory for these verbs, whereas it is optional for other psych predicates. Sentences (80a-c) demonstrate the necessity of obligatory psych movement ( c f . ( 8 0 a ) ) and obligatory extraposition ( c f . (8ob)) in the case of seem in comparison to the regular surprise sentences in ( 8 1 ) .
34
( 8 0 ) ( a ) *I seem that John is a fool. (b) »That John is a fool seems to me. (c) It seems to me that John is a fool. ( 8 1 ) ( a ) I am surprised that John gave up his earldom. (b) That John gave up his earldom is surprising to me. (c) It is surprising to me that John gave up his earldom. Thus, the flip/psych movement analysis of seem and appear not only incorporates the evils of Analysis I, it perpetrates some of its own; that is, it forces upon seem and appear a syntactic paradigm which they d o n ' t fit. 1.3
Analysis III: The Complements of SEEM Verbs as Sentential Objects
The most widely accepted analyses of the verbs seem, appear, happen and turn out, Analyses I and II, attribute a large degree of syntactic irregularity to these verbs: they involve marking the verbs with rule features which indicate that they must undergo certain transformations - either extraposition or extraposition and raising in the case of Analysis I or under Analysis II psych movement and then extraposition or raising and then psych movement. Several linguists have attempted to associate a greater degree of syntactic regularity with them by making different assumptions about the nature of their complement structures. Among these linguists are: Kajita 1967, Culicover 1976, Ruwet 1976a and 1976b,Emonds 1970 and 1976 and Bresnan 1972. The basic feature that links the works to be discussed in this section together is the assumption that the sentential complements to the SEEM verbs are generated in object position in the base phrase marker. Seem, appear, happen and turn out are then, in this sense, transitive and not intransitive verbs. 1.3.1
Kajita 1 9 6 7 , Culicover 1976 and Ruwet 1976a
These three works are concerned with offering straightforward syntactic arguments in support of the assumption that the complement clauses of these verbs are underlying object clauses. They examine the syntactic patterns of the verbs seem, appear, happen and turn out and their complement structures with respect to certain syntactic processes in the language that bear
35
on the nature of the sentential complement to these verbs and conclude from the syntactic data that the SEEM verbs behave differently from verbs taking subject complementation with respect to these processes. K a j i t a , Culicover and Ruwet more or less overlap with one another in the syntactic phenomena discussed. Culicover ( 1 9 7 6 , 256-257) is the least extensive of the three works as far as the problem at hand is concerned. He touches upon the problem briefly in conjunction with a discussion on the extraposition vs. intraposition of that-clauses. 9 Kajita - concerned directly with the semi-auxiliaries of English to which these four verbs belong - and Ruwet devote more discussion to the topic. Ruwet, however, is primarily concerned with French verbs like sembler 'seem 1 and in establishing his analysis by drawing on arguments from French syntax. Not all of his arguments are directly applicable to English; apparently the grammar of French displays a wider array of syntactic evidence toward the nature of these constructions than is uncoverable in English. Nevertheless, in compiling the syntactic arguments discussed in these three works, we come up with sufficient clues as to the nature of the complement sentence in seem-type structures. 1.3.1.1
Syntactic Tests
1. Selection of Nominal Subject It is generally recognized that the sentential complement to the SEEM verbs cannot occupy subject position in the sentence. This is true for a that clause, an infinitive construction as well as for a gerund. Kajita and Ruwet observe furthermore that an ordinary nominal cannot be a subject to SEEM either. The SEEM verbs simply do not select a subject, whether clausal or nominal. ( 8 2 ) ( a ) »His answer seems. (b) His answer makes sense. (83) (a) «The physical health a person is in turns out. (b) The physical health a person is in matters. ( 8 4 ) ( a ) *A subsequent terrorist attack appears. (b) A subsequent terrorist attack is likely. 9 Kajita (1967, 2O-3O). Kajita classifies happen apart from seem, appear and turn out because he considers only the sense
36
It might be thought that by subcategorizing the SEEM verbs for a subject clause only, the grammar could account for the fact that SEEM do not take nominal subjects. However, quite apart from the undesirability of obligatory extraposition of this clause, the syntactic paradigms to follow refute the assumption that the sentential complement to SEEM is a subject. If this is the case, it looks as though the SEEM verbs are either subjectless or have the pronominal subject _it in their underlying structure. 2. Anaphoric and Cataphoric Pronominalization Culicover and especially Ruwet scrutinize the pronominalization patterns of the SEEM verbs and observe that they do not pronominalize in the same fashion as verbs and predicates that take subject complementation. Whereas, for example, when a complement in a sentence such as (85) (85)(a) (b)
It is likely that it will rain tomorrow. That it will rain tomorrow is likely.
is replaced by a simple anaphoric pronoun (it or t h a t ) , the pronoun occupies subject position, ( 8 6 ) ( a ) It is likely. (b) It is probable. (c) It is possible. (d) That is obvious. (e) That is evident. this type of pronominalization does not yield grammatical sentences with SEEM, however. ( 9 7 ) ( a ) *It seems. (b) *It appears. (c) *It happens, (d) *It turns out. However, the clausal complements to seem and appear do pronominalize. ( 9 8 ) ( a ) It seems so. (b) It appears so. 9 cont.of happen which corresponds to the meaning Occur 1 or 'take place*. Happen does have a less concrete sense in which the meaning is paraphraseable by 'chances'. It is in this sense that happen is syntactically similar to seem.
37
This structure is reminiscent of pronominalization patterns with transitive verbs that take object clauses. ( 9 9 ) ( a ) I think so. (b) I believe so. (c) I guess so. (d) I ' m afraid so. In light of the structures shown in (98a & b ) , if one were to try to uphold the view that seem and appear take subject clauses, one would have to explain the source of the two pronouns in these sentences: it in subject position and so in object position. Subject clauses cannot be pronominalized after they have been extraposed. ( 1 O O ) ( a ) It is likely that it will rain tomorrow. (b) *It is likely fit. 4 that. Iso. But this explanation follows naturally under Analysis III. The i_t of (98a & b) is not an anaphoric pronoun referring to the clausal complement of seem or appear. It is an impersonal (nonreferential) pronoun which occurs as the subject of the SEEM verbs when they select a that clause as a complement. The complement clause which occurs as an object may be pronominalized by the pronoun so. Ruwet discusses a cataphoric pronominal construction as well and demonstrates that also in this case sembler does not exhibit syntactic behavior parallel to that of predicates or verbs taking subject complementation. Ruwet's example follows. ( 1 O 1 ) ( a ) ceci est probable: la flotte US coulera le Yamato. (b) *il est probable: la flotte US coulera le Yamato. ( 1 0 2 ) ( a ) *ceci me semble: le general Tojo est paranolaque. (b) il me semble ceci: le general Tojo est paranolaque. (Ruwet 1976a,17O) 1O The pronominalization facts are even more explicit in French as Ruwet demonstrates. The pronoun j/t in English is both an impersonal nonreferential pronoun (as in it is raining) and an anaphoric pronoun that can be coreferential with a sentence. French doesn't employ one pronoun with two functions as English it but has two distinct pronouns for these two distinct functions: il and ce, cela or ca, respectively. In a structure such as c^est possible, the pronoun ce is an anaphoric pronoun replacing a subject clause. The construction *cela me semble is not possible; the correct form of the sentence is il me le semble where the nonreferential jQ has subject function and the clitic pronoun le is an anaphoric pronoun in object function.
38
This construction - often referred to as right dislocation in linguistic literature - is exemplified for English in (1O3) with subject complementation and in (1O4) with object complementation. (103) It is not likely - that she will lift a finger to help. (104) I guessed it - that she wouldn't lift a finger to help. A construction parallel to (1O3) is not possible for seem. (105) *It seems - that she should want to help you. A dislocated sentence with seem takes the following form where an object pronoun is present in cataphoric function. (106) It seems so - that she should want to help you. These facts indicate that the complement to seem does not have the same function as the complement to likely. 3. Direct Questioning To carry our examination of the syntactic properties of the clausal complement to seem, appear, happen and turn out f u r ther, we notice that complements to SEEM do not accept an interrogative pronoun in subject position and can therefore not be questioned in the same manner as predicates such as likely, obvious, possible and so forth. ( 1 0 7 ) ( a ) It seems/appears/happens/turns out that John is a CIA agent. (b) *What seems/appears/happens/turns out? ( 1 0 8 ) ( a ) It is likely/obvious/possible that John is a CIA agent. (b) What is likely/obvious/possible? The clause that John is a CIA agent is therefore definitely not functioning as a subject to seem as it is to likely. It is important to observe that the following is also impossible. (109) *What does it seem? This would be the proper question under the assumption that the impersonal pronoun it_ is the subject to seem and the object clause being questioned is an instance of an NP. But ( 1 0 9 ) gives direct evidence against the latter assumption. An interesting result of Ruwet's study of the pronominalization possibilities of these complement structures in French is that the (object) clauses to sembler do seem to be NPs. They can be pronominalized by pronouns that ordinarily stand for NPs as in the following sentence. (110) Que te semble-t-il?
39
Here the interrogative pronoun que is obviously replacing the complement clause. The structure in English parallel to ( 1 1 0 ) , sentence ( 1 O 9 ) , is not good and therefore o f f e r s no such evidence. We can only say that the clause is not a subject to seem, therefore it must be an object. There is no evidence that this object complement is dominated by an NP node in the phrase marker in English. 4. Pseudo Cleft Sentences Completely in line with the discussion so f a r , we find that pseudo cleft variants of SEEM sentences are not grammatical. If a complement clause clearly functions as a subject to its matrix predicate, a pseudo cleft sentence is a possible variant. ( 1 1 1 ) ( a ) » W h a t seems/appears is that John drinks too much, (b)»What happens/turns out is that he never has any money when he needs it. ( 1 1 2 ) What is certain/obvious/likely is that he drinks too much. Kajita devotes considerable discussion to the relevance of the result of this pseudo cleft test with seem vs. certain type predicates in English (Kajita 1968, 28-30) and Ruwet comes up with the same results from French (Ruwet 1 9 7 6 a , 1 7 5 ) . 5. Parenthetical Phrases The parenthetical argument is due largely to Culicover ( 1 9 7 6 , 2 5 4 - 2 5 6 ) . He believes that verbs taking object clauses may form the type of parenthetical construction shown in ( 1 1 3 ) ( 1 1 3 ) The world is an oyster, I Tbelieve J suspect J realize (.hope ( 1 9 7 6 ,2 5 5 ) but that it is not possible to form this type of parenthetical with extraposed subject clauses. ( 1 1 4 ) *The world is an oyster, it Γ is obvious J bothers me j is surprising (.matters. ( 1 9 7 6 , 255) Thus, there is, he claims, a difference in the behavior of object clauses and extraposed subject clauses with regard to this construction. The latter once extraposed may not be reshifted back to topicalized position in the sentence. It is important to establish the truth of this claim because the verbs in question here pattern according to ( 1 1 3 ) .
40
(115)
The world is an oyster, it r seems I happens 1 turns out (.appears. (1976, 256) If paradigms ( 1 1 3 ) and ( 1 1 4 ) are indicative of verbs taking object and subject complementation respectively, then once again the SEEM verbs do not pattern like verbs taking sentential subject complementation but like other transitive verbs in English. This question is easier for Ruwet because French employs the referential pronoun in subject position of the parenthetical phrase (c'est evident) and the impersonal, nonreferential pronoun that replaces a subject clause that has been extraposed is not possible in a parenthetical (*il est evident). The would-be object pronoun accompanying a transitive verb in the parenthetical expression is deleted; in the case of croire 'to believe", for example, the parenthetical phrase is je crois. Sembler occurs parenthetically as il me semble parallel to je crois rather than *ca me semble like c'est evident, the latter being what we would expect if the complement clause of sembler were a subject. Therefore, the evidence in French is quite explicit that the complement to sembler does not function as a subject clause. Ross 1973, in an analysis of English parenthetical constructions which he calls suiting , discusses briefly in a footnote (p. 1 3 4 f . , f n . 4 ) the possibility that extraposed subject clauses can be reshifted to the front of the sentence and thereby build a parenthetical construction. He therefore postulates suiting as following extraposition in order of application in order to derive such sentences as ( 1 1 6 ) from the underlying structure ( 1 1 7 ) . 11 His analysis is called suiting (= S lifting) because the embedded S is lifted and Chomsky-adjoined to the left of the original matrix S, whereby the complementizer that is deleted. Max is a Martian/ I feel ( c f . ( i i ) ) is derived in this manner from (i) . (i)
£L·
(ii)
is Max a Martian
that is Max a Martian
41
(116) (117)
It rained, it is true. „[that it rained]
Ο
Ο
is true
This assumption directly contradicts Culicover's argument which rests on the belief that only object clauses can contain their original matrix clause as a parenthetical and not extraposed subject clauses. Ross is himself not sure about the relationship of extraposition to slifting. He classifies probable, likely, true, clear, evident and certain ( 1 9 7 3 , 136) as predicates that may undergo slifting
(once the subject clause has been extraposed) but
notices that an object clause, once extraposed, may not be slifted. [11 8)
C
He had only one left foot, f l fexplained} (*it) to Jo. (admitted i took (*it) for granted. (1973, 134f., f n . 4 )
Let us test Ross" claim that probable, likely, true, clear, evident and certain can appear in a parenthetical phrase. (119)
It rained, it is f«likely »probable «•certain ?evident •»clear true
These sentences are generally not good, although true yields a f a i r l y acceptable sentence. Nevertheless the question remains, if
some of the sentences in
( 1 1 9 ) are accepted by some
speakers, whether this is an equivalent construction to the one we are considering, for example that given in ( 1 2 0 ) . (120)
It rained, I think.
Predicates that take subject complementation other than the ones mentioned by Ross definitely yield bad sentences in a parenthetical construction, c f . ( 1 2 1 ) *It rained, it
is
( 1 1 4 ) and ( 1 2 1 ) .
irritating amusing horrifying possible ,being reported by the weather bureau.
Therefore this construction is not typical of subject clauses in general. At best there are a few predicates that allow Notice that a true parenthetical phrase like I thin]; has several variants.
it.
42
( 1 2 2 ) ( a ) It will rain again today, I think. (b) It will rain again, I think, today. (c) Will it rain again today, do you think? (d) Will it rain again, do you think, today? (e) It won't rain again today, I don't think. The putative parenthetical phrase it is true, which is the only clear case of ( 1 1 9 ) does not yield the f u l l paradigm of (122). ( 1 2 3 ) ( a ) It will rain again today, it is true. (b) *It will rain again, it is true, today. (c) *Will it rain again today, is it true? (d) *Will it rain again, is it true, today? (e) *It won't rain again today, it isn't true. The result of the comparison of ( 1 2 3 ) to ( 1 2 2 ) shows that the putative parenthetical phrase it is true is much more restricted than a parenthetical like I think. The fact that (123b) is not as acceptable as (123a) gives a clue, we believe, as to the nature of the construction. Notice that whereas it rained, it is evident might also be marginally acceptable, it too resists the word order of ( 1 2 3 b ) . ( 1 2 4 ) *It will rain again, it is evident, today. It seems that sentences like ( 1 1 4 ) and (123a) do not build the same type of construction as (122a) because the pronoun it is present which is coreferential with the clause that precedes it. The phrase it is true is complete and the tendency is to read it uncontracted, that is it + is + true and not it's true. The opposite is true of the parenthetical phrase I think of (120) Since I think isn't a complete phrase, it seems to be easier to incorporate into the sentence intonation span of the matrix clause. It is true/it is evident, as complete phrases, are not easily insertable into the preceding sentence as (123b) and ( 1 2 4 ) show. The parenthetical phrase it seems is equivalent to I think. It seems is not a complete sentence and therefore does not have sentence intonation of its own. It is an impersonal subject as we have shown and not an anaphoric pronoun coreferential with the clause to which it stands in a parenthetical relationship. The object is missing from it seems (=so) in the same way it is from I think (=scO. This object pronoun may be present in an at least parenthetical phrase.
43
(125)(a)
It rained, at least f I think so.
(b) (c) (126)(a) (b) (c)
Υ it (,* it
seems so. is true.
It rained, at l e a s t f I think that it did. V it seems that it did. l*it is true that it did.
In sentences
( 1 2 5 ) and ( 1 2 6 ) the verb seem functions d i f f e r e n t l y
from the predicate is true. It behaves like think. The parenthetical it
seems does not yield the entire paradigm
of the I think parenthetical. Nevertheless, it yields the (b-d) variants with considerable more ease than it (127)(a)
It will rain again today, it seems.
(b)
It will rain again, it
(c)
Will it rain again today, does it
(d)
Will it rain again, does it
(e)
is true.
seems, today. seem?
seem, today?
*It w o n ' t rain again today, it doesn't seem.
Thus, seem in a parenthetical construction exhibits syntactic behavior much more closely to think which takes an object clause than to a predicate like is true with a clause in subject function. Culicover 1976 and Ross 1973 both assume that the parenthetical construction is derived by means of a rule that lifts the subordinate that clause out from under the domination of the matrix clause and (Chomsky-)adjoins it to the left of the original matrix sentence. Emonds 1974 argues against this derivation and for a derivation that considers the two sentences involved to be "two successive independent clause S ' s concatenated without a coordinating conjunction, the second of which contains a proform referring back to the first" sentence (120) Emonds
(128)
1
(Emonds 1 9 7 4 , 2OO). For our
underlying structure would look like ( 1 2 8 ) E,
E
E
S.
it rained
I think so.
The pronoun so. under the second S carries the index i which "'" 1
shows that it
is coreferential with the f i r s t S ( = S . ) . Emonds
rejects Culicover and Ross' analysis because a) the source of parenthetical sentences like our ( 1 2 2 c & e) repeated here as (129a & b) would have to be structures like (13Oa & b) under Culicover or Ross' analysis which is obviously not correct
44
(129)(a)
Will it rain again today, do you think?
(b) (13O)(a) (b)
It won't rain again today, I don't think. Do you think whether it will rain again today? I don't think that it won't rain again today,
and b) there are certain parenthetical phrases that cannot be matrix sentences ( c f . Emonds 1974, 1 9 2 f . ) . Given an underlying structure ( 1 2 8 ) , Emonds postulates a rule of proform deletion to derive parentheticals. This rule deletes the referential pronoun in the second S and extracts the second S from the E node and attaches it as a sister node to the first S. (131)
E
Proform Deletion
it rained I think The nonrecursive symbol E in ( 1 2 8 ) is postulated to account for expressive elements such as yes, by God and a host of other possibilities that accompany sentences but cannot be embedded. In this fashion Emonds can relate parentheticals and similar semicolon sentences (It rained; I think so ( a n y w a y ) ) . In the derivation of the latter from ( 1 2 8 ) , the proform deletion rule would not apply. Emonds analysis predicts that parenthetical and semicolon sentences (of the type derivable from ( 1 2 8 ) )
are
related, the difference being that the coreferential proform present in the semicolon sentence is deleted in the parenthetical construction. Emonds does not explicitly address our question whether it is true is a parenthetical like I think is, but his analysis makes an interesting prediction which adds support to our negative conclusion. The subject it_ of it is true must be the proform that is coref erential to the preceding sentence it_rained_. Therefore the underlying structure of the putative it is true parenthetical must be ( 1 3 2 ) . (132) [it rained] PRO. is true b 1 i If this underlying structure is subjected to the same rule as the I think parenthetical is - proform deletion -, we derive (133)
*It rained, is true.
45
However, if proform deletion does not apply, we derive ( 1 3 4 ) (134) It rained, it is true. to which semicolon intonation is associated. Thus Emonds 1 analysis in effect says that this sentence ( 1 3 4 ) is not the same construction as our parenthetical ( 1 2 0 ) . Since proform deletion does not apply in the derivation of ( 1 3 4 ) , the second S is still dominated by E ( c f . ( 1 2 8 ) ) , which means that it may be accompanied by an expressive element. This is not possible under this analysis for a true parenthetical since proform deletion which creates the parenthetical phrase destroys the E dominating the second S. And this situation is exactly what we find. (135) It rained, fby God"!, it's true. 4 yes \ I shit J ( 1 3 6 ) *It rained, rby God), I think. I yes J ( 1 3 7 ) *It rained, fby God), it seems. \yes J Sentences (135) - ( 1 3 7 ) demonstrate once again the similarity of seem and its complement to think and its complement. Seem does not behave similarly to the predicate which takes subject complementation: is true. 6. As Parenthetical The discussion found under 1. - 5. exhausts the syntactic arguments offered in K a j i t a 1 9 6 7 , Culicover 1976 and Ruwet 1976a in favor of Analysis III; Ruwet discusses three other constructions, but they are relevant to French and not applicable to English. Nevertheless, once K a j i t a , Culicover and Ruwet have opened one's eyes as to what type of syntactic processes bear on the nature of the verbs seem, appear, happen and turn out and their sentential complements, other constructions come to mind which yield further evidence. For instance, the pattern of the SEEM verbs within the as parenthetical phrase indicates also that their complements do not function as subjects. (138) John, as is r obvious T , is lying. •f evident f I not hard to see J (139)
If, as is f ( e n t i r e l y ) possible), the devil prefers τ probable f I likely J sophisticated souls, then neither you nor I need fear hell's fire.
46
Sentences ( 1 3 8 ) and ( 1 3 9 ) contain an as parenthetical phrase construed with predicates such as obvious, evident, possible, probable, likely which overtly lack a subject within the phrase. The structure of the parenthetical phrase is ( 1 4 0 ) . (140) as is obvious. In each case the missing subject is understood as being identical to the matrix sentence which contains the parenthetical. For example, what is obvious in ( 1 3 8 ) is that John is lying. A pronoun may not occur in the blank space of ( 1 4 O ) . ( 1 4 1 ) »John, as it is obvious, is lying. Now let us observe the behavior of the SEEM verbs in the as parenthetical. ( 1 4 2 ) We thought we had enough money with us, but, as it turned out/happened, we d i d n ' t . ( 1 4 3 ) John, as it seems/appears, is not only a heavy drinker, but also a compulsive liar. The constructions in ( 1 4 2 ) and ( 1 4 3 ) require the subject pronoun it as opposed to those in ( 1 3 8 ) and ( 1 3 9 ) ; if the it is absent, the sentence is ungrammatical. ( 1 4 4 ) *We thought we had enough money, but, as turned out, we d i d n ' t . Under Analysis III it is not a coincidence that j.t is not possible with obvious and similar predicates but obligatory with seem, appear, happen and turn out. Let us apply Emonds' analysis of parentheticals to this type of parenthetical construction as well. The underlying structures of (138) and ( 1 4 2 ) would be roughly. (145)(a) [John is lying] as [Pro. is obvious] S x i (b) ... 0 [we didn't have enough money with us] as S i [it turned out P r o . ] The coreferential proform in the second successive clause has subject function in (145a) and object function in (145b) in accordance with the premises of Analysis III. In the process of the derivation of the parenthetical the proform which is coreferential with the previous S is deleted. This yields the desired results. Transitive verbs such as know may also occur in the as_ parenthetical phrase.
47
(146) John, as we all know, is a liar. The underlying structure of ( 1 4 6 ) would be similar to (145b) to which preform deletion would also apply. (147) [John is a liar] as [we all know P r o . ] 1 ι Thus, the behavior of the SEEM verbs in as parentheticals forces us to analyze them in a fashion parallel to transitive verbs in English. 7. Comparative Phrases If we compare our verbs to other predicates taking sentential subjects and sentential objects in comparison constructions headed by than, we find patterns similar to those of the preceding sections. (148) I have better nerves than it seems. (149) I have better nerves than is obvious. (150) I have better nerves than you think. The analysis of comparison constructions is certainly not this simple 1 2 , but let us assume for our purpose here that the than it seems, than is obvious and than you think phrases contain an embedded sentence that has approximately the form that I have χ good nerves (where χ stands for the degree to which the comparison phrase takes reference). This embedded sentence is then deleted from the than clause under identity with the matrix sentence. To account for the facts given in ( 1 4 8 ) - ( 1 5 0 ) , we are forced to postulate this embedded sentence in the following position. (151) (a) it seems S (b) S is obvious (c) you think S Once again the facts of the comparison construction in English indicate that the sentential complement to seem does not have a function comparable to that of obvious, a predicate taking subject complementation. 1 . 3 . 1 . 2 Conclusion The arguments that we have summarized in this section demonstrate that if we consider the SEEM verbs to be intransitive _ _ _ _ _ _ that ____ verbs are to be placed in a frame 0ΰ[ N P seem] where the 12 Cf. Bresnan 1972, Ch. 5 for a discussion of the difficulties involved in an analysis of this construction.
48
subject NP must be expanded to an S, we are complicating the grammar in more ways than just with respect to the rules of extraposition and raising; there are many syntactic processes that do not receive an adequate explanation. Under the assumptions of Analysis I or II, the only way the grammar can explain the d i f f e r e n c e between the putative intransitive predicates SEEM and other predicates which have subject clauses with regard to the constructions 1. - 7. of the preceding discussion is to mark the SEEM verbs with special rule features indicating that they cannot undergo pronominalization, wh-questioning and pseudo cleft formation as these other predicates may. But even with this ad hoc solution constructions 1. and 5. - 7. are not explained: the subject NP of the SEEM verbs must be realized as an S (that is, as a that clause or an infinitive) and may not be an ordinary nominal and the 3Ε1)Λ verbs take t a r t i n certain parenthetical, as and than phrases in a way not typical of verbs with subject complements. If, however, the SEEM verbs are fit into the underlying frame O i i t seems S] where the S is not dominated by an NP, the facts 1. - 7. fall into place naturally. Construction 1. is impossible because SELM do not select a subject, constructions 2. 4. are paradigms expected of subject complementation and therefore not applicable to the SEEM verbs and f i n a l l y , constructions 5. - 7. fellow on the premises of Analysis III without ad hoc stipulations by applying an analysis identical or similar to that proposed by Emonds 1974. l_>
1.3.2
Emonds 1970
In section 1 . 3 . 1 . 1 we discussed several straightforward syntactic reasons for classifying the SEEM verbs apart from predicates that take subject clauses. Emonds 1970 is another representative of this view for some of the reasons discussed in the preceding section - arguments 3. and 4. ( C f . Emonds 197O, 2O1) - as well as for another reason that is equally straightforward. Emonds observes that the SEEM verbs and their complement structures are very well accounted for within a system which considers all sentential complements to originate in VP final position of the phrase marker. In fact, Emonds 1970 rejects the idea of extraposition for all sentential predicate
49
complements because he feels that by making the inverse assumption to extraposition many of the irregularities that had to be patched up by rule features in Rosenbaum's extraposition account fall into place as regular syntactic behavior under this alternative assumption within the confines of his structurepreserving hypothesis. Instead of allowing that and infinitival complement clauses to be generated in subject and object positions in a sentence and then permitting them to be moved to sentence final position by means of extraposition, Emonds suggests generating all complement clauses in extraposed position to begin with. He intends to capture with this analysis the contrasting distribution between clauses on the one hand and NPs on the other, gerunds belonging to the category NP. All nonsubject NP positions exclude that and infinitival clauses 14 according to Emonds and NPs do not occur in extraposed position - since a gerund is considered an NP, the nonextraposibility of a gerund is thus explained. Nothing further need be said about oblique complement clauses as Emonds terms them (in Rosenbaum's terminology these are the verb phrase complements), but clauses that are interpreted as either subjects or objects (in Rosenbaum's terminology, noun phrase complements) require a means of representation at the deep structure level which ensures their correct interpretation since they themselves never hold these positions. Emonds calls these clauses antecedent clauses because, although they originate in VP final position, they are obligatorily accompanied by a coreferential pronoun (it) which occupies either subject or object position of the verb depending on how the 13 The structure-preserving hypothesis entails roughly a proposal to constrain the expressive power of transformations within transformational grammar to three well-defined and restricted classes of operations: root transformations, structure-preserving transformations and minor movement rules (later called local transformations). See Emonds ( 1 9 7 6 , 1-6) for a precise definition of these classes of transformations as well as for a statement of the structurepreserving constraint. 14 Cf. Emonds (1970, 75-78 and 85-91) where he tests the positioning abilities of Ss in direct and indirect object position, after prepositions, in possessive NP position of gerunds and NPs with head nouns, in apposition to NPs, in conjoined position with NPs, in focus position of cleft sentences and in the passive by_ phrase.
50
embedded S is to be interpreted. The deep structure markers (152) and (153) illustrate the structural difference between oblique and antecedent complements. ( 1 5 2 ) oblique complement ( 1 5 3 ) antecedent complement S
V
(Emonds 1970, 98) In ( 1 5 2 ) the embedded S is not coreferential with any NP in the phrase marker. The S . of ( 1 5 3 ) is coreferential with the object N P . ; it would therefore be interpreted as an object. If the index on the embedded S were i, it would be interpreted as coreferential with N P . , i.e. as a subject. Since subject clauses may optionally occur in subject position of a matrix sentence - this is the only NP position that a clause may hold according to Emonds ( c f . fn. 1 4 ) - Emonds postulates a root transformation of subject replacement that optionally replaces the coreferential it in subject position of the matrix sentence with its antecedent S. Under this analysis certain facts f a l l into line. Since subject replacement is a root transformation and therefore, by definition, can only apply in a root sentence, the only position that will be permitted by the grammar for a clause apart from its underlying position (=extraposition) will be subject position of a matrix sentence. Under the extraposition analysis, Ross tried to account for this distribution by formulating a constraint that blocked an S which is immediately and exhaustively dominated by an NP from sentence internal positions. This constraint resisted formulation, however, because gerunds and free relative clauses may occur in these NP positions. ( C f . Ross 1 9 6 7 , 33-34). Furthermore, the extraposition of a subject clause had to be considered obligatory if any constituent preceded subject position, i.e. by means of wh-movement, topicalization or one of any of the other preposing root transformations
51
has applied, then subject replacement may not and the clause will remain in extraposition being interpreted as a subject because of the coreferential it which occupies subject position. Seem, appear, happen and turn out display perfectly regular behavior within Emonds' system. Their complement clauses may not undergo subject replacement, therefore he classifies them as oblique complements and would introduce them into base structures similar to ( 1 5 2 ) with the subject NP, empty. Since the S in extraposition is not coreferential with the subject NP, subject replacement could never apply. Within the course of the derivation, the empty subject NP would be filled by the expletive it. Emonds 1 197O analysis of extraposition has been heavily criticized ( c f . Higgins 1973 and Postal 1 9 7 4 , 3 9 6 - 4 O 4 ) . Whereas his analysis seems to make some interesting predictions about the distribution of that and infinitival clauses, thereby eliminating a series of ad hoc conditions on the extraposition transformation, nevertheless, it contains some unnatural features of its own. Firstly, requiring an antecedent pronoun occurring in subject or object position to accompany a clause in extraposition in the case of antecedent complementation would entail extending the theory of grammar as far as the base rules are concerned as well as the rules of semantic interpretation that would have to reconstruct the function of the extraposed clause on the basis of an antecedent relationship between an NP and an S. But under the extraposition analysis the interpretation of a clausal complement as a subject or an object occurs naturally by virtue of the position the clause holds in deep structure. Secondly, there are verbs which are subcategorized for a subject and an object clause (prove, mean, imply, be, persuade, convince, show e t c . ) . Emonds has to 15 Ruwet 1 976a also demonstrates that Emonds 1 analysis does not handle the extraposition data in French in a convincing manner. Antecedent complements which are to be interpreted as subjects originate in extraposed position according to Emonds with a coreferential pronoun in subject position of their matrix sentence. On the basis of this coreferentiality subject replacement may apply. The referential subject pronoun in French is cela, ya or ce (c_") , but the pronoun which occurs with normal extraposed subject clauses is il, the nonreferential pronoun. If cela etc. occur with extraposed
52
handle these verbs in a very unnatural way. Since all
comple-
ment clauses originate in extraposed position, he allows the sentence final S position to be doubly filled in deep structure in these cases and requires that subject replacement apply within the course of the derivation of such sentences to yield a well-formed surface structure ( c f . (1970, 9 5 f . ) · This amounts to a situation where two semantically and functionally distinct clauses occupy one node in deep structure. The fact that these verbs may have a subject and an object clause can be handled very naturally by the extraposition analysis which assumes that clauses can occur deep structurally in the same environments as ordinary nominals. Although Emonds is forced to retract his analysis because of these and other points, his insight into the nature of the complements of the SEEM verbs remains valid: there is no need to attribute a large degree of irregularity to these verbs by classifying their complement structures as anything but VP complements, i.e. as embedded sentences that are base-generated in extraposed position. 1.3.3
Bresnan 1972
1.3.3.1
Bresnan's Fixed Subject Constraint
In discussing Kajita, Culicover, Ruwet and Emonds 1 objections to the traditional viewpoint of the SEEM verbs we have been summarizing fairly concrete syntactic arguments with which Bresnan is also familiar. These facts about the SEEM verbs correlate with the results of her study on complementizers. In particular it is the infinitival construction like that shown in (154) which leads Bresnan to believe that some of Rosenbaum's basic assumptions about the nature of the complement clause to the SEEM verbs are mistaken. ( 1 5 4 ) Mary happens to have once killed a man. ( 1 9 7 2 , 117) Bresnan does not believe that the for complementizer underlies these infinitival constructions. First, there are conditions 15 cont. clauses the construction is interpreted as a case of right dislocation (for example: c'est etrange, que Tanaka se maquille si outrageusementland not ordinary extraposition (il est etrange, . . . ) . ( C f . Ruwet 1976a, 183-184.) Therefore Ruwet postulates the following deep structure frame for verbs like sembler where the subject NP is empty and will be filled with il during the derivation: [ Ν Ρ Δ . ] V S X (Ruwet 1976a, 16ίΓΓ.
53 under these (155) (156)
which a complementizer does not delete, but even in cases for c a n ' t occur with happen. *It happens, unfortunately, for Mary to be too tall. *For Mary to be too tall happens. (1972, 122)
Secondly, if the embedded sentence of ( 1 5 4 ) were a for complement Mary shouldn't be able to be shifted out of the complement, because this would violate Bresnan's fixed subject constraint (FSC). The FSC asserts, in essence, that a subject NP adjacent to a complementizer is "fixed" and cannot be moved from that position ( 1 9 7 2 , 9 5 ) . Sentence (157b) exemplifies a violation of the FSC with a that complementizer, sentence (158b) with a for complementizer. In sentences (157b) and (158b) the subject of the complement clause has been moved to sentence initial position by wh-movement. ( 1 5 7 ) ( a ) You believe that someone fired on you. (b) «Who do you believe that fired on you. ( 1 9 7 2 , 96) ( 1 5 8 ) ( a ) You'd like for someone to accompany you. (b) *Who(m) would you like for to accompany you? (1972, 97) This is behavior peculiar to a subject NP next to a complementizer, if a complementizer has been deleted no violation ensues, cf. ( 1 5 9 ) . A nonsubject NP is also free to move, cf. (160). ( 1 5 9 ) Who do you believe fired on you? (160) What does he believe (that) you did? ( 1 9 7 2 , 96) The derivation of ( 1 5 4 ) under Analysis I would entail moving the subject of the embedded clause from a "fixed" subject position. ( 1 6 1 ) it happens [for Mary to have once killed a man] % J Λ = fixed position If these infinitival complements are for complements, that is, contain the for complementizer, then the whole set of subject raising data refutes the FSC, for each time a complement subject would have to be moved from a "fixed" position next to the complementizer for. If the FSC were stricken from the grammar, there would be no principled way of accounting for ungrammatical sentences such as (157b) and (158b). ^
^/
|
16 Bresnan terms this phenomenon "Subject Shift".
54
There is no way to reconcile sentences like ( 1 5 4 ) with the FSC. It is no good postulating complementizer deletion before subject raising, because subject raising can occur in sentences where a complementizer wouldn't ordinarily be able to delete. ( 1 6 2 ) ( a ) *It didn't seem to Mary John was very sick. (b)
John didn't seem to Mary to be very sick.
( 1 9 7 2 , 123)
( 1 6 3 ) ( a ) *It never happens John is here when the dishes need to be washed. (b)
John never happens to be here when the dishes need to be washed. (1972, 123) Bresnan concludes from this data that these clauses are not instances of for complements; in fact, there is no indication that they contain any complementizer at all. They are instances 17 of a "bare" infinitive unaccompanied by a complementizer. The consequence of the assumption that these are not for complements is, then, that these complements cannot be subject clauses. Subject clauses must contain a complementizer. ( 1 6 4 ) *John is here worries me. (1972, 123) As we have seen, two fundamental assumptions that Rosenbaum makes in the analysis of the infinitival constructions accompanying the SEEM verbs have been called into question by Bresnan's study of complementizers: that the for-to complementizer underlies such infinitival constructions and that these infinitival complements are underlying subject clauses. 17 It is Bresnan's thesis throughout Chapters 1-3 of her dissertation that verbs must be subcategorized for certain complement types (i.e. for certain complementizers) and that complementizers must therefore occur in deep structure. This is further substantiated, she believes, by the fact that complementizers aren't semantically empty but do have semantic function. Bresnan isolates in Chapter 2 the semantic function of the 3 complementizers of English, that, for and wh. By assuming that the complementizer for has semantic function, and that for doesn't underlie the infinitives resulting from subject shift, Bresnan is implying that there is a semantic difference between the 2 types of infinitive structures: the for and "bare" infinitives. In section c of Chapter 3 she discerns another large class of infinitives that do not contain an underlying complementizer - the believe type of verbs, or those verbs traditionally thought to undergo subject-to-object raising. She then postulates that the class of bare infinitives (seem/tend and believe types together) can be differentiated from for infinitives semantically in that the bare infinitive complement occurs with verbs that have "aspectual" or "perceptual" function ( 1 9 7 2 , 1 7 7 ) and the for infinitive has "inherent intentional
55
1.3.3.2
Bresnan's Solution
Bresnan is interested in providing a solution for the subject raising verbs which does not rest on the mistaken assumptions of Analysis I and which offers a better explanation of this phenomenon. The subject raising verbs are represented by the list in ( 1 6 5 ) . ( 1 6 5 ) sure, certain, likely, bound, apt, turn out, prove, happen, seem, appear, start, begin, continue, cease, tend, have, fail ... Analysis I does not explain what the difference is between these verbs that can raise the complement subject and those that c a n ' t ; this difference is simply coded into the transformational sequence by means of ad hoc markings on the individual verbs. Bresnan proposes to account for this difference by means of subcategorization. She would strictly subcategorize the verbs of ( 1 6 5 ) for the following configuration.
(166)
S, HP
I
£*.
t
V]
v Nl
(1972,129)
When one of the verbs from the list in ( 1 6 5 ) is placed under the V of ( 1 6 6 ) , this deep structure must undergo subject raising. Thus, Bresnan's analysis does not need to refer to ad hoc rule features. It also makes different assumptions about the nature of the complements to the verbs in ( 1 6 5 ) . As reflected in ( 1 6 6 ) , Bresnan assumes that characteristic of the subject raising verbs as a class is that they do not select a subject, their complement contains no complementizer and this "bare" infinitival complement is an underlying object clause. The verbs in ( 1 6 5 ) fall into two major subgroups. One subset exhibits subject raising structures only (prove, tend, fail, has, be bound, be apt, begin, start, continue, cease . . . ) . The other subset (sure, certain, likely, seem, appear, happen and turn out) allow in addition to the subject raising structures that complements. Bresnan handles this construction by a second subcategorization. 17 cont.meaning" ( 1 9 7 2 , 8 7 ) . Syntactically the difference in the presence or absence of for can be discerned also by d i f f e r i n g paradigms dictated by the restrictions placed on the for infinitives by the FSC.
56
1167)
I that
( 1 9 7 2 , 130)
This subcategorization cannot compete with ( 1 6 6 ) as a source for the subject raising structure - the presence of the complementizer that prohibits the complement subject from being raised (FSC) . An expletive _i£ will fill the Δ. under the subject NP within the course of the derivation. Thus, the difference between the following sentences (168)(a) (b)
May seems to be the nicest month of the year. It seems that May is the nicest month of the year,
is attributed to two different subcategorizations of the verb seem, those shown in ( 1 6 6 ) and ( 1 6 7 ) . The surface structure depends on which of the two underlying frames seem was inserted into. If seem was inserted into ( 1 6 6 ) , (168a) results, if seem occurred in ( 1 6 7 ) , the result is ( 1 6 8 b ) . Common to both ( 1 6 6 ) and ( 1 6 7 ) is that the underlying sentential complement is considered to be an object clause. Having established the analysis this far, Bresnan concedes that, although this is unusual for the subject raising verbs, a small class of exceptions (likely, sure, certain) can take a subject that clause as well. This calls for another subcategorization . (169) S
(it)
( 1 9 7 2 , 135) ( 1 6 9 ) like ( 1 6 7 ) cannot be a source for subject raising sentences, because the COMP prevents the complement subject from being raised.
57
1.3.3.3
A Shortcoming of Bresnan's Solution
We agree with Bresnan's observations as far as the transitive nature of the SEEM structures is concerned. Her total analysis, however, produces the following paradox. Likely has been subcategorized for an object that clause on the basis of its subject raising properties which accounts for sentence (170b) and for a subject that clause because of the grammaticality of sentence ( 1 7 0 a ) . ( 1 7 0 ) ( a ) That May is the nicest month of the year is likely. (b) It is likely that May is the nicest month of the year. Therefore, the relationship of sentences (170a & b) is not explained by means of extraposition in Bresnan's analysis but by means of verbal subcategorization. Now let us compare likely with obvious. Obvious is a predicate which does not belong to the subject raising class and must be subcategorized for a subject that clause. If obvious is subcategorized for a subject clause only, then sentence (171b) below must be derived by means of a movement transformation extraposition - working on the underlying structure (171a). ( 1 7 1 ) ( a ) That May is the nicest month of the year is obvious. (b) It is obvious that May is the nicest month of the year, The sentence pairs (170) and ( 1 7 1 ) are striking parallel but Bresnan's analysis does not capture this fact: (170b) is derived by a second subcategorization for likely but (171b) by the extraposition of a subject clause to obvious. Her analysis fails to give a plausible explanation for the possibility of sentence (17Oa) on the explicit assumption that likely as a subject raising verb takes object complementation. Simply adding the additional subcategorization ( 1 6 9 ) to the grammar is extremely ad hoc and unconvincing after she has given explicit reasons why the infinitival complement to these subject raising verbs cannot be a subject. The problem with Bresnan's solution centers around the LIKELY predicates which she has not characterized properly. Likely ends up with two different subcategorizations for the same that clause depending on whether it occurs superficially in extraposed or in subject position. This is especially unsatisfactory since the complement in (17Oa) has the same syntactic
58
function as in ( 1 7 0 b ) ; they are both subjects, cf. what is likely? The dilemma which this solution produces results from assuming that all verbs that subject-raise raise the embedded subject from an object clause. The fact that the predicates likely, sure and certain accept that clauses in subject function along with the results of the other syntactic tests applied in section 1.3.1.1 indicates that the that clause to these predicates is a subject clause and not an object clause. In fact, it is precisely the purpose of this chapter to show that seem, appear, happen and turn out differ from the other predicates of (165) and the predicates like obvious in exactly this point - there are no plausible syntactic arguments that the SEEM verbsexhibit sentential subject complementation, whereas this is not true for the LIKELY or OBVIOUS predicates. The arguments of this section speak for classifying the that clauses of seem, appear, happen and turn out - not, as Bresnan has done, together but - separately from those of likely, sure and certain. 1.4
Summary
In this chapter we have discussed three different approaches to analyzing the that and infinitival complements to the SEEM verbs that can be found in the linguistic literature since the beginning of transformational grammar. Analysis I. Rosenbaum 1967 lays the framework for this traditional analysis which still constitutes the most widely accepted analysis of the SEEM verbs in the linguistic literature. The key assumptions involved are the classification of the sentential complements to the SEEM verbs as subject clauses and the postulation of the obligatory application of the extraposition transformation to these subject complements. This is accomplished in terms of rule feature markings on predicates. Other typical samples of this traditional analysis are Kiparsky/Kiparsky 197O and ohring 1977 who try to explain the exceptional behavior of the predicates which require the obligatory application of extraposition on their sentential complements in terms of the notions "nonfactive predicate" and "raising". Their suggestions turn out to be equally as ad hoc as Rosenbaum's rule features because they cannot distinguish between SEEM and LIKELY pre-
59
dicates, both being nonfactive and both allowing raising, but it is only to the former that extraposition must apply. The fact that extraposition must be made to apply obligatorily to the SEEM constructions causes a great deal of syntactic irregularity in the grammar which indicates that there is something wrong with the analysis. Analysis II. The flip/psych movement analyses require assumptions about the nature of linguistic theory that are not justified within the strict syntactic framework of this topic. It is predominantly out of concern for clarity of exposition of Analyses I and III that this analysis is discussed here. Analysis II is ambiguous between Analysis I and Analysis III. If one looks beyond the actual initial structure, one has, in essence, the traditional Analysis I. However, the deep structures of Analysis II give the appearance of unanimity with Analysis III. This unanimity is spurious. The two analyses cannot be equated at the deep structure level (i.e. the feature they have in common) because they dissent in their definition of the notion deep structure. The former assumes a unity of syntax and semantics at this level, the latter upholds the autonomy of syntax thesis: whereas Analysis II is arguing for this deep structural configuration in semantic terms, Analysis III bases its premise on syntactic arguments exclusively. Analysis II then,for all practical purposes, is a variant of the traditional Analysis I. Analysis III maintains essentially that the basic assumption of Analysis I is mistaken: the sentential complements to SEEM verbs are not sentential subjects to which extraposition must apply, but they are sentential objects. With this solution the problem of obligatory extraposition and the resulting complexity to the grammar is nullified. K a j i t a , Culicover and Ruwet discuss convincing syntactic reasons for this conclusion. The arguments center around the paradigms of anaphoric and cataphoric pronominalization as well as pronominal patterns in direct questions, pseudo cleft constructions, parenthetical clauses and comparative clauses. These diagnostic "tests" for "subjecthood" and "objecthood" disclose d i f f e r e n t properties in the sentential complementation of SEEM predicates as opposed to LIKELY predicates, namely, when a that clause complement is
60
pronominalized in the manner just described, an object pronoun substitutes for it in SEEM sentences, whereas a subject pronoun occurs in LIKELY sentences. Emonds, working with the inverse assumption (intraposition) to the one endorsed by the remainder of the works discussed in this chapter (extraposition) - the extraposition-intraposition dichotomy will be a central topic of discussion in the next chapter -, concludes the object nature of the SEEM complements as opposed to the subject nature of LIKELY complements by the inability of the that clauses accompanying the SEEM verbs to undergo his subject raising transformation as all other subject complements do. Bresnan's argument is on a slightly more theoretical plane. She recognizes the vailidity of K a j i t a ' s and Emonds 1 arguments that that clause complements to the SEEM verbs are objects. Her contribution to Analysis III, however, lies in her observations concerning the infinitival complement to SEEM and LIKELY. She finds in her study of complementizers that this infinitival complement, which results in the subject raised infinitive, does not contain the complementizer for as Rosenbaum assumed. She postulates a fourth complement type to account for this, the bare infinitive complement, and shows how the subject raising process would not be able to apply to produce the correct infinitive if a true complementizer (that is, for) were present in this complement; such a process would violate her fixed subject constraint. Furthermore, a subject complement must always contain a complementizer. Therefore, out of shere syntactic necessity, the infinitival complement to SEEM as well as to LIKELY must originate within the VP. We have taken great pains in section 1.3.1.1 to examine the syntactic patterns of English that bear on the nature of a predicate sentential complement. From the empirical observations made in this section, we conclude that Analysis III is the only analysis that is in accord with the syntactic facts of the language. That is, the SEEM verbs are transitive in nature; they take finite sentential complements in object function.
61
(172) It (173)(a) (b) (174) It (175)
seems that times,are getting harder. *It seems. It seems so. ^ sentential subject is likely that times are getting harder. It is likely.
= sentential subject
The confusion that has arisen around the correct syntactic description of the SEEM verbs is undoubtedly due to the superficial similarity between SEEM constructions and constructions where the finite sentential complement is indeed a subject, in particular, between SEEM and LIKELY constructions as can be seen here by comparing ( 1 7 2 ) with ( 1 7 4 ) . The conclusion we have reached in this chapter is, however, that, inspite of coincidental superficial similarities, SEEM and LIKELY structures are not syntactic equivalents. The question that arises out of this conclusion to be explored in the following chapter is what ramifications this well-founded empirical observation has for the structure of the English predicate complement system.
2.
ENGLISH SYSTEM OF PREDICATE SENTENTIAL COMPLEMENTATION
The presentation of the differing views concerning the nature of SEEM constructions in the preceding chapter was organized into three general headings and touched upon a broad spectrum of linguistic opinion within the development of linguistic theory since 1965. In order to consider the consequences of our conclusions from Chapter 1 for the English system of predicate sentential complementation in a systematic way, we must first limit ourselves to a well-defined theoretical framework for our discussion. For this purpose we choose the most recent state of linguistic research in transformational grammar or what Chomsky terms the "revised extended standard theory" (REST),also called trace theory,which is clearly defined in such works as Chomsky 1973, 1975, 1976, 1977; Fiengo 1977; Chomsky and Lasnik 1977 and the like. 2.1
Derivation of SEEM Constructions within REST
In the following we will briefly introduce the framework of REST under the perspective of the topic at hand, namely how SEEM structures are to be derived. Motivated by the primary goal in linguistics of restricting the class of possible grammars in order to approach a theory of language acquisition, Chomsky has recently further restricted the framework of the extended standard theory to consist of a very simple core grammar which comprises the central and unmarked body of the grammar. The core grammar includes the base (i.e. the categorial component and lexicon) which produces the initial structures and only a few transformations, applying optionally, that derive surface structures from these initial structures. It is Chomsky's thesis that the core grammar can make do with two basic transformations i. NP-movement
ii.
wh-movement
63
accompanied by the assumption that a moved phrase·leaves behind a trace. That is, when a phrase moves by virtue of either i. or ii./ its empty labeled bracket remains behind with a referential index identical to that of the moved phrase. This is the trace of the moved phrase. With traces of underlying configurations present in surface structure, both the semantic component and the phonological component can work off surface structures. In the case of the former, rules of construal and interpretation will apply yield ing the logical form of the sentence. These rules are constrained by the principles of the cycle including subjacency; the propositional island constraint (PIC) and the specified subject constraint (SSC). Subjacency restricts the application of a rule of construal to one cyclic domain or adjacent cyclic domains. The propositional island constraint prohibits relating a constituent inside a tensed S to any element outside of that S. The specified subject constraint states that within a nontensed S, if a specified subject is present, it alone may be related to a constituent outside of that S. Under this conception of grammar, all major transformations that were previously thought to move or delete a phrase over a variable are now subsumed as instances of either i. or ii. This results in a radical reduction of the transformational component of the grammar. The earlier concept of an extensive transformational component containing a myriad of independent rules, obligatoriness conditions on rules, rule ordering, notions of clausemate, pruning etc. will be replaced by the simplier proposal just outlined. With this brief introduction, we will demonstrate how REST characterizes the derivation of SEEM and LIKELY complement structures based on the discussion in Chomsky and Lasnik ( 1 9 7 7 , 448-450). Since Bresnan 1972, the notion complementizer has taken on a more significant role in the theory of grammar. Lexical items are now subcategorized for the type of sentential complements they take and the appropriate complementizer (WH, that, for or 0) cooccurs under a special COMP node in initial structure. (1)
χ
s. COMP
64
Given the correct subcategorization for the relevant lexical items, the base will generate structures like the following. (2) [ A p likely [- that NP tense VP ] ] (3)
[ Ap illegal [·§ for NP to V P ] ]
(4)
[ Ap likely [- 0 NP to VP] ]
(5)
[ . _ unclear [^ WH PRO to do w h a t ] ] At
ο
( C f . Chomsky & Lasnik 1977, 448) These APs occur as predicates with the copula. That is,
they
occur in the AP position of the following configuration where the subject NP position is thought to be empty. (6)
NP is AP
Chomsky and Lasnik postulate a lexical rule to insert .it. into the empty matrix subject when the complementizer that, for or WH is present inside the S-complement to the predicate adjective. Lexical insertion, it will be remembered, takes place on the base phrase marker before the transformations apply. (7)
Insert _it in the position of NP in: ffor 1 NP be A [-4 that f S] 3 l+WH J
(Chomsky & Lasnik 1 9 7 7 , 4 4 9 )
Rule (7) inserts it into the empty subject position of the structures ( 2 ) , (3) and ( 5 ) . It is not applicable to structure (4) since no complementizer is present inside the S^ The following structures then enter the transformational component. (2) (a) it is [ ΛAir Τ , likely [οF that NP tense V P ] ] ( 3 ) (a)
it
is
( 4 ) (a)
[ N p e] is
(5) (a)
it
is
[ a D illegal [- for NP to VP ] ] Α-IT
Ο
[ A p likely [^ φ NP to V P ] ]
[ Λ Γ ) unclear Air
[- WH PRO to do w h a t ] ] ο
The transformational component carries only two optional instructions: move NP or move wh. (NP-movement is the only operation relevant to the immediate discussion. We will return to the scope of wh-movement in subsequent sections.) NP-movement moves the content of a category NP into an empty NP node present in the phrase marker and leaves its index behind as a trace of the movement. Consequently, NP-movement is applicable to structure
65
(4a) only, since in none of the other structures is there an empty NP node present at the point transformations apply. The subject NP moves up from out of the S-complement in ( 4 a ) to fill the empty matrix subject position leaving behind a trace of the original configuration. (8)
[Np
Bill] is [ A p likely [g [ N p
e] to like ice cream]]
If NP-movement shouldn't apply, structure ( 4 a ) will be rejected when it reaches logical form, since it has an open variable in matrix subject position and there is no rule of construal that can index this position. Thus, although transformations are formulated optionally, NP-movement applies, in a sense, obligatorily to ( 4 a ) . The specified subject constraint prohibits any NP other than the specified subject Bill from being moved. Thus, if NP-movement should move the NP ice cream out of the !3-complement into subject position yielding the following (9) [ N p ice cream] is [^ likely [^ 0 Bill to like t ] ] 1 this structure will be rejected at logical form on the grounds that it violates the SSC. In a nontensed sentence which contains a specified subject no other NP besides this subject can be affected by the rules of the grammar. Thus, from the initial structures (2) - ( 5 ) , Chomsky and Lasnik's it insertion rule derive ( 1 O ) , ( 1 1 ) and ( 1 3 ) and NP-movement applies to (4) to yield ( 1 2 ) . (10) It is likely that Bill likes ice cream. ( 1 1 ) It is illegal for Bill to steal ice cream. ( 1 2 ) Bill is likely to like ice cream. ( 1 3 ) It is unclear what to do. These same derivational processes apply to seem sentences. e (14) [ H V r> seems [— that Bill likes ice cream]] (15) [ p e ] [ p seems [·= 0 Bill to like ice cream]] The lexical item it will be inserted into the empty subject position of ( 1 4 ) since the complementizer that is present, deriving ( 1 4 a ) to which NP-movement cannot apply since there is no vacant NP in the sentence. However, it: cannot be inserted into (15) because no complementizer is present. NP-movement is in effect obligatory as it was in the case of ( 4 a ) . 1 The trace of a moved phrase is abbreviated frequently as ' t 1 .
66
(14)(a)
[ N p it]
(15)(a)
tNp
[seems [^ that Bill likes ice cream]]
B i l l ] f v p seems [^ 0 t to like ice cream]]
In this manner, REST derives the syntactic structures associated with SEEM and LIKELY sentences, but it makes no explicit statement about the syntactic function (subject vs. object) of the complement sentence. The parallelism of the derivation gives the impression that the following sentences are syntactically equivalent. (16) It is likely that Bill likes ice cream. (17) It seems that Bill likes ice cream. We have conclusively argued in Chapter 1 that this is not the case. Such misconceptions in the linguistic literature, even in recent linguistic literature, underline the importance of the topic of this dissertation. We conclude therefore, that the derivation of such structures as conceived in Chomsky and Lasnik 1977 is inadequate as it stands in that it does not capture the differing syntactic nature of the predicates seem and likely in sentences ( 1 6 ) and ( 1 7 ) which came to light in the discussion of Chapter 1. If the sentential complements to seem and likely sentences are to be derived in the manner Chomsky and Lasnik suggest, the question with which linguistic theory is now faced is how the sentential complements shown in ( 1 6 ) and ( 1 7 ) are to be assigned their syntactic function. These sentences have a similar origin and derivation in the present theory but they are not syntactic equivalents. The derivation of sentences ( 1 6 ) and ( 1 7 ) runs parallel within REST because the that clause as a s"-complement subcategorized by the lexical items seem and likely originates in both cases under the AP and VP node dominating the predicates likely and seem. In the following section we will j u s t i f y this specific assumption within REST on the empirical grounds of actual syntactic distribution of sentential complements within the English system of predicate sentential complementation as well as on the grounds of theoretical considerations. The task of section 2.3 will then be to provide the grammar with a means of differentiating the object clause of ( 1 7 ) from the subject clause of ( 1 6 ) in line with the results of Chapter 1.
67
2.2
The Distribution of Sentential Complements
2.2.1
Extraposition vs. Intraposition
2.2,1.1
Extraposition
The need for a rule of extraposition in Rosenbaum's 1967 analysis of predicate complementation arose from the assumption that certain sentential embeddings may bear subject or object relationships to a matrix predicate in the same manner as nominal complements do. To account for this such sentential complements were introduced into the phrase marker in the same place as the corresponding nominal complement would occur, i.e. under the subject or object NP node. Since, at this early stage of transformational grammar, deep structure was thought to contribute solely to the semantic interpretation of a sentence, this was the most plausible way for the grammar to account for the interpretation of such clauses as subjects or objects. Then, because embedded sentences are frequently found at the end of their matrix sentence, they were thought to be able to move to sentence final or extraposed position from these NP positions. But the extraposition transformation moved only the S node and the material it dominated leaving the NP node behind intact. Ss in their VP final position were not considered to be NPs. The extraposition rule responsible for this movement of embedded sentences away from NP positions in the phrase marker to sentence final position had to be annotated by a series of ad hoc conditions and restrictions stipulating when it must apply; that is, an embedded sentence always had the option of being extraposed, the stipulations on the extraposition rule were needed to explain when the embedded sentence could not maintain its underlying NP position in surface structure. Ross 1967 attempted to generalize the conditions under which a clause must undergo extraposition by formulating the following output condition for the purpose of prohibiting such phenomena exemplified by the sentences in ( 1 9 ) . (18) Grammatical sentences containing an internal NP which exhaustively dominates S are unacceptable. (Ross 1967, 33)
68
( 1 9 ) ( a ) *Did that John showed up please you? (b) *That that John showed up pleased her was obvious. (c) *I went out with a girl who that John showed up pleased. (d) *I want that Bill left to remain a secret. (Ross 1 9 6 7 , 33) However, Kuno 1973 gives examples of that clauses in sentence internal positions that would violate Ross 1 output constraint but that are grammatical. (20) Believing that grapes are sour gives one some solace. (Kuno 1973, 3 6 4 ) To permit sentence ( 2 O ) , the concept "internal" in ( 1 8 ) would have to be amended to be understood not as internal to the sentence as a whole, but, more explicitly, internal to the immediate parent constituent, in which case the that clause in (2O) would be the right-most constituent of its parent node and therefore a possible construction (Kuno 1973, 3 6 6 f . ) . However, with this emendation, Ross 1 constraint while allowing (2O) and prohibiting ( 1 9 a , c & d ) , would no longer rule out sentence ( 1 9 b ) , since that John showed up would be the left-most constituent and not an internal constituent with regard to its parent node S». Nevertheless, the intent of Ross 1 constraint is well-taken: the occurrence sentence internally of Ss in NP positions is highly restricted. The problem - in order to state the conditions on the extraposition transformation accurately - is to pinpoint an exact specification of the formal conditions under which such positioning is not permitted. In an attempt to do just this, Kuno 1973 postulates not one constraint as Ross did but four different constraints: 1. the internal NP clause constraint, 2. the self-embedding and conjunction-juxtaposition constraint, 3. the constraint on surface subject clauses and 4. the predicate deletion constraint, cf. Kuno ( 1 9 7 3 , 3 7 5 f . ) . The first constraint allows for ( 2 0 ) but rules out ( 1 9 a , c & d ) , the second prohibits ( 1 9 b ) , the third says that subject clauses may only appear in sentence initial position, excluding ( 2 1 ) and, redundantly, ( 1 9 a ) and the fourth accounts for the ungrammaticality of sentences like ( 2 2 ) . ( 2 1 ) *How likely is that John will come? (Kuno 1973, 371) ( 2 2 ) What is important? »That we work harder is. (1973, 3 7 4 )
69
Thus, the extraposition account of the distribution of sentential complements to predicates is an extremely complex analysis. The ad hoc nature of the constraints it requires obscures any explanation by the grammar as to how or whether embedded sentences are incorporated within a matrix sentence. 2.2.1.2
Intraposition
Emonds 197O rejected Rosenbaum's extraposition analysis on the observation that embedded sentences in reality do not occupy nominal positions in sentences. Emonds demonstrates the inability of sentential complements to occur in typical NP positions of a sentence - i.e. in direct object position, after prepositions, in the passive by_-phrase, in focus position of cleft sentences, in apposition to NPs, in conjoined position with NPs and even in subject position of all The following is a summary of Emonds
1
nonroot sentences.
data that Ss do not have
the distribution of NPs ( c f . 197O, 7 5 - 9 1 ) . 1. Direct Object Position *She won't tell she is sick to the doctor. (She won't tell the doctor she is sick.) 2. After
S
Prepositions
»He blamed it on that Bill was too strict. H e blamed i t o n Bill's s t r i c t n e
s
s
.
N
S P
3. Passive by_ Phrase •»The situation wasn't helped (by) that you spoke out of turn.
S
4. Focus Position of Cleft Sentences *It was to buy a new hat that I wanted. S » I t ' s that John has come too late that Bill realizes. S I t ' s the custard pie that I disliked. NP 5. In Apposition to NPs »All this constant activity, and to go out every night, All this constant activity, and going out every night,
to buy food twice a day is wearing me out. buying food twice a day is wearing me out.
S &S NP & NP
6. Conjoined with NPs »To pitch a tent everyday and outdoor bathrooms would bother me. S Outdoor bathrooms and pitching a tent everyday would bother me. NP
70
7. Nonroot Subject Position »That for Bill to smoke bothers the teacher is quite possible. S That Bill's smoking bothers the teacher is quite possible. NP »Although that the house is empty may depress you, it pleases me. S Although the house's being empty may depress you, it pleases me. NP The one exception Emonds found to the generalization that Ss do not have the distribution of NPs was that subject clauses may optionally occur in sentence initial position of a root sentence. (23) That Bill knows German thoroughly is obvious to all. ( 2 4 ) For the house to be painted would irritate him. (Emonds 197O, 79) Emonds maintains that it therefore only complicates the grammar to introduce Ss into NP positions in underlying structure and then be forced to extrapose them by means of a variety of ad hoc conditions in every case that either an NP position isn't 2 ambiguous with extraposed position or a subject clause is found in sentence initial position. Emonds chooses to generate all sentential complements in a sentence final S position in the base. The initial positioning of a subject clause in a root sentence can be handled simply by an optional root transformation moving the VP final clause to the front of the sentence. The extraposition and intraposition analyses d i f f e r in their claim as to how sentential complements to predicates are incorporated deep structurally into the overall sentence structure. Extraposition says there are two ways in which this integration occurs: an S can be the sole expansion of any NP node, i.e. can hold positions typical of nominals in a sentence, and can also occur in an S position as the final element under the VP node. Intraposition makes the much more restrictive general-
2 The following sentential complement to teach for example doesn't have to be extraposed only because its positioning is ambiguous between the object NP position and extraposed position. (i) He teaches that books are important. As soon as an indirect object is included into the sentence, however, it becomes apparent that the sentential object occupies the extraposed position in reality. (ii) »He teaches that books are important to his sons. 3 This is the source of the term "intraposition" since this rule is viewed as the inverse of "extraposition".
71
ization that there is one position only typical of such complementation in the base phrase marker and this is the VP final S position. Intraposition further embodies an interpretation mechanism in the base to account for the difference in subject and object function of these VP final clauses. Furthermore, the transformational component will contain an optional root transformation allowing a clause, on the basis of an anaphoric relationship with a subject pronoun (it) in the base, to be substituted for this pronoun. The extraposition analysis further presupposes a movement rule which removes any S dominated by NP out from under this NP position to the VP final S position of its matrix structure and a system of constraints on the extraposition rule as well as rule feature markings on several classes of verbs indicating when the implementation of this movement is imperative. The extraposition and intraposition analyses have different ramifications for the theory of phrase structure. Both extraposition and intraposition allow the VP symbol to be expanded to a V and, among other possible complements, an optional S. Extraposition, in addition to this, embodies the claim that the 4 NP symbol can directly dominate an S. (25) NP
I
S
2.2.2
Lexicalist Argument
One of the major concerns of the lexicalist position within transformational grammar, formulated initially in Chomsky 1970 and further endorsed by works like Jackendoff 1977, is the formulation of a theory for constraining the base component of the grammar, that is, the formal nature of the lexicon as well as the schema of the phrase structure rules.
4 In Rosenbaum 1967 the phrase structure rule responsible for introducing noun phrase complementation ( ( i ) ) generates the following structure when applying to sentential complements where the DET is empty and the features [+N, +PRO] refer to the pronoun it. (i) NP - DET N (S) (Rosenbaum 1967, 1) (ii) DET
72
2.2.2.1
Chomsky 197O
Chomsky 1970 examines the difference between gerundive nominale and derived nominale and comes to the conclusion that gerundive nominals as shown in ( 2 6 a ) can be correctly characterized
as
transforms of related underlying sentences, for example ( 2 6 b ) , but that a similar transformational derivation for derived nominals, as in ( 2 7 ) , would require such ad hoc complexity in the transformational mechanism of the grammar as to rob the analysis of any plausible explanatory power. (26)(a) (b) (27)
John's being eager to please John is eager to please John's eagerness to please
The hypothesis that certain nominals (e.g.
(Chomsky 197O, 187) ( 2 7 ) ) cannot be ex-
plained convincingly as the product of a nominalization transformation means that they will have to be accommodated in the phrase structure rules of the base. Thus, simplification of the transformational component leads to an extension of the categorial component. Chomsky argues that this is a motivated step by showing that there is another class of nominals which cannot be the result of a nominalization transformation because there exist no underlyingly related verbs from which they could be derived, cf. (28)(a) (b)
( 2 8 ) for two examples.
the author of the book a war of aggression against France
(Chomsky 1970, 1 9 6 )
With the recognition of the necessity of generating certain classes of nominals directly in the base, it becomes apparent that the phrase structure rules must be extended to accommodate all
the complement types that may occur with these nominals as
4 cont.Technically, then, an it accompanies, in deep structure, each sentential complement that is an expansion of NP. But clearly the pronoun it, subcategorized in a totally ad hoc fashion for the frame [+ S] N p - since no other nominal nor any other member of the class of pronouns can occur under the N when it accompanies S (this observation is due to Emonds ( 1 9 7 6 , 1 1 8 f . ) -, cannot be considered the head of a noun phrase construction allowing an optional S as its complement. The grammar is so constructed that if the S does not get extraposed away from this NP position within the course of the derivation, the it must delete leaving the NP dominating the S exhaustively. It is only when the S does not occur under the NP that the it may appear in surface structure, although this isn't always true of an it in object position. 5 Cf. Chomsky (1970, 188-193) for arguments that support his view. Jackendoff ( 1 9 7 7 , 7-18) reviews Chomsky's position.
73
well (NP, PP, S and fixed combinations of t h e s e ) . This step too has further motivation; there are noun phrases which cannot be adequately described as the reduction of a head noun plus relative clause construction, c f . (29)(a) (b)
(29).
the weather in England atrocities against civilians
(Chomsky 1 9 7 0 , 1 9 6 )
Once a nominal is looked upon as taking a wide range of complements in the base in a similar fashion to a verb or adjective, an obvious structural parallelism among the lexical categories V, A and N emerges. This parallelism can be captured in
the
phrase structure rules as a significant linguistic generalization which in turn places rigid constraints on the possible schema of phrase structure rules. Chomsky proposes the X notation to accomplish this. Under the X notation the permitted phrase structure schema conforms to the following pattern. The variable X stands for
the lexical categories (V, A or N) and
X stands for the phrase which contains X as its head (= a VP,
AP or N P ) . (30)
X * X ...
(Chomsky 197O, 210)
The dots ( . . . ) in (30) indicate that X ( i . e . any one of the lexical categories) is followed by a f u l l range of complement structures. X can also be preceded by syntactic elements which Chomsky terms specifiers and this yields the following generalized phrase structure pattern. (31)
X ·» [Spec, X] X
(Chomsky 1970, 210)
The term specifier subsumes basically the determiners associated with nouns, auxiliary elements with verbs and qualifiers with adjectives. The skeletal structure produced by the rules (30) and ( 3 1 ) is represented in the following tree diagram.
(32) SpecX
CompΛv
( C f . Chomsky 1970, 2 1 1 )
The X notation predicts that the basic lexical categories engage in structures with complements and specifiers to yield the possible class of higher syntactic categories, that is, the phrase structure, of the language. It can be looked upon as a theory constraining the formal nature of phrase structure rules, predicting that rewrite rules d i f f e r i n g from this pattern will yield nonoptimal linguistic structures in terms of the evalua-
74
tion measure of the grammar. Thus, the possible syntactic categories of a language are to a large extent defined concretely in terms of the lexical categories of the language and the specifier and complement constructions in which they take part. What began as a decision to extend the base component in a well-motivated way to derive certain classes of nominals that do not lend themselves to a transformational derivation has lead, in fact, to a highly constrained theory of phrase structure. The virtue of the X theory of syntax is that it captures the formal parallelism of the phrasal categories as an important feature of phrase structure. Once the possibility of stating this parallelism has been achieved, the structure of the lexicon can also be simplified. Lexical entries may take on a neutral character with respect to the syntactic features N , V and A. Selectional restrictions and subcategorization can be stated once for the neutral lexical entry, capturing the generalization that related lexical items, whether realized as a verb or noun or adjective (criticize, criticism or critical, for example) exert the same - or at least extremely similar - subcategorizational and selectional restrictions on their environments. This is possible because, with the highly constrained form of the syntactic phrase under the X theory of syntax, grammatical relations like 'subject of and Object o f 1 can be stated formally with considerable ease to generalize across categories. Under this viewpoint, the relationship between the lexical categories is no longer expressed by means of transformation but rather in the phrase structure rules, and above all, lexicon, hence the term lexicalist hypothesis. 2.2.2.2
in the
Jackendoff 1977
Jackendoff 1977 attempts to construct a f u l l theory of phrase structure for English by developing Chomsky's X convention to its fullest extent. He proceeds from the major and minor lexical categories and defines the further possible syntactic 6 That is V, N, A & P and M, Art, Prt, Det, Adv & Q respectively. For Jackendoff, the distinction between major and minor lexical
75
categories of the language in terms of these symbols in the prime notation: phrase structure rules have the form shown in (33) , but - and here he d i f f e r s from Chomsky 1970 - n may equal up to 3 primes for all lexical categories, which he terms his uniform three-level hypothesis, cf. Jackendoff ( 1 9 7 7 , 3337) . (33) X n * ... X n ~ 1 ... (Jackendoff 1977, 34) The restriction placed on a possible phrase structure rule by a schema such as ( 3 3 ) is that each phrasal category (X ) must rewrite as the same category with one prime less in value (X n ) . Specifiers and complements are optional elements that may occur at any of the levels of n. This predicts that each phrasal category must eventually contain a lexical head of the same category or X , where n equals 0. Such a theory of phrase structure as conceived in Chomsky 197O and extended in Jackendoff 1977 calls into serious question a number of phrase structure configurations that have been used by linguists in the linguistic literature. Under the X theory of syntax, phrase structure rules are so constrained that one major phrasal category cannot be rewritten as another. Specifically, a rule of the form ( 3 4 ) is impossible. ( 3 4 ) NP - S But this type of rule is in essence presupposed by the extraposition analysis as one possible type of sentential complementation to a predicate. On the other hand, it is Emonds' 1970 observation that this phrase structure rule finds no justification in the pure syntactic facts of English: Ss do not have the distribution of NPs. Thus Emonds' 197O observation concerning the nature of sentential predicate complementation adds some support to the lexicalist hypothesis in its attempt to constrain the possible class of grammars - specifically, the base component of grammar - in terms of the X theory of syntax. 6 cont. categories lies in the syntactic feature [+complement]; all major lexical categories take complements but the minor ones do not. 7 There are two types of rules which stand out as exceptions to this schema, rules introducing conjoined structures and a class of deverbalizing rules of which the gerundive nominale is an example, cf. Jackendoff ( 1 9 7 7 , 5O-53 and Chapter 9) for discussion.
76
Moreover, under the lexicalist position, the structural parallelism in the base of the lexical categories and their complement structures, as illustrated in Chomsky's schema (30) and in Jackendoff's somewhat altered conception ( 3 3 ) , is thought to constitute a significant linguistic generalization. That is, a nominal like necessity and its range of complements should reflect the structure of the related predicate necessary and its complements in order for the appropriate subcategorization to be stated once for the lexical entry independent of the syntactic features [ + N ] or [+V]. For a nominal like necessity and its sentential complement we have the form shown in ( 3 5 ) but not in ( 3 6 ) . (35) the necessity for John to leave (Jackendoff 1 9 7 7 , 76) (36) «for John to leave's necessity In order to maintain the parallelism between NP and S (or N and V, in lexicalist t e r m s ) , the sentential complement to the predicate necessary should be generated following it as well, c f . (37) .
in the base
(37) It is necessary for John to leave. (Jackendoff 1977, 76) The results of Emonds' 1970 study give Jackendoff independent reason to do this: Ss are not NPs and therefore cannot be basegenerated in the subject NP position. Here again Emonds' intraposition hypothesis and the lexicalist position complement each other. In Chomsky 197O before Chomsky had incorporated the intraposition analysis into his thinking, he was forced to derive the nominal shown in ( 3 5 ) from ( 3 8 ) because he considered ( 3 9 ) the correct underlying structure of the sentential counterpart in line with the extraposition analysis. (38) [for John to leave]'s necessity (Chomsky 1970, 205) ( 3 9 ) For John to leave is necessary. Chomsky accomplished the derivation of the nominal by assuming that the generalized form of the agent postposing transformation must apply obligatorily to ( 3 8 ) . Besides the valid objection as to how and why an embedded sentence can be considered an agent, the putative underlying form shown in ( 3 8 ) violates Chomsky's conception of phrase structure in terms of the X notation because it allows an NP node to be expanded directly as an S with no nominal head. But this violation of the restrictions placed upon phrase structure by the X
77
notation holds not only of the nominal construction
in ( 3 8 ) ,
it is also true of the sentence in ( 3 9 ) as well. Jackendoff considers ( 3 5 ) and ( 3 7 ) to represent the base order of the lexical items N and V and their complement Ss and assumes that a subject replacement transformation is free to apply to ( 3 7 ) to produce the sentence variant ( 3 9 ) . Jackendoff is only concerned with the phrase structure of English; he does not show how the subject replacement transformation would function to place the VP final clause optionally in sentence initial position although he assumes its existence ( c f . 1 9 7 7 , 2 2 4 ) . It is not at all easy, however, to see how this transformation would be a serious possibility for Jackendoff. The subject S would either have to be placed under the subject NP or would replace the NP node altogether. The f i r s t possibility would violate the spirit of Jackendoff's phrase structure rules: if NP cannot expand to S in the phrase structure rules for well-founded reasons, why should this structure be allowed later on in the derivation? The second possibility would amount to destroying structure in an arbitrary fashion. Therefore, it seems that a transformational analysis of the sentence initial placement of a subject clause is not the best solution. In the next section we will turn to a suggestion in Koster 1978 which allows a circumvention of these problems inherent in a transformational derivation of the sentence initial subject clause from sentence final position by means of subject replacement and thus improves upon one of the shortcomings of the intraposition proposal. Thus, the facts concerning the nature of phrase structure which have come to light under the lexicalist hypothesis and its attempt to restrict the options for transformational grammar speak against an analysis of sentential predicate complementation in terms of extraposition. Whereas Emonds argued for the VP final generation of all predicate sentential complements on the empirical groundsthat Ss never occur in nominal positions of a sentence, the lexicalist views of Chomsky 1970 and Jackendoff 1977 prefer such an analysis on the theoretical grounds that it allows a more highly restricted theory of phrase structure.
78
2.2.3
Sentential Subjects
2.2.3.1
Fullest Extension of the Intraposition Premise: Koster 1978
Like Emonds 1970 and Jackendoff 1977, Koster 1978 believes that there is no rule of extraposition: clauses are basegenerated in VP final position. By working within the framework of REST and especially by drawing on Chomsky's wh-movement analysis of topicalization in Chomsky 1977, Koster is able to carry Emonds 1 generalization that Ss are not NPs to its fullest extent, i.e. without the previous exceptional behavior of root subjects, and to refute at least one of the major objections to the intraposition analysis which have inhibited many linguists from preferring it to extraposition. In Roster's view the following sentences are instances of the same construction topicalization. ( 4 0 ) That he will come is clear. ( 4 1 ) That he will come I regret. Cf. Koster ( 1 9 7 8 , 5 7 - 6 1 ) . For Koster, Ss do not occupy NP positions in sentences. If a sentential subject occurs in sentence initial position - as in (4O) -, it does not occur under the subject NP node, but rather under the TOPIC node accompanying S. According to Chomsky ( 1 9 7 7 , 90-94) an initial symbol S can be expanded to the configuration TOP s" (S - TOP £5) yielding the base-generated source of a topicalization structure. Under Roster's assumption that sentence (4O) is a case of such topicalization, the sentence would have the initial phrase marker shown in ( 4 2 ) . (42) Jf.
that he will
>
I
VP is clear
That is, the sentential subject originates outside the matrix sentence and the subject NP of the matrix sentence contains a wh-element. The derivation of ( 4 2 ) proceeds in the following manner, cf. Koster (1978, 6 0 ) . The wh-phrase is moved into
79
COMP leaving behind a trace under the subject node. (43)
[|[TOp that he will come] [s[ C O M p w h ] [ N p t] is clear]]
The moved wh-phrase must delete. (44)
[|[TOp that he will come] [g[ COMp 0] t N p t ] is clear]]
The sentence S^ ( ' t is clear 1 ) is an open sentence, i.e. a sentence containing a variable ( = t ) that cannot be indexed since the antecedent to t has been deleted. A rule of predication which is necessary for relatives, left dislocations and pseudo cleft sentences as well, according to Chomsky's analysis of these constructions, will interpret the empty NP of the matrix sentence as being about the topicalized constituent. The derivation of sentence ( 4 1 ) , which contains a topicalized object clause, is parallel to that of ( 4 O ) . The clause that he will come originates as in ( 4 2 ) under the TOP node of the phrase marker, but the wh-element occurs in this case as the direct object to the verb regret. (45) I,
that ne will come Wh-movement applies placing the wh under COMP and leaving a trace under the object NP. The wh-phrase under COMP is then deleted. Because of the trace in the matrix sentence ( Ί regret t 1 ) , the rule of predication interprets the topicalized clause as being about this empty object. The consequence of Roster's analysis is that there is no such thing as a subject clause in the traditional sense. Subject clauses are either generated in extraposed position or are cases of topicalization, just as object clauses are. Q
8 This result of Koster 1978 was already hinted at in Emonds 1976. The outcome of the rules applying to clausal complements there predict that a nongerund clause will either end up in extraposed position or in topicalized position, the latter being found within the COMP node of S.
80
With this reanalysis of the position that a clause, interpreted as a subject clause, holds in sentence initial position, the generalization can be stated without exception that Ss do not hold NP positions in the sentence. There are specific S positions in a sentence - for example, VP f i n a l position or topicalized position. A phrase structure rule of the form NP ·* S has no j u s t i f i c a t i o n . Koster's account has invalidated at least one of the major objections to the original account in Emonds 197O, namely that for bisentential verbs ( i . e . verbs such as imply, mean etc. that may simultaneously take sentential subjects and objects) the VP f i n a l S node would have to be doubly filled. Bisentential complementation receives a natural explanation under Koster's analysis. A verb like imply has in underlying structure a topicalized subject clause and an object in extraposed position. ( 4 6 ) That it is snowing implies that winter is here. There is no need to postulate a doubly filled node to account for such sentences, in fact the assumption that a node cannot be doubly filled prohibits the ungrammatical ( 4 7 ) . ( 4 7 ) *It implies that winter is here that it is snowing. K o s t e r ' s analysis explains a sentence like ( 4 6 ) far better than either the extraposition analysis or the intraposition analysis: under extraposition, the optional extraposition rule would have to be constrained so as not to apply to the subject complement of a bisentential verb, otherwise ( 4 7 ) results and under intraposition, where a structure like ( 4 7 ) is the point of departure, the optional subject replacement rule would have to be obligatory, otherwise the grammar would predict ( 4 7 ) to be acceptable. K o s t e r ' s analysis incorporates the virtue of Emonds 1 analysis in explaining the inability of Ss to act like N P s . Furthermore, Koster points out, it o f f e r s an explanation for the well-known fact that subject clauses act like islands (see Ross* 1 9 6 7 ) sentential subject constraint). Nothing can be extracted from a subject clause occurring under TOP since the S dominating TOP has no COMP node. ( 4 8 ) »What did [that John saw t] surprise Mary. ( 1 9 7 8 , 62) There is really no evidence to believe that that clauses and infinitives occupy NP positions in a sentence and this includes
81
subject NP position. Koster 1978 o f f e r s the suggestion that initial position of a root sentence is syntactically ambiguous between subject position and topicalized position and shows that there is an extremely plausible derivation for sentence initial subject clauses in terms of base-generated topicalized structures. This approach allows for the total generalization that Ss d i f f e r from NPs in their distribution and alleviates the need for a dubious rule of intraposition (or subject replacement) to place a base-generated VP f i n a l clause under a root subject NP node. The conclusion that we have reached with the help of Emonds 1970, Jackendoff 1977 and Koster 1978 is that NP positions in sentences are distinguishable from S positions, the latter being restricted to the final constituent position under the VP or the special presentential TOPIC node. 2.2.3.2
A Counterargument to Intraposition and its
2.2.3.2.1
Higgins 1973
Solution
Before accepting an intraposition analysis, we must consider one strong anti-intraposition viewpoint. Higgins 1973 believes that there is a flaw in Emonds 1 1970 claim that all sentential complements originate in VP final position. Higgins purports to demonstrate that there is evidence in the grammar that such complements must occur at least deep structurally in NP positions and hence be subject to a rule of extraposition even though the grammar might exclude these complements from NP positions at the surface structure level; that is, require them to be extraposed by some means similar to Ross' output condition. Higgins believes that this can be most clearly demonstrated by viewing the transformational history of certain sentence types containing embedded sentential subjects. Emonds has shown that sentential subjects cannot occur in subject position of an embedded sentence. Higgins claims that this can't be a general condition on deep structure configurations; they must be able to be generated in such positions. That this needs be the case can be inferred from the fact that there are instances where, in order for certain transformational rules to apply properly, a sentential subject must be in embedded subject position at the time the rule applies.
82
Higgins argues that in order to derive the correct version of the pseudo cleft, topicalized, sentential relative and comparative construction, and not the (b)-version given below, the sentential subject must be in the embedded subject position at the time the respective rules apply. The ungrammaticality of the sentences (49b) - (52b) arises as a result of the sentential subject being extracted from extraposed position. Since under Emonds 1 intraposition analysis the sentential subject will never occur in the necessary position, sentences like ( 4 9 a ) - (52a) will not be derivable. ( 4 9 ) ( a ) What is most likely is that Susan said she would be late, (b) »What it is most likely is that Susan said she would be late. ( 5 0 ) ( a ) That Susan would be late John didn't think was very (b) (51)(a) (b) (52)(a) (b) Higgins
likely, »That Susan would be late John d i d n ' t think it was very likely. John hasn't arrived yet, which is odd. »John hasn't arrived yet, which it is odd. John's contributions are more important than was originally realized, »John's contributions are more important than it was originally realized. is working within a pre-REST framework that postulates
an extensive transformational apparatus. He assumes that sentences ( 4 9 ) - ( 5 2 ) are created by separate and independent rules, ordered with respect to other transformations, each with its own structural description and structural change: pseudo cleft formation is either an extraction rule or a deletion rule, topicalization is a movement rule, the transformation which forms a sentential relative is a special case of a relativization rule which is, in essence, a rule moving a whphrase, and the comparative construction results from deletion under identity, cf. Higgins (1973, 153-165). However, Chomsky's 1977 wh-movement analysis reduces pseudo clefting, topicalization, relativization and comparative constructions in sentences with nominal complementation to special cases of one more general process, wh-movement. If we can ex-
83
tend Chomsky's wh-movement analysis to deal with sentential as well as nominal constructions of these types, Biggins' argument would no longer seem cogent. If our supposition is correct, sentences ( 4 9 ) - ( 5 2 ) do not arise from the extraction or movement of a subject clause from an embedded subject position or the deletion of a subject clause in this position under identity with another S in the sentence but have another derivation which does not conflict with the hypothesis that all clauses originate phrase finally in the base. 2.2.3.2.2
Chomsky 1 9 7 7
Wh-movement moves a phrase marked with wh into the COMP node to the left of the complementizer. In English at least, this wh-phrase inside COMP may be shifted, subject to subjacency, from this COMP into further COMP positions under certain bridge conditions. ( 5 3 ) ( a ) who did Mary meet t (b) who did you tell Mary that she should meet t (Chomsky 1 9 7 7 , 84) With the process so conceived, its diagnostic properties, according to Chomsky 1977, are that it leaves a gap and that where COMP to COMP movement is possible, the construction seems to violate the three general constraints on the grammar: subjacency, PIC and SSC, cf. Chomsky ( 1 9 7 7 , 8 6 ) . It is precisely this paradigm of properties that the constructions under consideration share which leads Chomsky to believe that they can be explained in terms of wh-movement. In other words, all major transformations of movement and deletion that do not fall under the domain of the NP-movement rule are cases of wh-movement. For Higgins, four independent operations apply to produce sentences ( 4 9 ) - ( 5 2 ) . In each case he believes the subject clause is found in embedded subject position in the input structure that feeds the transformation. However, if we can extend Chomsky's proposal to these sentences, then the different sentence types ( 4 9 ) - ( 5 2 ) are not the result of independent operations but should all be explanable in terms of the more com9 prehensive process of wh-movement. In each case an underlying 9 Chomsky has only considered nominal complementation,nevertheless it is obvious how the analysis can be extended to incorporate sentential complementation.
84
wh-proform would be found in the target subject position not a sentential complement. The sentential complement occurs outside the matrix sentence domain in specially designated S-positions. The initial phrase structure rules assumed to apply in the derivations are given in ( 5 4 ) . (54) R1 : I - TOP 55 R2: S~ -* COMP S (Chomsky 1 9 7 7 , 91) R1 is the base-generated source of topicalized sentences. Pseudo cleft sentences can be derived by assuming that their basegenerated structure is a variation of this, namely = [ S is TOP] Ο which brings out an interesting relationship between the two constructions. The derivation of sentences ( 4 9 ) - ( 5 2 ) will proceed as follows. 1. Pseudo Cleft Sentence, underlying structure ( 4 9 * ) ( a ) [ = [ ^ [ C O M p J s [ w h a t is most l i k e l y ] ] is [ T Q p that Susan said she would be l a t e ] ] Wh-Movement (b)[|[^[ C O M p w h a t ] [ g t is most likely]
is [ TQp that Susan
said she would be l a t e ] ] 2. Topicalized Sentence, underlying structure ( 5 0 ' ) ( a ) [ = [ T O p that Susan would be l a t e ] [ - John d i d n ' t t h i n k ] [JT[ COMP ] g [what would be very l i k e l y ] ] Wh-Movement and Subsequent wh-Phrase Deletion (b) [|[τορ that Susan would be late] ^-[John d i d n ' t think] '•s'-COMP
what
H s t would be very l i k e l y ] ]
0 3. Sentential Relative, underlying structure
(51 ' ) (a) [ = [g John h a s n ' t arrived y e t Hcj[ C O M p
] [ g which is o d d ] ]
Wh-Movement ( b ) [ = [- John h a s n ' t arrived y e t Hs[ C O M P which] [ g t is o d d ] ] 4. Comparative Construction, underlying structure ( 5 2 ' ) ( a ) [ _ - John's contributions are more important] than[g-[ C ( ~ M p ] [·$ what was generally realized]] ο
Wh-Movement and (in most dialects) wh-Phrase Deletion (b) [ττ John's contributions are more important] than Ο [g-[ COMp w h a t ] [ g t was generally r e a l i z e d ] ] 0
85
We arrive at the surface structures shown in the (b)-variant where trace (t) indicates the gap left by wh-movement. Certain conventions apply to the surface structures derived by the base and transformational components to yield logical forms of sentences. They will introduce a variable in place of the nonterminal trace. Logical Forms ( 4 9 " ) [ = [ - [ C O M p what x ] [ s χ is most likely]]is [ TQp . . . ] ] ( 5 0 " ) [ mlUir ^ „ . . . ] [·= John d i d n ' t think[ οF * would be u n l i k e l y ] ] ο ( 5 1 " ) [ = [ £ . . . ] [g[ C O M p which x ] [ s χ is o d d ] ] ( 5 2 " ) [g· . . . ] than
[·ζ χ was generally realized]
The wh-movement analysis derives in all four cases open sentences. An open sentence has no interpretation of its own but refers to another constituent in the construction. Chomsky believes that the same rule of interpretation that applies to logical forms of relative clauses interpreting them as referring to their head NP can be used to interprete the logical forms of all constructions derived by wh-movement. In these cases the rule of predication will associate the variable χ with the propositions in TOP and S respectively. In none of the derivations of Higgins' examples ( 4 9 ) - ( 5 2 ) was it necessary to assume that a subject clause occurs in an embedded subject NP position. What has occurred in the position in question has not been a clause at all but rather an underlying wh-phrase in pronominal form, i.e. what or which. A trace eventually results from the removal of this wh-proform by wh-movement. The proposition to which this trace refers is present in the environment: in the TOP node of the pseudo cleft and topicalization structures and in the S^ to which the sentential relative and than comparative phrases refer. The property of a movement rule that causes it to leave a trace, a convention carrying surface structures with trace over to logical forms and a rule of predication working on logical forms ensure the proper interpretation of this underlying whproform. Chomsky 1977 shows that constructions of the types mentioned by Higgins 1973 all share the properties typical of wh-movement. If they are derivable in the manner proposed by Chomsky, the
86
transformational component will be significantly restrained. Since REST provides the means for an adequate analysis of the constructions illustrated in Biggins' examples ( 4 9 ) - ( 5 2 ) without entailing the assumption that a sentential complement must be generated in an underlying NP position (in this case, embedded subject position), we assume that Higgins 1973 does not refute our premise that sentential complements do not have the distribution of NPs. 2.2.3.3
Derived Sentential Subjects within an Intraposition Framework
In the discussion of section 2 . 2 . 3 . 2 we touched upon the scope of the transformation of wh-movement within Chomsky's revised version of the extended standard theory and the compatibility of such a theory with the premises of an intraposition analysis of sentential complementation. The second transformation of the core grammar, NP-movement,
is
thought to encompass - besides the previous subject raising rule - also the passive transformation. The rule NP-movement will, for instance, apply to the structures (53a) - (55a) below, moving the constituent [ _ . . . ] into the empty NP posiNf .
tion in the sentence. (53)(a)
[ N p e][ v p seems][ g [ N p
John][ y p to be a nice f e l l o w ] ]
(54)(a)
[ N p e ] [ y p be-en destroy][ N p
i
the c i t y ] [ p p by the enemy] k
(55) (a) [ N p e]^ be-en believe][ g [ N p
John][ y p to be a suit-
i able candidate]] The empty labeled bracket [ _
e] remains behind as the trace i of the movement. Thus, we derive (53b) - (55b) by NP-movement. (53(b) [ N p John][ y p seems][g[ N p e ] [ y p to be a nice f e l l o w ] ] (54)(a)
fNp
the city][ v p be-en destroy][ N p
(55)(b)
[Np
e][pp JL
J_
John][ y p be-en believe][ £ [ N p
by the e . ] Λ.
e ] [ y p to be . . . ] ] (Chomsky 1975, 1 1 3 f . )
10 I have labeled the agent phrase of the passive with PP, not NP as Chomsky does, because I believe that it derives in this position (not from NP-postposing) and the label PP will prevent it from being shifted by NP-movement yielding »by the enemy was destroyed the city t.
87
The similarity between these passive and subject raising structures is captured in this analysis. If the passive rule is to be subsumed together with subject raising under the rule of NP-movement , as seems desirable, it must be pointed out that it cannot apply to sentences containing object clauses transforming them into derived subject clauses under our thesis that clauses are not NPs. Thus, sentence ( 5 6 ) cannot have the derivation shown in ( 5 7 ) , parallel to sentences ( 5 4 ) and (55) . (56) That John has quit smoking is assumed by most people. ( 5 7 ) ( a ) * [ N p e][ v p be-en assume][ N p [ g that J. has quit smoking]; ( b ) * [ N p [„ that J. has quit smoking ] ] [
be-en assume]
[NP ± e]
NP-movement is involved in the derivation of ( 5 6 ) , but not in this manner. In section 2 . 2 . 3 . 1 we said that sentence initial clauses do not occur under the subject NP node, but they are instances of topicalization. ( 5 6 ) must derive from the underlying structure shown in (58a) with a passive verb and a clause in the TOP position. The topicalized clause is understood before the application of NP-movement (passive) as the object to assume, therefore the wh-proform generated simultaneously with the constituent TOP is found within the object NP. Both transformational rules, NP-movement and wh-movement, will apply to this configuration but NP-movement does not touch the S-complement. NP-movement applies to (58a) moving NP . which contains the wh-pronoun into the empty subject NP to derive (58b) . Then wh-movement applies placing the derived subject NP. into COMP yielding (58c) . The wh-phrase in COMP will be deleted and we are left with a surface structure containing traces of the two movement rules, cf . (58d) . (58) (a) [ T0p that J. has quit smoking] [^COMF H s f N p e] [ yp be-en assume ] [ N p
what]] i
TOP
t
quit
Sm
° king] C S [ COMP
[ yp be-en assume] [ N p e ] ] i
88
(c)
[ T0p that J. has quit smoking][-[ C O M p [ N p CNP
(d)
e]C
VP
be
what]
~ en assume H N p _ e ] ]
[ TOP that J. has quit smoking ] [ · * t is assumed t ] ]
The appearance of the two traces in (58d) has justification. Under the assumption that all semantic interpretation takes place at the level of surface structure, we need to know what grammatical relationships were present in the initial phrase marker. The device trace provides the surface structure with this information. In this case the clause was the object of assume before the application of NP-movement. More precisely, it is a case of a topicalized clause understood as the object of assume; the wh-word associated with the clause in TOP was originally found under the object NP node. The second trace in (58d) gives us this information. After NP-movement the whphrase became a derived subject. This information is supplied by the first trace in ( 5 8 d ) . Wh-movement into COMP and subsequent wh-phrase deletion in this position obliterates the original wh-proform, but the traces of it remain. Thus, the appearance of the two traces at the level of surface structure gives us two necessary pieces of information for the semantic interpretation of the sentence. 1) The relation between the that clause and assume is one of verb and object. 2) When NPmovement applies to (58a) the underlying object becomes a derived subject. Thus, underlyingly the sentence is understood as a topicalized object clause construction but after NP-movement applies the surface structure consists of a derived subject-predicate relationship between the topicalized clause and assume. The variable χ will replace the occurrences of t in logical form and be interpreted as being satisfied by the proposition inside the TOP node as discussed in the preceding sections. Passives of the following form are always acceptable. ( 5 9 ) It was 'reported ' that Carter is seeking reelection. assumed feared suggested .believed .
89 It was •complained' that NYC was on the verge of answered bankruptcy. muttered quipped ^protested that there are too many idiots in It was T f e l t (61 Washington this term. said thought found seen „ Sentences of the form [56) a r e more recalcitrant, c f . ( 6 2 ) ,
(60)
(63) and ( 6 4 ) . 1 1 ( 6 2 ) ?That Carter is seeking reelection was
reported assumed feared suggested believed.
( 6 3 ) *That NYC was on the verge of bankruptcy was
complained answered muttered quipped protested. ( 6 4 ) *That there are too many idiots in Washington this term is f f e l t said thought found ,seen. The restrictions on sentences of the type found in ( 6 2 ) - ( 6 4 ) can be reduced to restrictions on topicalization under our analysis. We have already pointed out that the following nonpassive sentences are d i f f i c u l t . (65)
[66)
That he can't control his emotions any better than that is possible. That there are too many idiots in Washington this term is clear.
!67) That the devil prefers sophisticated souls is probable. In these sentences a proposition has been focused in the topicalized position of the sentence but there is so little semantic material following the topicalized clause that such a focusing is hardly justified, i.e. slightly anomalous, unless the predicate receives the proper phonetic stress. 11 Jürgen Lenerz brought the importance of this question to my attention and supplied me with examples similar to ( 6 O ) , ( 6 1 ) , ( 6 3 ) and ( 6 4 ) . I am very grateful to him for his comments.
90
An analysis of passive and nonpassive sentences with initial subject clauses in terms of topicalization provides a basis to explain the intuitive similarity between the sentences ( 6 2 ) and ( 6 5 ) - ( 6 7 ) , namely as similar deviations from normal conditions on topicalized sentence structure. When the predicate in ( 6 5 ) - ( 6 7 ) receives a little more padding, the sentence is greatly improved. (65') (66") (67")
That he c a n ' t control his emotions any better than that is entirely possible. That there are too many idiots in Washington this term is clear to a large majority of voters. That the devil prefers sophisticated souls is highly probable.
This improvement can also be felt when the passive sentences in ( 6 2 ) receive additional semantic content in the form of an agent phrase or an adverb. (62')
That Carter is seeking reelection was f reported last night on the ABC news J feared by all Democrats everywhere ] believed by no one (.suggested to the press by Jody Powell.
To explain the d i f f e r e n c e felt between sentences ( 6 2 ) and ( 6 5 ) and ( 6 2 ' ) and ( 6 5 ' ) , that is, the fact that ( 6 2 ' ) and ( 6 5 ' ) are invariably more acceptable, we might appeal to a perception strategy that marks a TOP+S sequence as unbalanced if a f u l l clause makes up the focused constituent in TOP and is followed by an open sentence that is too stripped of semantic material to j u s t i f y focusing such a heavy constituent. The open sentence accompanying TOP in ( 6 5 ) contains the barest minimum possible on semantic content - a gap and an intransitive verb. Thus, a structure like ( 6 8 ) is unbalanced. (68)
[|[ TOP S ] [ s t V]
= unbalanced
If either appropriate stress is placed on the verb or other constituents are present to give the sequence 't V more prominence, the open sentence is perceived as being less stripped of weight and consequently the S is not so unbalanced, hence the greater acceptability of ( 6 5 ' ) . The fact that ( 6 2 ) is even worse than ( 6 5 ) under this perception strategy can be explained in terms of the two gaps present in the open sentence of ( 6 2 ) . The intransitive verb in ( 6 5 )
91
has lost its subject from the immediate S by wh-movement and subsequent wh-phrase deletion that derives a topicalized sentence. The transitive verb in ( 6 2 ) is missing its object by NP-movement which derives the passive and its subject by whmovement and subsequent wh-phrase deletion responsible for topicalization. (69)
[§t T O p S ] [ g t V t ] ]
= very unbalanced
The S in ( 6 2 ) (e.g. t was reported t) has been stripped of even more of the necessary semantic material to interpret its verb than the S in ( 6 5 ) (e.g. t is possible). However, if as in the case of ( 6 5 ) - the S in ( 6 2 ) receives additional constituents, as in ( 6 2 1 ) , the open sentence is perceived as less stripped and consequently the S is no longer quite so unbalanced. This perception strategy applicable to topicalized sentence structure enables us to explain why ( 6 2 ) and ( 6 5 ) are more marked structures than ( 6 2 ' ) and ( 6 5 " ) and why ( 6 2 ) is even less acceptable than ( 6 5 ) . The explanatory power that we have won by this analysis gives us reason to believe that the analysis and the premises on which it is based are correct. It is not clear why the verbs in ( 6 3 ) and ( 6 4 ) do not permit a topicalized clause at all, even when the passive agent phrase or an adverb is present. ( 6 3 ' ) *That NYC was on the verge of bankruptcy was complained yesterday by the mayor. ( 6 4 ' ) »That there are too many idiots in Washington this term is felt by a large majority of the voters. Thus, there is an unexplained difference between ( 6 2 ' ) on the one hand and ( 6 3 ' ) and ( 6 4 ' ) on the other. Our analysis emphasizes, however, that the reason for this distinction in grammaticality is not to be sought in the passive construction. The passive sentences ( 5 9 ) , ( 6 0 ) and ( 6 1 ) are all equally good; there is no discernible difference between ( 5 9 ) as opposed to (60) or ( 6 1 ) . It is the corresponding topicalized version of the derived subject clause that resists derivation. Sentences ( 6 3 ) and ( 6 4 ) constitute an absolute violation to the restrictions placed on topicalization - and do not fall under the perceptual strategy that we have just discussed - for reasons that presumably have to do with the semantic nature of the predicates involved. We can o f f e r no solution for this problem
92
at this time; however, it should be remarked that the ungrammaticality of ( 6 3 ) and ( 6 4 ) do not bear on the accuracy of our analysis of passivized subject clauses, but should be left to a more thorough study of topicalization phenomena. We have shown in this section how sentence ( 5 6 ) , repeated here for convenience, is to be derived under the hypothesis that Ss are not NPs.
(56) That John has quit smoking is assumed by most people. It is explained in terms of a base-generated topicalization configuration in which NP-movement applies to a wh-proform followed by wh-movement and wh-phrase deletion. In this section we have dealt with topicalized derived subject clauses. The nontopicalized version of ( 6 5 ) would be the following. (70) It is assumed by most people that John has quit smoking. The question of whether NP-movement is involved in the derivation of this sentence will be dealt with in the next section as an automatic consequence of attempting a theory of interpretation for all VP final sentential complements as subjects or objects. 2.2.4
Conclusion
With the goal in mind of attaining explanatory adequacy by adopting the most restrictive theory of grammar possible that follows from the syntactic facts of the language, we have been scrutinizing in this chapter the advantages of an intraposition analysis over an extraposition analysis of predicate sentential complementation. The intraposition hypothesis makes the more restrictive statement of the two analyses: predicate sentential complementation can only be introduced by the base component into phrase final 's position. This statement captures precisely the syntactic distribution of sentential complements within the English sentence and it is in accord with the lexicalist vision of a highly restrictive phrase structure schema, namely each syntactic phrasal category must contain a lexical head of the same category. Consequently, sentential complements under intraposition will never be able to occur in nominal positions in a sentence. All ^-complements that do not occur in VP final posi-
93
tion, are instances of topicalization. We exclude the possibility of a transformational rule placing an S-complement as sole dominee under the subject NP node of a root sentence. Since we reject the intraposition (or subject replacement) transformation in favor of the topicalization analysis discussed in section 2 . 2 . 3 . 1 , there is no sense in calling the revised hypothesis of the distribution of predicate sentential complementation under discussion 'intraposition'. Henceforth we will terra the hypothesis the 'nominal gap hypothesis 1 ( N G H ) , discarding the term intraposition, and turn now to the problem of interpreting S final clauses as subjects or objects. 2.3
The Nominal Gap Hypothesis
2.3.1
The Insufficiency of Chomsky and Lasnik's 1977 and Jackendoff's 1977 Analyses
We have shown the validity of the intraposition assumption underlying the initial phrase markers that Chomsky and Lasnik 1977 proposed for seem and likely structures as reviewed in section 2 . 1 . The complement clause will occur phrase finally in both cases. (71)
[
be likely [ S ] (V* = be, seem, . . . ) fa
UWH J
(Chomsky & Lasnik 1977, 4 4 9 ) Thus, ( 7 4 ) and ( 7 5 ) will be derived. (74)
it
[^ be likely [^ that Bill likes ice cream]]
(75)
it
[ vp seem [— that Bill likes ice cream]]
The question is, if all nontopicalized complement clauses occupy the VP final s" position, how does the grammar interpret the appropriate ones as subjects? Semantic interpretation within REST takes place at the level of surface structure. Jackendoff (1977, 100) suggests that a surface structure that contains the pronoun it
in subject position of the sentence and
94
a VP final clause, in his terminology the structure shown in ( 7 6 ) , would be subject to a rule of anaphora at the level of surface structure indicating that the S is understood as coreferential to the subject pronoun it and thus a subject clause. (76) V"1 N" '
it
V"
-
. . \ S
(Jackendoff 1 9 7 7 , 1OO)
The inadequacy of such an analysis in light of the facts that have come to light in Chapter 1 of this study is blatantly apparent. Chomsky and L a s n i k ' s it insertion rule accompanied by Jackend o f f ' s conjecture that a surface structure rule of construal will construct an anaphoric relationship between the it in subject position and the clause in VP final position is not a proper means of interpreting clauses as subjects for it cannot distinguish between the subject clause to likely in ( 7 4 ) and the object clause to seem in ( 7 5 ) , interpreting them both as subject clauses. The facts that have come to light in this study of the sentential complements to SEEM verbs strongly indicate that the pronoun it does not signal coreference of a VP final clause to subject position. The pronoun it signals nothing at all; it is merely a placeholder inserted into subject position - not under the conditions which Chomsky and Lasnik mention in their rule (73) but simply as a result of the fact that the syntactic structure of the English sentence does not allow a gap to begin a sentence. As the syntactic nature of SEEM constructions has become clear, it has also become clear that the interpretation of VP final sentential complements as subjects cannot be a superficial procedure as Jackendoff believes and has absolutely nothing to do with the pronoun it. In section 2 . 3 . 2 we propose a mechanism that will interpret S final clauses in terms of cooccurring nominal positions in the matrix sentence that are controlled by the VP final sentential complements. As a consequence of this proposal, Chomsky and Lasnik's rule for inserting the lexical item it into an empty subject position will need to be altered (= section 2.3.3) .
95
2.3.2
Control
Recall that wh-movement was a key feature in the analysis of topicalized clauses in section 2 . 2 . 3 . 1 providing the means for distinguishing between the interpretation of the clause that he will come as a subject or object in the following sentences. ( 7 7 ) That he will come is clear. ( 7 8 ) That he will come I regret. Looking at this derivation from the vantage point of surface structure, a gap in a nominal position of the matrix sentence (subject position in ( 7 7 ) and object position in ( 7 8 ) ) , cf. ( 7 9 ) and ( 8 O ) , yields an open sentence which is interpreted by the rule of predication as being satisfied by the presentential sentential complement. ( 7 9 ) that he will come + t is clear (80) that he will come + I regret t The gap occurs as the result of postulating an underlying element ( a wh-proform) that has the special property of being able to be shifted into COMP. All elements in COMP may be freely deleted - subject to the condition of recoverability. (In this case the wh-proform is recoverable since it is, in effect, synonomous with the topicalized sentential complement.) A nominal gap in the matrix sentence provides the grammar with the mechanism for interpreting the sentential complement. The element trace (t) is the remnant of the original wh-proform and the source of the nominal gap in the abstract phrase structure analysis of the sentence. Since the notion "nominal gap" will assume a crucial role in the following analysis of predicate sentential complementation, it will be appropriate to define it. A nominal gap is simply the appearance of the category NP mentioned in the subcategorization of a predicate with no lexical content at the level of surface structure. Wh-movement was also extended to the analysis of topicalized sentential complements in passivized sentences of section 2 . 2 . 3 . 3 . Recall that sentence ( 5 6 ) has the surface structure ( 5 8 ) . (56) That John has quit smoking is assumed by most people. (58) [ TOp that John has quit smoking][-[ C Q M p w h a t ] [ g t is assumed t by most people]]
96
The sentential complement in topicalized position will be interpreted by the rule of predication according to the nominal gaps present in the matrix as a derived subject. 2.3.2.1
Interpretation of S Final Sentential Subjects
The wh-movement analysis has provided a successful mechanism for interpreting topicalized sentential complements as subjects (or objects) in terms of nominal gaps in the matrix. Can the notion nominal gap be extended to accommodate sentential complements in S final position as well? We will turn our attention first to derived (i.e. passivized) sentential subjects in order to demonstrate the expediency of such an analysis and then show how it extendsto underlying (nonpassivized) sentential subjects in S final position yielding an adequate means of interpretation for all sentential subjects. 2.3.2.1.1
Derived Sentential Subjects
First, let us observe what happens if we assume nothing more than Chomsky and L a s n i k ' s and Jackendoff's suggestions in the derivation of sentence ( 8 1 ) . ( b ! ) It is assumed by most people that John has quit smoking. If the underlying structure is ( 8 2 ) , Chomsky and Lasnik's rule of it insertion will place it into the empty subject NP because of the presence of the complementizer that. ( 8 2 ) [ N p e ] [ v p [ y be-en assume][ p p by most people][^ that John has quit smoking]] This will allow J a c k e n d o f f ' s rule of construal to establish an anaphoric relationship between it and the S f i n a l S-complement indicating that it is a subject complement. But the derivation will block at this point. The verb assume is not subcategorized for a subject that clause, it must select a nominal subject. Therefore the rules of semantic interpretation applying to logical form will not be able to construct a meaningful sentence. The problem is that the grammar has missed a step. There has to be some device present in the logical form of ( 8 2 ) accessible to the rule of semantic interpretation indicating that they are not applying to the underlying version of the sentence, i.e. that version given in the subcategorization of assume in the lexicon, but to a derived version which has been
97
altered in a completely predictable and systematic way. Presumably the lexicon itself will carry a general redundancy rule instructing the rules of semantic interpretation how the underlying subcategorization of a lexical entry will be altered under passivization, but the rules of semantic interpretation will have to be cued from logical form that they are dealing with such a derived structure. As the analysis now stands, there is nothing present in the logical form of ( 8 2 ) to indicate to the rules of SI that they need to make this switch and, consequently, they will apply to the logical form of ( 8 2 ) with the same expectations as they do to the nonpassivized sentence ( 8 3 ) . (83) It is likely that John has quit smoking. Once agein, this analysis is obviously not sufficient. We suggest that the nominal object position to assume must be present in the phrase marker. We will assume for the moment that the sentential complement controls this NP position. (84) [Np
e][ v p [ v be-en assume][ N p
PRO][^ that J. has quit s . ] ]
NP-movement can apply to this structure as it applies in passive sentences with nominal complements and as it applied in the derivation of the topicalized counterpart of ( 8 4 ) , (that John has quit smoking is assumed by most people, discussed in greater detail in section 2 . 2 . 3 . 3 ) . An underlying phrase marker of the form ( 8 4 ) would capture the generalization that all passive sentences, whether they contain nominal complements or sentential complements, involve NP-movement in their derivation as well as provide the grammar with a mechanism for interpreting such S final sentential complements as derived subjects. The derivation proceeds as follows. NP-movement places the object NP into the empty subject NP node leaving a trace of the moved phrase behind in the object NP position. (85)
[Np
PRO][ V p [ v be-en assume][ N p
e ] [ ^ that J. has quit s . ] ]
The surface structure (85) under this analysis contains the device lacking under Chomsky and Lasnik's and Jackendoff's analysis to indicate that the initial relationship between the verb assume and the that clause complement was one of verbobject in accordance with the subcategorization in the lexicon
98
but that this relationship was altered by transformation to become one of subject-predicate. This information is supplied by the trace in the object position to assume affiliated with the sentential complement. Thus, the rules of SI will be able to construct the proper meaning for (81) interpreting the sentential complement as a derived subject on the basis of the empty nominal positions in the sentence and the notion 'control 1 Let us return to the object position of assume in ( 8 4 ) marked PRO. What the element PRO represents, in essence, is a position of control. Within REST, NP and s" represent different phrase structure positions linearly, but a class of sentential complements rely on an NP position in the matrix for their interpretation. We will term this class of NP affiliated complements 'NA complements'. NA complements control an NP argument belonging to their matrix verb. How is this accomplished? The subcategorization of assume, for example, must specify a subject and an object NP. (86)
[NP t v p [ v assume] NP . . . ] ]
(87) We assume responsibility. Assume may also be construed with a sentential complement in object function as we have been discussing, cf. ( 8 4 ) . (88)
[NP t v p t v assume] NP Q ...
(SQ)]]
The optional sentential complement in the subcategorization will be indexed with the same index as the object NP of assume. A sentential complement carrying the same index as an NP in the subcategorization of a predicate will be interpreted as a NA-complement that controls the relevant NP position. Therefore, if Ξ is chosen in ( 8 8 ) , the category NP with the same index must be generated concurrently in the phrase marker containing the nonphonetic feature [+PRO] indicating that this position is a position subject to an obligatory rule of control. ( 8 9 ) ( a ) [ N p w e ] f v p [ v assume][ N p P R O ] [ ^ that J. has quit s . ] ] The rule of obligatory control will apply at the level of surface structure coindexing the position PRO with the s final sentential complement on the basis of the information given in the subcategorization of assume. That is, the ability of a sentential complement to control an NP position in the matrix is not a property of the S~ complement itself, but a property of the matrix verb.
99
(89)(b)
[Np
w e ] [ y p t v assume][ N p
PRO][^
that J. has q u i t s . ] ]
This analysis is an extention of the notion control developed in Chomsky and Lasnik 1977 1 2 , but it lends itself well to the phenomenon under discussion. In particular, under this control analysis we need no further stipulation to prohibit such sentences as the following. ( 9 O ) * [ N p w e ] [ v p t v assume][ N p Bill][- that J. has quit s . ] ] According to Chomsky and Lasnik ( 1 9 7 7 , 4 4 1 ) it is a central feature of the analysis of control that lexical material cannot appear in a position susceptible to control. Control positions are anaphoric positions only. Thus, a structure like ( 9 0 ) is excluded on principled grounds. Thus, a nominal gap is accurately defined as an NP position specified in the subcategorization of a verb or predicate that is not occupied by lexical material in surface structure. The notion nominal gap has no meaning apart from the theory of predicate sentential complementation. If, for instance, the S in ( 8 8 ) is not selected and a nominal does not occur in object position, a violation of the subcategorization occurs, cf. (91 a ) . The NP may occur independently of the S in the form of a pronoun or it may contain lexical material, cf. (91d) respectively. ( 9 1 ) ( a ) *We assume. (b) We assume it/so. (c) What do we assume t? (d) We assume responsibility.
(91b & c) and
The notion nominal gap finds justification only within the theory of sentential complementation and it exists in such a theory to provide the means of interpreting sentential complements. The complements that are interpreted by this means are the NA complements (nominal affiliated complements). 12 Chomsky and Lasnik ( 1 9 7 7 , 4 3 9 - 4 4 4 ) developed the notion of control to handle the following positions: (i) it is unclear [·=· what[ c ~ " to d o ] ] O (ii) a. John persuaded Bill [^-[ 0 to leave]] O D —— b. John promised Bill [-[ to leave]] Ο Ο —' 13 " . . . P R O and lexical NPs (including trace) are in complementary distribution in surface structures. Where we find PRO, we can find neither a lexical NP nor a trace, and conversely. (Chomsky & Lasnik 1977, 4 4 1 ) .
100
A nominal gap can result in one of two ways according to discussion so far. 1) When the NA complement occurs in
the the
TOPIC node (i.e. outside the matrix), a wh-pronoun is generated simultaneously in the target position to which wh-movement then applies ( and subsequent deletion of the wh-proform). 2) When the NA complement originates in S final position (i.e. within the m a t r i x ) , the target NP position is present as a position of control containing a nonphonetic cataphoric feature matrix (PRO) which excludes the insertion of lexical material. In the first case the rule of predication interprets the open nominal position in the matrix as being satisfied by the topicalized S complement. In the second case, a rule of obligatory control assigns identical indices to the open nominal position and the affiliated S* complement. Thus, the presence of a position of control (PRO) in the underlying structure of (81) provides the structure upon which NPmovement can operate yielding a passive structure. Without the proper logical form which indicates that NP-movement has taken place, sentence ( 8 1 ) will not be derivable by the grammar. 2.3.2.1.2
Underlying Sentential Subjects
Sentential subjects are always NA complements; that is,
they
will always be affiliated with the subject NP. If the predicate likely has the following subcategorization in the relevant points, the initial phrase marker ( 9 3 ) will be derivable. (92) (93)
[NP g [^(^ be likely]] ... [g[Np
(S g ) ]
P R O ] [ A p [ A be likely]][- that John has quit smoking]]
This is a slightly different structure from the assume sentence ( 8 9 ) . The difference can be seen more clearly by comparing the corresponding tree diagrams.
(93')
that J.has quit smoking be likely
101
(891)
that J.has quit smoking T.Höhle has pointed out that it is necessary for a £3 comple14 — ment to c-command the NP position that it controls. The S of ( 8 9 ' ) c-commands the object NP since both are dominated by VP. A AP final £3 in structure ( 9 3 ' ) would not c-command the subject NP, however. The only plausible assumption, therefore, is that a NA complement in subject function occurs outside of the AP, as the final constituent of the matrix S. This solution has a further desirable consequence, namely it can accommodate the acceptable cases of bisentential complementation occurring on the right side of the verb. Jackendoff (1977, 96) mentions the following sentence. ( 9 4 ) It proves that Charley isn't guilty that he arrived so soon. If the subcategorization for prove reads as follows in the relevant points, the phrase marker underlying ( 9 4 ) will be ( 9 6 ) (95)
[NP s [ v p [ v prove] NPQ ...
er kauft sich regelmäßig Die Zeit. *Es ist wahrscheinlich,] 7. Parenthetical Phrase Simple parenthetical phrases are possible with verbs that take sentential subjects as shown in ( 3 1 ) . ( 3 1 ) Hans steht, fso erweist sich, \ auf intelligente so hat sich herausgestellt, | Frauen, so zeigt sich, so ergibt sich, so ist bekannt, so wird behauptet, ,so wurde festgestellt, The proform es may occur optionally as the subject to such verbs in the parenthetical phrase, although the constructions in ( 3 2 ) where the es occurs are not as well-formed stylistically as the parentheticals in ( 3 1 ) . (32) Hans steht, /'so erweist es sich, \ auf intelliso hat es sich herausgestellt,! gente Frauen, so zeigt es sich, so ergibt es sich, so ist es bekannt, so wird es behauptet, ,so wurde es festgestellt, In the case of the SCHEINEN verbs, on the other hand, the subject es must occur; it (33)
is obligatory.
Hans steht, /so scheint e s , l auf intelligente Frauen.
1 so heißt es, (34)
f
*Hans steht, r so scheint,) auf intelligente Frauen. \so heißt, J
Thus, the SCHEINEN verbs d i f f e r from verbs taking subject sentential complementation by requiring the subject pronoun es to appear. The syntactic difference between the SCHEINEN verbs and other verbs that take subject sentential complementation can be seen even more clearly if we construe the parenthetical phrase without the element so. In this case the proform e_£ is impossible with the STIMMEN verbs. (35) *Hans steht, (erweist es sich, auf intelligente Frauen. l zeigt es sich, l wird es behauptet, \ ist es bekannt,
121
auf intelligente Frauen. Hans steht, 'erweist sich, zeigt sich, wird behauptet, ist bekannt, Once again we see that the es subject to the SCHEINEN verbs (36)
must appear. (37) Hans steht, /scheint es,\ auf intelligente Frauen. \heißt es, j (38) »Hans steht, rscheint, 1 auf intelligente Frauen. (heißt, J 8. Wie Parentheticals In a wie parenthetical clause the pronoun es can optionally occur in construction with verbs that take subject sentential complementation, although the examples in question are stylistically better without it. (39) Geld ist, f a ) wie es sich gezeigt hat, \ die Wurzel wie sich gezeigt hat, [ allen Übels. b) wie es sich erwiesen hat, wie sich erwiesen hat, c) wie es sich herausgestellt hat,' wie sich herausgestellt hat, Once again the important difference between such verbs as those shown in ( 3 9 ) and the SCHEINEN verbs is that with SCHEINEN verbs the es is obligatorily present; this is the only option. (40) Ein Prophet gilt nichts, ( a ) wie es scheint, \ im eigenen *wie scheint, (_ Lande, b) wie es heißt, *wie heißt, The fact that the es cannot be missing with scheinen and heißen whereas it occurs optionally in the construction of ( 3 9 ) is significant. It points toward different syntactic behavior on the part of SCHEINEN verbs as opposed to STIMMEN verbs. 9. Comparatives In applying the comparative test we need to select verbs that will comply semantically with the construction - the examples of STIMMEN verbs which we have used so far are not very good in the als phrase, so we have chosen ist erlaubt and ist ver2 T.Höhle brought these facts to my attention. I am also indebted to him for appropriate examples. 3 The wie parenthetical phrase is not possible with all such verbs however. As H.Vater and T.Höhle have pointed out to me, stimmen itself cannot form a wie parenthetical and a verb like feststehen isn't usual in such a construction. These are most likely semantic restrictions that are dependent on the meaning of the predicate involved.
122
nünftig to represent this class. With these two verbs, the e_s subject is optional, although, once again, the variant with e_s seems stylistically not as well-formed as the variant without es. With the SCHEINEN verbs, on the other hand, es_ must occur. ( 4 1 ) Er verdient mehr Geld, f a l s von der Regierung erlaubt ist. | als es von d.Regierung erlaubt (42) (43)
3.1.2
Ich trinke mehr Bier, f a l s vernünftig ist. l als es vernünftig ist. Er trinkt mehr, f als es scheint. |*als scheint.
Summary
From this series of tests, it is apparent that scheinen and heißen display a different syntactic nature from stimmen or any other verb of the STIMMEN class. We assume that the daß clause of ( 2 ) , repeated here for convenience, (2) Daß er sich regelmäßig Die Zeit k a u f t , stimmt. is a subject clause on the grounds that it can be pronominalized by a sentential proform es or das as in ( 1 6 ) and ( 1 7 ) or by a w-proform in the direct question ( 2 4 ) and in the pseudo cleft construction ( 2 6 ) , in which case the respective pronoun occurs as the subject to stimmen. (16) (17)
Es stimmt. Das stimmt.
(24) Was stimmt? (26) Was stimmt, ist, daß er mit einer dicken Erbschaft rechnet Furthermore, the intransitive stimmen accepts a nominal subject. (8) Der Bericht stimmt. On the other hand, scheinen will not allow its daß clause to occur sentence initially and heißen requires the subject es to occur if the daß clause is preposed, cf. (7) and (5) of the previous section. Most importantly, the complement clause to scheinen and heißen cannot be pronominalized in a parallel fashion to the sentential complement of stimmen; direct questioning and pseudo clefting is impossible and when an anaphoric pronoun occurs instead of the full clause, it occurs as the object to scheinen and heißen. (25) *Was scheint? *Was heißt?
ist.
123
(27) «Was scheint, ist, daß er mit einer dicken Erbschaft rechnet. *Was heißt, ist, daß er mit einer dicken Erbschaft rechnet. (44)
Ist Hans ein Kleptomane?
(a) So scheint es. (b) So heißt es. In ( 4 4 ) the pronoun es is the subject to scheinen and heißen, but it is not anaphoric, the pronoun so is the anaphoric constituent. Thus, there is no evidence in the facts of the language that scheinen and heißen are intransitive verbs; neither do they occur intransitively with a nominal subject as stimmen does. Furthermore, they allow the complementizer daß of their complement to be deleted which is a characteristic of transitive verbs. On the basis of these facts centered around the pronominalization patterns of the language, we assume that the STIMMEN verbs take subject sentential complementation and that the SCHEINEN verbs do not - their sentential complement must be an object clause. The contrasting syntactic nature between scheinen and heißen on the one hand and the STIMMEN verbs on the other is further substantiated by the parenthetical and comparative phrase tests. Scheinen and heißen always require the occurrence of the subject es in these constructions. The STIMMEN verbs occur with or without the subject es_ in each case, although the absence of es is the stylistically preferred variant. The results of these three last tests indicate that the element es is a true subject to scheinen and heißen which cannot be missing and the object position to scheinen and heißen is open in the same way it is with other verbs that take sentential objects. ( 4 5 ) ( a ) so scheint/heißt es (b) so denkt/vermutet man ( 4 6 ) ( a ) *so scheint/heißt (b) »so denkt/vermutet In the case of verbs taking sentential subjects, however, it is the subject position that is open in these constructions. Even if the element e_s occurs, it doesn't seem to be anaphoric. ( 4 7 ) (a) so zeigt (es) sich (b)
so wird (es) vermutet
124
From the total picture projected from these 9 syntactic arguments, and especially from arguments 1. - 6 . , we conclude that scheinen and heißen do not belong to the STIMMEN 4 class of verbs : the daß clause complement to STIMMEN verbs is a subject clause and to SCHEINEN verbs an object clause. 4 The class of STIMMEN verbs is large; it consists of verbs and predicates like the following: stimmen, feststehen, auffallen, sich zeigen, sich herausstellen, sich erweisen, sich ergeben, sich gehören, ist klar, ist wahrscheinlich, (drohen) and so forth. Special mention should perhaps be given to the following verbs construed with the impersonal es subject and daß clauses in extraposed position, because the construction is somewhat more unusual than those we have been considering all along, ( i ) ( a ) Es geschah, daß ein Gebot von Kaiser Augustus ausging . (b) Es begab sich aber, ... (c) Es hat sich ereignet, ... (d) Es hat sich zugetragen, ... (H.Vater, pers.commun.) These examples also follow the pattern of STIMMEN verbs and consequently must be considered cases of sentential subject complementation, cf. (ii) . ( i i ) ( a ) Was geschah? (b) Was begab sich? (c) Was hat sich ereignet? (d) Was hat sich zugetragen? 5 At this point we would like to briefly mention the verb drohen. The grammatical status of drohen with a daß clause as shown in (i) is unclear: some speakers accept such a construction (this example was taken from Ohring ( 1 9 7 7 , 51) where it was judged grammatical) but a large number of speakers find (i) ungrammatical. (i) Es droht seit langem, daß die Lebensbedingungen in den Städten unerträglich werden. For those speakers who do accept ( i ) , drohen is a STIMMEN verb, since the same speakers find that the following tests applied to drohen yield results similar to stimmen constructions: the sentential complement to drohen pronominalizes in a parallel fashion to stimmen constructions and drohen can occur intransitively with a nominal subject. 1. Initial Positioning (ii) Daß die Lebensbedingungen ..., droht seit langem. 2. Nominal Subject (iii) Hungersnot droht. 3. Pronominalization (iv) Es droht seit langem/Das droht seit langem. 4. Direct Questioning (v) Was droht seit langem? 5. Pseudo Clefting (vi) Was seit langem droht, ist, daß die Lebensbedingungen ... 6. Deletion of daß (vii) *Es droht seit langem, die Lebensbedingungen ... If there were more general consensus on the grammatical status of drohen + daß, it would be an interesting sidelight for this study, since the German verbs drohen and scheinen would in this case enter into a similar opposition as manifested by likely
125
3.2
SCHEINEN + Infinitival
3.2.1
Definition of Class
3 . 2 . 1.1
Bech 1955
In order to discuss Gunnar Bech's 1955 treatment of scheinen in his study of the infinite verb in German, it
is necessary
to clarify three central notions Bech employs: status, orientation and coherent
field.
Status. For governed infinite forms
(i.e.
S u p i n a ) , Bech distinguishes
between three status: (48)
1. status 2. status 3. status
lieben zu lieben geliebt
( 1 9 5 5 , 12)
The status of an infinite verbal form is determined by its governing verb. The modal verb wollen governs the 1. status, versuchen the 2. status and haben (or sein)
the 3. status of
its verbal complements. (49)(a)
ich will sprechen
(b)
ich versuche zu sprechen
(c)
ich habe gesprochen (/ich bin gelaufen)
(1955,14)
Bech expresses the relationship between a ( f i n i t e or infinite) verbal form and the nonfinite V + V" (i.e.
V governs the status of V " ) . The status of V"
is placed behind it
in parentheses, cf.
which expresses the forms of (50)(a)
V
form it governs with the notation the notation in ( 5 O )
(49).
+ V"(1)
(b)
V + V"(2)
(c)
V
+ V"(3)
For the discussion to follow only the 1 and 2 status - the bare infinitive and the infinitive with zu - are of importance. Orientation. An infinitive form V" of a status relationship displays no subject on the surface although a logical subject is always under5 cont. and seem in English where the first representative of the pair (i.e. drohen/j.ikely) is intransitive and the second (scheinen/seem) transitive, but both drohen/likely and scheinen/seem have the subject raised infinitival complement in common.
126
stood. The orientation of V is the nominal of its field which is understood as the logical subject of V". The orientation of V is noted after it in parentheses. For the sentences given in ( 4 9 ) , V has, in each case, the orientation N ' : N " , i.e. ich ( N 1 ) is the subject of will and simultaneously the understood subject ( N " ) of sprechen. The same goes for versuche and habe/bin. Wollen (or versuchen, haben/sein as well) therefore has the orientation ( N ' : N " ) . An example of a V with the orientation ( A ' : N " ) is lassen where lassen governs the 1. status or drängen where drängen governs the 2. status. (51) er ließ den Diener abräumen V'(A':N") + V"(1) (52)
er drängte mich, an einer solchen Fahrt teilzunehmen
V'(A':N") + V"(2) (1955,
134 & 168)
In ( 5 1 ) and ( 5 2 ) the logical subjects of V" (abräumen and teilnehmen) are the accusative objects of lassen and drängen, namely den Diener and mich respectively. In (53) the logical subject of V" (beginnen) is the dative object of V (bedeuten), therefore V has the orientation ( D ' : N " ) . (53)
(er) bedeutete ihm, zu beginnen
V'(D':N") + V"(2) (1955, 182)
Coherent Field. Associated with each occurrence of V is a verbal field comprising all the items dependent on that V. Sentence ( 5 4 ) contains two such fields separated by the comma. ( 5 4 ) ich bitte ihn, morgen zu kommen (1955, 43) Ich bitte ihn is one field and morgen zu kommen the second. Two verbal fields held together by the hypotactic status relationship (V + V") can occur in one of two different forms when V" is in the 2. status: they can be coherent or incoherent; that is, they can amalgamate forming one topological entity or one can follow the other in a linear sequence. (55) daß er mir zu kommen versprach coherent field (56) daß er mir versprach, zu kommen incoherent field Sentence ( 5 5 ) is an example of two verbal fields conjoined into one coherent field. In sentence ( 5 6 ) the two verbal fields appear separately in a linear sequence; they are incoherent. Bech demonstrates the extent of the amalgamation of two verbal fields when they are conjoined into one coherent field. Senten-
127
ce ( 5 7 ) represents an incoherent conjunction of two verbal fields. ( 5 7 ) sie wagt nicht / ihn zu stören (1955, 72) 1 2 A union of these same two incoherent fields into one coherent field takes place so extensively that the elements nicht and ihn exchange places obeying the order of pronominal and negation elements in a single sentence. (58) sie wagt ihn nicht zu stören (1955, 71) 2 1 The negative adverb nicht and a positive pronoun etwas of two incoherent verbal fields in ( 5 9 ) (59) er vermag nicht, etwas Besseres zu tun (1955, 77) 1 2 will cohere into one negative pronoun when the two verbal fields in question are conjoined to form a coherent construction. (60) er vermag nichts Besseres zu tun 1+2 Bech classifies the German verbs that act as V ' (i.e. those that can take infinitive complementation) according to their orientation and the status of the V" which they govern. Bech's treatment of scheinen. Scheinen has the orientation ( N 1 : N " ) and governs the 2. status of V"; this is a V ' ( N ' : N " ) + V " ( 2 ) construction. ( 6 1 ) er schien seine Nägel zu betrachten (1955, 1O6) The subject of schien (= er) is the logical subject of betrachten and the zu must be present accompanying betrachten; the 1. status is not permitted when the governing verb is scheinen. ( 6 2 ) »er schien seine Nägel betrachten A further characteristic of scheinen is that V and V" must be conjoined in a coherent field ( c f . Bech 1955, 1 2 7 f . ) . ( 6 3 ) ( a ) daß er seine Nägel zu betrachten schien. (b) *daß er schien, seine Nägel zu betrachten. Scheinen with an infinitive complement has the following necessary characteristics, according to Bech, which will serve us as a temporary definition of the SCHEINEN class of verbs within the system of infinitival complementation in German.
128
(i)(a)
The subject of scheinen is the understood subject of the infinitive (in Bech's terras, scheinen has the orien-
(b) (c)
tation N 1 : N " ) . The infinitive governed by scheinen must be accompanied by zu (2. status). Scheinen together with its infinitive forms a coherent
verbal field. These three factors are instrumental in delineating the class of SCHEINEN + Inf verbs which will form the foundation of the discussion in this section. Besides scheinen; pflegen, drohen, versprechen and haben meet the defining criteria (ia-c). Pflegen of ( 6 4 ) has the orientation N ' : N " , the zu_ must accompany erweisen and the verbal fields of pflegen and erweisen must be coherent. ( 6 4 ) die Huldigungen, die man den ersten Lebensjahren zu erweisen pflegt. (1955, 1O5) ( 6 5 ) *die Huldigungen, die man pflegt, den ersten Lebensjahren zu erweisen. The same can be said for drohen and versprechen as exemplified by ( 6 6 ) and ( 6 7 ) . ( 6 6 ) die Spannungen, die seine Jugend zu zerspringen
drohten (1955, 103) ( 6 7 ) des Jahres, welches so unvergleichlich schön zu werden versprach (1955, 1O8) When drohen and versprechen mean literally 'to threaten 1 and 'to promise 1 and take a dative object, V + V" can be - in fact usually are - incoherent: (68) man hat mir gedroht, mich in ein Irrenhaus zu sperren (1955, 113) (69) selbstverständlich mußte ich Ihnen versprechen, bald zu kommen (1955, 108) Therefore drohen and versprechen in the sense shown in ( 6 8 ) and ( 6 9 ) (as opposed to ( 6 6 ) and ( 6 7 ) ) do not meet criterion (ic) above (i.e. V + V" can, in these cases, occur incoherently) and consequently do not belong to the SCHEINEN class. Drohen and versprechen in ( 6 6 ) and ( 6 7 ) do not literally mean 'to threaten 1 /'to promise 1 , but have a less tangible meaning closer to that of scheinen: they mean more 'to seem', 'to be about to' or 'to have the possibility of turning out to be 1 . Versprechen
129
can be neutral or the speaker can express with it positive overtones, whereas drohen carries with it slightly negative aspects. ( 7 0 ) ( a ) Das Spiel droht aus den Fugen zu geraten. (b) »weil das Spiel droht, aus den Fugen zu geraten. ( 7 1 ) ( a ) Das Spiel verspricht spannend zu werden. (b) *weil das Spiel verspricht, spannend zu werden. Although not discussed by Been, haben complies with the definition (ia-c) in the following sense. ( 7 2 ) Du hast zu arbeiten. ( 7 3 ) ( a ) Ich sage dir, daß du zu arbeiten hast. (b) »Ich sage dir, daß du hast, zu arbeiten. The class SCHEINEN + Inf is distinguished from all other verbs of German that require the 2. status of the infinitive they govern by the fact that they are not capable of occurring in incoherent constructions with this infinitive. Require has been stressed above because brauchen, when it occurs with the 2. status, must also form a coherent field with its infinitive ( c f . Bech 1955, 21O-212). (74) ich brauche es dir nicht zu sagen (75) »ich brauche nicht, es dir zu sagen However, brauchen may occur either with the 1. status or with the 2. status with no semantic difference and in this sense does not require the 2. status as SCHEINEN verbs do, cf. ( 7 4 ) and ( 7 6 ) . (76) ich brauche es dir nicht sagen Bech (1955, 212) stresses that brauchen is the equivalent of the modal verb müssen in negative or interrogative environments. Folsom (1966, 16 & 88) also classified brauchen in the relevant sense as a modal. Modal verbs have - in Bech's terminology - the orientation N ' : N " and require the coherency of the verbal fields belonging to V + V". Modals share, therefore, criteria (ia & c) above with the SCHEINEN verbs. 6 6 The verbs haben and sein, which together with the modal verbs make up the class of auxiliaries, have the orientation N 1 : N " and must occur in a coherent field with the 3. status they govern. They are not mentioned here since only the 1. and 2. status are relevant to this discussion.
130
Thus, the SCHEINEN verbs, scheinen, pflegen, drohen, versprechen and haben, have certain modal qualities. They d i f f e r from modals, however, in criterion ( i b ) : they govern the 2. status of their infinitive complement. The SCHEINEN verbs have criterion (ib) in common with nonmodal verbs (with the exception of the Acl verbs (lassen, machen, fühlen, hören, sehen, spüren . . . ) , henceforth LASSEN verbs) that can govern an infinitive. Namely, they require their infinitive complement to occur with zu. SCHEINEN verbs d i f f e r , however, from nonmodal, non-LASSEN verbs in criterion ( i c ) : they occur only in a coherent construction. Thus, scheinen, pflegen, drohen, versprechen and haben comprise an intermediate class between modals and pure verbs in German. They may only occur in a coherent construction but require zu to accompany the governed infinitive. The SCHEINEN verbs d i f f e r from the LASSEN verbs which have a similar intermediate position between modals and pure verbs in that the LASSEN class, besides having the orientation A ' : N " , govern the 1. status of their infinitival complement which accounts for their modal-like quality of coherency: the 1. status (V + V " ( 1 ) ) always results in a coherent construction. It is only when the 2. status occurs (V + V " ( 2 ) ) that there is a choice between a coherent or incoherent construction - if the V in question isn't a SCHEINEN verb. The SCHEINEN verbs are peculiar in that they have both the features (b) and (c) of definition ( i ) : they are the only German verbs which have feature (b) which also require feature (c) and, as a result, seem to be pure verbs with modal qualities. The following sketch represents - using Bech's terminology the extraordinary position of SCHEINEN verbs within the entire system of infinitive complementation in German.
7 There is a small class of exceptional nonmodal verbs that can govern either the 1. or 2. status with no semantic difference: heißen, helfen, lehren and lernen. Generally the construction is coherent when the 1. status is chosen and incoherent when the 2. status occurs; however, the 2. status can also engage in a coherent construction. ( C f . Bech 1955, 206)
131
(77)
Bech 1955
V(N':N") + V"(1)
coherent
Modals (including werden)
(tun, bleiben)
V(A': LASSEN
coherent
V ( N ' : N " ) + V"(2) SCHEINEN
coherent
ablehnen ...
coherency optional
V(A1:N") + V"(2) anklagen ...
coherency optional
V(D1:N") + V"(2) anbieten ...
coherency restricted 8
8 An incoherent field is the unmarked case for these constructions. Coherency may occur when D 1 is absent or when V has no objects, cf. Bech (1955, 1 8 7 f . ) .
132
Neither the expression es heißt nor es gilt plus an infinitive belong to the SCHEINEN class of verbs with infinitival complementation inspite of the fact that heißen groups with scheinen with respect to finite clause complementation, because being impersonal expressions they violate criterion ( i a ) . (78) es heißt, viel Geld zu verdienen (79) es gilt, viel Geld zu verdienen The subject of heißt and gilt in ( 7 8 ) and ( 7 9 ) is an impersonal, nonreferring es whereas the understood subject of verdienen is an arbitrary being (or beings) capable of earning money. Moreover, these constructions can be incoherent violating further criterion ( i c ) . ( 7 8 ' ) weil es heißt, viel Geld zu verdienen ( 7 9 ' ) weil es gilt, viel Geld zu verdienen We have shown that scheinen together with pflegen, drohen, versprechen and haben form a class of their own within the infinitive complement system of German differentiated from all other German verbs on the basis of criteria (ia-c) above. Within the framework of the standard theory of generative grammar, the SCHEINEN verbs are termed subject raising verbs, a classification which alone can't serve to differentiate them from modal verbs or such pure verbs as anfangen, beginnen or aufhören. It is for this reason - the purpose of a clear-cut definition of the class of SCHEINEN verbs with respect to infinitive complementation - that we began this discussion with Gunnar Bech's study of the infinitive in German. 3.2.1.2
Transformational Grammar
Returning to our temporary definition (ia-c) we see that within the greater class of subject raising verbs Bech's criterion (ib) serves to differentiate SCHEINEN from modals and criterion (ic) from verbs like anfangen, beginnen and aufhören. With this delineation we have achieved the desired result of excluding the modals and verbs like anfangen etc. from the discussion of SCHEINEN constructions. Criterion ( i a ) is, however, incorrect and needs to be altered. The theory of transformational grammar offers a more explicit description of this property. Bech's pre-transformational work groups scheinen, pflegen, drohen and versprechen together with verbs like ablehnen, be-
133
dauern, erwarten, wünschen and so forth under the heading ( N 1 : N " ) . In other words, there is no distinction in Bech's work between subject raising and equi constructions. Bech saw that the subject of the infinitive helfen in constructions like (80a) and (81 a) was understood as identical to the overt subject of scheinen or wünschen and expressed this by the orientation ( N ' : N " ) , but, having no access to the notion of deep structure, differentiated no further between the two constructions. ( 8 O ) ( a ) Hans scheint ihr zu helfen. ( 8 1 ) ( a ) Hans wünscht, ihr zu helfen. In terms of transformational grammar the constructions (80a) and (81a) have fundamentally different initial structures. In (8Oa) the term Hans occurs only once in the initial phrase marker, namely as the subject of the complement. (80)(b) [ 0 Hans ihr zu helfen] scheinIn the structure underlying (81 a ) , Hans occurs initially as subject to wünschen and also as the subject of the infinitival complement. (81) (b) Hans [·= 0 Hans ihr zu helfen] wünschHans is not the subject to scheinen in initial structure for 9 well-known reasons. Thus, we will restate (ia) of the definition of the class of SCHEINEN + Inf in the terminology of the extended standard theory of transformational grammar, since Bech's terminology is misleading. At the same time (ib & c) can be translated into transformational terminology; thus, we arrive at the following still tentative - definition of the class of SCHEINEN + I n f . ( i i ) ( a ) SCHEINEN verbs are the subclass of subject raising verbs which (b) require the zu complementizer to accompany their infinitive complement and (c) do not permit the extraposition of this infinitive complement.
9 For these well-known arguments cf. among others Postal ( 1 9 7 4 , 33-39 and 369-374) for English and Ebert (1975, 1 7 7 f . ) for German.
134 3.2.2
Derivation of SCHEINEN + Infinitival
3.2.2.1
NP-Moveraent
We reject from the start the subject raising analysis where scheinen is an intransitive verb with a subject clause to which subject raising should apply, inspite of the fact that this seems to be the prevalent conception in the literature, as was the case for English, cf.
Huber/Kummer ( 1 9 7 4 , 261 & 2 6 5 ) , Ebert
( 1 9 7 5 , 1 8 4 ) , Pütz ( 1 9 7 5 , 9 O f ) , Evers ( 1 9 7 5 , 7 ) , öhring ( 1 9 7 7 , 48-54) among others.
(82) *
S,
SCHEINThe motivation for such a conception in the earlier literature was the assumption that the finite (daß) clause complement to scheinen was a subject clause and that infinitive complements and finite clause complements were to be derived from the same underlying verb complement configuration by means of a series of different transformations, one of which would place the appropriate complementizer into the complementizerless sentential complement. We have offered enough proof in the previous section to disclaim the traditional view that the finite complement to scheinen is a subject clause. If there is no proof that the daß complement is a subject clause, there is even less evidence that this type of infinitival complement should be derived from ( 8 2 ) . That if
is,
( 8 3 ) is not an instance of sentential subject complementa-
tion, there is intuitively even less reason to assume the complement in ( 8 4 ) to originate in subject position. (83)
Es scheint, daß Geld die Wurzel allen Übels
(84)
Geld scheint die Wurzel allen Übels zu sein.
ist.
Because of these and additional arguments ( c f . sections 1.3 and 2 . 1 ) , we accepted the following NP-movement analysis of SEEM + Inf constructions for English: seem is generated with an empty NP subject and a sentential complement containing the null complementizer (0) in object position. The subject NP of the complement clause must move up to fill the [„p e]. The specified
135
subject constraint dictates that only the subject of the complement may move. ( 8 5 ) ( a ) e seems [·=· 0 Bill to like ice cream] (b)
Bill seems [- 0 t to like ice cream]
(85')
seem
0 Bill to like ice cream
The question is whether this conception of NP-movement can be applied to SCHEINEN constructions in German; that is, whether (87) is a plausible derivation of ( 8 6 ) . (86) Die Sache scheint ihm über den Kopf zu wachsen. ( 8 7 ) (a) e [·=· 0 die Sache ihm über den Kopf zu wachsen] scheint (b) die Sache [ — 0 t ihm über den Kopf zu wachsen] scheint O
(87')
t
0 die Sache ihm ü.d.Kopf l zu wachsen
scheint
The verb second rule will correctly place scheint after the first constituent in the sentence on the final cycle, in this case after the moved die Sache. The derivation shown in ( 8 7 ) yields the desired sentence ( 8 6 ) , but (88a & b) are also perfectly acceptable structural variants of ( 8 6 ) which the analysis, if correct, should account for. ( 8 8 ) ( a ) Ihm scheint die Sache über den Kopf zu wachsen, (b) Über den Kopf scheint ihm die Sache zu wachsen. In English an NP-movement rule that moves the complement subject up to matrix subject position accounts for the data in an intuitively correct manner. No other NP besides the subject can move and the rest of the S-complement remains intact after the movement of the subject. ( 8 9 ) ( a ) Bill seems to like ice cream. (b) »Ice cream seems Bill to like. (c) *Ice cream Bill seems to like.
136
The NP-movement analysis does not account for the German sentences in ( 8 8 ) , however. After it applies to ( 8 7 a ) to derive ( 8 7 b ) , the grammar would require another movement rule to move ihm or über den Kopf out of the complement as well. Note that a second movement rule is necessary; die Sache has to move first because it is a specified subject and as long as it is present within the complement clause, the constituents ihm and über den Kopf are not free to move by virtue of the specified subject constraint. If we want to do away with the SSC, saying that it doesn't apply to German,
(88a) could be derived since
NP-movement can now choose ihm instead of die Sache to be moved. We still cannot derive ( 8 8 b ) , though, where two constituents, über den Kopf and ihm, must move. The possibility of a double application of NP-movement is ruled out by strict cyclicity. Further complication arises when we add a dative experiencer to scheinen. In (90) it looks as though ihm is the complement constituent that has been moved. (90)
Mir scheint ihm die Sache über den Kopf zu wachsen.
The NP-movement analysis which entails the SSC will explicitly exclude this sentence. Ihm is not accessible to a movement rule because of the presence of the specified subject die Sache; in the complement. Sentence ( 9 1 ) will be ruled out on the same grounds. (91)
Es scheint ihm die Sache über den Kopf zu wachsen.
Another problem arises if NP-movement is to account for German scheinen constructions. We have said that NP-movement applies, in e f f e c t , obligatorily to an English structure like ( 8 5 a ) , because, if Bill is not moved to fill the empty subject NP to seems, the sentence will contain an open variable in subject position which cannot be interpreted; consequently, the sentence will be rejected. For the German structure ( 9 2 ) , however, there is no reason to believe that a movement rule has applied. (92)
Mir scheint die Sache zu scheitern.
The NP-movement analysis predicts ( 9 3 ) to be the initial structure of ( 9 2 ) . (93)
e
mir [·=· 0 die Sache zu scheitern] scheint
If no NP-movement applies to ( 9 3 ) and verb second places scheint after the first constituent e, we end up with a hanging open subject position.
137
(94)
e scheint mir [·=· 0 die Sache zu scheitern] O
If NP-movement is not applicable to ( 9 2 ) and the S constituent is retained, extraposition should be able to apply - but this is impossible. (95) *weil mir scheint, die Sache zu scheitern The inability of the complement to scheinen to undergo extraposition suggests that the original S constituent is no longer intact. The NP-movement analysis applied to SCHEINEN + Inf constructions is surrounded with complications. We are forcing a description of the English phenomenon onto German structures and the result is an extremely unenlightening analysis. For German, we need a description of the relationship between sentences ( 8 6 ) and ( 8 8 ) (a & b) which NP-movement cannot provide. SCHEINEN + Inf structures must have a dual clause source. (This can be seen from a sentence like (9O) where two datives are present, one belonging to scheinen and the other being a constituent of the complement clause.) The surface structures of ( 8 6 ) and (88a & b ) , however, have the reordering possibilities of a single clause, cf. these sentences to ( 9 6 ) where no scheinen occurs. ( 9 6 ) ( a ) Die Sache wächst ihm über den Kopf. (b) Ihm wächst die Sache über den Kopf. (c) Über den Kopf wächst ihm die Sache. The NP-movement approach is clearly counter-intuitive. Clause integration more accurately explains the relationship of ( 8 6 ) to ( 8 8 ) , namely as a parallel phenomenon to the data in ( 9 6 ) . These sentences simply display the variety of single clause word ordering possibilities in German. Ebert 1975 suggests such an analysis for scheinen as opposed to a subject raising analysis. His argument is that there are cases with all the features of raising constructions where no subject NP (i.e. a nominative element) is present in the complement to be raised. When passive applies on the embedded sentential complement in the following scheinen sentences, for instance, the following constructions arise. 1O Ebert 1975 is based on Reis 1 1973 observations as to the relevancy of a raising analysis for LASSEN constructions.
138
(97)(a) (b) (c)
Ihm scheint geholfen zu werden. Seiner scheint nicht mehr gedacht zu werden. An dem Wagen scheint noch gearbeitet zu werden. (Ebert 1975, 178) In these cases no subject NP has been moved, but rather a dative, genitive and prepositional object. But raising was defined only on subject NPs. Because of such data as ( 9 7 ) , Ebert (1975, 183) concludes: "...if raising is involved at all, then not just subjects but all initial constituents must be raised, ..." Our sentences (88a & b) show, however, that the raising of one constituent, the initial constituent, won't suffice, since in these sentences more than one constituent will have to be raised, die Sache and ihm in ( 8 8 a ) and die Sache, ihm and über den Kopf in (88b). Ebert continues in the same quote: "...but...topological facts indicate that it is not raising of a single constituent of the lower clause into the higher clause which is involved, but rather integration of both clauses." Because of sentences like ( 8 6 ) and (88a & b ) , this suggestion of achieving a surface integration of two underlying clauses seems to us to be the most viable solution. How would this process work formally? 3.2.2.2
V-Raising
Evers 1975 discusses evidence for the existence of a V-raising rule in German. V-raising takes a structure like ( 9 8 ) and transforms it into ( 9 9 ) . (98) S (99)
(Evers 1975, 3) A matrix sentence with an embedded clause is reorganized in such a manner that the verb of the original embedded clause regroups with the verb of the matrix yielding a verb cluster. As a result, the embedded S boundary is broken down (or pruned) and the constituents of the previous embedded S integrate into the matrix. This process explains our intuitions about sentences
139
(86) and (88a & b ) . If (1OO) is the phrase marker underlying ( 8 6 ) , then V-raising would yield (101).
(100)
S.
scheint ihm ü.d.Kopf (1O1)
zu wachsen
V-RAISING
scheint v/ach sen Verb second would place scheint after the first NP, die Sache, if no other movement is applied to the intermediate structure ( 1 O 1 ) , thus deriving ( 8 6 ) . Other movement is possible, though. The German single sentence allows almost any constituent to be fronted. If,
after V-raising
and verb second apply to ( 1 O O ) , the NP ihm is fronted by topicalization, we derive the following.
(102) NP ihm
V scheint
NP
I d.Sache
PP
I über
V
I
d.Kopf zu wachsen
This is sentence (88a). Sentence (88b) can be derived by topicalizing the PP (über den Kopf) instead of the second NP ( i h m ) . (103) PP
V
I
NP
NP
scheint d. Sache ihm über den zu wachsen Kopf The pronoun ihm will then be cliticized to the matrix verb. Thus, there is no need for an NP-movement rule in German to derive the SCHEINEN + Inf constructions. Evers derives his evidence for V-raising independent of subject raising data, which is no issue for him. His arguments include the following syntactic processes in German.
140
a. Gapping b. Nominalization c. Extraposition d. Clitic Placement e. Quantifier Hopping and f. Clause Negation ( c f . Evers 1975, 3 - 2 9 ) . We feel that his evidence is conclusive. The extraposition argument alone is enough to verify the expediency of V-raising in the grammar of German. The existence of a verb cluster derived by V-raising explains why (104) i s n ' t possible, inspite of the known cyclic nature of extraposition. (104) *weil wir Peter erzählen, daß der Bürgermeister sich weigerte zu bezahlen, hörten. (1975, 22) The correct form of ( 1 0 4 ) is ( 1 0 5 ) . (105) weil wir Peter erzählen ] hörten ] daß der Bürger^^~ O O meister sich weigerte zu bezahlen. ( 1 9 7 5 , 21) In (1O5) it looks as though the daß clause has skipped over two cyclic domains. Ordinarily this yields ungrammatical results: (106) »weil [daß [wir gesagt haben 0 ] nicht schlimm war c ] —"™"
o
ö
daß sie nicht kommen will In (106) daß sie nicht kommen will has been transported over two S boundaries which explains why the sentence is bad. In (105) - according to the V-raising analysis - two S boundaries are no longer present; the relevant structure after V-raising would be the following. ( 1 0 5 ' ) weil wir Peter __ [ erzählen hörten] g ] daß ... The verbs erzählen and hören cluster whereby the original S boundary after erzählen is pruned. Therefore extraposition, applying after V-raising on the cycle, can only place the clause after hörten and not after erzählen. V-raising does not apply to ( 1 0 6 ) ; it makes no sense to claim that gesagt haben and war cluster because of the semantic material between them. An extraposition analysis not supported by a V-raising analysis cannot account for such facts (cf. Lenerz 1974 for a detailed discussion of this problem without a V-raising solution). We have shown that an NP-movement analysis cannot account for all the relevant aspects of the SCHEINEN + Inf constructions. If a rule of V-raising exists in German, as Evers has argued that it must, an NP-movement analysis of SCHEINEN + Inf con-
141
structions is not only inaccurate but totally superfluous. The V-raising solution o f f e r s a formal mechanism for achieving the desired integration of the originally embedded constituents into the matrix. We will adopt, therefore, Evers' V-raising rule, which simultaneously accounts for many other features of the German sentence. At the same time, however, we do not adopt Evers 1 specific treatment of scheinen. First of all, Evers is mislead by the traditional view that scheinen is an intransitive verb with a subject clause and fails to see that scheinen together with pflegen, drohen, versprechen and haben builds a special class of verbs defined by criteria (iia-c) above. He classifies drohen with scheinen but treats pflegen as an equi verb and ignores altogether the special sense of versprechen and haben that makes them SCHEINEN verbs. Secondly, Evers (1975, 7 f . ) gives the following V-raising derivation of sentence ( 1 0 7 ) . (107) weil das Wasser die Kraniche zu vergiften scheint (108) (a)
I
NP
I
scheint
V
I
das Wasser die Kraniche zu vergiften (b) V-RAISING
NP
das Wasser
NP
die Kraniche
V
V
...
l
zu vergiften scheint We accept these two phrase markers, but reject Evers explicit claim that S. is a sentential subject. It is not clear, if Evers considers S1 a subject, why he doesn't place it under the subject NP as in ( 1 0 9 ) . (109)
scheint
142
Because he doesn't illustrate (1O7) with (1O9) but with ( 1 O 8 ) , our phrase markers coincidently coincide, although our analyses of scheinen don't. We maintain that S1 with scheinen is not a subject. If we add the dative experiencer to (107) the result is (110). (11O) weil mir das Wasser die Kraniche zu vergiften scheint If S1 were a subject clause, the dative experiencer mir would have to occur between S.. and V in (108a) assuming SOV order for German as Evers does. (108)(a1)
scheint d.Wasser d.Kraniche zu vergiften The presence of mir would block V-raising as Evers defines the rule (cf. 1975, 38) because an NP occurs between the V to be raised and the governing V. Consequently, the following sentence would arise. ( 1 1 1 ) *weil das Wasser die Kraniche zu vergiften mir scheint Sentence (11O) can only derive from ( 1 1 2 ) where S.. is an object 1 clause to scheinen. (112)
das Wasser die Kraniche zu vergiften After V-raising we derive ( 1 1 3 ) . (113)
zu vergiften scheint (11O) is a good sentence which is not derivable from Evers 1 assumptions about scheinen without further ad hoc rules correcting the placement of mir. (110) is derived without any such complication under the assumption that S.. is not a subject clause.
143
When the experiencer NP accompanying scheinen is not present, as in ( 1 0 8 ) , the embedded S is the initial constituent in the phrase marker due to the underlying verb final (SOV) topology of German. It is not an embedded subject clause, however, since it doesn't occur under the subject NP. We have the same situation with versprechen and possibly drohen when they take an experiencer phrase. ( 1 1 4 ) Das Spiel verspricht mir spannend zu werden. ( 1 1 5 ) Das Spiel droht mir aus den Fugen zu geraten. Such structures must go back to an initial phrase marker with the configuration shown in ( 1 1 2 ) where the embedded S is an object clause following the experiencer phrase and not to ( 1 0 8 ' ) where it is a subject clause. ( 1 1 6 ) ( a ) weil mir das Spiel spannend zu werden verspricht (b) »weil das Spiel spannend zu werden mir verspricht ( 1 1 7 ) ( a ) weil mir das Spiel aus den Fugen zu geraten droht (b) *weil das Spiel aus den Fugen zu geraten mir droht The following is the correct underlying structure of the SCHEINEN + Inf constructions given Evers'framework. (118)
(NP)
If the embedded S were a subject clause, the experiencer NP would occur between S and V in the matrix, thus blocking Vraising. V-raising is obligatory in the derivation of the SCHEINEN + Inf constructions. They derive from a dual clause source but have all the characteristics of a single clause on the surface - scrambling of NP and PP constituents, single sentence negation, etc. The fact that V-raising must apply to these verb-complement constructions explains why extraposition cannot: V-raising destroys the S node upon which extraposition is defined. Thus, we consider the SCHEINEN + Inf verbs subjectless transitive verbs which require the complementizer 11 The experiencer phrase in construction with versprechen and drohen is for a large number of speakers fairly restricted. Sentences ( 1 1 4 ) and ( 1 1 5 ) , nevertheless, demonstrate that such a construction cannot be ruled out entirely.
144
zu on their complement verb and obligatorily undergo Vraising, thus excluding extraposition. We can now revise our definition ( i i ) of SCHEINEN + Inf verbs to the more precise ( i i i ) . The SCHEINEN + Inf verbs ( i i i ) ( a ) obligatorily undergo V-raising and (b) require the complementizer zu on their complement verb. Statement (iiia) incorporates (iia & c) of the previous definition. v;e have shown that the SCHEINEN verbs in construction with an infinitival must derive from a sentential complement structure by means of V-raising - not by means of NP-movement. £.n analysis that predicts that the subject NP of the complement moves up to become the subject of the verb scheinen cannot account for the following paradigm. ( 1 1 9 ) ( a ) Die Sache scheint ihm über den Kopf zu wachsen. (b) Ihm scheint die Sache über den Kopf zu wachsen. (c) Über den Kopf scheint ihm die Sache zu wachsen. A V-raising analysis on the other hand explains this phenomenon in an intuitively satisfying way in terms of clause integration. The sentences of ( 1 1 9 ) begin as sentential complement structures to the verb scheinen. V-raising predicts that the complement verb wachsen and the matrix verb scheinen cluster, whereby the original clause boundary breaks down. We derive a superficial single clause structure in which the word ordering possibilities of a simple root sentence in German can be realized. The V-raising analysis therefore explains the modal-like property of the SCHEINEN + Inf verbs which we discussed at the outset of this section, namely that - inspite of a dual clause source - these constructions are always one topological (coherent) unit on the surface. Consequently, the infinitive structure in ( 1 1 9 ) will never need to be interpreted as a subject or object complement to the matrix verb (SCHEINEN) on the surface as other infinitives with zu will be: (120) Es ist anstrengend, in Köln Karneval zu feiern. ( 1 2 1 ) Oma beabsichtigt, Krimis zu schreiben.
145
The verbal constituents of ( 1 1 9 ) (scheint zu wachsen), rather, will be interpreted as one complex verbal constituent of a (derived) single sentence.
3.3
Summary
It is an important result of this study that the syntactic nature of seem and scheinen constructions has been clarified. Constructions like ( 1 2 2 ) vs. (123) and ( 1 2 4 ) vs. (125) have not been differentiated as syntactically distinct constructions in almost all of the linguistic literature dealing with the complement systems of English and German. (122) It is likely that money is the root of all evil. (123) (124)
It seems that money is the root of all evil. Es stimmt, daß Geld die Wurzel allen Übels ist.
( 1 2 5 ) Es scheint, daß Geld die Wurzel allen Übels ist. We have examined the SEEM/SCHEINEN constructions carefully and offered convincing evidence as to their transitive nature. When sentence pairs like ( 1 2 2 ) / ( 1 2 3 ) and ( 1 2 4 ) / ( 1 2 5 ) were put through the syntactic tests 1. - 7. of section 1.3.1.1 and 1. - 9. of section 3.1.1 respectively, clear-cut differences between ( 1 2 2 ) and ( 1 2 3 ) and between ( 1 2 4 ) and (125) emerged. The result of our comparison of SEEM to LIKELY in Chapter 1 and SCHEINEN to STIMMEN in this chapter is the realization that the superficial construction ( 1 2 6 ) ( a ) it V that ... (b) es V, daß ... is not the defining mark of sentential subject complementation. It is precisely this false conclusion that has contributed to the previous misunderstanding of SEEM and SCHEINEN constructions. A solution to the problem of when a sentence final complement is a subject complement must be sought on a deeper level in the grammar. A solution like Jackendoff's (1977,100) which interprets a sentence final S complement as coreferential to the subject pronoun _it/e_s in the superficial constructions of ( 1 2 6 ) makes inadequate predications for English complement constructions with the verbs seem, appear, happen and turn out as we discussed in section 2 . 3 . 1 . W e are now in a position to show that this situation is not simply a coincidental peculiarity of the English language. German complement structures
146
with the verbs scheinen and heißen support our conclusion as well that the interpretation of sentential complements as subjects or objects to their matrix verb is not simply a superficial matter that can be handled on the basis of the configuration shown in ( 1 2 6 ) . We attempted a theory in Chapter 2 (the nominal gap hypothesis) intended to describe the mechanism employed by the speaker of English in interpreting complement structures by means of which he possesses such knowledge. On the basis of the evidence brought to light in this chapter, we will attempt in the subsequent chapter to describe the necessary mechanism put to work by the German speaker in interpreting the complement structures of German. As a necessary by-product of the contrastive nature of this study, relevant differences in the English and German complement system will come to light along the way.
4.
GERMAN SYSTEM OF PREDICATE SENTENTIAL COMPLEMENTATION
4.1
Nominal Gap Hypothesis
4.1.1
Distribution of Sentential Subjects
Let us examine the distributional facts concerning sentential subjects in normal German complement sentences (i.e. nontopicalized, nondislocated and nonpseudo clefted sentences). 1. An embedded sentence with subject function can occur in sentence initial position, a position characteristic of NP subjects. (1) Daß Köln am Rhein liegt, ist bekannt. Such a sentential subject may also occupy sentence internal positions characteristic of NP subjects although such constructions are actually quite seldom. (2) Ist, daß Köln am Rhein liegt, auch in Amerika bekannt? (3) Mir ist, daß Köln am Rhein liegt, schon bekannt. That is, the grammar must allow for such sentences as (2) and (3) even though there is an alternative variant of these sentences which is preferred by speakers of German (which we will come to shortly). Thus, it seems that in German - as opposed to English - sentential subjects can occur in NP positions characteristic of subjects and are not limited to S~ positions which are distinct from NP positions. In Chapter 2 sentential subjects in English were shown to be possible in sentence initial positions like (1) but not in internal subject NP positions like (2) or ( 3 ) . The sentence initial placement of sentential subjects in English was then shown to be a topicalized position and not a true NP position. The different distribution of sentential subjects in German from that of their English counterparts can be further substantiated by considering the embedded subject position. Sen1 Cf.
(i) That Cologne is situated on the Rhine is well-known, (ii) »Is that Cologne is situated on the Rhine well-known?
148
tences like the following are also possible in German whereas such constructions are totally out in English. (4) Obwohl, daß Köln am Rhein liegt, in Deutschland allgemein bekannt ist,
wissen viele Amerikaner nichts davon.
2. Sentential subjects are found most frequently in extraposed position; that is, as the final constituent of the matrix sentence in which they are embedded. In such cases, although the constituent S occurs after the constituent VP in a position characteristic of Ss only, it is always associated with the subject NP position of the matrix. This occurs in one of two ways: either explicitly where the subject NP position is occupied by a specific nonanaphoric proform (i.e. es) or implicitly where the subject NP is marked by 0 (i.e. an NP position present in the underlying phrase structure representation of the sentence contains no phonetic material). Explicit association must occur where main sentence word order is involved and the subject NP is the constituent before the finite verb. (5) Es ist bekannt, daß Köln am Rhein liegt. When the word order is such that the subject NP is not the first constituent of the sentence, explicit association is always possible but not always obligatory (= (6) - ( 8 ) ) , although it seems to be obligatory in some cases (= ( 9 ) ) . (6) (7)
Mir ist (es) klar, daß wir ihm helfen müssen. Wichtig ist ( e s ) , daß du mit deinem neuen Mann auskommst.
(8) Sicher ist nur, daß Oma nicht so häufig tanzen soll. ( 9 ) ( a ) »Manchmal ist angenehm, daß du kommst. (b) Manchmal ist es angenehm, daß du kommst. (Pütz 1975, 56) When the sentential subject is not found in extraposition but actually holds an NP position of the sentence, it stands to reason that the element ejs cannot cooccur with it since the NP node would be doubly occupied in this case. (10) *Es, daß Köln am Rhein liegt, ist bekannt. ( 1 1 ) *Ist es, daß Köln am Rhein liegt, auch in Amerika bekannt? Therefore, because of the presence of es in ( 1 2 ) , the sentential subject must be considered extraposed. 2 Cf.
(iii)
»Although that Cologne is situated on the Rhine is well-known in Germany, many Americans are not aware of it.
149
(12)
Bekannt ist
es, daß die Römer Wasserleitungen bauten.
Sentence ( 1 3 ) verifies this. ( 1 3 ) »Bekannt ist es, daß die Römer Wasserleitungen bauten, gewesen. In the variant of ( 1 2 ) in which the optional es does not occur (= ( 1 4 ) ) the situation is not quite so clear, although the sentential subject most likely will have to be analyzed as occupying extraposed position since only (15) and not ( 1 6 ) is a possible construction. ( 1 4 ) Bekannt ist, daß die Römer Wasserleitungen bauten. ( 1 5 ) Bekannt ist gewesen, daß die Römer Wasserleitungen bauten. ( 1 6 ) »Bekannt ist, daß die Römer Wasserleitungen bauten, gewesen. In sum, we have the following situation in German. Either the sentential subject actually holds the subject NP position of the matrix itself or it is found in sentence final position and is affiliated with the subject NP position of the matrix in such a manner that prohibits the relevant subject NP position from being filled with other lexical material. In the latter case, the subject NP position will be either represented by the nonanaphoric proform es or it will be empty. 4.1.2
Distribution of Sentential Objects
An embedded sentence in object function can occur in one of two basic positions in the sentence: within the VP in the nominal object position or sentence finally in the so-called extraposed position. (17) Hans hat, daß er dabei sich selbst verrät, (18) (19)
(aus lauter
Dummheit) nicht gemerkt. Hans hat nicht gemerkt, daß er dabei sich selbst verrät. Hans hat es nicht gemerkt, daß er dabei sich selbst verrät.
When the sentential object is found in extraposed position as in ( 1 8 ) and ( 1 9 ) , it nevertheless is associated with the nominal object position of the matrix - that is, the position which the object clause has assumed in ( 1 7 ) - either implicitly as in ( 1 8 ) , where no lexical material occurs in this position, but an empty category NP will be present in the underlying phrase structure representation of the sentence, or explicitly
150 as in ( 1 9 ) . 3 Sentences like ( 1 7 ) where the sentential object holds the object NP position are not stylistically as good as those of the type found in (18) or ( 1 9 ) where the embedded senten-
ce is found in a strictly £J position of the sentence. Nevertheless, the grammar of German cannot rule out such sentences because they are basically grammatical and it can be easily left to a theory of performance to rule such constructions unacceptable when certain pragmatic conditions are not met. The following sentences, for instance, are not nearly as acceptable as sentence ( 1 7 ) . (20) »Wir nehmen, daß das Geiseldrama im Iran nicht unbedingt harmlos ausgehen wird, an. ( 2 1 ) »Paul vergaß, daß er die Rechnung schon bezahlt hat, nicht. The unacceptability of (2O) and ( 2 1 ) results from the fact that the placement of a lengthy sentential object in the object NP position separates verbal elements (or verbal and adverbial elements) which belong together thus leaving relatively light constituents isolated and dangling at the end of the sentence. The pragmatic factor that determines the relative acceptability of such constructions centers around the strategy of keeping constituents that must be processed together in close proximity, i.e. ( 2 0 ' ) Wir nehmen an, daß ... ( 2 1 ' ) Paul vergaß nicht, daß ... unless they are weighty enough not to loose their meaning in isolation ( c f . ( 1 7 ) ) . We will term this strategy a. A related performance strategy - strategy b. -, also connected with facilitating the processing of sentences, will prohibit repeated sentential embeddings of the same category. Whereas sentence ( 1 7 ) with one finite sentential complement in the object NP position is grammatical and to a certain degree acceptable, sentence ( 2 2 ) with a finite object clause embedded inside a clause of the same category achieves a lesser degree of acceptibility than that granted to ( 1 7 ) . 3 Explicit association is not always possible with object complementation. Pütz (1975, 68-7O) o f f e r s a list of transitive verbs that do not permit the e_s. In these cases the S complement can only be associated implicitly with the matrix. In all other cases, either explicit or implicit association is possible.
151
( 2 2 ) ? # W i r haben beobachtet, daß Hans, daß er dabei sich selbst verrät, aus lauter Dummheit nicht merkt. An infinite complement, on the other hand, embedded inside a finite complement is grammatical and acceptable ( 2 3 ) Wir haben beobachtet, daß Hans sich zu retten versucht. ( 2 4 ) ?Wir haben, daß Hans sich zu retten versucht, beobachtet, unless strategy a. is blatantly violated. (25) «Ich habe gehört, daß seine Frau ihn den Verwandten seine wertvollen Briefmarken zu zeigen zwingt. We see from sentences like ( 1 7 ) , ( 2 3 ) and possibly even ( 2 4 ) that the possibility of an S occupying the otherwise NP object position in the phrase structure of a sentence cannot be ruled out in German. By making such sentences grammatical, it can be left to a theory of performance to predict their acceptibility. Once again, contrary to the conclusion we came to in studying the English predicate complement system, the sentential object in German may hold the object NP position of a sentence. We conclude that within the German predicate sentential complement system an embedded sentence can occur in the subject or object NP position of a matrix sentence or it can appear as the final element of the matrix sentence. The latter option which we will refer to simply as the sentence final s" position is the unmarked position for a sentential complement: sentential complements may always occur in this position free of restriction. The former option - which we will term the NP position - is the marked case, since it is frequently the case that sentential complements which occur in NP positions are felt to be stylistically less well-formed than the sentence variant where the sentential complement occurs in the S position. 4.1.3
Lexical Redundancy and Control
4.1.3.1
Assigning the Syntactic Functions 'Subject' and Object 1 to Sentential Complements
When a sentential complement holds an NP position of a sentence, it will be attributed subject or object function by the same mechanism that accords NP complements their syntactic functions, i.e. the subject of the sentence is defined as the
152
NP which engages in the dominance relationship [NP, S] and the object of a sentence is the NP which engages in the relationship [ N P , VP], Here we are not deviating from the traditional notion in TG that an NP can be optionally expanded to an If and that given the appropriate subcategorization in the lexicon, a certain class of verbs can be placed into an initial phrase marker by lexical insertion where an 'S complement cooccurs in subject or object NP position. We would like to turn now to the mechanism that interprets sentential complements in the 'S position as subject or object clauses. We assume that no rule of extraposition exists in the grammar for predicate sentential complements and that Ss are freely generable by the base rules in either NP positions or in the S position. Given these two sentential complement positions produced by the base, a verb whose subcategorizational entry specifies that a sentential complement may occur as a possible expansion of its subject or object NP will be insertable into a phrase marker if the appropriate £3 complement is found that NP position or if the S complement has been written the sentence final S position. We will in the following how the base derivation of sentential complements in the
in into sketch S
position is accomplished and how such base-generated extraposed complements are accorded their syntactic functions. The verb stimmen, for example, can be placed into an initial phrase marker by lexical insertion rules where an S complement occurs in either the subject NP position or the S position. ( 2 6 ) ( a ) Daß Köln am Rhein liegt, stimmt. (b) Es stimmt, daß Köln am Rhein liegt. If the dual placement possibilities of the daß complement to stimmen in the above sentences had to be formulated into the subcategorization of stimmen, subcategorizations would turn out to be much more complex than hitherto supposed in the theory of grammar, robbing the analysis we are considering of its credibility. But they don't. A lexical redundancy rule which specifies the following would suffice to account for both (26a & b) given the normal subcategorization of stimmen;
153
Any sentential complement which is found in the subcategorization of a given verb as a possible expansion of the subject or object NP term to that verb can alternately occur in the S^ position of the sentence, in which case the NP position specified in the subcategorization is
'controlled* by the
sentential complement holding the S position. The lexical redundancy rule could be formulated as given in ( 2 7 ) ( 2 7 ) Lexical Redundancy Rule (LRR) (i)
Read the subcategorizational specification of a term