You are on page 1of 358

VERB COMPLEMENTATION

IN SWEDISH AND OTHER GERMANIC LANGUAGES

Sture Ureland
VERB COMPLEMENTATION IN SWEDISH
AND OTHER GERMANIC LANGUAGES

Studies in Comparative Syntax

AKADEMISK AVHANDLING

som med vederbörligt tillstånd av


Humanistiska Fakulteten vid Universitetet i U meå
för vinnande av filosofie doktorsgrad
kommer att offentligen försvaras
i H umanisthuset, Sal E,
onsdagen den 30 maj 1973, med början kl 10.00

av

STURE URELAND
Fil. lic., G.H.

UMEÅ 1973
Centraltryckeriet
E r r a t a t o P . S . U r e l a n d Verb Complementation in Swedish and Other
Germanic Languages. Studies in Comparative
Syntax.Umeå 1973: Skriptor.
p. 8, 1 ., + 2 8 Change century to centuries
p- 8, 1 ., - 1 3 C h a n g e is^ t o a r e
p. 12, 1 ., + 6 D e l e t e t h e p r e p . i_n
p- 26, 1 ., + 1 6 A d d i n g t o s m e l li n g
p. 27, 1 .. + 1 6 C h a n g e s u b c l as s e s ( 8 £) a n d ( 8_ f ) t o S u b c l a s s e s ( 8 £ ) a n d
p- 29, 1 ., - 2 3 C h a n g e s u b c la s s ( 4 b) t o s u b c l a s s ( 4 c: ) ( 8d_)
p. 4B, 1 .. - 1 7 Add with between: s e n t en c e s wit h active
p- 51, 1 .. - 7 S t a r t h e E n g l i s h t r a n s l a t i o n o f ( 5 3b )
p- 55, 1 ., - 1 9 Delete £ in Ellegård 1 971:93 and 1 35
p- 57, 1 ., - 2 2 C h a n g e ( 5J _ ) t o ( 5 2 )
p- 62, 1 ., - 2 C h a n g e ( 8_ 5 ) t o ( 8 , 3 )
p- 63, 1 ., + 2 8 Change his t o t ^h i s
p. 63, 1 ., - 6 Delete that before BE
p. 67, 1 ., - 1 Delete where
p- 79, 1 ., + 6 C h a n g e 50.^3 to 50.4
p. 79, 1 ., - 6 A d d D e l e t i o n t o P a s s i ve A u x i l i a r y De letion Rule
p- 83, 1 ., - 1 6 A d d / N o m i n a t i v e t o t h e t i t le A c c u s a t i v e / N o m i n a t i v e -
with-Predicative Adjective Con struction
p. 84 1 ., - 2 0 C h a n g e t h e t ra n s l a t i o n o f ( 1 1 2 ) ' t o ( x H e ^ 1 i s t e n e d
a s i f t h r o u g h t h e s t o r m h e . n o t i c e d h e r . h e a r o t he r
sounds 1 ^
p. 87, 1 ., + 3 C h a n g e 5^3 t o 7 _4
p- 93, 1 ., + 2 3 Change förrådit to förrått
p- 96, 1 ., - 6 C h a n g e 6 ^ 1 a n d 3 . 2 t o £ . 1 a n d 9 _.2
p. 99, 1 ., + 4 C h a n g e s e c t io n 8 _ t o s e c t i o n 9_
p. 103, 1 ., + 1 9 C h a n g e c if E x p e r i e n c e r t o t h e Experiencer
p- 107, 1 ., + 1 2 C h a n g e ( 1 4C H t o ( 1 4 2 ^
p. 107, 1 ., + 1 3 Change (142) to (143)
p- 112, 1 ., - 1 2 Change o to £ in +Pass in Figure 2, (152a)
p- 113, 1 ., - 1 3 C h a n g e c o re f e r e n c e i n d e x j t o i in Pro. namn
in Figure 3, (153a)and (153b) —
p- 121 , 1 ., - 2 Add which yields after sägas
p- 127, 1 ., - 1 D e l e t e f o o tn o t e 6 2
p- 150, 1 ., + 7 Add and ACI-Constructions (cf.(200)) before complements
p- 165, 1 ., - 1 0 C h a n g e 75.3^ t o 75.£3
p- 166, 1 ., + 2 Change ( 68b ) to (67b )
p- 166, 1. -5 Change 13.3^ to 13.4
p- 169, 1 ., - 1 2 A d d o r N o m i n a t iv e a f t e r A c c u s a t i v e
p- 170, 1 ., + 1 6 C h a n g e 5. 1 . 1_ to 5.1.2
p- 194, 1 ., + 1 3 C h a n g e t h e t ra n s l a t i o n o f z a k r y l a t o c l o s e d a n d s t ä n g d e
+14 in the English and Sw edish translation in footnote 31.
+17
p- 207, 1 ., - 1 3 Change 13.2 to 13.3
p. 214, 1. +21 Change +Dur under fa 11a t o + N o #n i n e x a m p l e ( 9 b )
p- 235, 1.-15 D e l e t e t h e s ta r i n ( 4 7 b)
p. 237, 1 ., + 6 C h a n g e s u b c l a s s e s ( 1 b ) a n d (2 b ) to subclass (2b)
p- 243, 1 .. + 1 7 Add (la) and before (2b)
p- 247, 1 ., + 7 Add i.e. (4a) after section 1
p. 250, 1 .. - 1 7 Change + Dur to +Stative u n d e r b e a l l w r o ng i n e x ( 9 5a )
p- 251, 1 ., + 9 C h a n g e + D u r t o + S t a ti v e u n d e r w a s a l l w r o n g i n ( 9 7 a )
p- 253, 1 ., + 1 4 C h a n g e (V K H t o ( 1 0 1 )
p- 266, 1 .. + 8 Change wel1- to i 11-formed
p- 2 7 9, 1 ., - 2 Add Step VI II is posited before in order to add
suffixes
p- 288, 1 ., + 1 6 Change mother's to mothers 9 c+\7l p.y
p- 294, 1 ., + 6 C h a n g e t h e s e c on d v e r b m a t r i x f ro m ! _ N | t o
N 1+N
297 1. - 1 1 Delete note 36
300 1. - 2 1 I n s e r t a f t e r c o m p a r at i v e - s y n c h r o n i c the following
passage: approach. H owever, not only are different
s u r f a c e s tr u c t u r e s . . .
305 1. - 6 Add if after that is
310 1. + 2 2 Add Raising to Subject Raising Rule
312 1. - 1 8 Change 5.4.2 t o 5.5_.2 in footnote 4
312 1. - 8 A d d S e c t i o n Two before (15) in footnote 7
313 1. + 1 Change 5.4.2 t o 5.5_.2
324 1. - 1 3 N o i t a l i c s of 7 n
336 1. -9 Change VII to VIII
VERB COMPLEMENTATION IN SWEDISH AND OTHER GERMANIC LANGUAGES
STUDIES IN COMPARATIVE SYNTAX

BY

STURE URELAND

Distribution:Språkförlaget Skriptor AB
Fack. 104 65 Stockholm 15
Copyright 1973 by Sture Ureland
All rights reserved

Centraltryckeriet, Umeå 1973


I

PREFACE

My interest in the sentential complement structures of Germanic


languages arose in 1968 when I wrote a report on the German
Accusative-with-Infinitive Construction for the Linguistics
Research Center at the University of Texas at Austin (cf. Ureland
1968). I dis covered that the number of verbs which took this
type of verb complementation was limited not only in German but
also in the other Germanic languages I knew. I decided then to
write an interlingual description of the embeddings which
occur after the given limited set of verbs in each of the three
Germanic languages, that is those of Swedish, German, and
English. The present investigation of verb complements in
Swedish and other Germanic languages is a product of this
intention.

Originally I wanted to include all verba sentiendi, verba dioendi,


verba putandi* and verba causativa. Such a comparative interlingual
description of all the four major verb classes mentioned here
would have extended beyond the limits of the framework of the present
dissertation. As time passed and insights into the linguistic
complexities grew I realized that more than a single colume
would have been necessary to encompass all the facts and observa­
tions on the general processes of embedding and their constraints
after all four classes of verbs. Because of space limitation
Part One and Part Two of the present investigation are focused
upon the sentential complements which occur after Swedish, German,
and English verba sentiendi. The syntax of these verbs turned
out to be so complicated that the first two parts had to be
devoted to describing the constraints on their rules of embed­
ding. Therefore, a second volume is planned which will describe
the sentential complements which occur after verba dioendi,
verba putandi, and verba causativa, that is Part Three and Part
Four.

Through the benevolent assistance of Prof.S. Allén and a grant


from the University of Umeå an important computerized corpus of
more than eight thousand verbs has been made available to me.
This corpus has not only given me valuable information on actual
verb complements after the verbs of interest, but has also served
as a starting point for a general classification of Germanic
verba sentiendi, verba dicendi3 verba putandi, and verba causativa.
The computerized corpus is referred to as the Gothenburg Corpus
in the present investigation. This corpus is a part of the larger
newspaper corpus computerized by the Gothenburg Research Group
as known from the works of Allén 1970a, 1970b and 1971.
Excerptation work from the Gothenburg corpus has given impetus
to a multitude of ideas so that the set framework of the
planned investigation could no longer be maintained but instead
the investigation expanded and gained in depth as far as the
verba sentiendi were concerned. Other sentential complements
besides Accusative-with-Infinitive were included to be able to
give an overall survey of the embedding process. Earlier descrip­
tions of verb complementation after the Swedish verba sentiendi
II

can all be said to be incomplete and atomistic. The empirical


contact with a large computerized corpus has provided the
investigator with facts that led to the discovery of hitherto
unknown constraints on the generation of various types of verb
embeddings. It is these constraints which are in the center of
interest here. The generative power of the transformational
embedding rules has to be limited by a number of deep, shallow,
and surface structure constraints. The lack of such constraints
in earlier traditional and generative descriptions of verb
complementation after Germanic verba sentiendi renders such
descriptions incomplete and in the case of generative descrip­
tions, too powerful, since they generate ungrammatical embeddings.
The present work is intended as a contribution to a more adequate
description of embedding processes in Germanic languages. The
interlingual approach developed here liberates us from any
monolingual bias concerning the verb complementation structure
of one given Germanic language. Instead the synchronic compara­
tive approach to Germanic complementation which has been developed
especially in Part Two gives us a broader perspective of general
embedding processes and their constraints in Germanic languages.

The writing of a monograph involves a long series of people and


institutions who have directly or indirectly influenced my
thinking in the course of my education. My studies at the Deutsche
Akademie der Wissenschaften zu Berlin, Arbeitsstelle Strukturelle
Grammatik, between 1964-1965 gave me important contacts with
Dr. M. Bierwisch, Dr. W. Mötsch, and Dr. K.E. Heidolph who in­
troduced me to generative grammar. The research at the Department
of Linguistics of the University of Texas at Austin some years
later in 1968-1969 was made possible through Prof. W.P. Lehmann
whose encouragement and advice were invaluable at a time when
linguistics was not yet established as an independent subject in
Sweden. To Prof. K.H. Dahlstedt I am equally indebted. His great
interest in young and daring ideas has stimulated me throughout
the present work. Many improvements have been possible thanks
to his insightful comments. To my colleague and wife Veronica
Bonebrake I owe not only stylistic improvements of my English
but also numerous suggestions for scientific improvements. I
also thank Mrs. Sigbritt Stenström whose great patience in typing
out the manuscript into the present readable format has been
admirable. Docent C.G. Söderberg, Archivist A. Bränström, Fil.mag.
M» Rahkonen, Fil.lic. S. Karlsson. Pastor 0. Korhonen, Fil.kand.L,
Lagerstedt and the Nomad P.M. Bals have provided me with informa­
tions on verb complementation after Russian and Finnish verba
sentiendi and dioendi. This information has been used in the
comparative synchronic approach developed in Part Two. To all
these people I exp ress my gratitude for the insights they have
provided me. Any errors or omissions are purely my responsibility.

Umeå, April 1973.

S.U.
III

LIST OF ABBREVIATIONS
Acc accusative
Achieve achievement
AC I Accusâtive-with-Infinitive
Adj adjective
Ahl Ahlen 1833
AWG Accusative-with-Gerundive
Caus-j^ permissive-causative Callow]
Caus2 hortative-causative (ask)
Causj instigating-causative (cause)
Caus4 instructive-causative (teach)
Causg causative-instructive (help)
Cod Bur Codex Burasanus
Col column
Compi.R. Complementizer Rule
Delib deliberate
DN Dagens Nyheter
Del deletion
Dem Pron demonstrative pronoun
DS deep structure
Dur durative
Ek Erikskrönikan
Fact factive
G Grimberg 1905
Gen genitive
GHT Göteborgs Handels- och Sjöfarts-
Tidning
Holm Holm 1952
Hum human
Infi Infinitive One in Finnish
Inf2 Infinitive Two in Finnish
Inf3 Infinitive Three in Finnish
Inf4 Infinitive Four in Finnish
Ingress ingressive
Instant instantaneous
IN; l symbol for the infintive suffix
IN2 symbol for the past participle
suffix
Inchoa inchoative
Iter iterative
Masc masculine
MELL Magnus Erikssons landslag
MES Magnus Erikssons stadslag
Mom aspectual feature for instanta­
neous Aktionsart
[N]S° the subject noun phrase of the
linjNP
matrix sentence
the subject noun phrase of the
complement sentence
N noun
IV

[N]VP object noun phrase


NP
NP noun phrase
NP; noun phrase with coreference
index
Olfact olfactory
Part partitive
Past Part past participle
Prep preposition
PrepP prepositional phrase
Pres Part present participle
Pro pronoun
R rule
Result resultative
S sentence
SS surface structure
50 matrix sentence
51 the first complement sentence
s2 the second complement sentence
S3 the third complement sentence
SA structural analysis
SAOB Svenska Akademiens Ordbok
SC structural change
SDS Sydsvenska Dagbladet Snällposten
SOV subject-obj ect-verb
ST Stockholms-Tidningen
Sub subordinator
Subj Rais Subject Raising
SvD Svenska Dagbladet
SvmL Svenska Landsmål
SVO subject-verb-object
SVÖ Svenska öden
sentence type node
jCop Copying (transformational) Rule
TExtra
Extraposition (transformational)
Posit
Rule
T,Nom
Nominalization (transformational)
Rule
rpPOSS
Possessive Suffix Spelling Rule
Suff
in Finnish
TPsych
Psychological Movement Rule
Move
^Reflex
Reflexivization Rule
TSubj
Subject Raising Rule
Rais
V verb
VgL I The Older Province Law of Väster­
götland (Den Äldre Västgötalagen)
of the 13th century
VgL II The Later Province Law of Väster­
götland (Den Yngre Västgöralagen)
of the 13th century
VP verb phrase
VSO verb-subject-object
ÖgL The Province Law of Östergötland
(östgötalagen)of the 14th century
v

TABLE OF CONTENTS
PART ONE
PREFACE I
LIST OF ABBREVIATIONS III
TABLE OF CONTENTS V
INTRODUCTION TO PART ONE 1
1.1 Four approaches to Verbal Complementation 1
1.2 A Combined Performance and Competence Study 2
1.3 Background and Goal 5
1.4 Synchronic and Diachronic Comparative Syntax 8
1.5 Theoretical Framework for Describing Surface
Structures 9
1.6 Intuition-Oriented Linguistics and Empirically-
Oriented Linguistics 13

2• Presentation of Modern Swedish ACI-Constructions


2.1 In Declarative Sentences 15
2.2 In Subordinate Relative Clauses 17
2.3 In Interrogative Sentences 18

3. Classification of ACI-Verbs in Modern Swedish


i
3.1 An Alphabetic List of ACI-Verbs in Modern Swedish 20
3.2 Semantic Subclassification of ACI-Verbs 21
3.3 Verbs of Perception (Class A) 23
3.4 Performative Verbs and Mental Verbs (Class B) 29
3.5 Instantaneous-Perfective Verbs (Class C) 32
3.6 Stative-Nonperfective Verbs (Class D) 33
3.7 Causative Verbs (Class E) 34
3.8 Reflexive Verbs (Class F) 36

4. The Concept of Syntactic Rule Domain


4.1 The Standard Theory and Level Constraints in Syntax 38
4.2 Level Constraints for Generating Well-Formed
Swedish ACI-Constructions 39
4.3 The General Process of Infinitivization in Germanic
Languages 41

5. The Subject Raising Rule in Swedish (SE)


5.1 The Verbal Complements of SE^ 'see visually1 46
5.1.1 The att and fcwr-Embeddings Occurring after SEi 'see
visually1 50
VI

5.1.2 Syntactic Evidence for Different Underlying


Representations of ACI-Constructions and Comple­
mentizer Embeddings 52
5.1.3 Semantic Evidence for Different Underlying
Representations of ACI-Constructions and Comple­
mentizer Embeddings 57
$.2 Subject Raising and Auxiliary Deletion or
Aktionsart Neutralization 60
5.3 Nominalized Complements Occurring after SE^ 'see
visually' and SE2 'realize' 67
5.4 Accusative-with-Gerundive Construction and att-
Deletion 68
5.5 The Surface Structures of FÄ SE 'catch sight of' 69
5.5.1 The Categorial Status of Swedish FÂ in FÂ SE
'catch sight of' 69
5.5.2 Comparative Syntactic Evidence for Treating FÅ as
an Aspectual Auxiliary 71
5.5.3 The Sentential Complement Structures of FÅ SE
'catch sight of' 75
5.5.4 Semantic Evidence for Treating FÅ^ as an Aspectual
Auxiliary 75

6. Results of the Linguistic Performance Study on


Verbal Complements Occurring after SE and FA SE
6.1 Aspects of the Productivity of Embedding Rules after
SEi 'see visually' 78
6.2 Aspects of the Productivity of Embedding Rules after
SE2 'realize, understand' 79
6.3 Aspects of the Productivity of Embedding Rules after
HELST SE3 'prefer to see' 80
6.4 Aspects of the Productivity of Embedding Rules after
FÂ SE 'catch sight of' 81

7. The Subject Raising Rule in Swedish (MÄRKA)


7.1 The Verbal Complements of MÄRKA 'notice' 82
7.2 Accusative-with-Predicative Attributes 83
7.2.1 Accusative-with-Past Participle Construction 83
7.2.2 Accusative-with-Predicative Adjective Construction 83
7.3 Nominalizations 84

8. Results of the Linguistic Performance Study on


Verbal Complements Occurring after MÄRKA 'notice'
8.1 The Productivity of Embedding Rules after MÄRKA
'notice' 87
VII

9• The Subject Raising Rule in Swedish (UPPTÄCKA)


9.1 The Verbal Complements of UPPTÄCKA^ 'discover^
visually1 88
9.2 The Verbal Complements of UPPTÄCKA2 'discover
cognitively' 94
9.3 The Verbal Complements of UPPTÄCKA3 'make a
discovery, find out1 96

10. Results of the Linguistic Performance Study on


Verbal Complements Occurring after UPPTÄCKA '"discover1
10.1 UPPTÄCKA^ 'discoveri visually' 99
10.2 UPPTÄCKA2 'discover2 cognitively' 100
10.3 UPPTÄCKA3 'make a discovery' 100

11• The Subject Raising Rule in Swedish (OBSERVERA)


11.1 The Verbal Complements of OBSERVERA1 'watch' 101
11.2 The Verbal Complements of OBSERVER/^ 'become
aware of' 102

12. Results of the Linguistic Performance Study on


Verbal Complements Occurring after OBSERVERA 'observe'
12.1 OBSERVERA-^ 'watch' 105
12.2 OBSERVERA2 'become aware of' 105

13. The Subject Raising Rule in Swedish (HÖRA)


13.1 The Verbal Complements of HÖRA^ 'hear1 auditorily' 106
13.2 A Specific Underlying Representation of ACI-
Constructions after HÖRA1 'hear-^ audi torily' 110
13.3 HÖRA-^ Occurring with Passivized Idioms 114
13.4 The att and /zwr-Embeddings Occurring after HÖRA 2
'hear2 of' 122
13.5 HÖRA^ Occurring with Nominalized Complements 127
13.5.1 The Unbalanced or the Balanced View for Describing
Nominalizations 127
13.5.2 The Problem of Paraphrases 132
13.5.3 Metaphors as Arguments for a Semanticist Solution
of Nominalizations 133
13.6 The Verbal Complements of FÄ HÖRA, 'hear-, auditorily'
and FÄ HÖRA2 'hear2 of, be informed' 134
13.6.1 The Categorial Status of Swedish FÄ in FÅ HÖRA 134
13.6.2 The Sentential Complement Structures of FÅ HÖRA-.
'suddenly hear-^' 136
13.6.3 The Sentential Complement Structures of FÂ HÖRA?
'hear2 of, be told' 138
13.6.4 Metaphorical Usage of FÄ HÖRA2 'hear2 of' 139
VIII

14. Results of the Linguistic Performance Study on Verbal


Complements Occurring after HÖRA and FA HÖRÅ

14.1 HÖRA^ 'hear^ auditorily' 140


14.2 HÖRA2 'hear2 of, be told' 141
14.3 FÅ HÖRA^ 'hear^suddenly or unexpectedly' 143
14.4 FÅ HÖRA9 'hear?of unexpectedly or be unexpectedly
told' z z 144

15. The Subject Raising Rule in Swedish (KÄNNA)


15.1 The Verbal Complements of KÄNNA^ 'feel-^tactually' 145
15.2 KÄNNA3 'feel3mentally' 147
15.2.1 Subject Raising and Passive Auxiliary or Copula
Deletion 150
15.2.2 Nominalization 151
15.3 KÄNNA2 'smell' 152
15.4 KÄNNA4 'feel4somatically' 154
15.5 FÂ KÄNNA 'feel and smell suddenly and unexpectedly' 158

16. Results of the Linguistic Performance Study on Verbal


Complements Occurring after KANNA
16.1 KÄNNA1 'feelx tactually' 160
16.2 KÄNNA2 'smell' 161
16.3 KÄNNA3 ffeel
3 mentally' 161
16.4 KÄNNA^ 'feel4 somatically' 162
CONCLUSION OF THE PERFORMANCE STUDY 164
CONCLUSION OF THE COMPETENCE STUDY 177
FOOTNOTES TO PART ONE 185

PART TWO
INTRODUCTION TO PART TWO 204
0. Presentation of ACI-Verbs in Three Germanic
Languages 207
1. Morpho-Syntactic and Semantic Aspectual Features
in Swedish, German, and English 209
2. Aspectual Constraints and The Problem of Lexical
Insertion 214
3. Aspectual Copying Rule and Variable Syntactic
Rules in Swedish and German 220
4. Overt and Covert Aspectual Markers in Germanic
Languages 224
5. Verb Complements Occurring After Tactual Verbs in
Swedish and German 232
IX

6. The Subject Raising Rule in English and the


Generation of ACI and AWG-Constructions (HEAR and
SEE) 236
7. English Verbal Complements after Durative Verbs
of Perception (LOOK AT, LISTEN TO, OBSERVE, WATCH,
and WITNESS) 243
8. English Verbal Complements after Nondurative Verbs
of Perception (BEHOLD, DISCOVER, and NOTICE) 246
9. English Verbal Complements after Tactual Verbs of
Perception (FEEL-^ and PERCEIVE-^) 248
10. FEEL.,, PERCEIVE,, and the Accusative-with-TO-
Infinitive 250
11. Reflexive Complement and Predicate Complements
after Verbs of Somatic Perception in Germanic
Languages 252
12. Towards a Theory of Comparative Synchronic Syntax
12.1 Common Germanic Deep Structures for Deriving Surface
Structures 256
12.2 The Synthesis and Analysis of ACI-Constructions 259
12.3 A Model for Describing Forward and Backward
Operating Transformations 263
12.4 Comparative Derivation of Verbal Complement
Structures 266
12.5 Comparative versus Contrastive Syntax 272
12.6 An Illustrative Example of Comparative Synchronic
Syntax 278

13. The Subject Raising Rule as a Formal Universal


13.1 Typological Evidence 285
13.2 Historical Evidence 285
13.3 Subject Raising in Finno-Ugric Languages 286
13.3.1 Verbal Complements after North Lapp verba eentiendi 286
13.3.2 Verbal Complements after Finnish verba sentiendi 287
13.4 The Two-Subject Hypothesis and Subject Raising 291
CONCLUSION TO PART TWO 296
FOOTNOTES TO PART TWO 312
APPENDIX
A. Glossary to Part One 321
B. Glossary to Part Two 322
REFERENCES 323
1

INTRODUCTION TO PART ONE


1.1 FOUR APPROACHES TO VERBAL COMPLEMENTATION
In the present investigation four different approaches will be
applied to describe the verbal complementation system in
Swedish and other Germanic languages, that is the sentential
complements which occur after Swedish, German, and English
verba sentiendi to be discussed in Part One and Part Two,and
verba dioendi to be discussed in Part Three.

The first approach will make use of our own intuition and
knowledge about the Swedish sentential complement system and
the transformational rules that generate embeddings. This know­
ledge constitutes the fundamental basis for describing the
process of embedding in written Standard Swedish.1

The second approach will take advantage of 8681 examples


mechanically excerpted by a computer from the newspaper corpus
of the Gothenburg Research Group on Modern Swedish (cf. Allén
1970aand 1971). This corpus has served as empirical verification
of our own intuitive judgements about the well-formedness of
Swedish verbal complement types.2

This collection of verba sentiendi (2284 examples) and verba


dioendi (3424) examples) has turned out to be extremely valuable
not only as a source of actual examples of Swedish embedding
types, but has also helped us to disclose a number of verb
complements which occur after the two major classes of verbs which
would otherwise have remained undescribed.

The third approach consisted of consulting handbooks of the


Swedish language such as the SAOB, östergren 1919-1968, Svensk
handordbok I960., and Illustrerad svensk ordbok 1958. Without
this lexicological research a number of fine semantic distinc­
tions between various subclasses of for instance verba sentiendi
could not have been made. Excerptation of the Gothenburg corpus
and consultation of the lexical handbooks were made simultaneously.

The fourth approach draws upon the insights into the Swedish
complementation system gained through both grammatical intro­
spection and empirical observations concerning a given corpus.
In this way a Germanic basis of syntactic description is
established for making interlingual comparisons between the
three Germanic languages.

The results of the empirical study of verbal complements,


cooccurrence restrictions, and transformational rule constraints
in Swedish can be used in describing similar structures in German
and English with their concomitant derivational constraints,
since the syntactic rules involved can be shown to be common
to Germanic languages.

Frequently, however, surface or deep structure constraints have


to be added to the grammar of this or that Germanic language
in order to constrain the rule of embedding. Such language-
specific constraints (e.g. the Aspectual Constraint on Subject
2

Raising in English discussed in Part Two, section 2) are to be


expected considering the wide variation of morphology among
the three Germanic languages. Consider for instance the lack
of an overt infinitival marker and the existence of the progress­
ive in^-Construction in English (discussed in Part Two, section
6), or the existence of a pseudoactive infinitive in German
which occurs in the surface structure rather than a past
participle plus a passive auxiliary (cf. Plötzlich hörte sie
ihren Namen rufen 'Suddenly she heard her name being called1),
discussed in Part Two section 12.4. It is not only such morpho­
logical idiosyncrasies which block the generation of a given
embedding type, but there are often syntactic idiosyncrasies
which are responsible for whether the potential rule of embedding
applies as for instance in Swedish, where the Subject Raising Rule
is free to generate ACI-Constructions under coreference linkage,
i.e. when a reflexive pronoun constitutes the object NP: Han
påstod (sade) sig vara sjuk 'He claimed (said) that he was ill'.

1.2 A COMBINED PERFORMANCE AND COMPETENCE STUDY


The primary goal of this description is to describe what a
native speaker of Swedish is assumed to know in order to master
the system of verbal complementation and the transformational
rules that generate embeddings. The secondary goal is to sketch
a theory of comparative synchronic syntax of Germanic languages
with respect to sentential complements which occur after a
given set of verba sentiendi and dicendi. Part One is focused
on the following verba sentiendi in Swedish: SE 'see', FÅ SE
'see suddenly or unexpectedly', MÄRKA 'notice', UPPTÄCKA
'discover', OBSERVERA 'watch', HÖRA 'hear', FÅ HÖRA 'hear
suddenlyor unexpectedly', and KÄNNA 'feel', (the ten subclasses
of these verbs of perception are omitted here).

In Part Two a synchronic comparative-syntactic description is


presented which treats verba sentiendi in Swedish, German,and
English, the generation of ACI-Constructions and other sentential
complements after perceptual verbs, and various level constraints
on embedding rules in Germanic languages. A synchronic comparative
theory is sketched in section 12 of Part Two for describing the
systems of verbal complementation in Germanic languages.

Part Three will treat the verb complements after verba dicendi
and Part Four will present a comparative study of verba dicendi
and their verb complements in Germanic languages.

An important dichotomy is maintained throughout this investiga­


tion between the usage of the Swedish verbal complementation
system as evidenced by the Gothenburg corpus and our own
linguistic knowledge about the system. Sentences which contain
verbs of perception have been excerpted from the Gothenburg
corpus and then these perceptual verbs have been classified
according to syntactic as well as semantic criteria. The
Gothenburg corpus has provided actual examples of Swedish verbal
3

complementation, i.e. of the following nine verbal embeddings


which have been in the focus of Part One:
1) the ACI-Construction, 2) the att-Embedding, 3) the hur-
Embedding, 4) the Accusative-with-Past Participle Construction,
5) the Accusative-with-Predicative Adjective Construction,
6) the Accusative-with-Adverbial Construction, 7) the Accusative-
with-Gerundive Construction, 8) Nominalization Structures, and
9) Passivized Idioms.

However, there is considerable difference between the actual


and potential distribution of embedding types as can be seen
from FIGURES 6 and 7 in the Conslusion of Part One. For each of the
eighteen Swedish verba sentiendi treated in Part One, there
is a disparity between actual and potential examples of the nine
embedding types.

For the att-Complementizer Rule, as demonstrated in Table 3 (FÅ


SEi), Table 10 (HÖRA^), Table 14 (KÄNNAi), and Table 15 (KÄNNA2)
no examples of att-Embedding have been found in the Gothenburg
corpus, although we know from our linguistic intuition that att-
Embeddings do occur after these perceptual verbs in Swedish.
Therefore, in the competence study of the present investigation
of the Swedish complementation system, such potential att-
Embeddings have been exemplified by freely-produced sentences.

Such a lack of data-coverage concerning the nine sentential


complements which occur after verba sentiendi does not surprise
a linguist who is familiar with the limitations of a corpus-
oriented description. Any corpus of a modern language is bound
to be finite and thus unreliable when used as the only basis
of syntactic research to describe linguistic competence. In order
to supplement the lack of data-coverage in a given corpus,
freely-produced embeddings are motivated (cf. Bierwisch 1963:
11 and Chomsky 1965:19).

Therefore the primary concern here is not the performance


study as such but the overriding goal of Part One is to make
deductions for a competence study of the Swedish sentential
complement system and the transformational rules that generate
embeddings after eighteen verba sentiendi.
Section 1.4 discusses the concepts Synchronic and Diachronic
Comparative Syntax whereby important research on the infinitive
in Germanic languages is mentioned. Section 1.5 presents the
theoretical framework for describing verbal complementation. A
combination of the syntacticist and semanticist models has
proved to be the most convenient framework for attaining the
goals of comparative syntax.

In section 1.6 the necessity of empirical observations of actual


sentences is stressed to gain reliable results in describing
linguistic competence. A one-sided intuition-based method is
therefore rejected.

Section 2.1 through 2.3 present Modern Swedish ACI-Constructions


after verba sentiendi, verba dioendi etc in declarative, sub­
ordinate and interrogative clauses. In order to gain an overall
picture of the syntactic domain of the Subject Raising Rule,
4

six different classes of verbs are then subclassified in sections


3.1 - 3.8 according to syntactic and semantic criteria (cf.
Subclassification Rules (i) - (viii)).

The concept of syntactic domain is further discussed in section


4, where a short history of the development of transformational
constraints within transformational theory is presented (cf.
especially sections 4.1 and 4.2).

In section 4.3 the general process of infinitivization is treated,


whereby the general character of the Subject Raising Rule is
demonstrated. It is a transformational rule which is common to
Germanic languages.

Section 5 through 16 are the core of Part One. Nine different


verbal complement types after verba sentiendi in Swedish are
systematically exemplified by excerpts from the Gothenburg
corpus. In such instances where this corpus lacks evidence of
a verbal complement type, our own linguistic intuition comple­
ments such insufficient data-coverage. (Cf. the discussion of
the Performance and Competence Study after the tables of
distribution and in the Conclusion of the Performance Study in
paragraph 3)•

Section 5 treats the Subject Raising Rule in Swedish and other


embedding rules which are applied to structures which contain
SE 'see' and FÅ SE 'catch sight of' as the matrix verb.

Section 7 treats the same rules operating on structures which


have MÄRKA 'notice' as the main verb; section 9 treats the
structures of UPPTÄCKA 'discover'; section 11 the structures
of OBSERVERA 'watch'; section 13 the structures of HÖRA 'hear'
and FÅ HÖRA 'hear suddenly', and section 15 the structures of
KÄNNA 'feel'.

Each description of the nine embedding types after the eighteen


verbs of perception is complemented by a Table of the Pro­
ductivity of Embedding Rules, i.e. in sections 6, 8, 10, 12, 14,
and 16. In the first paragraph of the Conclusion of the Per­
formance Study, FIGURES 6A and 6B summarize the actual
occurrence of embedding types as evidenced by the examples from
the Gothenburg corpus. In paragraph 3 of the same conclusion
the disparity between actual and potential occurrence of
embedding types after Swedish verba sentiendi is demonstrated
in FIGURES 7A and 7B.

The same conclusion also discusses the criteria for considering


a given corpus representative as a oorpirs fro m a syntactic point
of view, but not as a reflection of what a native speaker knows
about the sentential complement system of a given language.
5

1.3 BACKGROUND AND GOAL

The present stud/ of verbal complementation is the first part


of a trilingual synchronic and diachronic investigation of the
sentential complementation system in Swedish and other Germanic
languages.

The surface structures which occur as sentential objects after


two major classes of verbs have been focused upon, that is
after verba sentiendi and verba dicendi. Nine different
embedding types enumerated above in section 1.2 will be de­
scribed together with the transformational rules that generate
them. Special emphasis will be laid on a ve rbal complement
which bears the name of Accusative-with-Infinitive Construction
(the ACI-Construction) in traditional grammars. This type of
sentential complement which occurs after a number of specific
verbs in Germanic languages contains an object noun phrase and
an infinitival form of an embedded verb. Under the Two-Subject
Hypothesis ACI-Constructions are treated as being derived by
the Subject Raising Rule discussed in section 4.3 which
operates on underlying complement sentences. The Subject Raising
Rule is by no means a transformational rule which is specific to
Swedish. It is a syntactic rule which is common to Germanic
languages and which generates ACI-Constructions under a whole
set of varying conditions (cf. Part Two). For the sake of
convenience the term Accusative-with-Infinitive Construction has
been retained also for certain Swedish and English verbal
complements, although such a term may seem unmotivated from a
surface structure point of view, as these languages display no
overt object case makers.

However, since this investigation is a part of a larger tri­


lingual study of North and West-Germanic ACI-Constructions, and
since German and Icelandic have overt object case markers in the
accusative, it is convenient to keep the traditional term for
the description of Swedish and English constructions as well.

The abbreviation ACI-Construction which stands for the Latin term


accusativus cum infinitivo has been retained as a terminus
technicus also in the description of Germanic verb complements,
because it is an international designation for a well-known
surface structure and also because it is used in a great number
of traditional and historical grammars of European languages.

A second abbreviation used frequently in this investigation,


the AWG-Construction, which stands for the term Accusative-with-
Gerundive Construction, is a typical surface structure in
Modern English. The Latin abbreviation ACG, accusativus cum
gerundivo, is not found to be adequate since the Modern English
construction can be both an active and a passive verb complement,
whereas in Latin the gerundivum was only a passive construc­
tion cf. Prœ terea censeo Carthaginem delendam esse 'Moreover
I am of the opinion that Carthage should be destroyed').
6

It is not our intention to describe all the different kinds of


infinitives in Germanic languages which cooccur with (object)
noun phrases. This study covers only a limited type of verbs
taking sentential object complements in which the infinitve
occurs without an infinitival marker, i.e. in Swedish without
att 'to'.

(A) (1) Han sade sig vara trött.


(He said that he was tired)

(2) Han såg henne komma mot sig.


(He saw her approaching him.)

(3) Han hörde henne skratta.


(He heard her laughing.)

(B) (1) Han lovade henne att köra försiktigt.


(He promised her to drive carefully.)

(2) Han övertalade henne att köra försiktigt.


(He persuaded her to drive carefully.)

(3) Han fick henne att erkänna.


(He made her confess,)

We are concerned only with the (A) type of verb complements and
will disregard the (B) type. The present study is written with­
in the framework of generative grammar. However, on several
points in the discussion of the ACI-Construction in Swedish and
other Germanic languages the results of grammarians who represent
linguistic conceptions other than generative have brought insight
into various aspects of synchronic and historical syntax, e.g.
historical, traditional, and structuralist grammarians.

As an illustration of previous insightful approaches to the


description of verbal complementation in Germanic languages let
us first consider the treatment of infinitives and ACI-
Constructions by historical grammarians such as Paul 1920, Dal
1952, and Jespersen 1969, all three of whom speak of a "subject
of the infinitive" as exemplified by the following three
quotations:

(i) "Als Subj. zu dem Inf. ist entweder das Subj. des Verb,
fin. hinzudenken (Karl will reisen) oder ein von diesem
abhängige Kasus.(Er hiess ihn schweigen)"(Paul 1920:94)

(ii) "Das finite Verb wird nur durch den Infinitiv ergänzt;
das Subjekt des finiten Verbs ist das logische Subjekt
des Infinitivs". (Dal 1952:101)

(iii) "But extremely often the subject is left to be inferred


from the context. It may be identical with the subject
of the main verb: I want to sing SVO (S°I)".(Jespersen
1969:128)
7

Secondly, Beckman 1916 [1959] gives an analysis of the ACI-


Construction in his traditional grammar of Swedish which is
surprisingly modern and could almost be said to be transforma­
tional with some modifications of course.^

Thirdly, the structuralist and glossematic Danish grammarian


Bech 1955 has arrived at an analysis which is similar to that
of both traditional and generative grammarians in his in­
vestigation of the German nonfinite verb. In this interesting
study German ACI-Constructions are analysed by means of symbols
for underlying but deleted nouns (Nf' is used as a symbol for
the equivalent to the deep structure subject of the infinitivized
verb, and N' is the equivalent to the surface structure subject
of the finite verb (cf. pp. 31-47).

Bearing in mind the outstanding research of the European grammari­


ans mentioned here it is surprising to read the works on the
infinitive complement published by American grammarians who
seem to be unaware of earlier significant research on the
infinitive construction. In most cases they consider their own
results to be something completely new. A cliche reference to
Jespersen's 1954 standard work on E nglish syntax is what one
finds at most. It is true that the generative theory of language
constitutes a much more powerful tool for dealing with such
surface phenomena as t he ACI-Construction than did the histori­
cal or even the structuralist methods. It is,however, the belief
of this author that many insights into grammatical processes
have been consciously or subconsciously assimilated from the
work of earlier traditional grammarians and reformulated in
terms of the categories and concepts of transformational
grammar by generative linguists, without always recognizing or
acknowledging their debt to their forerunners. One important
exception is Chomsky, who has repeatedly emphasized his debt
to earlier traditional grammarians of the seventeenth,
eighteenth, and nineteenth centuries in a number of works,
particularly in his Cartesian Linguistics 1966. The same
intellectual honesty is found in Bierwisch's 1963 work on the
German verb.

The traditional grammarians' concept of the "logical subject


of the infinitive" and the treatment of auxiliaries as main
verbs by historical grammarians are evidence of their under­
standing of deep syntactic relations which is worthy of re­
cognition also in modern descriptions of infinitivization
processes and the syntax of auxiliaries.^

As examples of the negligence of earlier syntactic research


compare the "new" discovery of an underlying complement
sentence subject by Rosenbaum 1967a and Kiparsky & Kiparsky 1970
which is deleted in the infinitivization process, with Paul's
1920 "hinzugedachtes Subjekt";or the pseudodiscovery of
auxiliaries as m ain verbs by Ross 1967b with the common treat­
ment of the auxiliariès by L atin philologists, e.g. Woodcock
1959:16. The earlier understanding of these processes on the
part of traditional and historical grammarians has not always
been recognized by generative linguists.
8

1.1 SYNCHRONIC AND DIACHRONIC COMPARATIVE SYNTAX


In European linguistics and philology the process of infinitivi-
zation has been treated from various theoretical points of view
as li nguistic theories have appeared or fallen into oblivion.
The ACI-Construction was for a long time a popular object of
syntactic research among Scandinavianists, English philologists,
and Germanists who collected and classified various types of
ACI-Constructions excerpted from North, West, and East-Germanic
languages.

Åhlén 1833, Falk & Torp 1900, Nygaard 1905, Grimberg 1905,
Heusler 1921, and Wessen 1965 excerpted and collected ACI-verbs
from the ancient Scandinavian languages. They described the ACI-
Cons tructions of Old Norwegian, Old Icelandic, and Old Swedish
from a historical point of view. Nygaard and Grimberg give not
only lists of ACI-verbs and examples of ACI-Constructions from
Old Icelandic and Old Swedish respectively, but they discuss the
origin and the conditions of the use of the constructions as well.

In West-Germanic languages outstanding research on the infinitive


has been carried out by a great number of English and German
philologists, e.g. Callaway 1913, Bock 1931, Brunner 1962,
Visser 1956 and 1969, Mustanoja 1960 on the infinitive in Old
and Middle English; Erdmann 1874, Apelt 1875, Steig 1884, and
Behagel 1924 on the Old Saxon and Old High German ACI-Construc­
tions; Apelt 1875, Mönsterberg-Münchenau 1885, and Behagel 1924
on Middle High German ACI-Constructions.

The ACI-Constructions occurring in The Gothic Bible have been


treated by several German philologists of the past and present
century for instance by Ribbeck 1836, Köhler 1867, Apelt 1874
and v.d. Meer 1914.

The linguistic facts and data accumulated in all the research


on the infinitive in Germanic languages must be evaluated and
considered by anyone who intends to write a comparative study
of the infinitive in Germanic languages.

If the ruleswhich generate the ACI-Construction can be shown


to be rules common to Germanic,5 then observations of the con­
straints and derivational processes in any of the Germanic
languages is of interest for comparative Germanic syntax. The
Subject Raising Rule as f ormulated in section 4.3 is by no means
a language-specific rule, but a syntactic rule which is
characteristic of the Germanic languages (Swedish, German, and
English). It generates ACI-Constructions in each of the Germanic
languages (modern as well as a ncient). It operates in each
language, under a given set of conditions, on two postulated
deep structures of syntactic nature (cf. section 4.3). From
a synchronic as well as a diachronic point of view, comparisons
of level constraints and idiosyncratic properties of the ACI-
verbs between the three Germanic languages are an important
means in constructing a comparative model of syntax which is
also the goal of this investigation.
9

If synchronic comparative syntax is to reach the level of dia-


chronic comparative syntax in its descriptive rigor as
demonstrated by for instance Callaway's and Nygaard's dia-
chronic works on the English and Scandinavian infinitive, then
more interlingual studies of comparative nature have to be
carried out. To accomplish such a comparative ACI-syntax of the
Germanic languages, the present investigation considers the
results of earlier research on the infinitive, in not only
Scandinavian languages, but in other Germanic languages as well.
(Synchronic and diachronic perspectives of the infinitiviza-
tion process have been combined in a special article on the
change of syntactic rule domain between Old and Modern Swedish
(cf. Ureland 1972b(forthcoming).6

1,5 THEORETICAL FRAMEWORK FOR DESCRIBING SURFACE STRUCTURES


A linguist who wants to describe how a given set of surface
structures is understood and generated has today a great number
of linguistic models to choose from. The market is full of
theories for describing syntax, semantics, or phonology. A
young linguist may take a syntacticist approach and work within
the framework of the Standard Theory as sketched by Chomsky 1965
and as improved or modified into the Extended Standard Theory
by Chomsky himself 1970a, 1970b, and 1971a, his students, and
his colleagues. If our potential student chooses this framework,
he will work within an established syntactic theory with a host
of theoretical concepts. The applicability and descriptive
power of this theory of syntax have been tested for the last
ten years on various unrelated languages with varying results.
The descriptive adequacy of this theory has also made important
contributions to psycho- and neurolinguistics (cf. Slobin
1968 and 1971, McNeil 1970, Bever 1970, Weigl & Bierwisch 1970,
Whitaker 1971 etc).

However, our assumed linguist may also take a semanticist


approach and work within the framework of Case Grammar as develop­
ed by Fillmore 1968, 1971 and 1972, or he may work within the
framework of Generative Semantics as sketched by McCawley 1968
and 1972; Lakoff 1970 , 1971a, 1971b, and 1972; Postal 1970b and
others. Finally, if he is still interested in semantics he can
also read Chafe's 1970 book Meaning and the Structure of Language
in which Chafe treats semantic structures like those of the
generative semanticists as being the essential empirical basis
for the study of grammar.

Ellegård 1971 and Dahl 1971 have described the infinitivization


process in English and Swedish from two different points of
view. The former takes the syntacticist approach to the infini­
tive in English and postulates a syntactically motivated deep
structure (cf. p. 134) of a subject-verb-object order (SVO order),
whereas the latter takes the semanticist approach to the
Swedish infinitive along the lines of the generative semantic­
ists by positing a semantic but syntactically unmotivated
deep structure of a verb-subject-object order (VSO order)
10

(cf. p. 53). Dahl is here drawing upon the suggestion made the
first time by McCawley 1970 that English is a VSO language.
Both Eliegård and Dahl use the term Subject Raising for de­
scribing the process of infinitivization. Dahl's or more
accurately McCawley's 19 70 use of this term is, however,
different from the use of the same term coined by Kiparsky & Kiparsky
1970 and applied by Ellegård 1971.

For reasons discussed in Part Two section 12.3 there is no


motivation whatsoever for postulating a VSO order as a syntactic
basic order for English, Swedish, or German. A full range of
empirical evidence speaks against such an ordering of syntactic
elements. It is not certain whether McCawley intends the verb,
subject, and object categories to be syntactic categories or
whether they are to be semantic labels for predicate and
arguments respectively. Judging from the theoretical framework
in which they are presented, they must be semantic categories.
There are several empirical arguments against the VSO order.

First, Bever 1970 and other psycholinguists have demonstrated


that the SVO order is crucial in English for a whole set of
perceptual strategies by asking subjects to identify the
syntactic status of noun phrases (cf. for instance Bever 1970:
294). A VSO order would not reflect this psycholinguistic
evidence.

Secondly, Greenberg 1963 has shown that, from a language-


typological point of view, English cannot possibly be a VSO
language, nor can any of the modern Germanic languages, since
they do not show the characteristics of true VSO languages
such as Berber, Hebrew, Welsh etc. Such true VSO languages
display the adjective after the noun (Greenberg1s Universal 17),
the demonstrative follows the noun, the VSO order is the basic
order for declarative sentences etc. The Germanic languages
show no such syntactic characteristics. Therefore from a
comparative syntactic point of view McCawley's claim that English
is a VSO language seems ad hoc.

Thirdly, Lehmann 1970 refutes on historical and typological


grounds that the VSO order for English and German is the basic
Germanic order of éLements. Instead he classifies "the Proto-
Indo-European and the Early Indo-European dialects including
Proto-Germanic" as SOV languages (op. cit. p.2.). A syntactic
change has occurred in Germanic languages so that the former
Proto-Germanic SOV order has become the Modern Germanic SVO
order. McCawley's 1970 claim of a basic VSO order for English
is, in other words, rebutted also by historical evidence,be­
cause it is incompatible with the historical development of
Germanic syntax.

Fourth, even from an inherent generative semanticist point of


viewj McCawley's 1970 claim of a VSO order for Modern English
is inappropriate for generating a certain type of surface
structures which contain indirect objects.' Since no VP node
exists in a generative semanticist VSO representation of the
predicate (V) and the arguments (NP's), and since it is the
11

node S alone which directly dominates the V and NP's (cf. McCaw-
ley 19 70:295 and Dahl 1971:53), there is no non ad-hoc way to
block the movement of the indirect object NP {Bill) in (ii) by
the NP Fronting Rule which correctly moves the NP John into its
subject position:

^ I tycks John [ ha [ givit John Bill pengarna] ] ]


o Si S9 S9S,S„
12 2 1 o

After Subject and Predicate Raising in the generative semanticist


sense the following shallow structure results:

(ii) [»tycks ÏJohn i'[*ha givit 'Bill'pengarna] ]


So " S1 SlSo

The NP Fronting Rule now moves both John and Bill before the verbs
tycks and ha givit and the following ill-formed surface structure
is generated:

(iii) *John tycks Bill ha givit pengarna.

How can the NP Movement Rule be blocked from operating on the


indirect object NP node, if no dominating VP node is there in
the underlying representation to specify the difference between
the subject NP node {John) and the indirect object NP node
(Bill)?

The four arguments presented here, that is, the psycholinguistic,


the typological, the historical, and the inherent generative
semanticist arguments, are overwhelming evidence against the VSO
order as basic syntactic order of elements in Germanic languages.
For comparative syntax, then, little is gained by postulating
a semantic VSO order in Germanic languages.8

It is often a question of one's personal taste and previous


exposure to linguistic theories which makes one select a
given model of description for one's investigation of a given
surface structure. The problem with the current theories is
that no single one of them can be said to be exhaustive, un-
contradictory, or explanatory. We are convinced, however, that
working within a given theory yields better results than
working with no theory at all. Without a syntactic or semantic
theory of some kind, our observations are bound to become
superficial and will only meet the requirement of observational
adequacy. The different theories mentioned above are claimed
to meet the requirement of descriptive or explanatory adequacy
in the sense of Chomsky 1965. It is unsure whether they all meet
the latter requirements.

The present investigation of verbal complementation in Modern


Swedish was started within the framework of the Standard Theory.
As a great number of syntactic problems in connection with the
ACI-Constructions could not be adequately solved within that
12

theory (e.g. the ACI-Constructions which undergo relativiza-


tion in (20), section 2.2 and the Nominative-with-Passive
Infinitive Constructions in (36), section 4.2) a more accurate
theory was called for.

Through the theoretical contributions of both syntacticists and


s emanti eis ts a more usable theory for describing in verbal comple­
ment in Germanic languages has come to light. In the present
investigation a combined syntactic and semantic framework has
been used and has turned out to be the best model for de­
scribing ACI-syntax. A pragmatic approach to describe surface
structures is valuable for dealing with linguistic diversity
of surface structures. Here we are not so much involved with
the questions of whether there is such a th ing called Deep Struc­
ture,^ whether transformational rules have psychological
reality,10 0r whether there is d irectionality or not in the
process of derivation.il Rather, we are more concerned with an
exhaustive linguistic description of a certain type of surface
structure in Swedish with comparative aspects to other
Scandinavian and West-Germanic languages in order to contribute
to a comparative grammar of Germanic syntax on a generative
bas is.

The comparative study of surface structures must draw upon the


results gained in syntactic research during the past fifteen
years. There is no return to the slot-filling grammars of the
fifties. If we compare our approach with the current syntactic
theories, we feel that the Extended Standard Theory modified
by some of Chafe's 1970 concepts of semantics and Perlmutter's
1971 level constraints constitute the best theoretical frame­
work for dealing with the process of infinitivization and
embedding in general. On one hand, to use a one-sided semanticist
approach to ACI-syntax, starting with abstract semantic trees
of predicates and arguments containing semantic primitives has
not yielded the tangible comparative results which we are
striving for, i.e. to give an overall description of the
generation and use of ACI-Constructions in Modern and Old
Swedish with interlinguistic comparisons to other Germanic
languages. On the other hand, a one-sided syntacticist approach
in the sense of Chomsky 1965 to ACI-Constructions is also in­
sufficient for adequate descriptions of what native speakers
know about the generation and use of ACI-Constructions in their
respective Germanic languages.

Rather,we are constructing a comparative model of syntax which


will be applied to synchronic problems in the present de­
scription of verbal complementation in Germanic languages
(particularly Swedish, German, and English). Swedish will be
the point of departure for describing the complement structures
of the two other Germanic languages. The comparative goal of
describing verbal complements motivates an abstract approach
to syntactic surface structures in terms of common Germanic
deep structures and common Germanic transformations. The tasks
of this investigation are as fol lows: first, to demonstrate
what is common Germanic and what is language-specific from a
synchronic syntactic point of view; secondly, to demonstrate
the change in syntactic domain of the Subject Raising Rule
13

which has taken place between the ancient and modern Germanic
languages (cf. Ureland 1972b (forthcoming)). It is believed
that a comparative approach to syntax will give us results of
the same importance regarding the syntactic structures and
rules of the Germanic languages as did a comparative approach
to Germanic and Indo-European phonology and morphology in the
past century (cf. Rask 1818a and 1818b, Grimm 1819-1837,
Verner 1877, Bopp 1816, Paul 1920, Schmidt 1872, Schleicher
1848-1850 etc).

1.6 INTUITION-ORIENTED LINGUISTICS AND EMPIRICALLY-ORIENTED


LINGUISTICS
If syntactic research is to yield tangible and interesting
empirical results, aspects of the use of transformational rules
must be included in the study of a given language. Linguistic
research carried out within an intuitive framework of in­
vestigation constitutes an important first approach to the
study of syntax, but empirical observations based on sentences
used in natural linguistic and social situations are in-
dispensible for describing the syntax of natural languages.

One-sided intuition-oriented syntactic research leads to


artificial and lifeless hypotheses about the syntax of a given
language. Compare the unnatural and concocted sentences, for
instance, which are being published as crucial evidence for
theory construction in the avantegarde generative periodicals,
in which the majority of linguists avoid any aspect of language
use.I2 The overemphasis on describing language competence has
resulted in a neglect in one of the camps of American linguistics
in describing language usage. Chomsky's claim that a generative
theory is primarily a theory about a native speaker's linguistic
competence and not his use of the language has influenced a
great proportion of young linguists so much that aspects of
natural sentences have been forgotten or neglected. The
exploitation of intuitive judgements of grammaticality for
describing syntax was such a new approach to the description
of syntax that intuition-based research has been the primary
approach., to syntax during the last decade in generative
grammar.

However, there is another very important development within


American linguistics which promises more empirically-oriented
results. The works of such sociolinguists as Labov 1966 and
19-70, Wolfram 1969, Bailey 1970 and 1972 etc have demonstrated
the importance of including sooiolinguistia aspects for de­
scribing American-English phonology, morphology, and syntax.
These linguists constitute revivors of studies at a time when
such interest tended to be regarded as inferior because of the
predominant concern with constructing a theory of competence,
i.e. during the sixties.

Psycholinguists such as B ever 1970, McNeil 1966, Slobin 1968


etc belong to a group of American psycholinguists who are also
working with authentic sentences observed in various psycho-
linguistic experiments and test situations, that is linguistic
14

performance is their primary concern for constructing a psycho-


linguistic theory.

In neuro Unguis ties, observations of the native speaker's


linguistic behavior are indispensible for inferring various
brain functions, localizing the motor and sensory centers in
the cortex, studying the defects in linguistic competence which
must be claimed to be part of the central and peripheral
language systems (cf. the works of Luria 1966, Luria &
Tsvetkova 1968, Whitaker 1970 and 1971, Weigl & Bierwisch
1970, Gazzaniga 1970 etc).

Linguistic data of performance do determine the kind of hypo­


theses which these socio, psycho» and neurolinguists are
trying to construct to explain man's social, psychological, and
neurological behavior. For them data-oriented approaches are the
only acceptable approaches. However, if one reads the works of
Chomsky and his adherents, theory construction seems to be unr.
determined by such data, since most generative linguists are
not concerned with the psychological or sociological aspects
of language use, at least not in a direct way. (Cf. Chomsky's
1969 remarks on how linguistic data do not determine "a de­
scriptively adequate grammar" of a given language (op.cit. p.22,
footnote 8)).

By studying the process of embedding in a representative corpus


of Modern Standard Swedish, the corpus of one million words
computerized by the Research Group for Modern Swedish in Gothen­
burg (cf. Allén 1970 and 1971), relations between rules of
embedding have been disclosed which otherwise would have remained
undiscovered.14 we have in this corpus empirical evidence for
evaluating the productivity of the syntactic rules which are in
the focus of our interest, i.e. the rules of nominalization and
embedding. We are interested in the productivity of the Subject
Raising Rule for generating ACI-Constructions (Accusative-with-
Infinitive Constructions), in the att or hur-Complementizer
Embedding Rule for generating att or ftwr-embedded structures,
in the Nominalization Rule for generating nominals (nomina
actionis). How often does Subject Raising or Complementizer
Embedding occur in comparison to Nominalizations? The statistics
given in the sections of this investigation of course refer
only to the newspaper texts processed by the Gothenburg research
group. Since newspaper language is representative of a neutral
way of writing we believe that the distribution of embedding
types described in percentages here is representative of normal
usage of written Modern Standard Swedish subordination after
the verba sentiendi and dicendi.15

The combination of the two procedures of syntactic investigation,


that is the intuition-based approach and the data-based approach
"must be mastered by anyone who hopes to do significant linguistic
analysis" (cf. Labov 1970:41).
15

2, PRESENTATION OF MODERN SWEDISH ACI-CONSTRUCTIONS


We will describe and discuss such surface structures which are
represented in (1) through (25). Each native speaker of Swedish
has a subconscious knowledge of these sentences and their struc­
tures. What underlying representations and syntactic rules must
be postulated in order to explain a native speaker's competence
to produce and understand such surface structures? (For a detail­
ed discussion of the subclassification of the verbs involved
see sections 3.2 through 3.8).

2.1 IN DECLARATIVE SENTENCES


Class (A) Verbs of Perception (verba sentiendi)

SE 'see'
(1) Palme såg byggnadsarbetare demonstrera i New York City,
(Palme saw construction workers demonstrating in New York
City.)

HÖRA 'hear'
(2) Palme hörde dem ropa slagord mot Sverige.
(Palme heard them shouting slogans against Sweden.)

KÄNNA 'feel'
(3) Han kände pulsen slå fort.
(He felt his pulse beating fast.)

Class (B) Unguis tio-Performative Verbs (verba dioendi and


verba putandi)

SÄGA 'say' [+Performative]


(4) Palme sade sig göra en resa till USA som privatperson.
(Palme said that he made a journey to the US as a private
person.)

ANSE 'consider' [-Performative]


(5) Svenska radikaler ansåg Palme göra en fjäskresa till USA.
(Swedish radicals considered that Palme made a trip of
courtship to the US.)

TRO 'believe' [-Performative]


(6) Han trodde sig göra sitt bästa för Sverige.
(He believed that he did his best for Sweden.)

TYCKA 'be of the opinion' [-Performative]


(7) Han tyckte sig göra god reklam för Sverige.
(He thought that he made good propaganda for Sweden.)
16

Class (C) (verbs of instantaneous action)

FINNA1 'find'
(8) Palme fann en söt studentska sitta och lyssna på en radikal
talare.
(Palme found a beautiful student sitting listening to an
SDS speaker.)

Class (D) (actuality and stative verbs)

VETA 'know'
(9) Han visste sig ha gjort rätt mot sina föräldrar.
(He knew that he had been fair to his parents.)

Class (E) (causative verbs)

LÅTA^ 'allow' (-permissive-causative [+Caus^])


(10) Hon lät mannen komma in i sovrummet.
(She allowed the man to enter her bed room.)

BEDJA 'ask portative-causative [+CauS2])


(11) Kvinnan bad mannen hjälpa sig.
(The woman asked the man to help her.)

LÅTAq 'cause'(ins tigating-causative t+Caus^].)


(12) Han lät fångarna bygga ett palats åt sig
(He had the prisoners build a palace for him.)

LÄRA^ 'teach' (instructive-causative t+Caus^])


(13) Tysken lärde honom att tala tyska.
(The German taught him to speak German.)

HJÄLPA 'help' {auxiliary-causative^ Caus^].)


(14) Han hjälpte honom(att)tvätta bilen.
(He helped him to wash the car.)

Class (F) Reflexive Verbs

FÖRKLARA SIG 'say'


(15) Han förklarade sig vara införstådd med saken.
(He said that he agreed on the matter.)

LÄRA2 SIG 'learn'


(16) Han lärde sig att tala nordsamiska.
(He learned to speak North Lapp.)

UPPGE SIG 'say'


(17) Han uppgav sig vara tysk.
(He said he was German.)
17

TÄNKA SIG 'think, imagine'


(18) Han tänkte sig kunna utföra arbetet.
(He thought that he could do the job.)

VISA SIG 'turn out to'


(19) a. Han visade sig vara en bedragare.
(He turned out to be an imposter.)

VÄNTA SIG 'expect'


(19) b. Nixon väntade sig bli mottagen med jubel i Kina.
(Nixon expected to be received with cheers in China)

2,2 IN SUBORDINATE RELATIVE CLAUSES


The Relative Rule can operate on ACI-Constructions and embed
them into another sentence by moving the object noun phrase
to the front position of the sentence. We will not enumerate
all the possible relative clauses of the ACI-verbs mentioned
above, but will instead concentrate on those relativized ACI-
Constructions which occur after performative verbs {verba
dioendi) and verbs of opinion {verba putandi). The generation
of these surface structures will be discussed in more detail
in Part Three since they are heavy arguments against the
Chomskian claim that transformational rules operate locally on
contiguous trees in a derivation. In order to derive the
following relativized ACI-Constructionsa Global Rule Process
is assumed (cf.Global Rule II in Part Three).16

SÄGA 'say'
(20) a. Mannen^ som^ Peter^ sade vara sjuk hette Gustav.
(The man- w ho. Peter, said (that he.) was ill was:called
Gustav.) 1 3 1

PASTA 'assert'
b. Kvinnan^, som- Inger, påstod vara tyska var engelska.
(The woman, wno- Inger- claimed to be German was an
English lady-)

TRO 'believe'
c. Hjälten^ som- mannen, trodde vara en förrädare var
en sann patriot. 3

(The hero^ who: the man- believed to be a traitor was


a true patriot.) ^

TYCKA 'be of the opinion'


d. Pojken- som, flickan, tyckte vara för ung hette Gustav.
(The boy. who- the girl. thought was too young was
called Gustav.)
18

2,3 IN INTERROGATIVE SENTENCES

Similar movements of noun phrases are carried out by the Inter­


rogative Rule, if the noun phrase which is to be interrogativized
does not occur in the sentence-initial position. In the following
sentences it is the subject noun phrases which undergo the Inter­
rogative Rule, so no movements are necessary here. In Part Three
we will show some constraints on Wh-Movement which are due to
constraints on the Subject Raising Rule. Here only verba dioendi
and putandi are exemplified as occurring with ACI-Constructions
which have undergone Wh-Insertion:

SÄGA 's ay'


(21) a. Vem-sade sig. vara sjuk.
(Who^ said that he^ was ill.)

b. Vilken man^ sade sig. vara sjuk.


(Which mani said that he^ was ill.)

c. *Vilken man^ sade honom, vara sjuk.


(Which man^ said that héj was ill.)

PÅSTÅ 'claim'
(22) a. Vem- påstod sig. vara sjuk.
(Who^ claimed that he^ was ill.)

b. Vilken, man^ påstod sig^ vara sjuk.


(which man^ claimed that he^ was ill.)

c. *Vilken man^påstod honom, vara sjuk.


(Which man^claimed that nej was ill.)

TRO 'believe'
(23) a. Vem- trodde sig. vara sjuk.
(Who^ believed himself^ to be ill.)

b. Vilken man- trodde sig^ vara sjuk.


(Which man^ believed himself^ to be ill.)

c. *Vilken man. trodde honom.vara sjuk.


(Which man^ believed that^he^ was ill.)

TYCKA 'be of the opinion'


(24) a. Vem- tyckte sig. vara sjuk.
(Who^ thought that he^ was ill.)

b. Vilken mani tyckte sig. vara sjuk.


(Which man^ thought that he^ was ill-

c. *Vilken man. tyckte honom, vara sjuk.


(Which man^ thought that nej was ill.)
19

TÄNKA 'think'
(25) a. Vem. tänkte sig. kunna utföra arbetet.
(Who^ thought that he^ could carry out the job.)

b. Vilken man^ tänkte sig. kunna utföra arbetet.


(Which man^ thought that he^ could carry out the job.)

c. *Vilken man^ tänkte honom, vara sjuk.


(Which man^ thought that he^ was ill.)

All well-formed ACI-Constructions exemplified in (21) - (25)


contain reflexive object NP's. The verbs involved here are
subject to an Equal Subject Constraint on the generation of
ACI-Constructions which will be described in detail in Part
Three (cf. also Ureland 1970). Under special circumstances the
requirement of coreference linkage can be diminished, so that
also noncoreferent complement subject NP's can undergo Subject
Raising and occur in ACI-Constructions, although they are not
coreferent with the matrix subject NP (cf. for instance exampl
(20a)-(20d) above).

However, the verb ANSE 'consider' is not constrained by the


Equal Subject Condition on Subject Raising as noncoreferent
ACI-Constructions occur, e.g. Vem^ ansåg honom, vara sjuk 'Who
considered him. to be ill'. ^
20

3, CLASSIFICATION OF ACI-VERBS IN MODERN SWEDISH


In this section the 29 ACI-verbs of Modern Swedish will be
discussed and classified. First, an alphabetic list of Swedish
ACI-verbs will be given covering both reflexive and nonreflex-
ive ACI-verbs (section 3.1). Then in section 3.2, a subclassi­
fication of Swedish ACI-verbs will be carried out on the basis
of both semantic and syntactic criteria.^ The criteria used
for the semantic subclassification will be used not only for
Modern Swedish ACI-verbs, but for ACI-verbs of other Germanic
languages as well, both modern and ancient. We are here
developing a panchronic and a pandialectal mechanism for
handling the semantic subclassification of ACI-verbs in Germanic
languages in a comparative syntactic approach.

3,1 AN ALPHABETIC LIST OF ACI"VERBS IN MODERN SWEDISH

1) ANSE 'consider' 17) OBSERVERA 'observe'


2) FINNA1 ,find
1, discover' 18) PASTA 'claim'
3) FINNA2 'find2, think' 19) SE^ 'see visually'
4) FÖRKLARA (SIG) 'declare, say' 20) FA SE-^ 'catch sight of'
5) HJÄLPA 'help' 21) SÄGA 'say'
6) HÖRA^ 'hear auditorily' 22) TRO 'believe'
7) FÅ HÖRA^ 'hear suddenly' 23) TYCKA 'be of the opinion'
8) KANNAX •feel1 tactually' 24) TÄNKA1(SIG) 'imagine'
9) KÄNNA2 'smell' 25) UPPGIVA (SIG) 'state, say'
10) KÄNNA3 'feel3 mentally' 26) UPPTÄCKA 'discover'
11) KÄNNA^ *feel4 somatically' 27) VETA 'know'
12) LATAX 'allow' 28) VISA (SIG) 'turn out'
13) LÄTA2 'cause' 29) VÄNTA (SIG) 'expect'
14) LÄRA1 'teach'
15) LÄRA2(SIG) 'learn'
16) MÄRKA 'notice'

In the alphabetic list two occurrences of FINNA, four occurrences


of KÄNNA, two occurrences of LÅTA,and two occurrences of LÄRA can
be observed. We will treat these homonymous verbs as different
verb classes as there are syntactic as well as semantic reasons
for distributing them into various subclasses. KÄNNA, 'feel
tactually' is for instance associated with a different set of
syntactic and semantic features than KÄNNA^ 'feel mentally'.
21

3.2 SEMANTIC SUBCLASSIFICATION OF ACI-VERBS

In order to describe the total range of the semantic classes


of ACI-verbs- discovered in ancient and modern Germanic languages,
about"fifteen subclasses have turned out to be necessary as a
classificatory division. However, examples for all subclasses
have not been found in each Germanic language; for instance, the
two semantic subclasses of the verbs of auditory and visual
perception (la) and (2a) are not represented by any Old Swedish
ACI-verbs. Modern Swedish still lacks auditory verbs of purpose
on the part of the experiencer, i.e. ACI-verbs of subclass (la)
below.

For the sake of syntactic comparison and descriptive convenience


the 29 ACI-verbs of 3.1 (we count FINNA, LÅTA, and LÄRA as tw o
ACI-verbs each and KÄNNA as four ACI-verbs depending on the
linguistic context) can be subcategorized into five major
semantic classes and one syntactic class according to fundamental
semantic properties of the verbs and a surface structure con­
figuration, i.e. into Class (A), Class (B), Class (C), and Class
(D) by means of semantic properties and into Class (F) by means
of a syntactic surface structure configuration, i.e. the occurr­
ence of a reflexive pronoun.

We start the major subclassification of verbs by writing tfye


following rule:
SUBCATEGORIZATION RULE (i)
v * V
[+Subj Rais]

Rule (i) delimits the number of verbs to encompass those verbs


which are marked as undergoing the Subject Raising Rule and
which are in the focus of the present investigation. The feature
[+Subj Rais] is a rule feature and is here claimed to bean
inherent syntactic feature of the verbs in question, i.e. the
verbs enumerated in 3.1. Rule features have been shown to bean
adequate means to describe syntactic processes (cf. G. Lakoff
1965, Rosenbaum 1967a and 1967b, and R. Lakoff 1968). The verbs
marked for the major syntactic feature [+Subj Rais] are subclassi-
fied in their turn by a number of minor semantic feature.Thus,
in our description, four kinds of features are necessary for sub-
classifying the lexical items which are in the focus of our
interest: a) the categorical feature [+V], b) the rule feature
[+Subj Rais], c) semantic features such as [+Perceptive],
?+Performative], [+Causative] etc, and d) syntactic surface
features such as [+Reflexive] .19,20 ^he semantic feature [+Fac-
tive] is a precondition for the application of Subject Raising.^1
In the lexical entry of ACI-verbs this feature need not be
indicated, since a redundancy rule like [+ Subj Rais] [-Factive]
will make such a notation superfluous. The feature [Subi Rais]
must, however, be a fe ature marke d for each ACI - v e r b . 2 2 Without
the com bination of these four kinds of features, an adequ ate sub-
classification of Swedish ACI -verbs would not be po ssible.
22

Subcategorization rule (ii) expands each Modern Swedish verb


marked for Subject Raising into one of six further subclasses:

SUBCATEGORIZATION RULE (ii)

ACI-Class Examples in Modern


Swedish
+V
+Perceptual
I (A) HÖRA 'hear auditorily'
SEi 'see visually'
KÄNNAx 'feel'

(B) [+Perform]
PASTA 'claim'
+V SÄGA 'say'
UPPGIVA 'report'
+Linguistic
^(Performative
[-Perform]
ANSE 'consider'
TRO 'believe'
TYCKA 'be of the
[+Subj Rais] opinion'

+V
-Linguistic (C) FINNA1 'find1'
•••Instantaneous

•Linguistic (D) VETA 'know'


•Stative

LATAX 'allow'
•»•Causative (E) BEDJA 'ask'
^^Performative LÄRAi 'teach'

(F) LÄRA2(SIG) 'learn'


r+Reflexive
+v i TÄNKA, (SIG) 'imagine'
VISACSIG) 'turn out to
. be'

In the Modern Swedish lexicon each member of the six verb classes
is marked for Subject Raising, i.e. each verb expanded by rule
(i) and subclassified by rule (ii) takes object complements
consisting of an(object) noun phrase with infinitive construction
(=ACI-Complement ).

Class (A) contains verbs of physical perception {verba s entiendi)


such as Modern Swedish HÖRA^ 'hear', SE^'see', and KÄNNA^'feel'.
23

Class (B) is a cover term for the verbs expressing linguistic


acts of two kinds: a) performative acts^3 such as Modern
Swedish SÄGA 'say', UPPGIVA 'report1, and PÄSTÄ 'claim' ( verba
dioendi) and b) nonperformative linguistic processes as for
instance Modern Swedish ANSE 'consider', TRO 'believej and
TYCKA 'be of the opinion' (=verba putandi). Class (C) stands
for the verbs expressing an instantaneous mode of action
Aktionsart ), e.g. Modern Swedish FINNA^ 'find^, discover'
(=verbs of an instantaneous action). Class (D), on the other hand,
lists the actuality verb of a continuous (stative) mode of
action (=verba stativa), e.g. Modern Swedish VETA 'know'. Class
(E), contains verbs of various kinds of causation as for instance
Modern Swedish LÅTA^ 'allow', BEDJA 'ask', LÄRA^ 'teach' etc
{-verba causativa). Class (F), finally, is a cover term for
a class of ACI-verbs which is called the reflexive verb class.
The ACI-verbs of this class cannot cooccur with other object
noun phrases than those which are coreferent with the matrix
subject, i.e. the raised object noun phrases must undergo re-
flexivization: LÄRA«(SIG) 'learn', TÄNKA (SIG) 'imagine', VISA
(SIG) 'turn out to Be'.

3,3 VERBS OF PERCEPTION(CLASS (A))


A general survey of the ACI-verbs in various Germanic languages
reveals that each Germanic language has verbs of perception
cooccurring with ACI-Constructions as noun phrase complements
in the sense of Rosenbaum 1967. The verbs of Class (A) express
four of the five modalities or senses of man. The process of
hearing, seeing, feeling, smelling, and tasting is expressed in
all Germanic languages by means of verbs which are phonologically
different, but historically related. They have the common seman­
tic core of denoting the five modalities enumerated.24

Besides the phonological and semantic characteristics in common,


there is a common syntactic characteristic which all Germanic
verbs of perception share. These verbs can occur with ACI-Con­
structions as embeddings in all Germanic languages. The
complement structure of the modality verbs is often an ACI-
Construction, although other surface structures may occur as
embeddings as well, e.g. att-clauses, hur-clauses, (possesive)
pronouns plus gerundives, which are rare in Swedish and German
but extremely frequent in English, or nominalizations.
24

In Modern Swedish the processes of hearing, seeing, smelling,and


feeling are expressed by means of verbs which can be subclassi-
fied by the following rule:

SUBCATEGORIZATION RULE (iii)

ACI- Modern Swedish Examples


Subclass

+Process ~| (la) R-Purposel


•Auditory! [•Durative]
•iPurpose :
•IDurativej
(lb) -Purpose 1 pPurpose 1
•DurativeJ LrDurativel
HÖRA. 'hear HÖRA, 'hear
auditorily' auditorily'
FÅ HÖRA, 'hear
suddenly'
F+Process ~] (2a) (•Purpose 1
•Visual 1+DurativeJ
fcHPurpose IAKTTAGA*'watch'
jt|Durative OBSERVERA*
'observe'
(2b) -Purpose"] p-Purposel
•Durative] [^Durative]
SE,'see SE, 'see
visually' visually*
FA SE, 'catch
sight of'
MÄRKA *'notice'
UPPTÄCKA *
'discover
visually'
[•Process-] (3a) pPurpose ~] p-Purposel
[•TactualJ L+ Durative! b Durati vej
KÄNNAt 'fêel KÄNNA, 'feel,
tactuallv' tactually'
[•Process ~| (3b) pPurpose 1 [-Purpose *|
[+01factor)j L+DurativeJ bDurative]
KÄNNA2 'smell'KÄNNA2 'smell'
pProcessl (3c) p-Purpose I
[•Mental J L+DurativeJ
KÄNNA3(SIG)
'feelj mentally'

•Process (3d) PPurpose "1


+Somatic [•Durative],
KÄNNA4(SIG)
'feel somati­
cally'
\
25

The starred durative perceptual verbs IAKTTAGA and OBSERVERA


and the starred instantaneous perceptual verbs MÄRKA and UPP­
TÄCKA^ do not normally take ACI-Constructions as verb comple­
ments, although this kind of complement structure is possible
after these verbs.

HÖRA and SE can also express durative auditory and visual


perception respectively on the part of the experiencer, pro­
vided the complement sentence contains a durative Aktionsart
(cf. the Aspectual Copying Rule discussed in sections 2 and 3
in Part Two). To capture this additional durative usage of
SE-, and HÖRA, as exemplified by Han såg henne klä på sig 'He
saw her getting dressed' and tian hörde he nne komma i trappan
'He heard her climbing the stairs' the matrixP-Purposel has
[+Durative|
been placed in (lb) and (2b).

It is not quite true that Swedish SE is only used as a verb of


non-intentional passive perception which takes only an
experiencer as the human deep subject, because there are examples
where SE denotes intentional active perception as well as in
Jag såg och såg och vartân jag såg, så såg jag såg vid såg 'I
looked and looked and wherever I looked I saw one saw-mill
after another'. In English the verb LOOK must be used for
expressing visual perception which is intentional, active and
which takes a human agent.

The intentional verb SE could have been classified into Sub­


class (2a) together with IAKTTAGA 'watch' and OBSERVERA 'observe',
if it had been an ACI-verb. However, this verb complement can­
not occur after the intentional SE, so we need not include it
in our list under Subclass (2a).

The feature [Purpose] is positively specified for the subclasses


(la) and (2a) covering such verbs as for instance LISTEN TO-r
and OBSERVE in Modern English which take ACI-Constructions.
No Modern Swedish representatives of subclass (la) which take
ACI-Constructions have been retrieved. Two verbs of intentional
mode of perception on the part of the experiencer have been
registered for Modern Swedish as taking ACI-complements, i.e.
IAKTTAGA 'observe' and OBSERVERA 'observe'. Some native speakers
of Swedish accept only att-Embeddings after these two verbs,
however.

The mode of perception is negatively specified for the feature


CPurposeJ in subclasses (lb) and (2b) which contain HÖRA 'hear'
and FÄ HÖRA 'hear, overhear' in the list of perceptual verbs.
The feature [Purpose] is needed to distinguish various verbs
of seeing and hearing from each other in terms of presence or
nonpresence of intention on the part of the experiencer. It may
be an accidental gap in the Swedish language that subclass
(la) is not represented by an ACI-verb, e.g. by LYSSNA PÅ
'listen to' as is subclass (2a) in the dialects of at least
some speakers.
26

Two Modern Swedish verbs have been listed in (lb) and (2b) as
taking ACI-complements for which there are no direct corre­
sponding verbs in Modern English or German: FÄ HÖRA 'hear'
and FÅ SE 'see, catch sight of'. These two verbs of auditory
and visual process respectively can be used only in a non­
stative sense. They both denote a sudden nondeliberate mode of
auditory or visual perception which is in contrast with for
instance the two visual verbs IAKTTAGA and OBSERVERA just
discussed. In subclass (2b) two mòre nonstative verbs of perception
have been listed, i.e. MÄRKA 'notice' and UPPTÄCKA 'discover'.
(MÄRKA 'notice' might also be subclassified into (lb), since
something may also be noticed by means of auditory stimuli).
The inherent semantic aspectual feature [Durative] is in­
dispensable for blocking a great number of ACI-Constructions
in Germanic languages.

Since no verbs of smell [+01factory] or of tactual feeling


[+Tactual] are used in Swedish with a distinction between in­
tentional or non-intentional smelling of feeling, the feature
[Purpose] is irrelevant. As an illustration we shall compare
the following Swedish sentences with their English translations.
In English it is possible to use FEEL and SMELL in imperative
clauses, which is impossible with KÄNNA and LUKTA in Swedish.
In imperative clauses KÄNNA PÄ and LUKTA PÄ must occur as in
Känn på tygevl 'Feel the cloth!' but not *Känn tygetl 'Feel
the cloth'.'; and Lukta på den här blommani t Smell this flower!'
but not *Lukta den här blomman i'Smell this flower!'. 26,27

Besides the tactual and olfactory use of KÄNNA in Swedish,


third and fourth meanings of KÄNNA have been observed in Modern
Swedish, i.e. KÄNNA-(SIG) 'feel mentally' and KÄNNA4(SIG) 'feel
somatically).
The latter two verbs classes take reflexive pronouns plus in­
finitive or predicate adjective constructions as complement
structures. Since the reflexive pronouns are transformationally
derived from complement NP's we prefer to subclassify KÄNNA3
'feel mentally' and KÄNNA4 'feel somatically' together with
KÄNNAi 'feel tactually' and KÄNNA2 'smell'; there are semantic
reasons for classifying the four different types of the homonym
KÄNNA into the same class of verba sentiendi, i.e. Subclass (3).
It is only the specific kind of perception which distinguishes
them from each other. The syntactic characteristic which they
also share is t he fact that they all undergo Subject Raising
and take ACI-Embeddings. The semantic and syntactic properties
mentioned here are properties which belong to the deep structure.
There are, however, also surface structure reasons for classi­
fying KÄNNA^ and KÄNNA, into a reflexive subclass, i.e. Subclass
8 of section 3.8, which enumerates verbs which take only reflex­
ive ACI-Constructions, that is there must be coreference linkage
between the matrix and complement subject NP's before Subject
Raising can apply. Compare Han^ känner3 sig i vara inspirerad
'He feels inspired' with *Han± känner3 henne 1 vara inspirerad
'He feels that she is inspired' and Hani känae4 sigi vara illa­
mående 'He felt sick' with Hankände 4 hennevara illa-mående
'He felt her to be sick'. This coreference constraint on the
occurrence of the ACI-Construction is a constraint which also
27

other sets of verbs share, that is the performative verbs FÖR-


KLARA(SIGJ 'declare', and UPPGIVA(SIG) 'say, state', the mental-
probability verbs TÄNKA(SIG) 'imagine', and VÄNTA(SIG) 'expect',
the instructive-causative verb LÄRA(SIG) 'learn', and the
intransitive reflexive verb VISA(SIG) 'turn out'.

All these verbs will be enumerated by subcategorization Rule


(viii) in section 3.8, since they all require a reflexive
pronoun as an object NP in an ACI-Construction. The obligatory
reflexive verb complement after these verb classes in the surface
structure motivates a cross-classification, since we also want
to clarify that these verbs are reflexive verbs. A deep structure
subclassification would not be able to capture this surface
structure property. The verbs KÄNNA., and KÄNNA^ therefore belong
semantically, from a deep structure point of view, to Subclasses
(3c) and (3d), whereas from a surface structure point of view
they also belong to Subclasses (8e) and (8f), since they are
obligatorily reflexive verbs.

The verbs IAKTTAGA 'watch' and OBSERVERA 'observe' have been


starred in Class (2a), since not all native speakers of Swedish
accept ACI-Constructions as verb complements after these verbs.
The same idiolectal or stylistic difference is found in sentences
with MÄRKA 'notice' and UPPTÄCKA 'discover' as main verbs.

It is clear that the role of the subject noun phrase in sentences


which contain HÖRA 'hear', SE 'see', KÄNNA., 'feel tactually' etc,
on the one hand, is very different from the role of the subject
noun phrase of IAKTTAGA 'watch' and OBSERVERA 'observe' on the
other. If we follow Fillmore's 1971 or Chafe's 1970 terminology,
the former group of verbs can be said to take an Experiencer,
whereas the latter group of verbs takes an Agent as the subject
noun phrase. Such a classification in terms of the role which the
deep subject noun phrase plays is valuable in describing the
semantic difference between the two groups of verbs. In this
way the fact is stressed that in the case of SEE, HEAR, and FEEL
the deep subject NP is passively and indeliberately experiencing
something visible, auditory or tactual, whereas in the case of
OBSERVE, WATCH etc the deep subject NP more actively takes part
in the act or process of perception, that is as an Agent in
Fillmore's and Chafe's sense.

As indicated in Footnote 26 such a classification in terms of the


roles played by the subject NP's is not easy to formalize in the
subclassification of verbs as attempted in Subcategorization Rule
(iii). Some kind of a redundancy rule could be elaborated which
would automatically assign the Experiencer role to the Subject NP,
if the feature [-Purpose] is present in the verb of perception
(SEE, HEAR, FEEL etc). The Agent role would be automatically
assigned to the subject NP, if the feature [+Purpose]is present
in the verb of perception (OBSERVE, LISTEN TO etc).
28

REDUNDANCY RULE
S
+V o
•> [+Experiencer] /— +Perceptive
NP -Purpose
VP

REDUNDANCY RULE
S
o
S +V
(ii) [+N]° >[+Agent] / — +Perceptive
NP +Purpose
VP

Redundancy rules (i) and (ii) are written to operate on the


syntactic deep structure configuration, so that on the deep
structure subject NP, i.e. [+Nfso , the role of Experiencer or
NP
Agent is assigned according to the presence of the feature
[^Purpose] in the matrix verb. (For heuristic reasons the two
rules are formalized in a redundant manner).

A less redundant notation of the two redundancy rules would be


the following collapsed rule:

REDUNDANCY RULE
+V
(iii) [+N]° ExPeriencer] / — +Perceptive
aPurpose
NP
29

3,4 PERFORMATIVE VERBS AND MENTAL VERBS (CLASS (B))

The linguistic ACI-verbs of Class (B) can be further subclassified


by Subcategorization Rule (iv) into three major subclasses
according to the inherent semantic properties of being performa­
tive and declarative (4a); nonperformative, mental, and probable
(4b); or nonperformative, mental, and expressing an opinion (4c).
The latter two classes are also characterized as probability
and opinion verbs respectively.

SUBCATEGORIZATION RULE (iv)


ACI- Modern. Swedish
Class Examples

[ +Performativej
+Declarative J
(4a) FÖRKLARA(SIG)'explain*
PÅSTÅ 'claim'
SÄGA 's ay'
UPPGIVA(SIG)'say'

[+V

-1
I pPerformativël
+Linguistic r>\ +Mental
l+Prob able J
) (4b)
TRO 'believe'
TÄNKA^(SIG)'imagine'
VÄNTA (SIG) 'expect'

[ Performative
•Mental
•Opinion
ivej
(4c)
ANSE 'consider'
TYCKA 'be of the
opinion'
FINNA2'find, think'

Subclass (4a) contains linguistic verbs which are called overt


performative verbs, whereas subclass (4b) contains verbs of
opinion. In Part Three the use and generation of ACI-complements
after these verbs will be discussed in more detail. The verbs
FÖRKLARA (SIG), UPPGIVA (SIG), TÄNKA (SIG), and VÄNTA (SIG)
have been classified into both subclasses (4a) and (4b), and
subclasses (8a) and (8b) below, since they are both performative
and reflexive verbs. A cross-classification is necessary here.
One verb may belong to one or more ACI-Classes depending on
what subclassificational criteria are being used. What we need
is one subclassification for the deep structure where the
semantic and syntactic properties are specified, and another
subclassification for the surface structure which tells us that
certain verbs are obligatorily reflexive verbs if ACI-Construc-
tions occur as verb complements.

Recent linguistic research has provided us with insights into


the principles underlying various speech acts. A hypothesis
concerning the use of sentences and their illocutionary force
has been but forth by Austin 1962, Searle 1969, Ross 1970b,
Fraser 19 71, and Andersson 19 71. In the so-called Performative
Analysis of sentences (PA), it has been contended that the
native speaker is aware not only of the syntactic structure
and the lexical meanings of the morphemes of a sentence, but
also of the force-multiplicity of sentences. 3The surfa ce
30

structure may not reveal any evidence of possible forces which


it carries, i.e. one and the same surface structure may be
interpreted as having two, three or more distinct illocutionary
forces, e.g.

(26) a. Peter may leave now.


b. Why don't you open the door.

In sentence (26a) the semantic reading is threefold ambiguous:


i) prediction of the future, ii) permission, and iii) probability.
Sentence (26b) may be interpreted as a normal negated question
or as a request. There is no overt performative verb of
questioning or requesting in (26b) which tells the listener what
force is intended by the speaker, but the force is nevertheless
in most instances interpreted correctly by native speakers of
English, although foreigners may misunderstand the illocutionary
force intended by the speaker. The interpretation of force is
inferred somehow by the native speakers, although the
Illocutionary Force Marker, i.e. the higher performative sentence
in (26b) 'I ask you' or 'I request' has been deleted by the
Performative Sentence Deletion Rule in the sense of Ross 1970:
249. What remains after that deletion is only the Propositional
Content Marker under the performative verb, i.e. the surface
structure of (26a) and (26b).

We will draw upon the Performative Analysis as introduced into


linguistic theory by Ross 197(band improved by Searle 1969,
Fraser 1971, and Andersson 1971 for our description of the ACI-
Constructions in Swedish which are preceded by explicit per­
formative verbs expanded by Subcategorization Rule (iv) above,
i.e. class (4a).

The use of the term 'performative' has changed considerably since


Austin 1962 coined it for describing certain linguistic speech
acts which were associated with certain nonlinguistic performances
such as baptizing, inauguration, marriage etc. Ross 1970 and
other American linguists have adapted the concept 'performative'
to their syntactic/semantic analyses of sentences in describing
the use of sentences. In connection with imperative sentences
Ross called a verb 'performative' if it had a given number of
characteristics: 1) the subject of a performative verb is the
first person singular (I), 2) the indirect object is in the
second person singular (you), 3) performative sentence is
always affirmative, 4) the tense is the present tense, 5) the
performative verb can always cooccur with the modal adverb
hereby, 6) there is only one performative verb for each per­
formative expression, and 7) the performative sentence is always
a single top sentence.*8

Although Ross's performative analysis has been refuted and


modified by other linguists (see the criticism by Fraser 1971
and Andersson 1971), the basic idea of introducing performative
verbs in order to describe the illocutionary forces of sentences
has remained a useful tool in linguistic descriptions of how
sentences are used.29
31

If we use the term performative' for some classes of Modern


or Old Swedish verbs as meaning verbs which can be used
performatively, we can express different kinds of illocutionary
acts manifested by various types of surface structures. Law
texts abound in explicit performative verbs, i.e. verbs of
promise, threat, statement, suggestion, etc. In most sentences
the explicit performative verb remains in the surface structure,
but very often the structure has undergone Performative Sentence
Deletion. In a special investigation of ACI-verbs in Old Swedish,
no more than 14 such performative verbs have been retrieved from
the Old Swedish texts. The difference to Modern Swedish is v ery
large, since the number of performative verbs which can in
Modern Swedish take ACI-complements is very small, e.g. FÖRKLARA
(SIG) 'declare, say1, PÂSTÂ 'claim', SÄGA 'say', and UPPGIVA(SIG)
'state'. (Cf. Ureland 19 72b (forthcoming)on the historical change
of syntactic domain between Old and Modern Swedish).30

The ACI-verbs enumerated in Class (4b) are so-called mental verbs


which often take a complement sentence which implies probability,
e.g. the verbs TRO 'believe', TÄNKA-^(SIG) 'imagine', and VÄNTA
(SIG) 'expect'. In Class (4c) the mental verbs of opinion such
as ANSE 'consider', TYCKA 'be of the opinion', and FINNA2'find2,
think' are enumerated. The verb complement after each of these
verbs expresses something which is the opinion of the subject NP.

A classification in terms of the roles which the subject NP's


play may be practical if the inherent semantic properties of
the verbs of the subclass (4a) are considered as well. A classi­
fication only in terms of roles is not sufficient for distinguish­
ing between for instance Class (4b) and (4c), where no clear-
cut borderline can be drawn between the Experiencer and the Agent
roles of the subject NP's in Fillmore's 1971 and Chafe's 1970
sense. For instance we can ask if it is the Agent or Experiencer
role which underlies the semantic function of the human subject
NP when TRO 'believe' is the main verb and if it is the Agent
or Experiencer role which underlies the semantic function of
the human subject in the case of ANSE 'consider'. It seems as
if the Experiencer role is the most easily distinguishable
among the human subjects of Subclasses (4b) and (4c), however.

Class (4a) is the performative and declarative verb class


(SÄGA 'say', PÅSTÄ 'claim' etc) which demands a deep structure
Agent in Fillmore's sense as the subject NP; Class (4b) is the
nonperformative but mental probability verb class (TRO 'believe',
TÄNKA1(SIG) 'imagine' etc) which demands that an Experiencer
occurs as the deep structure subject NP; Class (4c) is the
nonperformative but mental verb class of opinion (ANSE 'consider',
TYCKA 'be of the opinion' etc) which also takes an Experiencer
as the deep structure subject NP.

A classification purely in terms of the Experiencer role is


inadequate for distinguishing between (4b) and (4c). The semantic
subclassification in terms of the features [Probable] and
[Opinion] is therefore necessary and motivated in Subcategoriza-
tion Rule (iv) above.
32

3.5 INSTANTANEOUS-PERFECTIVE VERBS (CLASS (c))

In Class (C) expanded by Subcategorization Rule (ii) one non-


performative and nonlinguistic verb is listed as a Swedish ACI-
verb. i.e. FINNA^ 'find1. The mode of action (Aktionsart ) of
this verb is inherently instantaneous. This property of being
instantaneous implies that the verb is also perfective which
is specified by the following rule:

REDUNDANCY RULE (v) ACI-Sub- Mod. Sw.


ry -iclass Examples
|+Instantaneous| > f-Perfective) / |-Linguistiçj (5) piNNAi • findi'

Redundancy Rule (v) states the fact that any verb marked as
being nonlinguistic and instantaneous is also perfective. Whereas
the two former features are purely semantic features, the
feature [+Perfective] is a morpho-syntactic feature which blocks
the cooccurrence of specific morphemes. In some languages like
the Slavic ones an obligatory outspelling of the morpho-syntactic
feature must take place so that an affix results.31>32 jn
Germanic languages no affixes are spelled out, but instead there
exist certain cooccurrence restrictions which will be dis­
cussed here.

Rule (v) is considered to be a kind of redundancy rule,


expressing the fact that the semantic feature [+Instantaneous]
implies a sudden mode of action.'3
The fact that Swedish FINNA^ and English FIND^ denotes action
of instantaneous Aktionsart delimits the kind of morphemes
with which the two verbs can occur. The Swedish auxiliary
HÅLLA PÄ ATT and the corresponding morpheme in English, the
gerundive affix -ING, are semantically incompatible with
FINNAi and FINDi respectively as the two following sentences
demonstrate:
FINNA1

(27) a. *Han höll på att finna pengarna i byrålådan, när hon


kom in,

b. *He was finding the money in the drawer when she


entered.

Not only can we claim the existence of the morphosyntactic


feature [+Perfective] on the basis of observations such as the
cooccurrence constraint on certain morpheme sequences as
exemplified in (27a) and (27b), the semantic constraints on
the generation of ACI-Constructions after FINNA^ give us further
motivation for positing this feature for FINNA^. Compare the well-
formed
33

(28) a. Där fann han eleven ligga avsvimmad


rsvimm på golvet.
ringl
(There he found the pupil 'lie*1 'subconscious on the
floor.) Le !

I matsalen fann jag alla vänta på mig vid brasan.


(In the dining room I found them all] I for me
by the fire place.) [ a j

with the ill-formed

(29) a. *Han öppnade dörren och fann pojken hoppa upp från
sängen.
(He opened the door and found the b°y^#c^ump^n^|out:

b. *Han öppnade dörren och fann mannen stjäla hans kläder.


(He opened the door and found the man|*^®^inglhis
clothes,) 1 J

In the examples in (28) the Swedish ACI-Constructions are well-


formed, since the complement verbs are verbs of a durative mode
of action. However, in (29) the complement verbs are verbs of
aninstantaneous mode of action. As the English translations
indicate, only the Accusative-with-Gerundives (AWG's) yield
well-formed translations of the Swedish ACI-Constructions. Thus,
in order to generate well-formed ACI-constructions after FINNAi
the matrix and complement verbs must stand in an asymmetric
semantic relation to each other. We are here touching upon a
semantic deep structure constraint on the Subject Raising Rule
which will be the topic of Part Two, sections 1 through 4.

3.6 STATI VE~NONPERFECTI VE VERBS (CLASS (D))

A similar redundancy rule will specify the morphosyntactic


characteristic of ACI-class (D). The only Modern Swedish
linguistic ACI-verb of this class, VETA 'know', denotes a
stative Aktionsart which is also nonperfective according to
the following rule:

REDUNDANCY RULE (vi)


ACI-Sub- Mod. Sw.
class Example
+V
[+Stative]—»>[-Perfeeti ve] / (6) VETA 'know
+Linguistic
34

In Swedish there is a constraint on the occurrence of the


durative auxiliary HÅLLA PÄ ATT, just as there is a similar
constraint on the occurrence of an Accusative-with-Gerundive
Construction in English, if the verbs are stative-nonperfective
as demonstrated by the following:

(30) a. *Han höll på att veta hennes namn.


(*He was knowing her name.)

b. Han höll på att lära känna henne.


(He was getting to know her.)

An asymmetric semantic relation must exist between auxiliaries


or suffixes denoting duration and verbs denoting stativeness as
in (30a). Where this requirement is not fulfilled a cooccurrence
restriction is violated. This is the motivation for positing
the feature [-Perfective] for the Swedish verb VETA and the
English verb KNOW, which indicates that the stative verb does
not take nonperfective auxiliaries or suffixes.

The morphosyntactic feature [-Perfective] blocks the outspelling


of the durative periphrastic auxiliary in Swedish HÅLLA PÄ ATT
or the English progressive suffix plus BE. Rule (vi) inserts
[-Perfective] into the lexical entry VETA and KNOW respectively
so that violations as exemplified in *(30a) cannot occur.

The perfective and nonperfective features are surface structure


morphosyntactic features which are associated with various
auxiliaries or affixes. For instance, if a verb like SE, 'see
visually' is durative then it cannot cooccur with the auxiliary
FÅ as in *Jag fick se henne medan hon spelade piano
[+Dur]
'"I suddenly saw her, while she was playing the piano' The
auxiliary fiok in the sense of instantaneous Aktionsart is ill-
formed here, as it is a perfective auxiliary.

3,7 CAUSATIVE VERBS (CLASS (E))

Class (E) expanded by Subcategorization Rule (ii) above contains


the different types of causative verbs which will be discussed
here. Class (E) is a cover term for several disparate types of
verbs. The reason that they have been classed together is the
fact that they all somehow express the causing of an action, a
state, or a process by an instigator, whether it is in a
permissive-causative manner [+Caus-,] as expressed by LÂTA^'allow'
(^Subclass (7a)), or in a hortative-causative manner [+Caus2]
as expressed by BEDJA 'ask' (=Subclass (7b), or in an
instigating-causative t+Caus^] manner as expressed by LÅTA2
(=Subclass (7c)), or in an instructive-causative manner [+Caus^]
35

äs expressed by LÄRA^ 'teach' and LÄRA2 (SIG) 'learn' («Subclass


(7d)), or in an auxiliary-causative manner [+Causs] as expressed
by HJÄLPA 'help' («Subclass (7e)J.

The features t+Causj^], [+Caus2] etc in Rule (vii) below are


not intended to be morphological features, but they are con­
venient labels subcategorizing a set of Modern Swedish verbs
which are felt to belong together in a common causative class
of verbs. The feature [Caus ] is a semantic feature which will
be shown to have syntactic consequences. The feature [Linguistic]
is inserted to distinguish between those ACI-verbs which denote
a linguistic action of some kind on the part of the instigator
e.g. BEDJA 'ask', LÄTA^ 'allow1. Where no such linguistic
action can be inferred from the context as in the case of LATA-
'cause' and HJÄLPA 'help' the feature [Linguistic]is negatively
specified.

SUBCATEGORIZATION RULE (vii)

ACI-Sub- Modern Swedish Examples


class
+Caus-. (7a) LATAj 'allow'
+Linguistic
+Performative
+Permissive
+Caus2
+Linguistic (7b) BEDJA 'ask'
+Performative
l±Hortative
rL+CausativeJ
+v 1 pCaus3
-Linguistic (7c) LATA2 'cause'
[•Instigating

[+Caus4
+Linguistic
•Instructive

+Caus5
(7d) LÄRA. 'teach'
LÄRA*(SIG) 'learn'

-Linguistic (7e) HJÄLPA 'help'


•Auxiliary
<r

Causative verbs are by no means a new concept in linguistics,


although one receives the impression from reading the works of
some generative semanticists who deal with causative verbs that
such verbs were quite recently discovered by them (cf. Lakoff
1965, McCawley 1968, and Postal 1970b). In hi storical linguistics
of the past century the transitive verbs of, for instance, sub­
class (7d) were derived by a historical rule which made under­
lying intransitive verbs transitive. (Cf. the derivation of
English LEARN as described by Klein 1971:415 from an underlying
historical structure paraphrased as 'to make known ' or 'cause
to follow'.
36

3,8 REFLEXIVE VERBS (CLASS (F))

In Class (F) eight reflexive ACI-verbs can be listed: FÖRKLARA


(SIG)'declare1, UPPGIVA (SIG) 'say, state', TÄNKA, (SIG) 'imagine',
VÄNTA(SIG) 'expect', KÄNNA3(SIG) 'feel mentally KÄNNA4(SIG)
'feel somatically', LÄRA2(SIG)'learn', and VISA (SIG) 'turn out'.
The first two verbs have also been listed as performative
declarative verbs belonging to Subclass (4a) above. They both
denote a performative speech act on the part of the agent,the
human subject. There are cross-classificational reasons for
classifying them both as reflexive and performative verbs.

The same cross-classification can be claimed for TÄNK/i(SIG)


'imagine' and VÄNTA (SIG) 'expect' which are also subclassified
as mental nonperformative verbs belonging to Subclass (4b).
However, in order to stress the similar surface structure
characteristics of both subgroups of ACI-verbs as verbs which
take only reflexive noun phrases as objects in ACI-Constructions,
they are treated together with the last two of the eight
reflexive ACI-verbs enumerated above: LÄRA2(SIG) 'learn' and
VISA (SIG) 'turn out'. The ACI-Constructions of these verbs are
not as easily accounted for in terms of Subject Raising as the
four other reflexive verbs just mentioned.

The verb LÄRA2(SIG) can also be subclassified as a causative


verb in Subclass (7d). By doing so there is no evident need,
however, to distinguish .it from LÄRA^ 'teach', if we work
with the goal of describing only Swedish ACI-Constructions.
The generation of ACI-Constructions can be carried out by
Subject Raising in sentences which contain either LÄRA^ or LÄRA2.
We want to accomplish more than a description of Swedish ACI-
Constructions. We want to make a distinction between the two
verbs in order to find a usable subclassification within compara­
tive Germanic syntax. In both German and English a different
verb is inserted according to the roles of the subject noun
phrase. If the subject noun phrase is the experiencer and
receiver of the instruction, and the object of the instruction
is a sentential object, then LERNEN or LEARN is inserted in
German or English respectively. If the object of the verb of
instruction is a human noun phrase, and the subject noun phrase
is an Agent in Fillmore's sense, then LEHREN or TEACH is in­
serted, provided the subject noun phrase is not the receiver
of the instruction.

Arguments for cross-classifying KÄNNA3 and KÄNNA4 into the


reflexive verb class were discussed above in section 3.3.

The most problematic verb enumerated among the reflexive verbs


above is VISA (SIG) 'turn out', whose ACI-Construction has a
different derivational history than those of the other ACI-
verbs discussed so far.34
37

SUBCATEGORIZATION RULE (viii)

ACI-Sub­ Modem Swedish


class Examples

t it i veT
Performat:
tive J
•Declarat:
(8a) FÖRKLARA(SIG)'declare*
UPPGIVA(SIG)'say,state*

pPerformative"]
+Mental (8b) TÄNKA,(SIG) 'imagine'
L+Probable J VÄNTA (SIG) 'expect'

-*<
E Perception]
•Mental J
(8c) KÄNNA3(SIG) 'feel
mentally'

[ «•Perception]
+Somatic J
(8d) KÄNNA4(SIG) 'feel soma­
tically'

[+Caus4
+Linguistic (8e) LÄRA2(SIG) 'learn'
1+Instructive

[^Transitive] (8f) VISA (SIG) 'turn out'


38

4. THE CONCEPT OF SYNTACTIC RULE DOMAIN


4.1 THE STANDARD THEORY AND LEVEL CONSTRAINTS IN SYNTAX

Since the introduction of a deep structure and transformational


rules into linguistic theory, it has continually been a problem
to delimit the generative power inherent in the generative
transformational theory which describes and explains syntactic
structures and processes. The goal of generative grammarians
has always been to describe the generation of all well-formed
and only the well-formed surface structures.

In early transformational grammar certain rules were therefore


said to be either optional or obligatory (cf. Chomsky 1957:45);
certain conditions of identity between noun phrases were set up
to determine whether syntactic rules such as the Relative,the
Passive, and the Reflexive Rules were to apply or not, (cf.
Chomsky 1957:43, Lees 1966:89, and Bierwisch 1963:92 and 95).
Then the concept of rule ordering, either disjunctive or
conjunctive, was stressed as being crucial for the derivational
histories of surface structures (cf. Chomsky 1957:42 and 44,
Chomsxy 1965:133, Chomsky & Halle 1968:30 etc). The cyclical
nature of certain transformational rules was postulated as a
means of blocking or generating various surface structures
(cf. Chomsky 1965:143, Chomsky & Halle 1968:26 and 29).

Recently, that is since the publication of Aspects, several of


Chomsky's students and colleagues have developed still more
important mechanisms for constraining the generative power of
transformational grammar as sketched in the Aspects-model by
Chomsky 1965.

Lakoff 1965 developed a theory of exceptions to general syntactic


processes by making the important distinction between "major and
minor rules" in syntax. Ross 1967a introduced the concept of con­
straints on various types of movement transformations. Postal
19 71 reformulated some of these constraints as being based on
a more fundamental principle, i.e. his Cross-over Principle,
which allowed or blocked the Passive, Tough, and Psych Movement
Rules by means of coreference or noncoreference linkage between
subject and object noun phrases. Psycho- and neurolinguistics have
lately interpreted such constraints on movement transformations
as constraints on our perceptual apparatus which reflect the
limitations of neurological processes (cf. Bever 1970 and
Whitaker 1971). Perlmutter 1971, working within the framework
of the Standard Theory, showed the necessity of an output
filter, or surface structure constraint component, to generate
or block Spanish and French surface structures which contain
pronominal clitics.

Lakoff 1970a, finally, demonstrated the need of global rules to


generate certain surface structures in Greek and English which
cannot be generated by locally operating transformational rules
in the sense of Chomsky 1965. In the Aspects-model Chomsky
claims that transformational rules work only on contiguous trees
39

in a derivation (cf. op. cit. p. 138). Lakoff, on the other hand,


contends that transformational rules must operate between non­
contiguous trees as w el l . 3 "

4,2 LEVEL CONSTRAINTS FOR GENERATING WELL-FORMED SWEDISH ACI~


CONSTRUCTIONS
There is consequently no lack of attempts to improve the genera­
tive machine and to constrain the domain of various transforma­
tional rules within different syntactic theories. The problem
inherent in the Standard Theory as presented by Chomsky 1965
lies not in the generative nature of its transformational rules,
but rather in the unconstrained power of these rules to generate
well and ill-formed surface structures alike in the same un­
constrained way. Chomsky himself has been forced to acknowledge
this fact several times and has also pointed out the need for
various conditions on transformations as mentioned above. '
Therefore, in his latest three papers on syntax, he presents a
more highly constrained generative transformational theory,
the so-called Extended Standard Theory, 'EST' (cf. Chomsky 1970b,
1971a, and especially 1971b.

The syntactic domain of transformational rules must be delimited


by various sorts of constraints or conditions on their applica­
tion, so that well-formed surface structures result. These con­
straints can be set up as filters on deep, intermediate (shallow),
or surface structure levels.38

Three sorts of level constraints will be demonstrated to be


necessary in order to generate well-formed Accusative-with-
Infinitive constructions in Modern Swedish.

Part Three discusses in detail the Coreference Linkage Constraint


(The Equal Subject Constraint) on the Subject Raising Rule in
Swedish which is a semantic local constraint on the deep structure
level. The constraint is needed to block the generation of
sentences like

(31) a. *John sade flickan vara sjuk.


(John said that the girl was ill.)

b. *Vem sade flickan vara sjuk,


(Who said that the girl was ill.)

c. *Mannen sades av John vara sjuk.


(The man was said by John to be ill.)

Another deep structure constraint is the Aspectual Constraint on


the Subject Raising Rule which has been discovered to constitute
a blocking device against the cooccurrence of such matrix and
complement verbs whose inherent modes of action (Aktionsarten)
are semantically incompatible, e.g.
40

(32) a. *Han observerade barnen tappa mjölkflaskan.


(He observed the children drop the bottle of milk.)

b. *Hon upptäckte en man köra på hennes bil.


(She discovered a man driving into her car.)

c. *Han märkte Peter falla från taket.


(He noticed Peter fall from the roof.)

Both the Swedish and English sentences are ill-formed in (32)


because of the violation of the Aspectual Constraint on Subject
Raising. In Part Two, section 2 this deep structure constraint
will be treated in more detail.

In the discussion in section 5.1 and 13.1 of various surface


structures occurring after the ACI-verbs SE 'see' and HÖRA 'hear1
at least three Intermediate (Shallow) Structure Constraints
will be treated which are involved in the generation of ACI-
structures or other surface structures which also undergo
Subject Raising:
1) the Coreferenoe Constraint on the Reflexive Rule which blocks
the generation of sentences of the following surface structure
in Swedish and English

(33) *Peter^ hörde sig^ tala om skatter och löner.


(Peter^ heard himselfj speaking about taxes and salaries.)

The Reflexive Rule has operated here after Subject Raising,


although the structural description of this rule is not fulfilled,
i.e. there is no coreference between the subject noun phrase
and the object noun phrase.

2) the Non-Coreferenoe Constraint on the Passive Rule which is


needed to avoid sentences like

(34) *Sig hördes av FNL:arna tala om Vietnam.


(Themselves were heard by the FNL adherents to speak of
Vietnam)

3) the Indefinite Subject Noun Phrase Constraint on the Swedish


Passive Rule after Subject Raising in sentences containing per­
formative verbs (verba dicendi) or some verbs of perception
(verba sentiendi) e.g. HÖRA 'hear'.

(35) *Palme hördes av LO tala om löneförhöjningar och skatter.


(Palme was heard by the LO speaking of pay raises and taxes.)

This is an example of a shallow structure constrain^ treated


from the viewpoint of the Subject Raising Rule. It is evident
that we cannot describe a given syntactic rule in isolation,
for instance the Subject Raising Rule, without taking other
pertinent syntactic rules into consideration as well, such as
the Passive Rule, the Reflexive Rule, and the Agent Phrase
Deletion Rule. The specific processes of these syntactic rules
will be discussed at some length in Part Three in connection
with the Subject Raising Rule.
41

Finally, as an example of a surface structure constraint of


phonological nature we will discuss in Part Two and Part Three
the generation of two Swedish passives within one and the same
sentence. The conditions of application on the Swedish Passive
Rule must be specified in such a way that the occurrence of an
s-passive in both the matrix and the complement sentences is
blocked. The following sentences, which contain s-passive$ in
the matrix and complements sentences, are unacceptable:

(36) a. *Den turkiske revolutionären uppges skjutas i dag.


(The Turkish revolutionary is reported to be going to
be shot today.)

b. *Den turkiske revolutionären anses hängas i dag.


(The Turkish revolutionary is considered to be going
to be hanged today}

c. *Den brasilianska studenten hördes piskas i cellen intill.


(The Brasilian student was heard being lashed in the
cell next door.)

In order to avoid the generation of such ill-formed surface struc­


tures as exemplified above in (31) through (36), a less power­
ful syntactic theory than the one sketched in the Aspects-model
by Chomsky 1965 will be developed in the present investigation
of Swedish ACI-Constructions. It will be the task of the next
few sections to describe the deep structure, intermediate
structure, and surface structure constraints which can be set
up as filters of well-formedness in order to constrain the un­
limited generative power of the Standard Theory. There is no
way to block ill-formed structures such as t hose of (36) unless
certain constraints are built into a theory of syntax which is
intended to describe what a native speaker knows about his
language.

4.3 THE GENERAL PROCESS OF INFINITIVIZATION IN GERMANIC


LANGUAGES

In order to generate such surface structures as represented in


(1) through (25) in section 2, infinitivization, interrogation,
reflexivization, and relativization processes are postulated
here as being somehow represented neurologically and mentally
in the native speaker's brain. In a synchronic grammar the
goal is therefore to simulate what constitutes these trans­
formational processes in a given language.

The following two internal structures, (37a) and (37b), are


posited as underlying all ACI-Constructions in Modern Swedish:
42

VP

+N +V
+Hum Subj Rais

Condition:X^NP coreferent
With NPg

Condition: a) NP2 = NP^

b) NP1 f. NP2 or NP3

The matrix verb in (37a) and (37b) is marked for the syntactic
rule feature [+Subj Rais] .39 The complement subject, i.e. the
NP3 under the complement sentence node S^, must be raised into
the matrix sentence S0. Subject Raising m this sense is here
indicated by arrows. Raising occurs if certain conditions are
fulfilled, that is if the structural description in phrase
markers (37a) or (37b) is met. The phrase marker (37a) is the
internal or underlying structure of the ACI-Constructions
occurring primarily after performative verbs{verba dioendi) as
exemplified above in subsection 2.1).40

However, a similar syntactic rule operates on the phrase marker


in combination with the Equi NP Deletion Rule in (37b), where
the object noun phrase (NP2) is coreferent with the subject noun
phrase of the complement sentence (NP3). Phrase marker (37b) is
posited to underlie ACI-Constructions after verbs of perception
(verba sentiendi) such as SE 'see', and KÄNNA^1feeli', instanta­
neous verbs such as FINNAj 1findi, discover1, and causative
verbs such as LÅTAj 'allow', BEDJA 'ask', LÅTA2 'cause' etc.
43

The complement subject of Si, i.e. NP3, is considered to be


raised to the coreferent object noun phrase in the matrix
sentence (NP2) and deleted there, leaving only NP2 intact.
This deletion is ordinarily called Equi-NP Deletion but will
be treated here together with the Subject Raising Rule.
Compare (38a) and (38b) between which the infinitivization
process has operated, i.e. Subject Raising and Noun Phrase
Deletion:41

(38) a. Jag fannl konom ^an satt där.

(I j^ouridj He was slttlnê there.)

b. Jag fannj konom där.

CI {found} him sittin8 there.)

By t reating the infinitives in (38b) and all of the infinitives


of (1) through (25) in section 2 above as products of one and
the same major syntactic process, i.e. Subject Raising, we can
describe the same surface structures in terms of one general
operation. The conditions or constraints on this operation will
be explicitly discussed in the following sections.

There are several heavy constraints on Subject Raising which


operate on both phrase marker (37a) and phrase marker (37b).
First, there is the special syntactic characteristic of the
matrix verb involved that delimits its capability to allow
Subject Raising. It may simply not be marked as having the
rule feature [+Subj Rais] in its lexical entry, as for instance
in Modern Swedish SVÄRA 'swear' or FÖRBINDA (SIG) 'pledge'.
Corresponding Old Swedish verbs did allow Subject Raising, as
demonstrated by the following sentences:
SVÄRA 'swear'

(39) a. *då skall nämnd med ed svära honom vara femton år


gammal.
SW/ERIA 'Old Swedish swear'

b. j>a skal nœ md swce ricehan va?œ femton œ arœ gamblan.


[+Acc]I+InfJ
(VgL II, Add 7, § 29: G 340)
(then the jury shall swear an oath that he is fifteen
years old)
44

FÖRBINDA (SIG) 'pledge

(40) a. *och vi förbinda oss själva skola hålla alla förut­


nämnda artiklar obrottsligt.

BINDA SIK 'Old Swedish 'pledge'

b. ok bindum ui os siœ luœ alice for nee mda artìculos


[+Acc]
ubrutlika Haida skula .(MELL Kg 31: G 343)
[+Inf]
(and we pledge that we will ourselves obey all the
paragraphs mentioned before.)

Secondly, there may be idiosyncratic constraints on the Subject


Raising Rule that delimit its rule domain. Compare the following
Modern Swedish and Modern Icelandic ACI-Constructions:

SÄGA 'say'
(41) a. *Han säger mannen vilja betala för skadorna.

SEGJA 'Modern Icelandic say'


b. Hann segir manninn vilja bae ta skemmdirnar.
f+Acc] [+Inf]
(He says that the man wants to pay for the damages.)

We shall show that Modern Swedish has developed an Equal Subject


Constraint on Subject Raising which Modern Icelandic lacks as
demonstrated by (41b). An exact translation into Modern Swedish,
which is unacceptable, is given in (41a). For some reason
Modern Swedish has acquired an idiosyncratic constraint on
Subject Raising in sentences which contain performative verbs
(verba dicendi). In a special investigation on the change of
the syntactic rule domain from Old to Modern Swedish, the
reasons for this syntactic rule change have been treated (cf.
Ureland 1972b(forthcoming)).42

Such syntactic characteristics inherent in the lexical entries


as demonstrated for the Old Swedish verb in (39b) and (40b)
and the idiosyncratic coreference constraint on Subject Raising
as demonstrated for Modern Swedish in (41) vary from language
to language, and even from one chronological period to the next
within one and the same language.

We know, for instance, from syntactic research that the domain


of the Subject Raising Rule was much wider in Old Swedish than
it is in Modern Swedish. The same statement can be made about
Ancient Greek and Latin in comparison with Modern Greek and
Modern Romance languages. First, a far larger number of verbs
could occur in Ancient Greek, Latin, and Old Swedish with ACI-
Constructions. Secondly, certain nouns, adjectives, and even
impersonal verbs like Latin APPARET 'it is clear' allowed
Subject Raising and therefore took ACI-Constructions as
complements. In the latter sense, too, the Subject Raising Rule
45

can be said to have been more extensive in the two classical


languages than in any of the modern Germanic lariguages.lt is
the more extensive type of Subject Raising which we will claim
to be formally common to Germanic and Romance languages although
the specific rule marking or the constraints may vary.
46

5, THE SUBJECT RAISING RULE IN SWEDISH (SE)


The verb SE 'see' is defined in Swedish dictionaries by a large
number of different synonyms. See for instance Svensk illustre­
rad ordbok 1958 which gives 12 meanings of the homograph SE. All
classes of SE will not concern us here, but only those three
classes of SE which take embeddings as verb complements, that
is SEi 'see visually', SE2 'realize', and SE* 'prefer to see'.
These three types cf SE are represented in the Gothenburg corpus
by a large number of authentic examples some of which will be
discussed in this section.

All three types of SE focused upon here express a passive


perceptual process on the part of the human deep subject, i.e.
the Experiencer in Fillmore's 1971 sense.

In section 5.1 the underlying syntactic representation of the


ACI-Construction after SEi 'see visually' will be treated.
In 5.1.1 the att and fcwr-Embeddings are presented. Section 5.1.2
and 5.1.3 discuss syntactic and semantic evidence for a
different underlying syntactic representation for the ACI-
Construction than for complementizer embeddings.

In section 5.2 the Subject Raising Rule and the Auxiliary or


Aktionsart Marker Deletion Rule is demonstrated to generate
the Accusative-with-Past Participle Constructions.

Nominalizations as verb complements after SE^ and SE2 are treat­


ed in section 5.3. A lengthy discussion of the status of the
homograph FÅ then follows in section 5.5. Finally, in section
6, the result of the present linguistic performance study of
the verbal complements occurring after SE 'see' is presented
in tables showing verbal complementation as exemplified in
the Gothenburg corpus.

5.1 the verbal complements of sei 'see visually'

By means of the Passive Rule Test and the Cleft Sentence Rule
Test one can easily show that the italicized noun phrases in
(42) through (44) are object noun phrases of the matrix senten­
ces. Before doing so, we will set up an underlying syntactic
representation (£S J) from which we will generate (42). In
contrast to Ureland 19 70 a different underlying representation
is assumed for the ACI-Construction after SE than after HÖRA.
For ACI-Constructions after HÖRA the same underlying representa­
tion is assumed as for ACI-Constructions after verba putandi
such as ANSE 'consider' in (44).
47

(42) Palme såg byggnadsarbetare demonstrera i New York City.


(Palme saw construction workers demonstrating in New York
City.)

(43) Palme hörde dem ropa slagord mot Sverige.


(Palme heard them shouting slogans against Sweden.)

(44) Svenska radikaler ansåg honom göra en fjäskresa till USA.


(Swedish leftists considered him to be making a trip of
courtship to the US}

DS I

NP- ^rep P

NP NP
pv 1
+Past
,+ Subj Raisj
!,NP" IP PrepT NP

Palme sag ba. ba. demonstrer- i NYC

The Subject Raising Rule in our syntactic sense, which raises


the third coreferential noun phrase (NP^) into the matrix
sentence (SQ) to be deleted there, is now applied to the phrase
marker DS J, whereby the morphological feature [+INi] for the
infinitive suffix is inserted into the complement verb matrix.
Surface Structure I (SS I) below will result after Subject
Raising takes place. This structure bears the name of the
Accusative-with-Infinitive construction in traditional as well
as historical grammars:

SS I

(46)

reoP

+Subj Rais Prep^NP

Palme sag ba. demonstrera i NYC


48

The morphological feature t+IN^] is here no ad-hoc feature.


Rather, it is necessary in order to represent the infinitive
ending in such processes as the Suffix Agreement Rule in some
(dialectal) Swedish. According to this rule the index for fhe
infinitive ending [+INi]is changed to agree with the index for the
ending of the past participle I + IN23 in sentences which contain
verbs in the perfect or pluperfect tenses, e.g. Har du hört
honom sagt det 'Have you heard him say so1, Han har fått go'ort
det 'He has been allowed to do it1, and Har du sett honom
ätit (äte) surströmming 'Have you seen him eat fermented herring1.

After the Subject Raising Rule has generated SS I, the Passive


Rule Test may be performed to show whether the raised noun phrase
has been deleted so that it is no longer a coconstituent of the
complement verb {demonstrera), or whether the (object) noun
phrase occurring after the matrix verb belongs to the matrix
sentence. The constituent structure of SS I above can also be
represented by labelled bracketing:

(47) SS I [ [Palme] såg [ba.] [demonstrera] i NYC]


S0NPX NPX NP2 NP2 VP VP SQ

(48) [ [ba.] sågs av [Palme] [demonstrera] i NYC]


So NF-,i NP~
l NPt
11 NPTVP VP So

As can be seen from (48) the order of noun phrases is reversed


and the constituent structure of SS I is destroyed by the
operation of the Passive Rule. There is no doubt that the
second noun phrase (NP2) is a constituent of the matrix sentences
in SS I, since it has Become the surface structure subject in
(48) in the cross-over operation which the Passive Rule con­
stitutes.

However, there exist in Swedish passive sentences active


infinitives like (49), in which the agent has been deleted:

(49) Byggnadsarbetare sågs demonstrera i NYC.


(Construction workers were seen to be demonstrating in NYC.)

If such surface structures occur, the deleted agent must be in­


definite. Since the semantic interpretations of (48) and (49)
differ considerably, two different underlying structures must be
postulated to explain the difference in meaning. The underlying
representation of (49) cannot be DS J^because DS I contains a
definite matrix subject (=Palme). A different underlying re­
presentation containing an indefinite matrix subject (=någon
'somebody') is posited for (49). An empty dummy symbol under
S0
[NP^] is probably the most adequate representation of the subject
node in DS I (45) for generating passive sentences like (49).
The obligatory deletion is then automatically accounted for.

That NP2 is not a coconstituent of the complement verb is further­


more demonstrated by the ill-formed sentence (50) in which NP2,
49

together with the infinitive, has been moved to the subject


position by the Passive Rule:

(50) *Byggnadsarbetare demonstrera sågs av Palme i NYC.


(Construction workers demonstrating were seen by Palme
in NYC)

Thus the Passive Rule Test demonstrates that the second noun
phrase in DS I belongs to the matrix sentence in SS I after
Subject Raising and Equi-NP Deletion have operated. These two
transformational rules explain adequately why it is that the
second noun phrase is felt as being both object and subject
at the same time. This fact has puzzled philologists and
historical grammarians,43 as well as normative traditional
grammarians.44 By positing an abstract underlying structure
{=DS I) and applying the Subject Raising and Equi-NP Deletion
Rules a description is accomplished which meets the require­
ment of descriptive adequacy, i.e. explains in a linguistic
way the double interpretation of the syntactic status of NP2
in our example (42) above.

The former object clause of DS I dominated by NP no longer


exists, which motivates tree-pruning, so that the NP and Si
nodes need not be indicated in SS I. Of the constituent sentence
only the verb phrase remains, i.e. the infinitive phrase.
(Cf. the tree-pruning convention as suggested by Ross 1969:299).
The treerpruning blocks the recoverability of the deleted nodes,
however (see Part Two, section 12.3).

The Cleft Sentence Test will yield a similar result and thus
confirm our observation that the raised and deleted noun phrase
(NP3) no longer belongs to the constituent sentence but rather
to the matrix sentence. The noun phrase which appears here is
the second noun phrase (NP2)• since (50') is ill-formed after
the Cleft Sentence Rule has applied to SS I:

(50*) *Det som Palme såg i NYC var byggnadsarbetare demonstrera


(What Palme saw in NYC was construction workers demon­
strating.)

It is also possible to apply the Cleft Sentence Rule so that the


object NP of the complement sentence, byggnadsarbetare
'construction workers', is moved as in the following example:

(50'') Det som Palme såg demonstrera i NYC var byggnads­


arbetare.
(What Palme saw demonstrating in NYC was construction
workers.)

An emphatic effect is of course obtained by moving the comple­


ment object from the position after the verb of perception.
A similar effect is obtained by moving the same object NP to a
position before the matrix sentence under Relativization as in
the following sentence:
50

(50''') Det vax. byggnadsarbetare som Palme såg demonstrera i


NYC.
(It was construction workers that Palme saw demonstrat­
ing in NYC,)

In traditional grammars such a construction was called an


"emphatic construction" (cf. Beckman's 1959 term "utbrytning"
for the ]atter Emphatic Movement Rule (p. 240)).

A whole series of Passivization and Subject Raising is possible


if the matrix verb is SE^ 'see visually'. The Passive Rule may
apply not only to the matrix sentence as in (48) and (49)above
after Subject Raising, but also to the complement sentence be­
fore Subject Raising as in the following sentences:

(51) a. Palme såg ägg kastas av byggnadsarbetare i NYC.


(Palme saw eggs being thrown by construction workers
in NYC.)

b. Palme såg ägg bli kastade av byggnadsarbetare i NYC.


(Palme saw eggs being thrown by construction workers
in NYC.)

It is unsure however, whether the s-passive can occur in both


the complement and matrix sentences contemporaneously as it does
in the following sentence:

c. *? Ägg sågs av Palme kastas av byggnadsarbetare i NYC.


(Eggs were seen by Palme being thrown by construc­
tion workers in NYC.)

There is a surface structure constraint on the occurrence of


the double s-passive under the same sentence node which requires
that the second passive must be a BLIVA passive as demonstrated
by the following well-formed sentence:

d. Ägg sågs av Palme bli kastade av byggnadsarbetare i NYC.


(Eggs were seen by Palme being thrown by construction
workers in NYC}

As the English translations and the Swedish passives of (51d)


reveal there is nothing wrong with the double application of
the Passive Rule, but rather it is the phonological forms of
the two passives in *(51c) which are unacceptable.(Cf. Part Two,
section 12.2 on the surface structure constraint on the
application of the Swedish Passive Rule).

5.1.1 the att and hur-embeddings occurring after se x ' see


visually'

On the other hand, if we apply a different rule of embedding,


the att-Complementizer or Tzwr-Complementizer Embedding Rule,
the coconstituency of the constituent verb NP3 of DS I will
51

remain intact, even though the Passive and Cleft Sentence Rules
have applied. Comparé (a) with (b) and (c) with (d) in the
following sentences:

(52) a. [ [Att ba. demonstrerade] 3 sågs av Palme i NYC.


NP Sx S1NP

(That construction workers demonstrated was seen


by Palme in NYC)

b. Det som Palme såg i NYC var [ [att ba, demonstrerade] ]


NP S1 S1NP
(What Palme saw in NYC was that construction workers
demonstrated)

c. [ [Hur ba. demonstrerade] ] sågs av Palme i NYC.


NP S1 SXNP
(How construction workers demonstrated was seen by
Palme in NYC.)

Dét som Palme såg i NYC var [ [hur ba. demonstrerade] ]


NP Sx S1NP
(What Palme saw in NYC was how construction workers
demonstrated.)

When no Subject Raising occurs, the constituent configuration of


the underlying representation of S-, is un changed, that is the
complement subject agrees with complement verb which carries tense.
The Si and NP nodes under VP in the phrase marker must therefore
be preserved, which is indicated by the labelled brackets in the
sentences above.

Complementizer embedding is obligatory if the verb of perception


is SE2 'see cognitively, realize', since ACI-Constructions can
occur only after SEi 'see visually', which is a verb of visual
perception as opposed to SE2 which denotes cognitive perception.

(53) a. HAN SER HUR SAMTAL OMEDELBART LEDER IN I ROLLER.(945:160:6/


426434)(DN)
(He sees how conversations immediately lead to role-
playing.)

b. *Han ser samtal leda in i roller.


(He sees conversations lead to rolë-playing.)

(54) a. TJA, JAG HAR LITE SVÅRT ATT SE HUR DET SKALL GÂ TILL(1027:1
78:9/1144070)(GHT)
(Well, it is difficult for me to see how that can
come about.)

b. *jag har lite svårt att se det skola gå till.


(*It is difficult for iiié to see that come about.)
52

It is clear that Subject Raising cannot be allowed to operate


on the underlying structures of (53a) and (54a). First, because
it is physically impossible to see visually that conversations
lead to role-playing. SE2 'see cognitively1 must be marked as not
allowing Subject Raising. Secondly, because in (54b) there is
a constraint on the generation of the ACI-Construction, when
the complement sentence must express tense. Both the fact that
SE2 'realize' is the matrix verb and the fact that the complement
sentence is a tensed sentence (expressed here by skall 'will1)
block the Subject Raising Rule from applying. (Cf. section 5.1.2
where more examples of tense blocking of Subject Raising are
discussed).

Both from syntactic and semantic points of view there are reas ons
to derive the sentences which contain the Tzwr-Embedding from a
different deep structure than the ACI-Constructions.

5,1.2 syntactic evidence for different underlying representa­


tions of aci-constructions and complementizer embeddings

The status of the complementizer-embedded sentences in (51)


through (54) is consequ ently very different from that of the
sentences embedded by the Subject Raising above in (42) through
(44). This is one of the three syntactic motivations for treat­
ing the ACI-Constructions as derived from a different under­
lying structure than the sentences embedded by complementizers.
The following underlying structure is therefore posited for
embedding by means of att and /zur-complementizers, the in­
sertion of which is triggered by the rule feature [+attiCompi R]
and [+ ftur-Compl R]:

+V
J+att-Compl R|
p-ftwr-Compl RI
+Past

fatt; I
Palme sag i hur1 ba. demonstrerade i NrC

Sentences (52a) through (52d) result from the operation of either


the att or the /zur-Complementizer Insertion Rule which replaces
the sentence type node Sub, introduced by the base rules with
either att 'that' or hur 'how' respectively. (The transformational
replacement of the type node symbol is indicated by the dotted
square under the type node).
53

Complementizer embedded sentences can thus undergo the Passive


and Cleft Sentence Rules in a much more unconstrained way than
embeddings by the Subject Raising Rule, a fact demonstrated by
the examples (52a) through (52d) as compared with the examples
*(50) and *(501). This is the first syntactic argument against
deriving complementizer-embedded constructions and ACI-Con-
structions from one and the same underlying syntactic deep
structure.45

The second and more serious argument deriving ACI-Constructions


and att-embedded structures from the same underlying representa­
tion is, however, the fact that there is no non-ad hoc way to
avoid the generation of ill-formed ACI-Constructions in cases
like the (b) sentences of (56) through (58) below. The (b)
sentences have been converted into unacceptable ACI-Construc-
tions from authentic (a) sentences excerpted from the Gothenburg
corpus

SEX 'see'
att-Embedding
(56) a. DÅ MANNEN SÅG ATT HUNDEN HADE SKOTTSKADOR I HUVUDET.
(973:81:6/511640)(SDS)
(When the man saw that the dog had shot-wounds in its
head)

A CI-Construction

b. *då mannen såg hunden ha skottskador i huvudet.


(*when the man saw the dog having shot-wounds in its head.)

att-Embedding
(57) a. MEN NÄR MAN SÅG ATT I NGEN MÖTTE CHURCHILL.
(302:65:8/515530)(SDS)
(but when they saw that no one came to meet Churchill.)

ACI-Construction

b. *men när man såg ingen möta Churchill.


(*but when they saw nobody coming tó meet Churchill.)

att-Embedding
(58) a. PROFESSOR WERKÖ SKULLE FÖR SIN DEL GÄRNA SE ATT DE SÖKANDE FICK
GENOMGÅ EN F ÖRBEREDANDE KURS.
(393:179:5/111019)(GHT)
(Professor Werkö would for his part like to see that
the applicants had the chance to take a preparatory
course.)

ACI-Construction
b. ?Professor Werkö skulle för sin del gärna se de
sökande få genomgå en förberedande kurs.
(?Professor Werkö would for his part like to see the
applicants having the chance to take a preparatory
course.)
54

The ill- formedness of the (b) sentences call for an explanation.


One might argue that SE^ in (56)-(58) is not a verb of immediate
perception and this should be the reason that ACI-Constructions
cannot appear in the (b) sentences. Such a claim is false as
far as (56a) and (57a) are concerned where direct visual per­
ception is implied. No, we have to look deeper into the matter.
In (57b) an ACI-Construction is simply impossible because the
complement sentence does not correspond to the truth value.
Since nobody came to meet Churchill, it is impossible to see
this "nobody" meeting him. In the following sentence the truth
requirement is not violated, where "somebody" occurs as the
complement subject:

(57) b'. men när man såg någon möta Churchill.


(but when they saw somebody meeting Churchill.)

If the Negation Placement Rule operates so that the scopus of


the negation refers to såg as in the following sentence, then
an ACI-Construction is possible:

(57) b''. men när man inte såg någon möta Churchill.
(but when they did not see anybody come to meet
Churchill.)

There are some unknown constraints on negated ACI-Constructions


which cannot be explained here. Whereas *(57b) is experienced
as ill-formed by most native speakers there are some speakers
who accept the negated ACI-Constructions also in the following
sentence:

(57) bT''. ?Han såg ingen komma emot sig på vägen.


(?He saw nobody approaching him on the road.)

However, if the scopus of the negation is on the verb of visual


perception, then again an ACI-Embedding is acceptable:

(57) blv. Han såg inte någon komma emot sig på vägen.
(He did not see anybody approaching him on the road.)

As far as the ill-formedness of *(56b) and *(58b) is concerned


one can refer to the semantic properties of the complement verbs.
It seems as *if the v erb of possession HAVA 'have' blocks its
subject from being raised into the matrix sentence to yield an
ACI-Construction. A similar blocking function occurs with FÅ
'be allowed to'. An att-Embedding is definitely a more appropriate
type of embedding here than is an ACI-Construction.

However^the blocking of the Subject Raising Rule is due not only


to t he semantic properties of the complement verb as d emonstrated
by (56)-(58), but also to the semantic properties of the verb of
perception. The matrix verb is also of importance for the choice
of complement type. Where SE is a verb of cognition in the sense
of 'realize', i.e. SE£, a similar semantic constraint to the one
just exemplified in (53)-(54) can be observed. Compare the
following (a) and (b) sentences:
55

SE9 'realize'
a. VI SER ATT LAGERKRANTS HAR FUNNIT TRÖST I DANTES VERK.
(257:49:6/412505)(DN)
(We see that Lagercrantz has found comfort in the work of Dante)
b. *vi ser Lagercrantz ha funnit tröst i Dantes verk.
(*we see Lagercrantz having found comfort in the works of Dante.)
(60) a. EFTERSOM DE SETT ATT REDAN FARTYG PÅ 20 å 22 KNOP IN TE ÄR LÖNSAMMA.
(278:178:2/414484) (DN)
(since they have seen that ships of 20 or 22 knots are not profitable.)
b. *eftersom de sett redan fartyg på 20 â 22 knop inte vara lönsamma.
(*since they have seen ships of 20 to 22 knots not be profitable.)
(61) a. JAG KAN I NTE SE ATT LAGEN HADE VISAT HENNE MERA AKTNING.
(1049:103:3/122049) (GHT)
(I cannot see that the law had shown her more respect.)
b. *jag kan inte se lagen ha visat henne mera aktning.
(?I cannot see the law having shown her more respect.)

The reason for the ill-formedness of the ACI-Constructions in the


(b) sentences is also that the tense property of the complement
sentences is lost after the Subject Raising Rule operates. The loss
of tense is not acceptable as evidenced by the ill-formedness of
the (b) sentences.47
It is this kind of evidence which speaks against the common deriva­
tion of att-embeddings and ACI-Constructions from one and the same
syntactic deep structure. We are involved both with a TENSE CONSTRAINT
and a SEMANTIC CONSTRAINT on the choise of the embedding type.Chomsky
1971b points out some types of constraints on certain movement
transformations in so-called tensed sentences.48
If both ACI-Constructions and complementizer embeddings are generated
from one and the same underlying representation as has been claimed
for different Germanic languages (for English by for instance Jacobs
& Rosenbaum 1968:175 and 195, R. Lakoff 1968:25-29, etc, and for
Swedish by Ellegård 1971:93 and 1935, and Dahl 1971:44 and 48), then
it will be extremely difficult, if not impossible, to state the
conditions under which Subject Raising can apply.The linguists
mentioned above have not discovered and therefore have not considered
the tense constraint in their descriptions of English and Swedish
verbal complementation. Such a constraint contradicts the postula­
tion of the same deep structure for deriving ACI-Constructions and
att-Embeddings.
Following the descriptions of the generation of the ACI-Construction
by these linguists, one receives the impression that the choice of
embedding type is always optional, which is certainly not true for
the sentences in (56) through (61). The choice of the att-comple-
mentizer is obligatory here, although it is conceivable that there
are cases where the choice of embedding may be optional too, e.g.:
SEi 'see'
(62) a. NI SER ATT MINA LÄPPAR RÖR SIG.(1416:12:2/126018) (GHT)
(You see that my lips are moving.)
b . Ni ser mina läppar röra sig.
(You see my lips moving.)
56

(63) a. ögonvittnen har förklarat att de såg att planet slog


i va ttnet med kraftig rök kring vingarna.
(eyewitnesses testify they saw that the plane hit the
water with smoke emitting from around the wings.)

b. ÖGONVITTNEN HAR FÖRKLARAT ATT DE SÅG PLANET SLÅ I VATTNET MED


KRAFTIG RÖK KRING VINGARNA.(439:29:3/211192)(SvD)
(eyewitnesses have testified that they saw the plane
hit the water with smoke emitting from around the
wings.)

(64) a. hörde att hon skrek på hjälp, såg att hon kröp blödan­
de på trottoaren.
(heard that she screamed for help and saw that she crept
bleeding along the sidewalk.)

b. HÖRDE HENNE SKRIKA PÅ HJÄLP, SÅG HENNE KRYPA BLÖDANDE PÅ TROTTOA­


REN.(83:97:6/222136)(SvD)
(heard her screaming for help and saw her creep bleeding
along the sidewalk.)

Rather than developing an extremely complicated mechanism for


stating the obligatory or optional choice of embedding type, we
prefer to set up two different underlying syntactic representa­
tions, i.e. DS I and DS II for the ACI-Construction and the
att-Embedding respectively (cf. (45) and (55)).

In (65) and (66) there is an additional rule, the Wh-Movement


Rule, which moves the object noun phrase to sentence-initial
position after Passivization. Here no Wh-Movement is possible
at all after att-Embedding:

(65) a. SOM VÅRT FÄDERNESLAND NÅGONSIN SETT FÖDAS. (1 042:108:2/122025)


(GHT)
(whom our country has ever seen being born.)

b. *som- vårt fädernesland sett att! de. 'föds.


-L* % !
(*whom^ our country has seen that they^ are born.)

(66) a. SOM MAN STÄNDIGT SER STÅ OCH HÄNGA. (1133:65:4/215144)(SvD)


(whom one always sees hanging around.)

b. *som^man ständigt ser attde^ {s tår och hänger.


(*whom| one always"*sees that they^ hang around.)

That the constraint on the choice of the att-Embedding Rule


together with the Wh-Movement Rule is not caused by the Passive
Rule can be seen from (67a) and (67b) where active infinitives
occur in the complement sentences:

(67) a. SPANJORER SOM MAN SER LÄSA TIDNINGAR.(247:221:3/315320)(ST)


(Spaniards whom one sees reading newspapers.)
57

b. *spanjorer som. man ser att•de• !läser tidningar,


. _ • ?•_'
(•Spaniards whom^ one sees that they^ read newspapers.)

The choice of embedding in connection with Wh-Movement is


therefore obligatory. Only the Subject Raising Rule can be
applied to yield structures like (65a) and (66a), that is to
a syntactic deep structure like (45) and not (55). Deep struc­
ture (45) must be posited as the underlying representation of
the passive ACI-Constructions that have undergone Wh-Movement
and Passivization. This is the third syntactic argument for
positing different underlying representations for att-Embeddings
and ACI-Constructions. (Cf. also Bresnan 1972 who argues for a
deep structure complemnetizer node which blocks the movement of
wh-marked noun phrases to sentence-initial position).

5.1,3 semantic evidence for different underlying representa­


tions of aci-constructions and complementizer embeddings

There is also semantic motivation for claiming two different


underlying representations for ACI-Constructions on one hand,
and complementizer- embedded structures on the other. Not only
in the case of the ftwr-Embedding does it seem counterintuitive
to claim the same underlying structure as for the ACI-Construc-
tion, but also for the att-Embedding. One need only take an
example of ACI-Construction or att-Embedding to seethe semantic
difference as far as focus and tense are concerned. In (51)
through (67) above three syntactic reasons were presented for
not deriving ACI-Constructions and att-Embeddings from one and
the same deep structure.

In order to support the claim that the ACI-Constructions obtain


different semantic readings than the hur or att-embedded
structures the reader is referred to the examples of (62) through
(64) above/ where the raised object NP's of the ACI-Constructions
are more in the focus of the speaker's intention than are the
same NP's as subject NP's of the complement sentences under
hur or att-Embedding. Compare also the following ACI-Construc­
tions (b) with similar att-Embeddings (a) as further evidence
that the semantic readings differ:

SE-, 'see' / \

seJr om
att
(68) a. Folk hade gått dit för att fFaglum, Gösta
I hur I
Petterson alltså, skulle vinna.}
•that
(People had gone there to see, < if Fåglum Gösta
how
Petterson, would win.) ^ '
58

b. FOLK HADE GÅTT DIT FÖR ATT SE FÅGLUM, GÖSTA PETTERSON, ALLTSÅ,
VINNA.(1252:63:7/415409) (DN)
(People had gone there to see Eå^glujn, that is, Gösta
Petterson, win.)

(69) a. han har sett j näckrosorna skinit (sken) vaxvita


i brunblanka tjärnar.
(he has seen {u0!jwater lilies sh ne waxen-white in
brown-lustrea torest lakes.)

b. HAN HAR SETT NÄCKROSORNA SKINA VAXVITA I BRUNBLANKA TJÄRNAR.


(719:88:3/115075)(GHT)
( He has seen water lilies shine waxen-white in brown-
lustred forest lakes.)

affi
(70) a. plötsligen såg jag • kurl tre tyska heinkelplan dök
upp vid horisonten. ,
(suddenly I saw/j^** three German Heinkel planes appear-
ed over the horizon.)

PLÖTSLI GEN SÅG JAG TRE TYSKA HEINKELPLAN DYKA UPP VID HORISONTEN.
(568:65:2/212198)(SvD)
(Suddenly I saw three German Heinkel planes appear over
the horizon.)

(71) tills jag såg j^urj långa rader av bomber började falla,

(until I saw long series of bombs started to fall.)

b. TILLS JAG SÅG LÅNGA R ADER AV BOMBER BÖRJA FALLA.(568: 70:3/212198)


(SvD)
(until I saw long series of bombs starting to fall.)

That the semantic readings differ between the (a) and (b)
sentences is unquestionable both in the Swedish sentences and
in the English translations. It is a mystery how this semantic
difference could have been ignored by generative grammarians (cf.
for instance the works of the two American and the two Swedish
linguists mentioned above, in section 5.1.2).49

We claim instead that both the hur and a£t-complementizers yield


different semantic readings than do the ACI-Constructions for
three major reasons: First, the subject- (finite) verb coconsti-
tuency (and possibly number agreement) is preserved in comple­
mentizer-embedded sentences; second, the presence of
tense in att and hur- mbeddings yields more explicit semantic in­
formation than does the tenseless infinitive of the ACI-Construc-
tions; third, the focus of the perceptual (visual, auditory, or
tactual) process is concentrated on the entire fact that some­
thing is being seen, heard, or felt in complementizer-embedded
structures, whereas it is doubtful that ACI-Constructions yield
such a holistic focus on the perceptual process. We claim that
59

it is the object noun phrase following the verb of perception


in the surface structure that is being focused upon in an
ACI-Construction more than the circumstances or the time under
which or at which the act or process of perception is experienced
on the part of the experiencer.(Compare also Anward 1972 (forth­
coming).
Chomsky and others have recently suggested that certain semantic
rules should be formulated in such a way that both deep and
surface structures are involved in the operation of semantic
interpretation, especially such semantic rules which are associated
with focus, scope of negation, quantifiers, and presuppostion
(cf. Chomsky 1970b, 197la, and 1971b,Jackendoff 1968 and 1970,
Katz 1972 etc).50

As further evidence for the semantic difference between comple-


mentizer-embedded structures and ACI-Constructions one can refer
to such sentences where Passivization and Agent Deletion have
operated in connection with the Subject Raising Rule.

After Passivization and Indefinite Agent Deletion occur in the


complement sentence the new subject NP is raised into the
matrix sentence yielding the following passive ACI-Constructions
in the (b) sentences. In the (a) sentences att-Embedding occurs
after Passivization and Indefinite Agent Deletion instead of
Subject Raising:

SE-^ 'see'

(72) a. sett att Södergatan gräves i ett enormt dike.


(seen that the Södergata is being dug into an enormous
ditch.)

b. SETT SÖDERGATAN GRÄVAS I ETT ENORMT DIKE (1282: 3: 3/41549 3)(DN)


(seen the Södergata being dug into an enormous ditch.)

(73) a. dagens uländer som ser att de fattiga raserna obarm­


härtigt hejdas vid USA:s gränser.
(today's developing countries which see that the poor
races are unmercifully being stopped at the borders
of the US.)

b. DAGENS ULÄNDER; SOM SER DE FATTIGA RASERNA OBARMHÄRTIGT HEJDAS


VID USA:S (GRÄNSER)(1239:43:2/323354)(ST)
(today's developing countries which see the poor races
being stopped unmercifully at the borders, of the US)

(74) a. sett att sådan förbättringar förverkligas


(seen that such improvements are being realized)

b. SETT SÅDANA FÖRBÄTTRINGAR FÖRVERKLIGAS (982:92: 2/533664)(SDS)


(seen such improvements being realized)
60

(75) a. vi såg att de överfördes till ambulanserna


(we saw that theyfwere carried over to
t' the
Iwere being carried'
ambulances)

b. VI SÅG DEM ÖVERFÖRAS TILL AMBULANSERNA (13 61:93:7/531664)(SDS)


(we saw them being carried over to the ambulances)

It is clear that the object noun phrases SÖDERGATAN, DE FATTIGA


RASERNA,'the poor races', SÅDANA FÖRBÄTTRINGAR 'such improvements',
and DEM ' them' in (72b) through (75b) are more closely focused
upon than in the corresponding passive <z£t-Embeddings in (72a)
through (75a).

There is for instance in (75b) also a noticeable aspectual


difference to (75a) where on one reading the tense feature
signals a more completed mode of action. The ACI-Construction
in (75b) obtains a durative interpretation, which is rendered
into English by the gerundive construction.

Through the passivization of the complement sentence together


with Subject Raising, a surface structure results that is a
passive ACI-Construction which is semantically different from
the surface structure where an att-complementizer and a passivized
tensed verb are present.

In (72a) it is uncertain whether a present tense is to be


selected, that is GRÄVES 'is being dug' or a past tense as in
GRÄVDES 'was being dug'. The tense feature is neutralized in
the ACI-Construction and must be supplied from the matrix verb
or the linguistic context.

5,2 subject raising and auxiliary deletion or aktionsart


neutralization
In the Gothenburg corpus several cases of Subject Raising have
been found which result in Accusative-with-Past Participle
Constructions in the surface structure. There is no infinitive
form of a passive auxiliary (BLIVA 'become' or VARA 'be') in
the surface structure. One might argue that in (76) through (77)
below no passive auxiliaries have ever been present in the
derivational histories of these sentences.

Acousative-witk-Past Participle
SE^ 'see visually'
(76) a. MAN HADE NOG VELAT SE DEM VERIFIERADE I ANDRA KÄLLOR
(1379:51:5/1250 26) (GHT)
(One would have liked to have seen them verified in
other sources.)
61

b. OCH SE BUTIK EFTER BUTIK FULLPROPPADE AV KÖPMEDVETNA KÖPSTARKA


TONÅRINGAR(1321:25:6/515683) (SDS)
(and see shop after shop stuffed full with eager
affluent teenagers.)

C. FÖRST NÄR HAN SER FAKTORERNA KARTMÄSSIGT FRAMSTÄLLDA


(954:48:3/521684) (SDS)
(First when one sees the factors plotted on the map.)

SE2 'realize'

(77) a. HAR SONEN TILL ÖSTERRIKES KEJSARE FÖRKLARAT ATT HAN SER SIG
TVINGAD TILL ATT VIDTA SPECIELLA ÅTGÄRDER (975:175:2/533629)
(SDS)
(the son of the Austrian Emperor has declared that he
sees himself forced to take special measures*)

b. EN UNG FLICKA,,S0M DÅ HON SER SIG ÖVERGIVEN LÅTER PROSTITUERA SIG


(560:41:1/226174) (SvD)
(a young girl who prostitutes herself, when she sees
herself abandoned«)

C. MEN SER SIG OMGÅENDE VERDERLAGD (1318:108:5/522551) (SDS)


(but sees himself immediately contradicted}

The assignment of Aktionsart to these Accusative-with-Past Participle


Constructions is crucial for correct semantic interpretations. The
two potential passive auxiliaries BLIVA 'become' and VARA 'be',
which reflect two underlying Aktionsarten of the complement
sentences, can be said to constitute the morphological manifesta­
tions of semantic characteristics which are involved here, i.e.
the ingressive or the stative mode of action ('procedural'). In
the Accusative-with-Past Participle Constructions of (76) through
(77), it is not possible to spell out correctly either BLIVA
'become' expressing an ingressive mode of action, or VARA 'be'
expressing a ZustandspassiVj without access to the full linguistic
context. Theoretically one could paraphrase each of the sentences
in (76) through (77) by inserting either BLIVA or VARA in the
following way:

SE^ 'see visually'


?
(76) a'. Man hade nog velat se dem (bli)/'vara verifierade i
andra källor.
(One would have liked to have seen them (*be) verified
in other sources»)
?
b', och se butik efter butik (bli)/ 'vara fullproppade av
köpmedvetna köpstarka tonåringar.
(and see shop after shop (*be) stuffed with eager
affluent teenagers.)
?
c'. Först när man ser faktorerna (bli)/ 'vara kartmässigt
framställda.
(First when one sees the factors (*be) plotted on the map.)
62

SE2 'realize'

(77) a!. har sonen till Österrikes kejsare förklarat att


han ser sig (bli)/ vara tvingad att vidta speciella
åtgärder.
(the son of the Austrian Emperor has declared that he
sees himself (being) (*be) forced to take special
measures.)

b'. en ung flicka som då hon ser sig (bli)/ vara övergi­
ven låter prostituera sig.
(a young girl who prostitutes herself, when she sees
herself (*be) abandoned.)

c'. men ser sig (bli)/ vara vederlagd omgående.


(but sees himself (*be) immediately contradicted)

The Accusative-with-Past Participle Constructions in (76) through


(77) can be regarded as being generated by the operation of a
Passive Auxiliary Deletion Rule in the sense of Chomsky 1965:
10 4-107 which is applied to the same shallow structures that
underlie the ACI-Constructions of (76') through (77'). However,
such a solution is too syntactically oriented. Instead, one could
claim that the plain Accusative-with-Past Participle Construc­
tions are the result of a blocked Passive Auxiliary Spelling
Rule in the sense ef Jacobs & Rosenbaum 1968:108-118.

What is essential for the problem at hand is the assignment of


the underlying Aktionsart which is decisive for the morphological
outspelling of either BLIVA 'become' or VARA 'be'. A careful
study of the linguistic contexts of (76) through (77) results
in a straight-forward stative aspectual assignment to VERIFIERADE
'verified' in (76a), but an ingressive aspectual assignment to
FULLPROPPADE 'being stuffed' in (76b) and FRAMSTÄLLDA 'being plotted'
in (76c). The same ingressive aspectual interpretation must be
assigned to VEDERLAGD 'being contradicted' in (77c), whereas there
is some doubt about the correct aspectual assignment to TVINGAD
'being forced' or 'forced' in (77a) and ÖVERGIVEN 'being abandoned'
or 'abandoned' in (77b).51

In the latter two cases the stative point of view is probably


the Aktionsart which is intended by the writer, but the linguis­
tic contexts do not give clear-cut information here. Compare
the following full linguistic context of (77a) and (77b) as
evidence of the claimed aspectual unclarity. (Because of space
limitation only the contexts of the two problematic sentences
are given in the full text here; for the context of the clear-
cut cases in (76)through (85) the reader is referred to Foot­
note 51):
63

(77) a''. I ett brev till partiets ordförande och koalitions-


kabinettsvicekansler, Dr. Bruno Pitterman* har sonen
till Österrikes siste kejsare förklarat att han ser
sig tvingad till att vidta speciella åtgärder som
kan möjliggöra hans återvändande.
(In a letter to the party chairman and coalition vice­
president of the cabinet, Dr. Bruno Pitterman,the
son of Austria's last Kaiser has declared himself
forced to take special measures which can make his
return possible.)

b,f. Grundtemat är en adelsmans förförelse av e n ung


flÌQkat som då hon ser sig övergiven låter prosti­
tuera sigj medverkar ofrivilligt i ett rånmord,
deporteras till Sibirien och försonas i döds­
ögonblicket under färden till fängelset med sin
nu uppriktigt ångerfulle förförare.
(The basic theme is a noblemans seduction of a
young girl, who prostitutes herself when she sees
herself abandoned, becomes an unwilling accomplice
in a murder for theft, is deported to Siberia and
is pardoned at the moment of death on the journey
to the prison with her now truely remorseful seducer.)

Considering the hesitation on the part of the reader to assign


Aktionsart markers in the form of aspectual auxiliaries to such
Accusative-with-Past participles as (77a) and (77b), it is
doubtful that the Accusative-with-Past Participle Constructions
can be derived via the shallow structures which are exemplified
by the paraphrases in (77a') and (77b'). To his hesitation of
derivational history of the Accusative-with-Past Participle
one can also add the fact that some of the VARA paraphrases in
(761) and (77') are ill-formed or grammatically dubious ACI-
Constructions (see more on this below).

For the sake of descriptive convenience we will treat the


relation between Accusative-with-Past Participle Constructions
and their ACI-paraphrases in terms of an Aktionsart Neutraliza­
tion Rule.

Aktionsart Neutralization seems always to be optional when


BLIVA constitutes the potential passive auxiliary (symbolized
by the parentheses around this auxiliary in (76') and (77')),
whereas it is sometimes obligatory when VARA is the potential
passive auxiliary. Compare for instance the paraphrases (76a1),
(76b1) and (76c'), where VARA as the complement verb of the
passive ACI-Constructions yields semigrammatical sentences, with
the ACI-paraphrases (77a'), and (77bf) which do not seem to be
subject to this surface structure constraint which makes
the passive ACI-Constructions ill-formed.

As the English translations in (76') and (77') reveal that BE as a


passive auxiliary denoting an ingressive mode of action is
ill-formed. An Accusative-with-Gerundive Construction is obliga­
tory here in the form of BEING plus the past participle of the
verb. (Cf. Part II, section 6 on this surface structure con­
straint in English). If BE denotes a Zustandspassiv, then it
64

must be obligatorily deleted. These two facts are symbolized


in (76') through (77') by means of the starred English BE within
parentheses.

The whole issue of whether the auxiliary is to be generated by


the base rules as in Chomsky's Aspects-Theory, or whether it
should be introduced transformationally in the sense of Jacobs
& Rosenbaum 1968 or morphologically in the sense of European
generative linguists, is not only a notational issue. It
seems more adequate and plausible to elaborate constraints on
the morphological rules which spell out passive auxiliaries
under a set of given conditions, than always to generate the
passive auxiliaries in the base and then delete them. However,
space does not allow a detailed investigation of such con­
straints on the outspelling of the passive auxiliaries in
Swedish.

By using the terms Subject Raising and Auxiliary Deletion or


Aktionsart Neutralization we simply indicate that the Accusative-
with-Past Participle Construction is generated, without claiming
that in each example of this surface structure a passive
auxiliary must necessarily be posited to have existed in its
derivational history. A more adequate term than Auxiliary
Deletion is therefore Aktionsart Neutralization.

SE^ 'see' and Ingressive Aktionsart Neutralization (BLIVA


De letion)
(78) a. 1$6k FICK EKIPAGET SE SIG SLAGET MED ENDAST EN NOSLÄNGD
(1567:49:7/215272) (SvD)
(In 1964 the team saw itself beaten by, a nose.)

b. 1964 fick ekipaget se sig bli slaget med endast en


noslängd.
(In 1964 the team saw itself being beaten by a nose.)

(79) a. UPPREPADE PÅ ONSDAGEN ATT DE ÖNSKAR SE KONFERENSEN UPPSKJUTEN


OCH ATT DE I NTE KOMMER ATT.;. (874:13:5/333371) (ST)
(repeated on Wednesday that they wish to see the
conference postponed, and that they will not....)

b. Upprepade på onsdagen att de önskar se konferensen


bli uppskjuten och att de inte kommer att....
(repeated on Wednesday that they wish to see the conference
to be postponed and that they will not ...)

As stated above, native speakers cannot always judge from such


short contexts as those given in (79) through (82) which of
the two passives is intended by the writer, BLIVA or VARA. A
more complete linguistic context is necessary to disambiguate
the two types of Accusative-with-Past Participle. It is evident
that we cannot limit our scope to sentence-oriented descriptions
but must include text-theoretical considerations as well. (Cf.
Isenberg 1968, Petöfi 19711^ Ihwe 1971 etc on text-theoretical
65

aspects for describing sentences). The choice of the passive


auxiliary or Aktionsart marker is crucial for the interpreta­
tion of the sentences given:

SE^ 'see' and Stative Aktionsart Neutralization (VARA Deletion)

(80) a. MEN A LD RI G HAR H A N V ÄL S ET T NÅGO T S Å V ÄL KL ÄT T


(577:81:6/326294) (ST)
(But one has never seen anything so well dressed)

b. men aldrig har man väl sett något vara så välklätt.


(but one has never seen anything (*be) so well-dressed.)

SE2 'realize' and Ingressive Aktionsart Neutralization (BLIVA


Deletion)

(81) a. BAISSESPEKULANTERNA SE R S I G TVUNGNA AT T KÖPA P U N D


(1385:84:1/124022) (GHT)
(stock-market speculators see themselves forced to
buy pounds.)

b. baissespekulanterna ser sig bli tvungna att köpa pund


(stock-market speculators see themselves (*be) forced
to buy pounds.)

SE^ 'realize' and Stative Aktionsart Neutralization (VARA


De letion)
(82) a. FISKAR NA, SOM S E R UPPE HÄL LET H O TA T A V NÖJES FI SK E T
(48:40:5/215208) (SvD)
(the fishermèn who see their living threatened by
sports fishing.)

b. fiskarna som ser uppehället vara hotat av nöjesfisket


(the fishermen who see their living (*be) threatened
by sports fishing.)

(83) a. IN DU ST RI E R SO M SER SI N A EKONOMISKA IN TRESSE N HOT ADE


(1780:47:1/511621) (SDS)
(industries which see their financial interests
threatened.)

b. industrier som ser sina ekonomiska intressen vara


hotade.
(industries which see their financial interests (*be)
threatened.)

A closer investigation of the linguistic contexts of (82a) and


(83a), reveals that in (82a) sports fishing seems to have been
going on for some time in the community so that a stative
Aktionsart can be inferred as underlying the Accusative-with-
Past Participle Construction.
66

The linguistic context of (83a) also yields a stative inter­


pretation of the past participle, which may motivate selecting
VARA as the complement auxiliary in the shallow ACI-Construc-
tion preceeding the Accusative-with-Past Participle Con­
structional

In order to solve the problem of Aktionsart assignment, an


Aspectual Copying Rule will be demonstrated which accounts
for the transformational process which is assumed to operate
between clauses of the same sentence or between the sentence
sequence of a given text (cf. Part Two, section 3). Text-
theoretical considerations are indispensable for describing
phenomena like the semantic interpretation of a surface
structure such as the Accusative-with-Past Participle Con­
struction.

Between phrase marker (84) and phrase marker (85) the Aspectual
Copying Rule transfers the ingressive feature from the time-
adverbial sentence MEDAN POPTONER FY LLDE R UMMET 'while pop-music
tones filled the room' into the dominating complement sentence
TONÅRINGAR FULLPROPPA B UTIK EFTER BUTIK 'teenagers stuff shop after
shop'. The transfer and copying of the ingressive feature is
here symbolized by the broken arrow:

S
(84)

NP VP

och se butik efter butik en påminnelse


S
T Aspectual
Copying
Rule
tonåringar ful" -.ut­
efter butik

medan poptoner fyllde rummet


67

After Passivization and Subject Raising have applied on the


S2 and S^ cycles, the following phrase marker results:

S
(85)

en påminnelse

och se butik efter butik fullproppade a.V i vi n rea -r

Ingressive
medan poptoner fyllde rummet

The Passive Auxiliary Segmentalization Rule may now apply,


spelling out BLIVA to yield (76b'): O CH S E B UT IK EFTER B UT IK BLI F ULL­
PROPPADE AV K ÖPMEDVETNA TONÅRINGAR, MEDAN POPTONER F YLLDE RU MMET. If the
assumed Aspectual Copying Rule does not apply as demonstrated
between (84) and (85), an ill-formed ACI-Construction may result,
which contains VARA as the complement auxiliary before the past
p a r t i c i p l e : * O CH S E B UT I K EFTER B UT I K VA RA FU LL PR OPP AD E A V TO N ÅR I NG A R,
MEDAN POPTONER FYLLDE RUMMET. '*and see shop after shop be stuffed with
teenagers, while pop-music tones filled the room'.

•5.3 nominali zed complements occurring after se x ' see visually'


se 2 ' realize'

As marginal verbal complement structures, nominalized complements


can occur after both SEi 'see' and SE2 'realize'. Sometimes it is
a verb which is nominalized, sometimes it is an adjective which
undergoes nominalization. In connection with the description of
verbal compléments which occur after HÖRA the balanced and un­
balanced views for deriving nominalized verbs from underlying
structures will be discussed in more detail, as nominalizations
seems to play a greater role as verbal complements after HÖRA
than after SE. We refer to the discussion in section 13.5.1 on
the nominalization process in deriving nominalized complement
structures.

It is clear that somebody had been fighting in (86), but by using


the nominalized verb STRIDEN 'the fighting' the writer avoids go­
ing into the details of who where and when.
68

SEX 'see'

(86) EN UNG KVINNA SOM FRÄN BUSSENS ÖVERVÅNING SETT STRIDEN VID HÅLL­
PLATSEN (850:95:3/222209) (SvD)
(a young woman who from the top deck at the bus had seen
the fighting at the bus stop.)

(87) I VARJE FALL VORE DET INTRESSANT ATT SE BELÄGG PÅ VAD DENNA
INSTITUTION... (1684 :124:4/314325) (ST)
(In any case it would be interesting to see evidence of
what this institution....

SEq 'realize'

(88) OCH HAN HAR LIKA SVÅRT ATT SE MÖJLIGHETERNA TILL EN FREDLIG
SAMEXISTENS (10 31:26:3/1260 74) (GHT)
(and it is just as difficult for him to see the possi­
bilities for peaceful coexistence.)

A similar avoidance of explicitly expressing who is going to prove


something can be observed in (87) and who is going to offer the
possibilities fora peaceful coexistence in (88). We postulate
underlying structures which contain human indefinite subject
noun phrases for deriving (86) through (88). These subject NP's
are deleted in the nominalization process.

5,4 accusative-with-gerundive construction and att-deletion

In the Gothenburg corpus only one single example of the Accusative-


with-Gerundive Construction has been retrieved.

(89) LÄNGS NORGES LÅNGA KUST KAN M AN UNDER ALLA ÅRSTIDER SE TIOTUSENTALS
FISKAR HÄNGANDE TILL TORK (960:33:3/515606) (SDS)
(along the long coast of Norway one can see tens of thousands
of fish hanging to dry in all seasons)

Whereas the AWG-Construction is extremely frequent in English, where


it often is obligatory for semantic reasons, it is quite rare in
Standard Swedish (cf. the discussion of ACI and AWG-Constructions
in Swedish, German, and English in Part Two).

Examples of czt£-Deletion after SE, and SE- are also very rare in
the Gothenburg corpus. Only one example of att-Deletion has been
found after SE^ 'see':

(90) WIENOPERAN HADE GÄRNA SETT HAN BLIVIT DÄR (1495:17:3/212147) (SvD)
(The Viennese Ope ra would have liked to have seen him there.)

The Gothenburg corpus gives us a representative picture of the


use of the AWG-Construction and att-deleted sentences. It is by
no means an accident that these two constructions occur only once,
69

whereas the ACI-Construction occurs 43 times and the hur-


embedded sentences 38 times. The table of the frequency of
verbal compléments presented in section 6.1 indicates the
productivity of the embedding rules which are encompassed
in this investigation. Of all the embeddings observed in the
Gothenburg corpus, less than one percent are carried out by
the att-Deletion Rule and the Subject Raising Rule in connection with the
Gerundive Outspelling Rule. Our intuitive feeling about the peripheral
nature of the AWG-Construction and att -deleted embeddings is
verified by the low frequency with which these surface structures
occur in a representative corpus like the one computerized in
Gothenburg (cf. the discussion of lexical and syntactic repre-
sentativity in the Conclusion of the Performance Study, para­
graph 2).

5,5 the surface structures of få se 'catch sight of'


5.5.1 THE CATEGORIA!. STATUS OF SWEDISH F A

In the lexical classification of the perceptual verbs enumerated*


in section 3.1 the verb compound FÅ SE 'catch sight of' is in­
cluded for practical reasons in the list of the ACI-verbs
under the simplex verb SE^ 'see visually', that is as [FÅ SE].
V V
However, from a syntactic and semantic point of view the
labelleing ^ j^FÂ ^ ^ might be better motivated for

reasons to be discussed in this section as far as FÄ SE, FÅ HÖRA,


and FÂ KÄNNA are concerned. To assign the category status of the
verb FÅ is problematic in Swedish. The category assignment of
this verb depends on the kind of theory one is working within.
(We will disregard the meaning 'few' of the adjective FÅ and all
four meanings of the main verb FÅ as described in the SAOB 1928:
cols- 1936-1937 as 'catch' (fånga), 'recover' (få sig),'obtain'
{erhålla), 'give' {överlämna)). There are at least six major
types of the function word FÂ which are here treated as a s ingle
auxiliary and which must be distinguished from one another on
the one hand, açid the four different main verb meanings on the
other.A special investigation would be required to describe the
over-all usage of the Swedish auxiliary FÅ in the six meanings
touched upon here, i.e. the perfective aspectual FÅ^ as exempli­
fied in (91), the future modal FÄ£ as exemplified in (92), the
two modality usages of FÅ3 expressing possibility or necessity
as exemplified in (93) respectively, the permissive auxiliary
FÂ4 exemplified in (95), and finally FÄ5....att 'cause' as
exemplified in (96). (Reasons will be given below for treating
FÂ5 as a main verb, in section 5.5.4)
70

FÂ^ SE 'perfective-aspectual auxiliary1

(91) NÄR JAG KOM GATAN FRAM OCH PÅ AVSTÅND FICK SE ETT UNGT ÄLSKANDE PAR
STÅ DÄR BLICKSTILLA (721:155:1/115070) (GHT)
(when I came up the street and at some distance I saw a
loving young couple standing there without moving}

FÂ2 'future auxiliary'

(92) S E DA N FÅR JAG SE H U R DET BLIR I FORTSÄTTNINGEN(856:68:2/215194)


(SvD)
(then I will see what happens in the future.)

FÂ3 SE 'modal auxiliary'


a) Possibility
(93) a. VARJE ÄRENDE FÖLJS UPP EN TID EFTERÅT SÅ ATT MAN FÅR SE HUR DEN
HANDIKAPPADE HAR KUNNAT ANPASSA SIG (44 7:154:5/214165)(SvD)
(each case is carefully examined so that one can see how
the handicapped person has been able to adapt himself-)
b) Necessity
(94) STUDIERNA I STOCKHOLM FÅR SES MOT BAKGRUNDEN ATT HERRARNA KOMMER
FRÅN... (1290:22:4/414508) (DN)
(The studies in Stockholm must be seen as the result of
the fact that the gentlemen come from )

FÅ4 V 'permissive auxiliary'

(95) INGEN UNDER 21 ÅR FÅR KOMMA IN PÅ ETT FRANSKT KASINO (684:24:4


/515663) (SDS)
(no one under 21 may enter a French casino.)

FÅg ... att V 'cause'

(96) DÅ HOPPAS MAN PÅ ALLVAR KUNNA FÅ ÖKNEN ATT BLOMMA (34:134:5/115069)


(GHT)
(then one seriously hopes to get the desert to bloom.)

We are concerned only with FA^, the perfective-aspectual auxiliary,


in this investigation and will therefore disregard entirely the
usage of FÅ2, FÅ3, FÅ4, and FÂ5. We cannot limit our scope to
this auxiliary without considering the other types of the auxiliary
FA in Swedish, however, as the problem of treating FA^ as a part
of the lexical entry FA SE or as an independent verb is similar
to the problem of how to treat FÅ2, FÄ3, FÂ4, and FÅ5 and their
succeeding main verbs. Shall FA in all these cooccurrences be
treated as a main verb in the sense of Ross 1967b or as an
auxiliary in a traditional sense? (Cf. also the treatment of
Swedish FA by Wagner 1972).
71

5.5.2 C O M P A R A T I V E S Y N T A C T I C E V I D E N C E F O R T R E A T I N G F Â A S AN
ASPECTUAL AUXILIARY

Before going into the notational question of whether to treat FÄ^


as an auxiliary or a main verb in the sense of Ross 1967b,
we want to point out the unique status of Swedish FÄ^ as an
aspectual marker. Comparisons with closely related languages
like Danish, Icelandic, German, and English reveal that no overt
aspectual auxiliary exists in these languages which have any
resemblance to the Swedish FÅ^ (on the existence of FÅ^ in
Norwegian see footnote 52).

Swedish

(9 7) a. när jag kom gatan fram och fick se ett ungt älskande
par.
Danish

b. når jeg kom gaden frem og så et ungt elskende par.


Ice landio

c. jbegar ég gekk niàur götuna og sa ungt åstfangiä par.


Norwegian

d. da jeg kom bortover gaten ok fikk 0ye på et ungt.


elskende par (fikk se et ungt elskende par.)
German

e. als ich die Strasse hinunterging und ein junges liebe­


volles Paar sah.
English

f. when I went down the street and saw a young loving


couple.

Danish informants do not accept fik se in (97b) and suggest


fik 0ye på instead to express the instantaneous indeliberate
perception. Icelandic informants do not accept fêkk sjâ in the
sense of 'caught sight of'. Norwegians hesitate to accept
fikk se, though they do not consider it to be ill-formed.52
German native speakers often replace sah by erblickte. In
English the phrase 'caught sight of1 can be used instead of saw.

Since the perfective aspectual status of the Swedish FÂi in


(91) is unquestionable and since it therefore is an oveit aspec­
tual marker, it is descriptively motivated to regard FÅ^ as an
auxiliary. Aspectual markers like the perfective aspect prefixes
in the Slavic languages or the perfective or imperfective aspect
auxiliaries in Germanic languages (e.g. Swedish FÂ^ and the
English ^n^-construction) are surface manifestations of semantic
Aktionsarten or procedurals (cf. Forsyth 1970 and Isaéenko 1960).
From a comparative point of view it is more practical to regard
both affixes and auxiliaries which express semantic Aktionsarten
as surface phenomena. It therefore suffices to propose morphologi­
cal outspelling rules which segment out either affixes, as in
Slavic languages, or auxiliaries as in Germanic languages. If
72

such an approach to the surface differences between the languages


mentioned here is accepted, the syntactic base rules and deep
structures can be kept universal as fa r as the notation and
generation of aspectual markers are concerned. The surface struc­
ture differences between languages which take affixes and
languages which take aspect auxiliaries can then be expressed
in terms of language-specific transformational rules or morpho­
logical rules depending on which transformational theory one is
working with. (Cf. footnote 31 and Part Two, section 4 on overt
aspectual markers).

Chomsky's 1965:107 base rule: (xvi) Aux—*Tense (M) (Aspect) is


too surface-structure oriented and toe specific only for English
to be useful for a comparative approach to syntactic structures
and to capture the interlingual generalization which can be ex­
pressed in terms of different aspectual morphological rules. Ross
1967b is also too syntactically biased in his discussion of
auxiliaries as main verbs. He posits a syntactic deep structure
which contains the modal auxiliary as a dominating node over
the verb which is to be infinitivized. His postulation of such
deep structures also conflicts with the goal of constructing a
comparative syntax, that is to d escribe surface structure
differences in terms of common syntactic deep structures but
language-specific transformational or morphological rules.
Following Ross's 1967b notation (cf. p. 10) we would get the
following language-specific deep structure representation of the
Swedish ACI-Construction which occurs after fick se in (91):53

(98)

jag
+Mom
-Delib
+Past

jag se de't'

ett ungt älskande par stå- där

Under the assumption that two coreferential subject NP's exist,


the second coreferential NP^ is deleted by t he Equi-NP Deletion
Rule and after It-Replacement in the sense of Ross on the
cycle sentence (91) would result. Ross's suggestion is, however,
inadequate for comparative purposes because it is too language-
specific. Underlying structure (98) is so language-sepcific for
Swedish that the corresponding structures as exemplified in
(97b) through (97f) would hardly be describable in terms of a
common underlying structure, since Danish, Icelandicj German,
and English lack an overt aspectual perfective marker.
73

Instead of such a language-specific solution we suggest a


derivation of FÅ^ in the verb compounds FÄ SE, FÄ HÖRA,FÅ KÄNNA
etc as a result of a language-specific morphological rule which
operates on a structure which can be claimed to be common to
the Germanic languages discussed.

(99)

NPi

+V ett ungt älskande par


Jag +Mom
-Deliberate
+Pas t

ett ungt älskande par


stå- där stilla

After Auxiliary Outspelling and under Tense and Aktionsart


Copying onto the new auxiliary segment, the following shallow
structure results before Subject Raising takes place:

(100)

'i
I Aux
Jag
+Aux
+Pas t
+Mom
-Delib

fick ett ungt äls­


kande par stå-
där blickstilla

It is difficult to say whether the semantic features for Aktions­


art in the main verb should be retained in the shallow structure
representation (100) after they have been copied into the
auxiliary segment. This is a notational question and is of no
concern for the issue at hand. It is the existence of the morpho­
logical process demonstrated between phrase structure (99) and
phrase structure (100) which distinguishes Swedish from the
other Germanic languages. There is a similarity to the Slavic
languages, which comparative generative syntax can describe
in terms of a general morphosyntactic process. For instance in
Russian and Polish a morphological rule can be worked out which
spells out a specific prefix instead of an auxiliary as in Swedish.
74

After Aktionsart Copying of the feature [+Mom] (standing for


instantaneous mode of action) and [-Delib] (standing for in­
deliberate mode of action) a prefix results, i.e. /u/ before
the main verb, slyèal or videi, which does not carry tense or
gender, because the latter features of the verb are not copied
from the verb segment. (Compare the Swedish phrase marker in
(100) where the tense feature is copied onto the auxiliary
segment, leaving the main verb tenseless). The phrase markers
(101a) and (101b) demonstrate the operations of the two morpho-
syntactic processes discussed here:

(101a) (101b)

VP j Aktionsart
1 Copying
y
+
"+V ••Prefix
+Mom Aspectual •»Mom +Past
-Delib Prefix ^Delib ••Mas c
+Pas t Segmentali-
+Masc zation
y'
On Is ly5 all
[videi 1

(He (heard]) (He (heard))


saw 1 t
In the light of the comparative evidence presented here from
Germanic and Slavic languages an underlying representation in
the form of (98) appears extremely ad hoc. In postulating under­
lying representations for various surface structures one must
consider what consequences such underlying structures will have
for the goal which a linguist chooses in his description of a
given set of surface structures.

There are no underlying structures which are predetermined to


suit all purposes. It may very well be the case that one under­
lying representation is appropriate for describing semantics,
if this happens to be the linguist's concern, and another under­
lying representation is more appropriate for describing compara­
tive syntax. Since the goal outlined in this investigation is
to make a comparative description of verbal complementation in
Germanic languages, we are convinced that the underlying re­
presentation suggested in (99) is more adequate for comparative
syntax than the Ross type of representation indicated in (98).
75

5.5.3 THE SENTENTIAL COMPLEMENT STRUCTURES OF FÅ SE, 'CATCH


S I GH T O F 1 1

ACI-Cons truotion

(102) N ÄR J AG KOM G AT AN FRA M OCH PÅ A VS TÅ ND F I CK SE ETT U NG T ÄLSKANDE


PA R STÂ DÄR BLICKST ILLA ( 7 2 1 : 1 5 5 : 1 / 1 1 5 0 7 0 ) ( G H T )
(when I came up the street and at some distance saw a
loving young couple standing there without moving.)

att-Embedding

(103) när jag kom gatan fram fick jag se att han kysste henne.
(when I came up t he street I suddenly saw that he kissed
her.)

hur-Embedding

(104) när jag kom gatan fram fick jag se hur han slog till
henne i ansiktet med knytnäven.
(when I came up the street I saw how he hit her face
with his fist.)

Accusative-with-Past Participle Construction

(105) 19 64 FICK E KIPAGE T SE S IG S L AG E T M ED EN DAS T EN N0SLÄNGD(1567:49:7/


(In 1964 the team saw itself beaten by a nose.) 215272)(SvD)

5.5.4 S E M A N T I C E V I D E N C E F O R T R E A T I N G F Â . A S AN A S P E C T U A L
AUXILIARY 1

As evidence for the assumption that there are semantic features


of Aktionsart which are associated with FÄ} one can point out
the fact that these semantic properties delimit the range of
main verbs which can cooccur with FÂi- Both Modern Swedish SE^
|see', HÖRA 'hear1, and KÄNNA]- 'feel tactually' may convey an
instantaneous perceptual process and can therefore cooccur with
the perfective auxiliary FÄ^ :

(106) a. Han fick höra henne tala illa om sin väninna.


(He heard her slandering her girlfriend.)

b. Han fick se henne hoppa till.


(He saw her give a hop.)

c. Han fick känna pulsen slå hastigt.


(He felt the pulse beat fast.)
76

However, Få^ cannot cooccur with other verbs of visual or


auditory perception which are of inherent instantaneous (e.g.
MÄRKA 'notice'^and UPPTÄCKA 'discover') or inherently durative
(e.g. OBSERVERA 'observe'5and IAKTTAGA 'watch') Aktionsart:

(107) a. *Han fick plötsligt märka henne i rummet.


(He suddenly noticed her in the room.)

b. *Han fick plötsligt upptäcka sin vän på bryggan.


(He suddenly discovered his friend on the bridge.)

c. *Han fick observera henne på gården.


(He observed her in the yard.)

d. *Han fick iakttaga honom på gården.


(He watched him in the yard.)

The ill-formedness of *(107a) through*(107d) is so great that the


translation into English must be based on the simplex verbs
in the past tense. In*(107c) and *(107d) there is semantic
incompatibility between the durative Aktionsart implied in OBSER­
VERA 'observe' and IAKTTAGA 'watch' on the one hand, and the
perfective aspectual auxiliary FÅ^ on the other, which has been
spelled out here incorrectly. The cooccurrence of FÅ-^ and these
two inherently durative verbs of intentional perception is there­
fore automatically blocked by a semantic constraint. Such seman­
tic blocking demonstrates that the instantaneous Aktionsart is
associated with FÂ^. (FÄ is naturally possible as auxiliary in
the sense of 'be allowed to' in (107c) and (107d), that is FÂ4.)

In *(107a) and *(107b) the cooccurrence of the perfective


auxiliary FÅ^ and the inherently instantaneous verbs of non-
intentional perception like MÄRKA 'notice' and UPPTÄCKA 'dis­
cover' also produces ill-formed sentences, probably because of
a redundant constellation of two verbal categories through
which both are associated with instantaneous Aktionsart.

Such cooccurrence restrictions as exemplified for Swedish in


(107) can easily be stated in terms of constraints on the
morphological rule which spells out FÂi as a perfective auxiliary.

Similar constraints can be stated on the outspelling of aspectual


prefixes in Slavic languages, where for instance the perfective
aspectual prefix /u-/ must not be segmented out from the verb
segment, because it is ill-formed as an overt aspectual marker
in the position before verbs of instantaneous Aktionsart, i.e.
in Russian examples such as otjnuloja 'woke up 1 in (108a)
and priexal 'entered' in (108c) as compared to the ill-formed
examples *uotjnulcja 'woke up' in *(108b) and *upriexal in
*(108d) :

(108) a. Ivan otjnulcja.


(Ivan woke up.)

b. * Ivan uotjnulcja.
(Ivan woke up.)

c. Ja tjital knigu, kogda on priexal.


(I was reading the book, when he entered.)
77

(108) d. *Ja tjital knigu, kogda on upriexal.

As a further argument for not considering auxiliaries like the


perfective aspectual FA.9and the modality FA3 for example, as
main verbs in sense of Ross 1967b one could refer to the
passive structure of FÂ3 SE 'must see1 in (109). It is not FÂ3
'must1 that obtains the passive /-s/ marker, as one would
expect if FA3 'must' were a main verb, but the real main verb,
i.e. SE 'see^. Compare the following well-formed (a) sentence
from the Gothenburg corpus with the ill-formed constructed (b)
sentence :

FA J 'must'-

(109) a. STUDIERNA I STOCKHOLM FÅR SES MOT BAKGRUNDEN ATT ...


(1290:22:4/414508) (DN)

(The studies in Stockholm must be seen as the result


of the fact that )

b. *studierna i Stockholm fås se mot bakgrunden att....

There is only one example of a passivized FA in the Gothenburg


corpus that can take the passive /s/, that is FÂ5 'cause',
which is therefore considered as a true main verb :

FÂ5 'cause'

(109) c. VÄRMDE UPP DJURET IGEN, KUNDE DET FÄS ATT ÖVERLEVA OCH T.O.M.
REPRODUCERA SIG (1347:28:3/521584) (SDS)
(if the animal were warmed upp again, it could be
made to live and even to reproduce.)
78

6. RESULTS OF THE LINGUISTIC PERFORMANCE STUDY ON VERBAL


COMPLEMENTS OCCURRING AFTER SE AND FA SE
Of the total 1502 examples of the verb SE 'see' retrieved in
the Gothenburg corpus ten per cent take some kind of an embed­
ding, i.e. 123 embeddings occur after SE^ 'see visually',
30 embeddings occur after SE2 rrealize, understand', and only
2 examples after SE3 'prefer to see'. The total number of
embeddings is therefore 155 occurring after the homograph SE.

A closer examination of the 155 embeddings which occur after


the three verbs óf visual and cognitive perception in three
tense forms (the present tense SER 'see(s)', the past tense
SÅG 'saw'), the past participle SETT 'seen'and the infinitive
form SE 'see' reveals that the Subject Raising Rule is a
very productive embedding rule in sentences which contain SE^
'see visually' as the main verb. The distribution of the types
of embedding can be seen in Tables 1 and 2 :

6.1 aspects of the productivity of embedding rules after se}


'see visually'

Table 1. THE PRODUCTIVITY OF EMBEDDING R ULES AFTER SE]


1 see v isually1

Type of Embedding Number Distribution of Embed-


ding Types after the
Verb SE1 'see visually'
in the Gothenburg
corpus
1) ACI-Constructions with
active or passive 43 34.9 %
infinitives
2) fcur-Embeddings 38 30.9 %
3) Accusative-with-Past
Participle Constructions 19 15.5 %
(VARA or BLIVA Deletion)
4) att-Embeddings 17 13.8 %
5) Nominalizations 4 3.3 %
6) Accusative-with-Gerundive
Constructions 1 0.8%
7) att-Deletion 1 0.8 %

Total number of embeddings : 123


Total number of examples
of the homograph SE 'see'
in the Gothenburg corpus : 150 2
79

The statistics presented in Table 1 refer only to the news­


paper texts processed by the Gothenburg Research Group. It is
a striking fact that the ACI-Constructions together with the
structures that have also undergone Passive Auxiliary Deletion con­
stitute slightly more than half of all the embeddings which occur
after SE^ 'see visually', i.e. 50.3 4. The Subject Raising
Rule thus generates ACI-Constructions which are more frequent
as embeddings after SEi than the complementizer embedded struc­
tures (hur and att-Embeddings). The latter type of verbal
complementation encompasses 44.7 % of all embeddings (30.91
for ftwr-Embeddings and 13.8 % for att-Embeddings).

Three peripheral verbal complements which occur after SEi


are Nominalizations, Accusative-with-Gerundive (AWG-) Con­
structions, and the embedding which has undergone att-Deletion.
(The complementizer att is considered here to have been in­
serted, but later deleted (cf. example (90)). Together these
surface structures barely constitute 5 % of all the embed­
dings which occur after SE-^. For co mparative purposes it is
important to note the very low frequency of the Swedish AWG-
Construction which is known to be such an important type of
verbal complementation in English. However, in written Modern
Standard Swedish it plays an insignificant role as is shown
by its distribution of 0.8 % of all embeddings in the Gothen­
burg corpus in Table 1.

6,2 aspects of the productivity of embedding rules after


se 2 realize, understand

Table 2 reveals that SE2 'realize, understand' takes no ACI-


Construction as verbal complementation on Modern Standard
Swedish as evidenced by the newspaper texts of the Gothenburg
corpus. (The semantically related INSE 'realize' can take
neither ACI-Constructions nor Past Participle Constructions,
but only att and Tzwr-Embeddings). The Subject Raising Rule
consequently does not embed sentences after SE2 in order to
yield AÇI-Constructions. This fact is surprising in light of
the numerous ACI-Constructions which occur after SE-^ 'see
visually'. (There should be no semantic blocking of ACI-Con-
structions after SE2, because there are verbs of cognition
in Swedish which do take ACI-Constructions, e.g. VETA 'know',
TRO 'believe', T ANCA 'think' etc (cf. Part Three)). However,
the Subject Raising Rule generates Accusative-with-Past
Participle constructions together with the Passive Auxiliary Rule.
These two rules in coordination yield 20 % of all embeddings
which occur after SE2 'realize, understand'.

The most frequent embedding rules are, as expected of this


kind of verb, the att- and hur-Complementizer Embedding Rules
which embed almost 75 % of all verbal complements after SE2.
80

Nominalizations occur marginally as verbal complements after


SE2 'see cognitively', i.e. 6.7 % of the occurring embeddings
consist of nominalizations. The Nominalization Rule can there­
fore be said to apply infrequently in order to embed sentences
into each other.

Table 2. THE PRODUCTIVITY OF EMBEDDING R ULES AFTER SE^realize'

Type of Embedding Number Distribution of Embedding


Types after the Verb SE2
in the Gothenburg corpus

1) att-Embeddings 15 50 %
2) ftwr-Embeddings 7 2 3.3 %
3) Accusative-with-Past
Participle (VARA and 6 20 %
BLIVA Deletion)
4) Nominalizations 2 6.7 %
5) ACI-Constructions 0 -

6) Accusative-with-Gerundive 0 -
Constructions

Total number of Embeddings : 30


Total number of examples
of the homograph SE in
the Gothenburg Corpus : 150 2

6.3 aspects of the productivity of embedding rules after


helst ses 'prefer to see'

Since only two examples of SE3 'prefer to see' have been found
in the Gothenburg corpus and since they both take att-Embeddings,
no table of distribution is necessary. Instead we prefer to
present the two examples as follows :

SE3 HELST 'prefer to see'

(109) C. O CH H E LS T SÅ G HERRARNA A T T ABC FOR TSATT E SI N VERKS AMHE T


(1755:13:3/512698) (SDS)
(and the gentlemen would prefer to see the ABC
continue its activity.)

d. E FT ER DEN P UB LI CI TET SA KEN FÅT T S Â G R EG E RI N G EN FÖRMODLIGEN


HELST ATT MIHAJLOV FÖRBLEV (921 :308 :3/422507) (DN)
(after the publicity which the matter has received
the Government would rather see Mihajlov remained..)
81

6.4 aspects of the productivity of embedding rules after


FA se x ' catch sight of'

The examples of the homograph FÅ in the Gothenburg corpus are


the most numerous of all the verbs investigated here, that is
4382, of which 174 are examples of the adjective FÅ 'few'. The
4208 other examples cover all five different classes of FÅ
discussed in section 5.5, i.e. the perfective-aspectual auxiliary
FA]_, the f uture auxiliary FÄ2, the modality auxiliary FÅ3, the
permissive auxiliary FA^ the causative main verb FÅ5 'cause',
and the main verb FÄ 'obtain'.

As no detailed investigation of the distribution of other


types of FA than FAi is intended here, only the statistics
which are relevant for FA SE 'catch sight of' have been included
in Table 3 below :

Table 3. THE PRODUCTIVITY OF EMBEDDING R ULES AFTER FÅ SE1


'catch sight of1

Type of Embedding Number Distribution of Embedding


Types after the Verb FÅ SE
in the Gothenburg corpus
0
0

1) fcur-Embeddings 3
<*=>

2) ACI-Constructions 1 20 *
3) Accusative-with-Past
Participle Constructions 1 20 %
4) att-Embeddings 0 -

5) Nominalizations 0 -

6) Accusative-with-Gerundive 0 -

Constructions

Total number of Embeddings : 5


Total number of examples
of the homograph FA SE-,
'catch sight of' in the
Gothenburg corpus: 53
82

7. THE SUBJECT RAISING RULE IN SWEDISH (MRRKA)


7,1 the verbal complements of märka 'notice'

In Svenska akademiens ordbok (the SAOB) more than a dozen


definitions are given for the lexical entry MÄRKA 'notice'
(cf. op. cit. Volume 17:cols. 1973-1982). We will disregard
the first twelve meanings of this word given there and con­
centrate on the meaning of MÄRKA which is defined as 'to obtain
impressions through the senses in an intellectual or intuitive
way' (col 1978). The stimuli implied in the usage of MÄRKA
can be of either an acoustic or visual nature, but as far as
UPPTÄCKA is concerned mostly visual stimuli are implied. The
SAOB gives three verbal complements as possible embeddings after
Swedish MÄRKA, i.e. a) the ACI-Construction, b) att-Embedding,
and c) the Accusative-with-Past Participle Construction. Due
to incomplete data-coverage in the Gothenburg corpus only the
second of the verbal complements has been found. Since the ACI-
Construction and the Accusative-with-Past Participle Construc­
tion do occur after MÄRKA, these verbal complements will also
be described here as verbal embeddings after MÄRKA.

The Gothenburg corpus does, however, contain a type of embedding


which is not exemplified by the SAOB under the thirteenth de­
finition of MÄRKA in col. 19 79, that is the att-Embedding as in
(111) :

att-Embedding

(110) a. BRYTER HAN UPP FRÅN LAGR ET. KAMRATERNA MÄRKER KNAPPT A T T H A N
GÅR . DE H A R H ÄNDE RNA FULLA ( 1 6 5 3 : 4 4 : 6 / 3 2 6 2 9 4 ) ( S T )
(he starts to leave the magasine. His companions
hardly notice that he is leaving.)

b. V E M MÄ RKER ATT ET T PA R L ÅN G A FINGRAR S AMT ID IGT STICKS N ED


(1205:40:1/215156) (SvD)
(Who notices that a couple of long fingers are being
stuck down at the same time.)

C. VI MÄRKTE A T T Ö L ET BORNERADE M I ND RE Ä N V A D M AN Ä R V A N
£999 :28 :7/515621) (SDS)
(We noticed that the beer foamed less than one is
accustomed to.)

hur-Embedding

(111) H ÄR MÄ RKER HAN B Ä ST H UR MY CKET HAN LÄ RDE S IG UND ER S I N A. . .


(932 :139 :1/423416) (DN)
(Here he notices best how much he learned during
his...)
83

From the SAOB the following example of the ACI-Construction has


been excerpted, as there is no evidence of this verb complement
type after MÄRKA in the Gothenburg corpus :

ACI-Construotion

(112) HAN LYSSNADE UTÅT , S OM OM HA N G EN OM S TO RM EN MÄRKTE SI G H ÖR A


ANDRA LJ UD (St rindberg Sv. ö. 2 :292 (1883): SAOB
1945 :col. M 1979)
(He listened as if through the storm he noticed that
he heard other sounds.)

The example (112) is undoubtedly literary. The high stylistic


value of the ACI-Construction after MÄRKA is probably the
reason that it does not occur in newspaper texts, although one
can imagine contexts where a literary style is not necessary
for using ACI-Constructions after MÄRKA (cf. Part Two, section 2)•

A second verbal complement type, not documented in the Gothen­


burg corpus but discussed by the SAOB, is the Accusative-with-
Past Participle Construction which is generated by Subject
Raising in combination with the Passive Auxiliary Rule (BLIVA
or VARDA Deletion). Compare the two following examples given
by the SAOB:

7.2 accusative-with-predicative attributes


7.2.1 ACCUSATIVE-WITH-PAST PARTICIPLE CONSTRUCTION

(113) a. INGEN S AG SI G FÖ RSM ÅD D, OCH INGEN MÄRKTE SI G G Y N NA D


(Runeberg 1 :200 (1836): SAOB1945 : col. M 1979)
(no one felt neglected, and no one noticed that he
was favored.)

7.2.2 ACCUSATIVE-WITH-PREDICATIVE ADJECTIVE CONSTRUCTION

b. MÄRKS DEN LI LL A KROPPEN SL AP P / LI FV AS HAN MED GISSELRAPP


(Rydberg: Dikt. 2:81 (1891): SAOB 1945 : col. M 1979)
(if one notices that the little body is loose, then
it is livened with lashes.)

The literary flavor to the Accusative-with-Past Participle


Constructions in (113) is also indisputable after MÄRKA, al­
though no modern speaker of Standard Swedish would experience
(113a) to be so literary, that it could not be used in a normal
speech situation, formally, however, an att-Embedding without
Passive Auxiliary Deletion is selected as verbal complement in
sentences like (113a') ,and( 113b1):

(113) a', ingen såg sig försmådd, och ingen märkte att han
var gynnad
(no one felt neglected, and no one noticed that he
was favored.)
84

(113) b1, märktes det att den lilla kroppen var slapp, då
livades den med gisselrapp
(if one noticed that the little body was loose, then
it was livened with lashes)

The Subject Raising and Auxiliary Deletion Rules seem to carry


a considerable stylistic effect in (113) as far as the per­
ceptual verb MÄRKA 'notice' is concerned. Such a stylistic effect
is not accomplished by deleting the passive auxiliary alone
after SE 'see'. But by deleting the passive auxiliary after såg
as well as märkte in (113a), the writer leaves it open to the
reader to assign the Accusative-with-Past Participle either a
perfective passive reading (BLIVA) or a stative passive reading
(VARA)fZustandspassiv.

There is also an Equal Subject Constraint which blocks Subject


Raising from applying if the matrix and complement subjects
are not coreferential. Strindberg's sentence (112) would have been
impossible if the two subjects referred to two different per­
sons :

(112) ' *Han lyssnade utåt, som om han^ genom stormen märkte
hennej höra andra ljud
(He- listened as if through the storm she. heard other
sounds) ^

A third verbal complement which has been discovered in the


Gothenburg corpus, but which has not been treated as an embed­
ding type by the SAOB is the nominalization type of verbal comple­
ment after MÄRKA .

7.3 nominalizations

(114) a . HA N MÄRKER FÖRÄNDRINGEN INTE MI N ST P Å DEN D EL A V P ER SO N AL EN .. .


(1006 :135 :2/51581) (SDS)
(One notices the change, not least from the reaction
of the section of the employees)

b . P Å LÄNSSTYRELSEN HA R MAN D OC K INTE MÄ RKT NÅG ON N ÄMNVÄR D Ö K N I N G


AV A NT AL ET LÅNEÄRENDEN ( 5 0 5 : 3 7 : 1 / 2 1 3 1 4 4 ) ( S v D )
(At the county government one has not noticed any
significant increase worth mentioning of the number
of applications for loans.)

C . O C H PÅ I D EON MÄR KER M AN ING EN AVMAT TNING I PUB LIKI NTR ESS ET
(1755:43:3/512698) (SDS)
(and at the Idêon Theater one notices no decrease in
the interest of the audience.)
In generative literature the derivational histories of such
nominalizations as exemplified in (114) have been discussed.
It is clear that in these sentences a clear-cut recategorization
rule can be formulated which operates on underlying complement
sentences. By positing such a recategorization, the relation­
ship between (114) and (115) can be linguistically explained
in terms of a transformational process:

(115) a. man märker att något förändras inte minst på den del
av personalen
(one notices that something is changing, not least
from the reaction of the employees)

b. på länsstyrelsen har man dock inte märkt att antalet


låneärenden ökat nämnvärt.
(at the county government one has not noticed that
the number of applications for loans have increased
significantly)

c. och på Idéon märker man inte att publikintresset


avmattas.
(and at the Idéon Theatre one does not notice that
the interest of the audience is decreasing.)

The underlying complement structures for deriving the nominalized


structures of (115) can be postulated to something like the
following :

(115) a1. man märker [Pro förändras]


si si

(one notices [Pro is changing])


si si

b'. ...man har märkt [antalet låneärenden ökar]


si si

(...one has noticed [the number of applications


si

for loans is increasing])


si

man märker inte [publikintresset avmattas]


si si

(one does not notice [the interest of the audience


S1
is decreasing])
Sn

For further discussion of the balanced and unbalanced views on


nominalizations see also section 13.5.1. It is unclear whether
there is syntactic motivation for claiming a transformational
recategorization of the posited complement verb in all those
86

cases where a nominal occurs. A transformational derivation of


nominals seems motivated where we have corresponding verbs as
e x e m p l i f i e d f o r ( 1 1 4 a ) b y F Ö RÄ N D RA S ' c h a n g e 1 , f o r ( 1 1 4 b ) Ö K A
'increase', and (114c) AVMATTAS 'decrease'.

However, nominals often occur which obtain a metaphorical inter­


pretation (cf. (116a) or nominals may express something without
reference to the person involved or to th e time or place (cf.
(116b)).

(116) a. TROTS D E T TA MÄRKER MAN EN MYCK ET B EST ÄM D VILJA I FÖRESTÄLLNINGEN


(39 :110 :2/226253) (SvD)
(Inspite of this, one notices a very determined will
in the performance.)

b. INOM BA R NS JU K VÅ RD E N H AR MAN MÄRKT DETTA BE H O V


(1896:145:2/123020) (GHT)
(in medical care for children one has noticed this
need.)

In (116a) as well as (116b) one cannot judge what the under­


lying deep subject KP would be. Underlying structures like
(116a') and (116b') appear far-fetched and construed for de­
scribing the semantic interpretation of (116a) and (116b):

(116) a'. trots detta märker man [ Pro will något i före­
set*Hum]
ställningen]

(despite this one notices [ Pro wants something in


S^Hum]
the performance])

b'. man har märkt [ Pro behöver något]


S^t+Hum]

(one has noticed [ Pro needs something])


S^t+Hum]

A more balanced approach to the nominals in (116a) and (116b)


would be to regard such structures as a result of word forma­
tion rules within the lexicon. We need not postulate transforma­
tional derivations for all nominals as did Lees 1960, rather
we can regard some nominals to be lexical entries as suggested
by Chomsky 19 70a and others. In describing the occurrence of
nominalization in the Gothenburg corpus, we have only con­
sidered such cases where the transformational history is un­
questionable, as in (114) above. The metaphorical usage of cer­
tain nominals like VILJA 'will' in (116a) motivates us to exclude
such lexicalized items from being treated as products of
embedding.
87

8, RESULTS OF THE LINGUISTIC PERFORMANCE STUDY ON VERBAL


COMPLEMENTS OCCURRING AFTER MÄRKA 'NOTICE7
Of the 53 examples of MÄRKA 'notice' retrieved in the Gothenburg
corpus the distribution of embedding types is as follows :

8,1 THE PRODUCTIVITY OF EMBEDDING RULES AFTER MÄRKA 'NOTICE'

Table k. THE PRODUCTIVITY OF EMBEDDING R ULES AFTER MÄRKA


1 no t i ce 1

Type of Embedding Number Distribution of Embed­


ding Types after the Verb
MÄRKA 'notice' in the
Gothenburg corpus

1) att-Embeddings 25 59.5 %
2) Nóminalizations 13 30.9 %
3) /zur-Embeddings 4 9.5 %
4) ACI-Constructions 0 -

5) Accusative-with-Past
Participle Constructions 0 -

6) Accusative-with-Gerundive
Constructions 0 -

Total number of embeddings : 42


Total number of examples
of the homograph MÄRKA
'notice' in the Gothenburg
corpus : 74
88

9, THE SUBJECT RAISING RULE IN SWEDISH (UPPTÄCKA)


Both UPPTÄCKA 'discover' and MÄRKA 'notice' belong to a class
of verbs which expresses an instantaneous passive process of
perception on the part of the human subject, the experiencer,
i.e. Class (2b). All verbs treated so far are so-called
passive perceptual verbs in the sense that the human experiencer
passively registers events or states visually or cognitively
(cf. SEi 'see visually', SE2 'realize', and MÄRKA 'notice').
Also UPPTÄCKA 'discover visually, realize', requires a deep
subject which has the role of an experiencer. There are, how­
ever, perceptual verbs which can be classified as active- verbs
of perception such as OBSERVERA 'observe' or IAKTTAGA 'watch'.
The latter verb category, Class (2a), requires an agent in the
sense of Fillmore 1971 as a deep subject and the whole process
of visual perception is an active visual process symbolized
by the feature |+PurposeJ. There is also a significant difference
between the durative Aktionsart inherent in the OBSERVERA and
IAKTTAGA verb category on the one hand, and the instantaneous
Aktionsart of the UPPTÄCKA and MÄRKA verb category on the other.

The two groups of verbs thus differ as far as the inherent


semantic properties of Aktionsart are concerned as well as
the kind of roles their human deep structure NP's play.

Since UPPTÄCKA, OBSERVERA, and IAKTTAGA are all verbs of visual


perception and therefore/ belong closer together semantically
with SE 'see' than with HÖRA 'hear', they are treated in con­
nection with SE.

9,1 the verbal complements of upptäcka^ 'discover visually*

A semantic survey of UPPTÄCKA 'discover' as it occurs in various


examples of the Gothenburg corpus reveals that this homograph
is a cover-term for at least four major meanings : UPPTÄCKA^
'discover visually', UPPTÄCKA? 'discover cognitively', UPPTÄCKA,
'find out', and UPPTÄCKA4 'discover a talent', östergren 1968:
cols. 330-331 enumerates all four meanings of UPPTÄCKA. Since
only UPPTÄCKA!, UPPTÄCKA2, and UPPTÄCKA3 take embeddings, the
last type of the homograph will be touched upon marginally.

östergren 1968 gives two synonymous expressions to describe the


meaning of UPPTÄCKA1 'discover visually', that is BLI VARSE and
FÅ ÖGONEN PÅ 'catch sight of'. By g iving these synonyms the
instantaneous Aktionsart as well as the experiencer role of the
deep subject are implied but not directly stated in the dic­
tionary. OBSERVERA and IAKTTAGA are paraphrased as VARA VARSE 'keep
the eyes on'. With the latter synonym the dictionary expresses
the durative Aktionsart, but omits an express statement of the
agent role of the human deep subject. (Cf. östergren 1931).
89

Traditional dictionaries like the SAOB and östergren's Nusvensk


ordbok are valuable sources of information on the Swedish
language, but they suffer from being inexplicit and informal as
far as Aktionsart and subject roles are concerned. They require
that their users supply the necessary grammatical information
to understand a lexical item. This subconscious knowledge is
in the focus of a theoretical linguist's interest. A linguist
who wants to explicitly describe verb classes in terms of deep
subject roles, inherent semantic properties, and Aktionsarten
relies less than a traditional dictionary does on such sub­
conscious knowledge of the native speaker when he describes
lexical entries like verbs.

UPPTÄCKA'discover visually'

att-Embedding

(117) a. N ÄR J A G F I C K SYN PÅ TJUR EN OCH UPP TÄ CKT E A T T DET V A R EN S VÄ R


TJURC364:135:5/115065) (GHT)
(when I caught sight of the bull and discovered that
it was a troublesome bull.)

b N ÄR AMERIKANARNA F Ö R NÅGOT ÅR SED AN UPPTÄCKTE A T T DE T HOS D ER AS


AMBASSAD I MOSKVA F ANNS (58 7:2 2 3: 3:/311379) (ST)
(When the Americans discovered about a year ago, that
there was in their embassy in Moscow)

C. SA M IN VÄN L . O CH FAKTISK T UPPTÄCKTE HAN IN TE AT T H ON HA D E


LITET HOSTA OCKSÅ (1375:100:1/115054) (GHT)
(my friend L. said and he actually discovered that
she also had a slight cough)

d. UPPT ÄCKTE VI TILL V ÅR FÖRSKRÄCKELSE AT T DE T SA TT EN


( 1 3 4 9 : 5 6 : 3 / 5 3 1 6 6 9 ) (SDS)
(we discovered to our horror that there was an
X sitting )

The most frequent use of UPPTÄCKA in the Gothenburg corpus is in.


the sense of 'discover visually', that is UPPTÄCKAi. In this
corpus the verb has been found to take only att-Embeddings and
Nominalizations as verbal complements. Both ACI-Constructions
and Accusative-with-Past Participle Constructions are, however,
possible.

hur-Embedding

No ftur-Embedding after UPPTÄCKA^ 'discover visually' has been


retrieved in the Gothenburg corpus, although verbal complements
embedded by the /zwr-Complementizer Embedding Rule are conceivable
as demonstrated by the following sentences:
90

(118) a. när jag fick syn på tjuren upptäckte jag hur (oför­
siktigt) en liten flicka kom springande mot mig på
ängen-
(when I cau ght sight of the bull I disc overed how
(carelessly) a little girl came running towards me
on the meadow.)

b. när amerikanarna för något år sedan upptäckte hur


(skickligt) ryssarna hade byggt in mikrofoner i
deras ambassad i Moskva
(when the Americans discovered about a year ago how
(ingeniously) the Russians had implanted microphones
in their embassy in Moscow)

c. sa min vän L. och faktiskt upptäckte han inte hur


(illa) hon var klädd.
(my friend L. said and he actually discovered how
(badly) dressed she was)

d. upptäckte till vår förskräckelse hur (dumdristigt)


han satt på kanten av stupet.
(discovered to our horror how (stupidly) he was sit­
ting by the precipice.)

The complement sentences often require a manner adverb to yield


a well-formed /zwr-Embedding, even though sometimes the manner
adverb may be optional (indicated in (118) by parentheses). As
demonstrated above for the verb SE^ in section 5.1.3 also the
hur-embedded compxement sentences after UPPTÄCKA^ 'discover
visually' convey a completely different meaning than do att-
embedded sentences. Two different underlying structures must
therefore be posited to account for the two different semantic
interpretations of att and /zur-Embeddings.

There is a verb constraint on the Subject Raising Rule in (117)


which blocks this rule from embedding the complement sentences
into ACI-Constructions, if the complement sentence verb is not
an intransitive existential verb, i.e. SITTA 'sit', STÅ 'stand',
LIGGA 'lie' etc. As evidence for this claim compare the following
sentences which have undergone Subject Raising and been trans­
formed into ill-formed ACI-Constructions under violation of the
EXISTENTIAL VERB CONSTRAINT on Subject Raising:

ACI-Construction

(117) a'. *när jag fick syn på tjuren och upptäckte det vara
en svår tjur.
(*when I caught sight of the bull and discovered him
to be a troublesome bull.)

b'. *när amerikanarna för något år sedan upptäckte det


hos deras ambassad i Moskva finnas
(*when the Americans discovered about a year ago
there be(ing) in their embassy in Moscow )
91

(117) c'. *sa min vän L. och faktiskt upptäckte han inte hen­
ne också ha lite hosta.
(*my friend L. said and he actually discovered her
also to have a slight cough.)

d'. *upptäckte vi till vår förskräckelse det sitta en...


(*we discovered to our horror there an X sit(ting)...)

(117d') contains an intransitive complement verb sitta 'sit',


but the ACI-Construction is nevertheless ill-formed. Of all
the ill-formed ACI-Constructions exemplified above, (117d')
is the least ill-formed construction due to the fact that it
does contain an intransitive existential verb, sitta. We must
add a further constraint on the Subject Raising Rule to account
for the ill-formedness of (117df), since the EXISTENTIAL VERB
CONSTRAINT is inadequate to block constructions like (117d')

The transformationally introduced surface structure subject


det, the so-called informal subject det, cannot be raised by
the later Subject Raising Rule to yield (117d'). Let's call
this constraint the INDEFINITE SUBJECT CONSTRAINT on Subject
Raising, which can be claimed to block also (117a') and (117b'),
but not (117c'). The last of these sentences is the least ill-
formed ACI-Construction, due to the lack of an informal subject
det. (117c') is blocked by the EXISTENTIAL VERB CONSTRAINT,
whereas (117a') and (117b') are blocked by the INDEFINITE
SUBJECT CONSTRAINT. Subject Raising must be ordered after the
insertion of the indefinite subject det 9 otherwise sentences
like (117a'), (117b'), and (117d') cannot be blocked by the
INDEFINITE SUBJECT CONSTRAINT.

If a full subject noun phrase or its pronominalized product


occurs in the complement sentence, an ACI-Construction can occur
as a verbal complement after UPPTÄCKA^ 'discover visually':

(119) a. Nixon upptäckte Mao sitta på fel sida om honom vid


banketten.
(Nixon discovered Mao on the wrong side of him at
the banquet.)

b. Martin upptäckte Eva stå där ute på vägen vid hans


Volkswagen.
(Martin discovered Eva standing there on the road by
his Volkswagen.)

c. Han upptäckte bilen komma emot sig med stor hastighet.


(He discovered the car approaching him at a great speed

The claim of an EXISTENTIAL VERB CONSTRAINT on Subject Raising


is weakened by the example (119 c) which contains a durative
verb of movement. It seems as if durative Aktionsart is necessary
for Subject Raising to apply. Inherently durative verbs like the
92

copula VARA 'be' in (117a), FINNAS 'exist' in (117b), HAVA 'have'


in (117c) etc are, however, not capable of occurring in ACI-
Constructions. Subject Raising must be blocked in such sentences
if these verbs occur as complement verbs and if the subject is
the indefinite pronoun det 'it'.

Native speakers of Swedish react differently to the ACI-Construc-


tions exemplified in (119). Some accept all three constructions,
but some reject (119a), but accept (119b) and (119c), whereas
others accept only (119b). A performance study would be necessary
to show the different type of constraints on the Subject Raising
Rule as it is structured by native speakers of Swedish. Space
does not allow such a detailed treatment of idiosyncratic
constraints on syntactic rules which are known to vary from one
speaker to another in a given speech commuity.

Accusative-with-Past Participle Construction

Neither the Gothenburg corpus nor östergren 1968 give any examples
of the Accusative-with-Past Participle Construction, although
sentences which contain such a verbal complement can be easily
produced, e.g.:

(120) a. Man upptäckte honom blodig och sönderslagen på gatan.


(He was discovered bloody and bruised on the street.)

b. På majstången upptäckte han några grenar felaktigt


instuckna.
(He discovered some branches which were wrongly
attached to the May Pole.)

The deletion of the existential verb LIGGA 'lie' in (120a) and


VARA 'be' or SITTA 'sit' in (120b) is optional, since ACI-Con-
structions which contain these existential verbs as complement
verbs are also possible:

(120) a'. Man upptäckte honom|*^?ggaj blodig och sönderslagen

på gatan.
(He was discovered lying bloody and bruised on the
street-)

b'. På majstången upptäckte mannen några grenar^?sittal


felaktigt instuclcna. v. J
(The man discovered some branches which were wrongly
attached to the May Pole.)

The Existential Verb Constraint discussed above requires that


the complement verb be a semantically full verb like LIGGA 'lie'
or SITTA 'sit', for existential VARA 'be' yields dubious or ill-
formed ACI-Constructions after UPPTÄCKA^ 'discover visually' (cf.
(120a») and (120b').
93

Nominalization

Only one example of Nominalization which occurs as a verb


complement after UPPTÄCKA^ has been found in the Gothenburg
corpus. However, many abstract nominals are imaginable as
embeddings after UPPTÄCKAi as demonstrated by the following
examples:

(121) a. M AMMA UPPTÄCKE R SLARVET OCH SY NDA REN SÄ T T S AT T SKRIVA


(720:132:2/112090) (GHT)
(Mother discovers the carelessness and the offender
is put to writing}

b. SÄPO upptäckte Wennerströms förräderi mot Sverige.


(the Swedish CIA discovered Wennerström's treason
against Sweden)

c. först dagen därefter upptäckte man förstörelsen i


sommarstugan där tjuvarna brutit sig in.
(not until the following day did one discover the
destruction in the week-end house, where the thieves
had broken in.)

Each linguistic context of (121) tells us what deep subject NP is


to be posited for associating the correct semantic readings with
the nominal. It is clear that in (121a) the offender has been
careless (SLARVAT), that in (121b) a person called Wennerström
has betrayed Sweden (FÖRRÅDfT), and that in (121c) thieves have
destroyed the week-end house (FÖRSTÖRT), although in (121a) and
(121c) no explicit morphosyntactic information indicates what NP
is to be understood as the complement subject NP. In order to
describe the interpretation of the nominals of (121), the follow­
ing underlying structures are posited which undergo a nominaliza­
tion process:

(121) a'. mamma upptäcker [ Pro har slarvat]


S^ +Hum S^

(mother discovers [ Pro has been careless})


S^ +Hum S^

b'. SÄPO upptäckte [Wennerström har förrått Sverige]


si si

(the Swedish CIA discovered [Wennerström has betrayed


Sweden] ) S1

si

c'. först dagen därefter upptäckte man [ Pro har för-


S^ +Hum
stört något i sommarstugan]
si
(not until the day after one discovered [Pro has
SiC+Hum]
destroyed something in the week-end house]
94

By nominalizing the complement verbs in (121a') through (121c'),


the speaker need not specify the time of action in terms of tense.
(For the sake of simplicity the complement sentences have been
assumed to contain the perfect tense, to make the difference in
time to the matrix sentence clear). By using a nominalized ver­
bal complement the speaker furthermore avoids any direct state­
ment of the person who was careless in (121a1) or the actor
of the destruction in (121c1). Such subject NP information is,
however, explicitly signalled in (121b') by the genitive attribute
WENNERSTRÖMS 'Wennerström's 1. The human Pro-dummy posited in
(121a1) and (121c') as the complement sentence subject symbolizes
this fact. Through a surface interpretation rule in the sense
of Katz 1972, the full semantic readings of (121) are obtained
for the empty complement subject NP's, so that SYNDAREN 'the offend­
er'and TJUVARNA 'the thieves' are associated with the Pro-
dummies of the complement sentences. It is doubtful, from syn­
t a c t i c a s w e l l a s s e m a n t i c , p o i n t s o f v i e w , t o c l a i m t h a t SYNDA REN
in (121a) and TJUVARNA in (121c) are the underlying subjects of
the complement sentences.

9.2 the verbal complements of upptäcka 2 'discover cogniti vely'

The embeddings after the cognitive verb of perception UPPTÄCKA^


are almost three times as numerous in the Gothenburg corpus as
compared to UPPTÄCKA^ 'discover visually'.

att-Embedding

(122) a. DET ÄR DÂ HON U PPT ÄCK ER, A T T HAN Ä R FULLSTÄNDIGT OKÄND F Ö R


H EN NE (1018:200:9/126074) (GHT)
(That is when she discovers that he is completely
unknown to her.)

b. HA N U PPTÄ CKER A T T DET FINNS MÅNGA I S AM MA PR ED IKA ME NT SO M


(377:78:5/126051) (GHT)
(He sees that there are many people in the same
predicament who )

hur-Embedding

(123) a:l H URSO MHE LST S Å UPPTÄCKTE J AG BU M S H U R R OLIG T D ET K A N VARA ATT


(1888:30:3/115075) (GHT)
(Anyhow I immediately discovered how fun it can be
to )

b ;1 S Å D AVI D NÅ R EN KR IS AR T AD SLUT PUN KT, N Ä R H A N UPPTÄC KER, H UR


H A N S S E XU E LL A BETEENDE ÖV ER G ÅR T IL L (1030:343:5/
126074) (GHT)
(David comes to a point of crisis, when he discovers
that his sexual behavior changes into )
95

The cognitive verb UPPTÄCKA2 must be marked as not allowing


Subject Raising to occur. Only att or ftwr-Embeddings seem to
be well-formed as verb complements after UPPTÄCKA2 'discover
cognitively', since both ACI-Constructions and Accusative-
with-Past Participle Constructions are impossible after this
verb of cognition.

ACI-Construotions

(122) a'. *det är då hon upptäcker honom vara fullständigt


okänd för henne.
(*that is when she discovers him to be completely
unknown to her.)

b'. *han upptäcker det finnas många i samma predikament


som....
(*he sees many people be in the same predicament who..)

In connection with SE2 'realize' we indicated that ACI-Construc­


tions do not occur after verbs of cognitive perception (cf.
section 5.1.2). The constraint on the Subject Raising Rule can
be said to be repeated here in the case of UPPTÄCKA2 'discover
cognitively', and we need not formulate a special constraint
for UPPTÄCKA2, since the property of being a cognitive verb
of perception is sufficient for stating a general constraint.
As we stressed in section 5.1.2 the syntactic rule feature
[-Subj Rais] is needed to block the generation of ill-formed
ACI-Constructions as exemplified above in (122a') and (122b').

Accusative-with-Past Participle Construction

(Impossible after UPPTÄCKA2 'discover cognitively').

Nominalizations

Some few examples of abstract nominals as verbal complements


after UPPTÄCKA2 'discover cognitively' have been retrieved in
the Gothenburg corpus:

(123) a. T IL L S M AN UPPTÄCKER D EN SE MA NT IS KA GLIDNINGEN I M O RA V I AS


ANVÄNDNING (1517:74:3/226245) (SvD)
(until one discovers the semantic shifting in the
use of by M.)

b. i tid upptäcktes den felaktiga tolkningen av grund­


fördraget från DDR:s sida.
(the wrong interpretation of the treaty by the GDR
was discovered in time.)

It is implausible to derive the abstract nominal in (123a) from


an underlying structure which contains the verb GLIDA 'shift' as
the complement verb. Such a structure would not yield the
metaphoric interpretation which is implicit in (123a). Instead
96

the nominal GLI DNING 'shifting' must be treated as a lexicalized


entry which cooceurs with SEMANTISK 'semantic'. Metaphors are
subjected to other rules of interpretation, because they do not
have a direct symbolization but rather trade on the interpreta­
tion of other lexical items (cf. Chafe 1970:44). However, a
derivation may be possible for the abstract nominal in (123b),
where the complement verb -TOLKA 'interpret' and the subject NP
DDR 'the GDR' can be made to undergo the nominalization and
gen itivization processes respectively, as indicated by the
following underlying structuré:

(123) b'. Pro upptäckte i tid [ DDR tolkade grundfördraget


[+Hum][+Pâss] S1

felaktigt]

After Nominalization and Genitivization have operated an


obligatory outspèlling of a preposition occurs before the deep
structure object NP, GRUNDFÖRDRAGET, 'the treaty':

(123) b''. Pro upptäckte i tid DDR:s tolkning av grundför-


[+Hum][+Pass]
fördraget felaktigt.

After Recategorization of the manner adverb into an adjective


and the Passivization of the matrix sentence under Agent Dele­
tion the following surface structure is generated:

(123) b'''. DDR:s felaktiga tolkning av grundfördraget upp­


täcktes i tid.
(The wrong interpretation of the treaty by the GDR
was discovered in time.)

After another transformational rule, the Topicalization Rule,


and a change of the genitive phrase into a prepositional phrase
the surface structure of (123b) results. The last transforma­
tional rule which changes the genitive phrase into a prepositional
phrase is optional in Swedish but obligatory in English.

9.3 the verbal complements of upptäcka3 'make a discovery,


find out'
A third meaning of UPPTÄCKA must be exemplified here which is
different from the two meanings just treated in sections 6.1 and
6.2. The verb UPPTÄCKAj 'make a discovery' is of course closely
related to UPPTÄCKA^ 'discover visually'. Whereas the latter is
a pure verb of passive perception which takes an Experiencer
as a deep subject NP, the former is an active transitive verb
of purpose which takes an Agent as a deep subject NP.
97

Object NP as Verb Complement

(124) a. H A N S LANDSMAN PR IES TLY D EL AR ÄR AN A V A TT H A UP PT Ä CK T SYRET


MED S VEN SKE N S CH E EL E C1902:9:6/122071) ( G HT )

(His countryman Priestly shares together with the


Swede Scheele the honor of having discovered oxygen.)

b. AJ-LTSÅ EFT ER DET ATT BEHRENS HADE UP PTÄCKT P Å SK Ö N


(635:349:3/412427) (DN)
(after Behrens had discovered the Easter Islands.)

att-Embedding

(125) a. EINTH OVEN, SO M R E DA N FÖ R EN MANSÅL DER S ED A N UPP TÄ CKT E A TT


FENO MEN ET H A R SAMBAND MED PU PI LL CE NT RU M ( 1 6 1 5 : 1 1 2 : 2 / 2 2 1 2 4 5 )
(SvD)
(Einthoven, who more than a generation ago discovered
that the phenomenon is correlated with the pupil
center of the eye-)

b. INT E UT AN EN LÄ TT RYSNING UPPT ÄCKER J AG A T T DE A L L A ARB ETA R I


EN TÄMLIGEN LIKA (955:183:3/526424) (SDS)
(Not without a slight shiver did I discover that they
all work in a rather similar....)

Indirect Quotation

(126) MEN NO RR OM T ORNE T R Ä S K FINN S DEN, UPPTÄCKTE S T E N BRAND ER NU


(259:87:6/415517) (DN)
(But north of Lake Torne it exists, Sten Brander
discovered now)

ACI-Constructions after UPPTÄCKA3 'make a discovery, find out'


don't seem to be well-formed as verb complements, even though
existential verbs occur as complement verbs. Compare the att-
Embedded structures in (125) with the following constructions:

ACI-Construction
?
(127) a. '•Behrens upptäckte Påskön ligga mellan Chile och
Pitcairn.
?
('•Behrens discovered the Easter Islands to be
situated between Chile and Pitcairn Island.)

b. •Priestly och Scheele upptäckte syre finnas som


komponent i luften.
(•Priestly and Scheele discovered oxygen be as a
component in the air.)
The embedding by the /zur-Complementizer Rule is of course
possible, even though no such verb complement has been found in
the Gothenburg corpus:
98

hur-Embedding
?
(128) a. 'Behrens upptäckte hur Påskön var belägen mellan
Chile och Pitcairn.
(Behrens discovered how the Easter Islands were
situated between Chile and Pitcairn Island)

b. Priestly och Scheele upptäckte hur mycket syre det


finns i luft.
(Priestly and Scheele discovered how much oxygen
there is in the air.)

For the correct use of verb complements after UPPTÄCKA3 one


has to know what the presuppositions are. If an object NP is
used as in (124b),then BEHRENS discovered the Easter Islands
without any surprise being implied. If an English ACI-Construc­
tion is used, then Behrens knew about the existence of the
Easter Islands but not their location. Their location was a
surprise to him. If a complementizer embedding is used Behrens
knew the approximate location of the Easter Islands and the
fact is stressed how he discovered them.
99

10. RESULTS OF THE LINGUISTIC PERFORMANCE STUDY ON VERBAL


COMPLEMENTS OCCURRING AFTER UPPTÄCKA 'DISCOVER'
The three different types of the homograph UPPTÄCKA 'discover'
discussed in section 8 are covered by 108 examples in the
Gothenburg corpus, of which 41 examples have embedded sentences
as verbal complements. The first type of perceptual verb UPP-
TÄCKA^ 'discover visually' has a distribution of embeddings as
illustrated by the following table:

10.1 upptäcka, 'discovervisually'

Table 5- THE PRODUCTIVITY OF EMBEDDING R ULES AFTER UPPTSCKAj


1 di s c o v e r v i s u a l 1 y 1

Type of Embedding Number Distribution of Embedding


Types after the Verb UPP-
TKCKAi fdiscover visually'
in the Gothenburg Corpus

1) att-Embeddings 9 90 *
2) Nominalizations 1 10 %
3) ftwr-Embeddings 0 -

4) ACI-Constructions 0 -

5) Accusative-with-Past
Participle Constructions 0 -

6) Accusative-with-Gerundive
Constructions 0 -

Total number of embeddings: 10


Total number of examples
of the homograph UPPTÄCKA
'discover' in the Gothen­
burg corpus: 108
100

10.2 up ptäcka 2 'discovercogni tively'

Table 6. THE PRODUCTIVITY OF EM BEDDING RU LES AFTER UPPTÄCKA2


'discover cognitîvely'

Type of Embedding Number Distribution of Embedding


Types after the Ve rb UPP-
TÄCKA2 'discover oognitively'
in the Got henburg Corpus
1) att-Embeddings 20 71.4 %
2) /zwr-Embeddings 4 14.3 %
3) Nominalizations 4 14.3 %
4) ACI-Constructions 0 -

5) Accusative-with-Past
Participle Constructions 0 -

6) Accusative-with-Gerundive
Constructions 0 -

Total number of embeddings: 28


Total number of examples of
the homograph UPPTÄCKA
'discover' in the Gothen­
burg corpus: 108

10,3 upptäcka3 'make a discovery/ find out'

Table 7- THE PRODUCTIVITY OF EM BEDDING RU LES AF TER UPPTÄCKAj


'make a disco very, find ou.t '

Type of Embeddin g Number Distribution of Embedding


Types after the Verb UPP­
TÄCKA z 'make a discovery>
find out' in the Gothen­
burg Co rpus
1) att-Embeddings 2 66.7 %
2) Nominalizations 1 33.3 %
3) fcur-Embeddings 0 -

4) ACI-Constructions 0 -

5) Accusative-with-Past
Participle Constructions 0 -

6) Accusative-with-Gerundive
Constructions 0 -

Total number of embeddings: 3


Total number of examples of
the homograph UPPTÄCKA
'discover' in the Gothenburg
corpus: 10 8
101

11. THE SUEJECT RAISING RULE IN SWEDISH (OBSERVERA)


11.1 the verbal complements of observera]: '.watch.'

For comparative reasons OBSERVERA 'observe' has been included in


the list of ACI-verbs in section 3.3 labelled together with
IAKTTAGA 'watch' as Class (2a). This verb class ±s characterized
as an active durative class of visual perception by the semantic
features [+Perceptual], [+Process], [+Visual], [+Purpose], and
[+Durative]. No explicit lexical indication of the Agent role
which the deep subject plays is necessary in the subclassifica­
tion of this verb, since a semantic redundancy rule (RR ii) in
section 3.3 has been elaborated to account for the characteristic
role of the subject NP of verb class (2a).

Swedish standard dictionaries like the SAOB ( 1 9 4 9 ) , Illustrerad


svensk ordbok (1958) andSvensk handordbok (1960) give two major
definitions of OBSERVERA which reflect the two most important
meanings of this verb, i.e. OBSERVER/^ 'watch' and OBSERVERA2
'become aware of'. The first of the two meanings of OBSERVERA
is the one which has already been defined by Subcategorization
Rule (iii) in section 3.3, that is as the durative verb of
active visual perception, OBSERVERAI, which in Swedish dictionar­
ies is paraphrased as iakttaga 'watch' or ge akt på 'observe'
(cf. for instance Illustrerad svensk ordbok 1958:1067), whereas
OBSERVERA? 'become aware of' is a verb of passive visual percep­
tion which takes an Experiencer as a human deep subject (cf.
section 11.2 below) and is of an instantaneous Aktionsart.

Object NP

(129) M AN BÖR OBSERVERA P R I SE T PÂ PH I LI PS KA SE T TE RN A ( 5 8 4 : 1 : 3 / 3 2 6 2 8 7 )


(ST)
(One should keep an eye on the price of Philiphs
casettes.)

att-Embedding

(130) OBSE RVE RA A TT DET M DEN OINFORMERADES DI LEM MA ( 1 1 1 7 : 1 6 2 : 6 /


222246) (SvD)
(Notice that it is the dilemma of the person who is
uninformed)

No ACI-Construction after OBSERVERA^ 'watch' has been found,


neither in the Gothenburg corpus, nor in any of the Swedish
handbooks mentioned above. Since most informants hesitate to
accept ACI-Constructions after this verb both OBSERVERA and
IAKTTAGA have been starred in section 3.3. This surface struc­
ture would not, however, be impossible as verb complement also
in Swedish. It is probably a syntactic idiosyncracy that ACI-
Constructions do not occur after OBSERVERA-^ in some idiolects.
It may be the case that ACI-Constructions belong to a literary
style if they occur after OBSERVERAI 'watch' (cf. the literary
evidence of ACI-Constructions after OBSERVE in Part Two, section
7).
102

ACI-Construction
?
(131) a. 'Mannen observerade henne sitta på en bänk i parken.
(The man observed her sitting on a bench in the park.)
?
b. 'Han observerade en man promenera fram och tillbaka
på gatan.
(He observed a man walking up and down the street.)

It is interesting to note that the aspectual status of OBSERVERA


is unclear in sentences like (132) which are isolated from their
natural linguistic context:

(132) P Â S LU S SE N MI TT FÖ R KATARINAHISSEN. HA N OBSERVERA DE EN MA N


SOM P RO ME NE RA DE FR A M OCH TI LL BA KA ( 1 0 7 9 : 1 8 : 1 / 2 1 5 2 4 8 )( S v D )
(at Slussen opposite to Katarinahissen. He observed
a man who fwalked Iback an forth.)
îwas walking]

If the observation is of durative Aktionsart, then the ing-form


is obligatory in English. If, however, OBSERVERA in (132) has
the meaning of 'become aware of suddenly' then the simple con­
struction walked is obligatory. This motivates us also to classi­
fy OBSERVERA into another class of verbs which denotes immediate
perception.

11.2 the verbal complements of observera 2 ' become aware of'

The second meaning of OBSERVERA is paraphrased in the Illustrevad


svensk ordbok as varsna 'become aware of' or lägga märke till
'notice' (op.cit. p. 1067). This type of verb belongs to a
completely different verb category than OBSERVERA^ 'watch', in
that OBSERVERAi takes an Agent as a deep subject and expresses
an active visual perception, whereas OBSERVERA2 'become aware
of' takes an Experiencer as the deep subject and expresses a
passive instantaneous visual perception. The Swedish dictionaries
indicate this semantic difference by giving different synonyms.

Object NP

(133) a. R U NS TE N , SO M INGEN ANNA N OBSERVERAT ( 1 0 0 7: 5 4 : 1 / 5 1 5 6 8 1 )(SDS)


(a rune stone which nobody else had observed)

b. MEN INGEN OBSERVERADE DEM VID AVRESAN ( 9 2 0:8: 7 / 4 3 5 4 8 8 )(DN)


(But nobody noticed them at the departure.)

att-Embedding
(134) MAN HA D E NOG OBS ERVER AT AT T P L AN T OR I S EN A RE GENERATIONER
S K I L DE . . . . ( 3 3 3 : 1 0 0 : 7 / 5 1 1 6 7 0 ) (DN)
(One had observed that plants were different.... in
later generations.)
103

Nominalizations

(135) a. S VE NS K A BARNLÄKARFÖRBUNDET H AR REDAN OB SERVE RAT TEN DEN SER NA A T T


TI LL SK AP A DAGHEMSPLATSER (1896:184:2/123020)( G H T )
(The Swedish Pediatric Association has already observed
the tendencies to create places in day-care centers.)

b. D EN S OM S EG L AR B ÂT H AR S Ä K E RT O B SE R VE RA T S K I L L N A D E N I R Ö R E L SE
UNDERSEGEL (23:76:1/125052)(GHT)
(Whoever sails boats has certainly noticed the difference
in movement while sailing.)

The passive perceptual meaning of OBSERVERA^ 'become aware of'


exemplified in (133) through (135) is reminiscent of the passive
perceptual meaning of MÄRKA 'notice' and UPPTÄCKA^ 'discover
visually' discussed in sections 7 and 9 respectively. Traditional
dictionaries of the Swedish language do not explicitly characterize
these verbs as passive perceptual instantaneous verbs each of which
requires the deep subject to be an Experiencer. Instead the
reader is presented a number of synonyms from which he is to in­
fer the syntactic arid semant ic distinctions of passive and active
perception in the verb or of Experiencer and Agent roles of the
deep subject. Modern linguistic theory provides us with deeper
insights into the semantics of verbs and we can therefore make
more explicit the semantic distinctions which are necessary to
distinguish verb classes such as OBSERVERA^ 'watch' and OBSER-
VERA2 'become aware of'.

The syntactic status of the nominals T E ND E NS E R NA 'the tendencies'


and SKILLNAD 'difference' in (135) is problematic. In the balanced
view of describing nominals discussed in sections 5.3 and 13.5,
there is some motivation for deriving these nominals from under­
l y i n g c o m p l e m e n t s e n t e n c e s w h i c h c o n t a i n t h e m a i n v e r b s T EN D ER A
'tend to' and SKILJA SIG 'be different':

(136) a. svenska barnläkare har redan observerat [ Pro


S^[+Hum]
tenderar att tillskapa daghemsplatser]
si

(Swedish pediatricians have already observed [ Pro


S^[+Hum]
tend to create day-care places] )

b. den som seglar båt har säkert observerat [ Pro skiljer


S^-Hum]
sig i rörelse under segel]
si

(Whoever sails boats has certainly noticed [ Pro


S^ -Hum
is different in movement while sailing] )
si
104

In order to assign (136a) and (136b) the correct semantic


reading, contextual information is necessary which associates
semantic contents with the subject Pro-dummies One does not
know from the limited context presented in (136a), if it is the
Swedish State, the community, or some private institution which
tends to create nurseries or kindergartens. In (136b) one may
infer from the matrix sentence that the complement subject is a
sailing ship.

No ACI-Construction or /zwr-Embedding has been found to occur in


the Gothenburg corpus or any dictionary, although at least hur-
Embeddings are possible after OBSERVERA2'become aware of' as
demonstrated by the following sentence:

hur-Embedding

(137) man hade observerat hur plantor skilde sig i senare


generationer.
(one had observed how plants differed from one another in
later generations.)

12, RESULTS OF THE LINGUISTIC PERFORMANCE STUDY ON VERBAL


COMPLEMENTS OCCURRING AFTER OBSERVERA 'OBSERVE'
The two different semantic classes of the homograph OBSERVERA
'observe' treated in section are exemplified by 23 excerpts
in the Gothenburg corpus, six of which take some kind of an
embedding. The distribution of embedding types can be seen from
Table 8 and Table 9 below.
105

12,1 observera, 'watch'


Table 8. THE PRODUCTIVITY OF EMBEDDING R ULES AFTER OBSERVERA.
1 watch 1

Type of Embedding Number Distribution of Embedding


Types after the Verb
OBSERVERAj 'watch' in the
Gothenburg Corpus

1) att-Embeddings 2 100 %
2) ftur-Embeddings 0 -

3) ACI-Constructions 0 -

4) Nominalizations 0 -

5) Accusative-with-Past
Participle Constructions 0 -

6) Accusative-with-Gerundive
Constructions 0 -

Total number of embeddings: 2


Total number of examples
of the homograph OBSERVERA
'observe' in the Gothenburg
corpus: 23

12.2 observera 2 ' become aware of'


Table 9. THE PRODUCTIVITY OF EMBEDDING RULES AFTER 0BSERVERA2
'become aware of1

Type of Embedding Number Distribution of Embedding


Types after the Verb
OBSERVERA2 'become aware of
in the Gothenburg Corpus
1) a£t-Embeddings 2 50 %
2) Nominalizations 2 50 %
3) ftur-Embeddings 0 -

4) ACI-Constructions 0 -

5) Accusative-with-Past
Participle Constructions 0 -

6) Accusative-with-Gerundive
Constructions 0 -

Total number of embeddings: 4


Total number of examples
of the homograph OBSERVERA
'observe' in the Gothen­
burg corpus: 23
106

13, THE SUBJECT RAISING RULE IN SWEDISH (HÖRA)

In section 5 through 12 the verbs of visual perception which


take ACI-Constructionsand other verbal complements have been
treated and exemplified by excerpts from the Gothenburg corpus.
In those cases where no authentic examples were found in this
corpus, due to natural insufficient data coverage, lacking
verb complements were supplemented by our intuitive knowledge
of syntactic surface structures in Modern Swedish. The empiri­
cally-oriented corpus approach to the study of verbal comple­
mentation in Modern Swedish must thus be complemented by
intuition-based knowledge of the syntactic structures of a
living language like Modern Swedish. The present investigation
is just as much a competence study as it is a performance study.
Therefore, when in some cases syntactic evidence was lacking in
the Gothenburg corpus and our own language intuition failed to
give definite answers as to the existence of given surface struc­
tures, standard dictionaries of the Swedish language have also
provided information on the existence of verbal complements.

A combination of corpus-based observations (the Gothenburg corpus),


our own intuition-based knowledge of Modern Swedish, together with
information retrieved from standard dictionaries on semantic and
syntactic structures was the approach used to describe the verbal
complements after the verbs of visual perception. The same
approach will be applied to the verbal complements of the audi­
tory and tactual verbs of perception, i.e. HÖRA 'hear', FÅ HÖRA
'hear suddenly', and KÄNNA 'feel'.

Svensk illustrerad ordbok 1958:590 and Svensk handordbok 1960:


336 give together seven major meanings of the homograph HÖRA
'hear', of which only two meanings will concern us here, that
is HÖRA2 'hear auditorily' and HÖRA2 'hear of, be informed',
which take ACI-Constructions or other embeddings. The meanings
of HÖRA3 'listen to', HÖRA4 'examine, ask', HÖRA5 'belong to'
and two other obsolete meanings of HÖRA will be disregarded
here, since these latter types of HÖRA do not take embeddings.

13,1 the verbal complements of höra! 'hearjauditorily'

For the syntactically and semantically related verb HÖRA 'hear',


we can observe the same syntactic restrictions as found for SE,
regarding the Passive and Cleft Sentence Rules. As the Passive
Rule is the more interesting rule in this context, we will
concentrate our discussion on the surface structures which
result after its application. Compare the following structures
which result after Subject Raising, Passivization, and Agent
Deletion, applied at (138):

(138) Palme hörde [ [ba. ropade slagord mot Sverige] ]


NP S1 S1 NP
Palme heard C cw. shouted slogans against Sweden-] )
NP NP
107

(139) Palme hörde ba. ropa slagord mot Sverige.


(Palme heard construction workers shouting slogans against
Sweden»)

(140) *Byggnadsarbetare hördes av Palme ropa slagord mot Sverige*


(Construction workers were heard by Palme to be shouting
slogans against Sweden.)

(141) Byggnadsarbetare hördes ropa slagord mot Sverige.


(Construction workers were heard to be shouting slogans
against Sweden.)

That there is no coconstituency in the surface structure of (139)


between the object noun phrase and the infinitive after Subject
Raising is demonstrated by the ill-formed passive sentence *(140)
and the ill-formed cleft sentence *(142):

(142) *Byggnadsarbetare ropa slagord mot Sverige hördes av Palme.


(•Construction workers shouting slogans against Sweden was
heard by Palme.)

(143) *Det som Palme hörde var byggnadsarbetare ropa slagord.


mot Sverige
(*What Palme heard was construction workers shouting
slogans against Sweden.)

However, the Agent Deletion Rule which has not operated in *(140),
but which has done so in (141), seems to be an obligatory rule
in all structures after Subject Raising in sentences containing
HÖRAi 'hear auditorily' as the matrix verb. We are confronted
with an ill-formed structure in *(140) which seems to become
well-formed only after Agent Deletion in (141). It would be a
mistake to derive (141) via the shallow structure of *(140),
since the semantic interpretations are considerably different.
(The same statement was made regarding the passives of SE above
in subsection 5.1).

For (141) we would have to set up an underlying structure con­


taining an indefinite deep structure subject. One could also
claim that there are no passives of sentences like (139), since
after Subject Raising, sentences which contain the verb HÖRA as
matrix verb cannot undergo the Passive Rule, if the matrix
subject is definite. If a passive form of HÖRA plus an infinitive
does occur, then the deep structure subject of the matrix
sentence must be indefinite and obligatorily deleted by the
Agent Deletion Rule, , the so-called Empty Subject Deletion' Rule.
A plausible derivation of (141) is therefore as follows:
108

FIGURE 1

(i) [ Pro hörde [ [ba. ropade slagord mot Sverige] ] ]


S [+Hum][+Pass] NP S, S.NP S„
o 1 1 c

( Pro heard c.w. shouted slogans against Sweden)


[+Hum]

l|TSubj
v Rais

(ii) [ Pro hörde ba. [ropa slagord mot Sverige] ]


SQ[+Hum][+Pass] VP VP SQ

(Pro heard construction workers shouting slogans


[+Hum]
against Sweden)

jj-pPass
(iii) [Ba. hördes av Pro [ropa slagord mot Sverige] ]
Srt [+Hum]VP VP S
o o

(Construction workers were heard by Pro shouting


[+Hum]
slogans against Sweden)

Il-Agent
^ Deletion

(iv) [ Byggnadsarbetare hördes ropa slagord mot Sverige]


S Sn
o o

(Construction workers were heard shouting slogans


against Sweden.)

There is a general constraint on the Passive Rule in connection


with the verb HÖRAi also when no Subject Raising is involved.
Compare sentence (144) with *(146), and (145) with *(147a).The
passive rule can operate freely without constraints on structures
containing the verb SE 'see visually', but not on structures
containing the verb HÖRA 'hear auditorily'. Passivization of
sentences which contain HÖRAi as the main verb requires that no
agent phrase occurs:

(144) Föreställningen sågs av 400 människor.


(The performance was seen by 400 people}

(145) Tjuven sågs av två förbipasserande.


(The thief was seen by two passers-by«)
109

(146) *Konserten hördes av 400 människor.


(The concert was heard by 400 people.)

(147) a. *Tjuven hördes av två förbipasserande.


(The thief was heard by twc passers-by.)

However, the verb HÖRA can occur with an agent phrase if it means
'interrogate1 as in

(147) b. Tjuven hördes av två poliser om inbrottet.


(The thief was interrogated by two policemen concerning
the burglary.)

We are of course dealing with a different class of HÖRA in (147b),


that is HÖRA4 'interrogate'. The Passive Rule is unrestricted
in its application to sentences which contain this verb.

Ross 1967a, Perlmutter 1971, Postal 1971, and Lakoff 1970a suggest
various types of constraints on transformational rules to limit
the generative power of transformational grammar. The constraints
suggested by these linguists act as filters which block certain
ill-formed surface structures like *(140) and *(146) above.

In order to block ill-formed sentences like *(140)one may wonder


whether a Surface Structure or a Deep Structure Constraint should
be developed. The constraint cannot be formulated as a filter on
the DEEP Structure level, because Subject Raising irust operate
first to yield the coconstituency of stage (ii) in FIGURE 1 be­
fore the conditions of the Passive Rule are met, i.e. the
existence of an object NP in the matrix sentence. There is, how­
ever, no raised object NP in the matrix sentence of step (i), the
posited Deep Structure level. Therefore, the constraint on the
generation of *(140) cannot be formulated as a Deep Structure
constraint.

The alternative of a Deep Structure constraint, a Surface Struc­


ture constraint, is technically not excluded. One could formulate
a Surface Structure Constraint on Passivization of sentences
which contain HÖRA^ 'hear auditorily' and which contain definite
Experiencer NP's. But since Surface Structure constraints in the
sense of Perlmutter 1971 are very late blocking devices which are
designed to filter out ill-formed phonological sequences such as
for instance pronominal clitics in the Romance languages or the
occurrence of two s-passives in Modern Swedish (cf. Part Two,
section 12.2) we feel motivated to regard the blocking of sentences
like *(140) as a result of an earlier constraint on the Shallow
Structure level in the sense of Postal 1971, i.e. on stage (ii)
in FIGURE 1. Such a constraint on the Shallow Structure level
would block the crossing-over of the raised object NP and the
subject Experiencer NP, if the latter NP is definite.

However, a Shallow Structure constraint of this type still pre­


supposes previous information about the outcome of the Passive
Rule, i.e. one has to know that the Pro-Experiencer dummy is
deleted after the Cross-over Operation.
110

It has been demonstrated here that any type of constraint,


whether it applies to the Deep Structure, the Surface Structure,
or even the Shallow Structure is inadequate in accounting for
the generation of (141), and the blocking of *(140). We have
therefore developed a process completely different from the
filters suggested by both Perlmutter 1971 and Postal 1971 in
order to generate indéfinite passives like *(140) and those of
verba dioendi and putandi described in Part Three, and suggest
instead Global Rule I.(Cf. Ureland 1970:241:247 where two
Global Rule Processes are discussed in detail).

13,2 a specific underlying representation of aci"constructions


after hörax 'hear-lauditorily'

In sentence (139) above, the ACI-Construction which occurs after


HÖRAi 'hear auditorily1 is not treated as being derived from
the same underlying syntactic representation as the ACI-Construction after SEi
i.e. from DS I(=(45)) which contains two occurrences of preferential noun
phrases, one of which is deleted in the process of infinitiviza-
tion. All of the ACI-Constructions of SEX 'see visually' were
treated in this way (although the ACI-Constructions of SE2
'realize' must be seen as derived from DS III as indicated for
l
HÖRA- below).
However, there may be doubt as to whether DS I (-(45)) is an
adequate underlying syntactic representation for the ACI-Con-
structions also after HÖRAi, since DS III below is an under­
lying structure which can be motivated just as well:

(148)

DS III

NP VP

NP
+V
+Past
+Subj Rais VP

NP

Palme hörde ba. rop- slagord

There is a distinct difference between the ACI-Constructions


which occur after SE^ 'see visually' on the one hand, and those
which occur after HÖRAi 'hear auditorily' on the other. Whereas
one can see crickets and clocks without any implication of their
being associated with an act of producing light or their being
in a state of shining, one cannot hear crickets and clocks with­
out some sort of sound being transmitted. In order to use the
Ill

verb HÖRA^ 'hear auditorily' one must presuppose that a sound


source exists, whereas in the use of SEi 'see visually' no
visual source is presupposed.

(149) a. OCH MED ENS HÖR JAG DE SENA SYRSORNA SPELA I GRANNSKAPET
(1647:139:4/322399)(ST)
laiiu suddenly
(and auuucuiï I± Ji cai the
hear
+1,0 neighborhood.)
une late crickets
(.chirpingJ
f
u 1 x ujn.c 1» a i1 * • ~ • _ _ 1 in
the P g

OCH JAG HÖRDE EN KLOCKA DÅNANDE SLÅ ETTUSENNIOHUNDRASEX'


(273:67:1/423464)(DN)
(and I hea rd a clock fejJiJJingj resoundingly one-
thousandninehundredsixtyfive.)

To represent in a description of the native speaker's use of


ACI-Constructions after HÖRAi 'hear auditorily', the presupposi­
tion of a sound source NP as being directly associated with the
complement verb of sounding or making a sound is incorporated
into the underlying syntactic structure, i.e. DS IJI(=(148)).
This proposal reflects the physical situation more adequately
than does DS I (=(45)).

In sentences like (149a') and (149b') the complement verb


denoting the act of making a sound is therefore regarded as
having been deleted:

(149) a'. och med ens hör jag de sena syrsorna-


(and suddenly I hea r the late crickets.)

b'. och jag hörde en klocka någonstans.


(and I heard a clock somewhere.)

There is also an aspectual point of view involved in (149a) and


(149b). Without the more complete linguistic context of these
two sentences there is no way to interpret the Swedish ACI-Con-
structions accurately, since the chirping of the crickets in
(149a) can be either of an instantaneous mode of action (trans­
lated as CHIRP) or of a durative mode of action (translated as
CHIRPING). The same aspectual ambiguity of the Swedish ACI-Con-
struction is present in (150) and (151) below, which also have
to be rendered into English in terms of two different morpho­
logical forms of the verb.

(150) HÖRDE HENNE SKRIKA PÅ HJÄLP (83:96:6/222136)(SvD)


(heard her f^££ingj £or helPJ

The derivation of ACI-Constructions after HÖRA from DS III(-(148))


rather than from DS I(=(45)) is further supported by the deriva­
tional histories of passive ACI-Constructions, where the deriva­
tion in terms of a deleted indefinite object noun phrase in the
matrix sentence seems implausible. Compare the following authen­
tic sentences:
112

(151) a. Då han kom mitt för berget, hörde han sitt namn ropas.
(Bankeryd, SvLm III, 2:375: Holm 122)
(When he got to a point on the middle of the mountain,
he heard his name being shouted*)

b. När jag hörde nyckeln vridas om i dörrlåset


(Kläckeberga o. Dörby, SvLm IX, 2:66: Holm 122)
(When I hea rd the key being turned in the lock of the
door)

c. Både jag och Svärd hörde dem uppläsas.


(Alseda, SvLm B 30:36:Holm 122)
(Both I and Svärd heard them [the words] being read
aloud.)

In order to derive these sentences, which are excerpted from Holm's


1952 corpus on Swedish dialects, in the most plausible manner, we
cannot posit DS I for (151a). In such case the following inconsis­
tent derivational history would arise:

FIGURE 2

(152) a. Han- hörde Pro. [ [ Pro-j rop- Proi namn] ]


1 [+Hunr] NP S,[+Hum'][+Poss][+Hum] S, NP
1 [+Poss] 1

Uy>ass

b. Han- hörde Proj [ [ Proi namn rop-+s av Proj ] ]


[+Humj NP S-,[+Hum] [+Hum ]STNP
[+Poss] 1

11TSubj TAgent
y Rais Del

c. Han- hörde Proj Pro-; namn ropas


[+Hum ] C+HumJ
[+Poss]

||TIndef Obj NP + TReflexive


V- Del Possessive

d. Han^ hörde sitt^ namn ropas

(He^ heard his^ name being called)

If the underlying representation (152a) and the derivational


history described in (152b) through (152d) is assumed to be
correct, then one must take recourse to unnatural syntactic
rules and violations of the conditions of rule application. One
has to claim that the Passive Rule operates before Subject
113

Raising, a plausible claim as such. One must further claim that


Subject Raising operates under new conditions, so that the new
subject noun phrase sitt namn 'one's name' is raised after
Passivization into the matrix sentence. The agent phrase av nå gon
'by someone1 must then be deleted. If all this is accepted as
adequate, one still has the problem of getting rid of the object
noun phrase in (152c) after hörde'heard' which denotes the one
who calls the name, after the new subject of the complement
sentence sitt namn has been raised into the matrix sentence.
Subject Raising in this sense would never be possible, since the
new complement subject NP is not coreferent with the posited direct
object of the matrix sentence.

Instead, in order to derive the sentences (151a) through (151c)


above, we posit DS III (=(148)) as the more adequate underlying
syntactic structure for these passive impersonal sentences and
establish the following derivational history:

FIGURE 3

(153) a. Han- hörde [ [ Pro-j rop- Proj namn] ]


1 NP S, [+Hum] t+Hum ] S, NP
1 [+Poss] 1

||TPass TAgent
i}L Del

b. Han.: hörd e [ [ Proj namn rop+s] ]


NP S1 [+Hum] S1 NP

jj TSubj
Rais
+ TPoss
Reflexive

c. Han^ hörde sitt^ namn ropas.

(He^ heard his^ name being called.)

Since the DS I (-(45)) has been demonstrated here to be an in^-*


adequate underlying syntactic representation for deriving the
passive ACI-Constructions in(151), we may conclude that it is
also inadequate for deriving taotive ACI-Constructions in (149),
particularly if one wants to claim that passive and active surface
structures are derived from one and the same underlying con­
ceptual structure.
114

A third derivational possibility is worth mentioning here which


has been referred to in earlier descriptions òf ACI-Construc-
tions after the German verb HÖREN 'hear' (cf. Duden 1959:922
and Bierwisch 1963:123).

(153) a'. Han. hörde Proi namn [ Proj rop- Proi namn]
[+Hum] S-,[+Hum] [+Hum] S-,
[+Poss] [+Poss]

I TPass rpAgent
V- Del

b'. Han- hörde Pro^ namn [ Proi namn rop+s]


[+Hum] S,[+Hum] S,
[+Poss] [+Poss]

TSubj TPoss

II Rais Reflexive

c'. Han hörde sitt namn ropas.

The problem is whether one should allow an ad-hoc coconstituency


in the Deep Structure (153a'). Shall we accept a coconstituency
which would result in an ill-formed construction if it were not
embedded? The sentence Han hörde sitt namn 'He heard his name'
sounds ill-formed without the name being associated in some way
with a complement verb which denotes a sound source of some kind,
i.e. ropas 'being called' or uttalas 'being pronounced'. Compare
also the ill-formedness of the following relative construction
Han hörde sitt namn som ropades*He heard his name which was being
called'. Such a constituency violation does not occur after SE,
however: Han såg sitt namn 'He saw his name' and Han såg sitt
namn som skrivits på väggen 'He saw his name which had been
written on the wall'.

13,3 höra x occuring with passivized idioms

From the very beginning of transformational grammar linguists


have tried to describe and explain grammatical phenomena in terms
of statements which were as general as possible. The greater the
degree of generality was, the greater the degree of descriptive
or explanatory adequacy was felt to be. To be forced to admit
exceptions to a major rule meant some kind of a capitulation in
the early days of transformational grammar. In order to save
the holy principle of generality of syntactic rules, linguists
concocted all sorts of ad-hoc structures and fantastic deriva­
tions. Consider the transformational treatment of all English
nominalizations by Lees 1960 or the derivation of all attribu-
115

tive adjectives from predicative adjectives, although no


authentic predicative adjectives exist (e.g. criminal law as
derived from * law whic h is criminal or ein kupferner Türgriff
from *ein Türgriff, der kupfern ist).55

There are a great number of surface structures, however, which


do not lend themselves to being treated as products of deriva-
.tional processes. This does not by any means mean a capitulation
in describing a natural language. On the contrary, the insight
into the complexity of language is increased by postulating a
complex lexicon which contains long lists of lexicalized idioms
and phrases. We will discuss one lexicalized item here which is
better treated as an idiom than as a derivational product, i.e.
the passive infinitives SÄGAS and TALAS OM which occur after
HÖRA^ 'hear auditorily1.

Lakoff 1965 was the first American generative linguist who called
attention to the need of including exceptions to syntactic rules
within the theory of syntax. He worked out a theoretical notation
for dealing with major and minor rules in terms of markedness
or unmarkedness. This was an important contribution to the de-
scriptional adequacy o f generative grammar.

Weinreich 1966 pointed out the weakness of generative grammar to


deal with several surface structures which must be described as
lexicalized idioms.56 He showed that well-known phrases like to
kick the bucket for 'to die' or to pull somebody's leg f or 'to
joke with somebody' could not be described in derivational terms
within an interpretative syntactic/semantic theory, but must be
entered into the lexicon as idioms. Fraser 1970a has given further
support to Weinreich's claim in a recent article.

The status of idioms such as take advantage of and see the light
within the theory of grammar has, among other things, motivated
Chafe 1970 and 1971 to elaborate a completely new semantic
conception for generative grammar by claiming that there is
directionality in the derivation of sentences (cf. especially
Chafe 1970:68-70).

In computational linguistics the postulation of long chains of


coherent morphemes as constituting lexicalized idioms is a
necessary and a common practice for writing computer programs
for machine translation and other linguistic purposes (cf. Allén
1972 (forthcoming) on lexicalized phrases).

From a psycholinguistic point of view, it is by no means self-


evident that lexicalized idioms must be transformationally derived.
Evidence from both psycho-and neurolinguistics proves the contrary
(cf. Whitaker 1971 who has found that phrase units are so-called
"ictal" speech automatisms and must be considered separately
from the central language system „p. 143)). The idioms or lexicalized
phrases have to be postulated as separate lexical underived units,ànd
inserted as a coherent block into the phrase markers.57

In Swedish verb syntax there are examples of coalescence of two


verbs into one verb. In Modern Swedish the cooccurrence of HÖRAi
116

plus a passive infinitive of the verbs TALA OM 'speak of1 and


SÄGA 'say' is extremely common. Holm 1952:123 points out that
the construction HÖRA TALAS OM is extremely frequent in all
Swedish dialects. The passivized infinitives of the latter
two verbs can be adequately treated as some kind of a set idiom
constituting a main verb, although the set idiom may undergo
various transformations such as Wh-Movement. Compare (154a) and
(154b):

(154) a. A TT VI INTE H ÖRT T ALA S OM HUNGERN I VÄRLDEN


(929:12:1/423452)(DN)
(that we have not heard of the hunger in the world)

b. BADORTER, SO M MAN ALDR IG HÖ R T ALA S OM ( 5 5 1 : 1 7 3:4 / 2 3 5 1 2 3 )( S v D )


( S e a - s i d e p l a c e s w h i c h o n e n e v e r h e a r s of.)

c. vi har hört sägas att han gift sig.


(we have heard that he has gotten married)

The lexicalized character of HÖRA plus TALAS OM or SÄGAS is un­


questionable in (154). We consider the occurrence of verb clusters
in these sentences to be a consequence of the following set lexi­
cal idioms:

(155) C [HÖRA] [TALAS OM] J'hear of'


V V V V V V

(156) [ [HÖRA] [SÄGAS] ] 'hear it said'


V V V V V V

The arguments for considering (155) and (156) as lexicalized


entries are threefold. First, the second part of the idiom, i.e.
TALAS OM and SÄGAS can be dispensed with in (154') to express
the same semantic content as (154) which corresponds roughly
to the syntax of the English translations:

(154) a1, att vi inte hört något om hungern i världen


(that we have not heard of hunger in the world)

b', badorter som vi aldrig hört något om.


(sea-side places which one never hears of.)

c', vi har hört att han gift sig.


(we have heard that he has gotten married.)

Such a paraphrase relationship as is visible between the sentences


of (154) and (154') can be explained in terms of two different
lexical entries of HÖRA. In (154) where the idioms occur which
contain passive infinitives, HÖRA^ 'hear auditorily' is considered
to be the verb of perception, whereas in (154') the simplex verb
HÖRA2 'hear of, be told' can be claimed to be the verb of per­
ception.
117

The situation seems to be similar if the verbum dioendi SÄGAS


or TALAS OM takes an att-Embedding as a verbal complement:

(157) a. Jag hörde


talas omJ att han kom

(I heard that he came yesterday.)

Jag hörde att han kom igår.


(I heard that he came yesterday.)

Sentence (157a) can easily be paraphrased by (157b) without any


major cognitive change of meaning. In (157a) HÖRA^ 'hear auditor­
ily' occurs together with the passive infinitive, but in (157b)
it is HÖRA2 'hear of, be told' that occurs as the verb of per­
ception. (We disregard here the second possible interpretation
of (157b), that is 'I heard acoustically that he came yesterday
(as he parked the car outside my house)).

There is one verb complement, however, which does not allow a


similar paraphrase substitution, that is, when verb nominals
occur after HÖRA TALAS OM 'hear of':

(158) a. Jag hörde talas om hans ankomst.


(I heard of his arrival.)

b. *Jag hörde om hans ankomst.


(I heard of his arrival)

As indicated in *(158b) HÖRA2 'hear of' cannot replace the idiom


HÖRAi TALAS OM 'hear of'.

The second argument for treating HÖRA TALAS OM or HÖRA SÄGAS


as a lexicalized idiom is the fact that the BLIVA Auxiliary
Passivization Rule cannot be applied to the phrase marker under­
lying structures as those of (154) as the following ill-formed
BLIVA-passivized ACI-Constructions demonstrate:

(154) a''. *att vi inte hört hungern bli |Q^aladm| * värlc*en.

(•that we have not heard the hunger spoken of in the


world.)

b''. *badorter som man aldrig hör bli jomtalade™']

(*sea-side places which one never hears spoken of.)

c1'. *vi har hört bli sagt att han har gift sig.
(*we have heard said that he has gotten married-)

The ill-formed structures here will have to be blocked somehow,


if our claim as to the idiomatic status of HÖRA^ plus TALAS OM
or SÄGAS is not accepted. The burden falls upon those who reject
such a claim to elaborate a means to block *(154a'') through
*(154c'f) with a device which would be non ad hoc.
118

A third argument can be presented to support the claim for a


lexicalized status of the Swedish verb HÖRA^ TALAS OM and HÖRA^
SAGAS from comparative syntactic evidence. Similar lexicalized
idiosyncracies occur in both German and English.

In Modern German for instance, pseudoactive infinitives occur


after HÖREN^ 'hear auditorily', where one would expect passive
infinitives in the following sentences excerpted from German
literature:58

(159) a. Man hört ... einen Stuhl rücken


(E. Dwinger Weiss. 93: Bech 141)
(One hears ... a chair scraping)

b. plötzlich hörte sie ihren Namen rufen (R. Huch


Triumph. 120:Bech 141)
(suddenly she heard her name being called)

c. bei Eva Sorel3 habe ich Verse lesen hören


(J. Wassermann Wahn. II 209: Bech 142)
(at Eva Sorel's place, I have heard verses being read)

d. Nur leise hörte man sagen:„Dus ich habe noch einen


Hasen" (E. Glaeser Fried. 95: Bech 142)
(Only softly one heard somebody saying:"You there, I
have one more hare")

The German pseudoactive infinitives are rendered into English


(and also Swedish) by passive constructions in order to be
adequately translated in (159b) and (159c). Compare also the
following translations of (159) into Swedish:

(159) a'. Man hör....en stol flyttas.

b', plötsligen hörde hon sitt namn ropas.

c', hos Eva Sorel har jag hört verser läsas»

df . Bara tyst hörde man det sägas: "Du där, jag har en hare
till."

In (159a) an intransitive English verb, scrape, is more accurate.


In (159d) an active infinite verb form is to be preferred in
English, the gerundive saying.

In colloquial American English a similar active-like infinitive


of TELL also occurs, when a passive infinitive or a past
participle with a deleted auxiliary is used in the standard
language, if the second verb is TELL, but not if the second
verb is SAY.

(160) a. I hav e heard tell he is dead-

b. *1 have heard say he is dead.


119

In order to unify the description of the use of infinitives


(passive or active) of the two verba dioendi (Sw. TALAS OM,
SÄGAS, Germ. SAGEN, and Eng. SAY, TELL) in Germanic languages,
we claim that the choice of active or passive infinitive is
language specific and should be treated as a lexicalized
phenomenon.

However, judging from the numerous potential pseudoactive


infinitives other than SAGEN in Modern German which occur after
HÖREN (and SEHEN), the lexicalized status of such constructions
might be doubted for German, since their occurrence is very
frequent in the literary texts from which the excerpts in (159)
are taken.

Also in Swedish there are instances where the lexicalized status


of HÖRA plus a passive infinitive may be doubted and where a
derivational process seems more motivated:

(161) a. D E TT A MÄSTERSTYCKE S OM MAN NÄ ST A N AL L TI D H Ö R TOLKAS A V MANLIGA


SÅNGARE (1555:80:5/226110)(SvD)
(this masterpiece which one hears almost always
interpreted by male singers.)

b. VI H ÖR HO NOM BA RA O MTALA S (1698:36: 2/426464)(DN)


(We only hear him spoken of.)

There are two strong syntactic arguments for claiming that the
passive infinitives tolkas 'interpreted' and omtalas 'spoken of1
are derived and not part of idioms. First, we cannot delete the
passive infinitive in (161) while the cognitive meaning is
preserved as we could do in (154'). Secondly, the auxiliary BLIVA
can be optionally inserted without any consequences for the well-
formedness of (161a') and (161b'). (As to the Aktionsart
Neutralization visible in the BLIVA-deleted Accusative-with-Past
Participle Construction see section 5.2)

(161) a', detta mästerstycke som man nästan alltid hör (bli)
tolkat av manliga sångare
(this masterpiece which one hears almost always (being)
interpreted by male singers)

b', vi hör honom bara (bli) omtalad.


(we only hear him (being) spoken of.)

As a third argument one could point out that in (161a) and (161b)
two different lexical items appear after HÖRAi, that is TOLKA
and OMTALA as opposed to TALA OM and SÄGA exemplified in (154a)
and (154c) above.

These three facts force us to claim that the passive infinitives


of tolkas and omtalas in (161) have a different status than
sägas and talas om in (154).
120

The alternative solution to the idiom-hypothesis suggested above


for treating HÖRA SÄGAS and HÖRA TALAS OM would be to regard
the passive infinitives as transformationally derived ACI-
Constructions, especially if verbs other than the two verba
dioendi constitute the passive infinitives. The passive infini­
tives which occur in (159') after HÖRAi 'hear auditorily1 are
not regarded as being idioms, but are transformationally derived
surface structures by Passivization and Subject Raising. Since
we are involved with different verbs than SÄGAS and TALAS OM,
the idiomatic status of HÖRA SÄGAS and HÖRA TALAS OM is not
challenged.

A human empty-subject Pro-form, the Agent, can be posited as


the underlying deep subject of flytta 'move1 in (159a'), which
after Passivization and Subject Raising is deleted as shown in
FIGURE 4:

FIGURE 4

(162) a. [ Man hör [ [ Pro flytt- en stol] ] ]


S NP S,[+Hum][+Pass] S, NP S„
Ol 10
(One hears Pro move- a chair)

irPass
T

b. [ Man hör [ [ en stol flytt+s av Pro ] ] ]


S„ NP S n [+Hum] S, NP Sn
Ol IO

TSubj + TAgent

11 Rais Deletion

c. [ Man hör en stol flyttas]


SÄ So

Similar derivational histories can be posited for (159b'),


(159c'), (161a) and (161b). Only (159d') seems to be a problem
here, which we repeat as (163) for the convenience of the
reader:

(163) Bara tyst hörde man det sägas: "Du där, jag har en hare
till."
(Only softly one heard somebody saying: "You there, I have
one more hare.")

It is not possible to delete the verb of saying here under


preservation of the cognitive meaning as was the case in (154')
above. Compare for instance (154a') and the following sentence:
121

(163') Bara tyst hörde man (det): "Du där, jag har en hare till"
(Only softly one heard: "You there, I hav e one more hare")

Before a full quotation, the verb HÖRA2 !hear of' cannot replace
HÖRA^ 'hear auditorily' plus the verb of saying (SÄGAS), A
transformational history must therefore be posited to underlie
the generation of sentences like (163) as demonstrated by the
following figure:

FIGURE 5
(164) a. Man hörde [ Pro säg- [ det [Du där jag har en hare
S,[+Hum] NP S,
till] ] ]
S2 NP Sx

||TPass rpAgent
v1 + Del

b. Man hörde [det [ Du där jag har en hare till] ] äg+s


s
NP S2 S2NP

TSubj TExtra-
Rais Posit

c. Man hörde det sägas: Du där, jag har en hare till

(One heard somebody saying: "You there, I hav e one more


hare!1)

Since the dummy det 'it' occurs in the surface structure of (163),
it is motivated to posit it in front of the complement sentence
of säga 'say' in the underlying structure (a).

Swedish impersonal passives like det sägs1it is being said', in


(163) are considered to be derived from deep structures which
contain impersonal deep human subjects, which are obligatorily
deleted.59 Therefore, after Passivization and Indefinite Human
Agent Deletion, Shallow Structure (b) results. Subject Raising
now applies, raising the entire object [ det[ Du där jag har en
NP S2
hare till] ] into the matrix sentence, whereby infinitiviza-
S2 NP
tion of säg+s to sägas occurs. Now either det-Deletion may occur
yielding

(165) Man hörde "Du där jag har en hare till" sägas.

or Extraposition of the S2 sentence occurs which moves S2 to a


position after the passive complement verb säant* -1
(164c).
122

13,4 the att and hur-embedded sentences occurring after höra 2


'hear 2 0f/ be told'

The problem of embedding has been the topic of lengthy discussion


among generative grammarians. Several hypotheses have been pre­
sented for dealing with the mechanism in language which embeds
sentences into each other. According to one of them embedding
is carried out by the base rules (cf. for instance Bierwisch
1963:38, Rosenbaum 1967b:3, and Lyons 1968:225).

(165) S >(T) NP VP

The T stands for a potential embedding symbol which can be re­


written as SUB by a further rule of the base. SUB stands for a
subordinator.

Another hypothesis of embedding has been presented for English


and Latin by R. Lakoff 1968 who claims that all subordinators
or complementizers are introduced transformationally as that in
English and quod in Latin. These complementizers are then changed
by so-called Complementizer Changing Rules into other types of
complementizers (into how, if, although etc).

It can easily be shown that the latter transformational hypothesis


for generating all complementizers suggested by R. Lakoff is
inadequate for describing the semantics of t/zat-embedded sentences
as compared to t he semantics of how, if, and although-embedded
sentences. Subordinators like how, if, and although are carriers
of meaning and the choice of a that or /zsw-complementizer is
of great consequence for the semantic readings of sentences. A
more adequate way of introducing complementizers than comple­
mentizer changing rules is by means of lexical insertion rules
which operate on the deep structures containing the type node
SUB. R. Lakoff's derivation of complementizers is synchronically
as well as diachronically implausible.

It is true that the most frequent subordinator att in Swedish


or that in English does not have any independent meaning like
hur 'how' or fastän 'although'. Att or that is only a grammatical
function word that signals the status of a complement sentence
as being embedded within a matrix sentence. By assuming that
such subordinators like att or that are introduced by lexical
insertion rules, we postulate a different deep structure for att-
embedded clauses than for the ACI-Constructions. The postulation
of two different deep structures agrees with our intuitions about
the semantic difference between the readings of the (a) and (b)
sentences in (166).

(166) a. ÄN IDA G FÅR MAN HÖR A Ä L D R E P ER SO NER BERÄTTA O M OMÄNSKLIGA


(1036:87:6/123084)(GHT)
(still today one can hear elderly people telling us
of inhuman....)
123

(166) b. än idag får man höra att äldre personer berättar om


omänskliga
(still today one can hear that elderly people tell us
about inhuman )

It is clear that (166a) and (166b) are semantically closely


related. To ignore the semantic differences, i.e. the focus on
elderly people and the lack of explicit tense in (166a) on the
one hand, and the holistic interpretation and explicit tense
of (166b) on the other, would be a mistake, however, which is
often made by linguists who are primarily interested in under­
lying structures and the derivational processes, and who
consequently neglect the accurate description of semantic
distinctions revealed in varying surface structures. In connec­
tion with the ACI-Constructions occurring after SE we pointed
out the semantic differences which exist between ACI-Construc-
tions and at£-embedded surface structures. The arguments for
deriving these two different surface structures from two differ­
ent underlying structures are also the same for HÖRA,and we
refer the reader to section 5.1.2 and 5.1.3 where these arguments
are discussed.

A further argument can be made for setting up a deep structure which


is different for att-embedded sentences than for ACI-Constructions.
When HÖRA2 means 'be told' as in (167), then only att-embedded
structures (=the (a) sentences) are possible, since ACI-Construc-
tions (=the (b) sentences) are completely ungrammatical:

HÖRA,, 'be told, hear of'

(167) a. NÄR ÄG AR EN HÖ RDE A T T JAG H ETT E RAGNAR


(65:118:3/235204) (SvD)
(when the proprietor heard that my name was Ragnar.)

b. * när ägaren hörde mig heta Ragnar


(when the proprietor heard that my name was Ragnar.)

(168) a. J A , SE , VI HÖRDE A TT S JÄL VA ÖVERDJ ÄVULEN S KU LL E K OM M A


(1113:83:12/215225) (SvD)
(yes, we heard that the head devil himself was to
come.)

b. * ja, se, vi hörde själva överdjävulen skola komma.

(169) a. N ÄR H A N A LLT SÅ HÖ RD E , ATT EDWA RD K E NN E DY FUNDERADE P Å ATT BE


OM HJÄLP (308:169:4/533664) (SDS)
(when he heard that Edward Kennedy was thinking of
asking for help)

b. * när han alltså hörde Edward Kennedy fundera på att


be om hjälp.
124

The HÖRA in the (a) sentences is a different kind of verb from


the verb of auditory perception which we have been concerned
with above in sections 13.1 through 13.3 Only the verb denoting
auditory perception can take ACI-Constructions. The HÖRA in (167a),
(168a), and (169a) can be subcategorized as a linguistic, mental,
nonperformative verb belonging to subclass (4b) as described in
section 3.4, but must be marked as [-Subject Rais], since it does
not take ACI-Constructions. For the sake of distinguishing
between the two kinds of HÖRA, we have called the ACI-verb HÖRAi
'hear auditorily1, and the mental linguistic att-embedding verb
HÖRA2 'hear of, be told'.

The latter verbs can also take another complementizer for embed­
ding a complement sentence, i.e. hur 'how'.

HÖRA ^ 'hear auditorily'

(170)a. OCH HAN HÖR HUR ROLIGT DE HAR (720:136:3/1112090) (GHT)


(and he hears how much fun they have.)

HÖRAo 'be told'

b. Jag har inte hört hur han mår


(I have not heard how he feels.)

In order to give an adequate description of the three different


surface structures exemplified here, i.e. the att-embedded
structures in (167) through (169), the ACI-Constructions in
(151), and the hur-embedded structures in (170), we posit three
different underlying representations. The type node symbol T
is expanded by the following base rule:

(171) T *SUB

The embedding symbol SUB for Subordinator is converted by a


segmental rule into the feature [+Sub]. Into the syntactic
matrix of this complex symbol strict subcategorization features
of the verb (or the noun) will be inserted by some kind of
transformational rules for government. These rules are similar
to the transformational rules which assign case features to
object noun phrases as for instance in German, where each
transitive verb has to be subcategorized in the lexicon in terms
of case rule features, i.e. each transitive verb has to be marked
as to whether it takes genitive, dative, or accusative object
noun phrases (cf. Bierwisch 1967). As a parallel to this solution
for handling case feature government in German, complementizer
feature government can be treated as some sort of copying rule that
operates from the main matrix verb onto the preterminal complex
symbol marked as [+Sub] whereby the complementizer features are
assigned. Compare the two government rules proposed in (172a)
and (172b)which assign case and complementizer features to
the noun and the subordinating symbol respectively:
125

The Dative Case Government Rule in German

(17 2) [ C (Det) (Adj) N ] ]


a. VP + —— NP NP NP VP

Ê si-««-
SA: (2) (3)

SC: (2) (3) 4


f+Obi]
I+GQVJ Condition:Obligatory

The [+Obl] and C+Gov] features stand for Bierwisch's plus oblique
and plus governed morphosyntactic features which denote the
dative case. The abbreviation Rgs for Latin Regens symbolizes
that we are involved with case rule features which trigger the
dative case feature in the noun. (It makes little sense to speak
of case features within the verb matrix). The later distribu­
tion of the dative case features to the complex symbols of the
determiner (Det) and the adjective (Adj) is carried out by
Case Agreement Rules.

The Subordinator Government Rule in Swedish

(172)[ [ [+Sub] NP VP ] ] ]
b. vi> • NP(S) . NP S NPVP
. att-Rgsl
(: • hur-Rgs!

SA:

SC: 2
Uatt\
i +hurf
Condition:Optional

Att-Rgs and hur-Rgs are rule features which represent the poten­
tial complementizers which the verb in question can take as sub-
ordinators after copying and lexical insertion have taken place.
Such rules which copy complementizer features onto the subordinat­
ing symbol [+Sub] are therefore similar in nature to the Case
Government Rule as formalized in (172a).In order to generate
surface structures containing att-clauses or fcur-clauses the
embedding symbol T is selected and is rewritten by base rule (171)
as SUB.(When an ACI-Construction is to be generated no such
selection of a SUB symbol is made). By a s trict subcategorization
rule the categorial symbol SUB is converted into the preterminal
complex symbol [+Sub]. A subsequent complementizer government rule,
126

i.e. (172b), copies complementizer features into the [+Sub]


segment. Then a lexical insertion rule applies replacing it
by the conjunction in question which meets the condition of
lexical insertion.

Some verbs like SE and HÖRA have rule features for complementizer
government which allow them to take various sorts of comple­
mentizers. For instance, SE and HÖRA can take att> om> and när
as subordinators, whereas PÄSTÅ 'claim' or ANSE 'consider'
only takes att as subordinator after negated matrix sentences:

(173) a. Han såg (hörde) inte att hon var sjuk.


(He did not see (hear) that she was ill.)

b. Han såg (hörde) inte om hon var sjuk»


(He did not see (hear) if she was ill.)

c. Han såg (hörde) inte när hon kom hem.


(He did not see (hear) when she came home.)

(174) a. Han påstod (ansåg) inte att hon var sjuk-


(He did not claim (consider) that she was ill.)

b. *Han påstod (ansåg) inte om hon var sjuk.


(*He did not claim (consider) if she was ill.)

c. *Han påstod (ansåg) inte när hon var sjuk.


(*He did not claim (consider) when she was ill.)

We are here making syntactic statements about the cooccurrence


of certain verbs and complementizers. It is clear that in order
to rule out the ill-formed (b) and (c) sentences in (174) seman­
tic conditions of well-formedness would be necessary in the sense
of Karttunen 1971, who has suggested a solution based on predi­
cate logic. He claims that each predicate (=verb) is lexically
marked as having a certain set of arguments (sentential or nominal).
A predicate in this sense can be a two-place or three-place
predicate. He presents a list of implicative verbs, negative
implicatives, £/-verbs and only-i/-verbs, which require a given
set of semantic if-then-conditions to be used in sentences. Since
no predicate-logic approach is intended here in describing the
cooccurrence of complementizers and verbs of perception, a
discussion of the semantic well-formedness conditions for the
choice of subordinators in (173) and (174) must be omitted.
There is in our opinion a clear borderline between what one can
do with such a syntactic description as this and what one can do
with a description of the use of sentences based on predicate
logic in the sensé of Kartunen 1971. A classification of verbs
in terms of their syntactic properties continues to be valid
(cf. op. cit. p. 1). For our purpose of describing the syntactic
difference between ACI-Constructions on the one hand, and
127

complementizer-embedded structures on the other, it is sufficient


to state the syntactic conditions for the generation of the
former surface structures and the conditions of complementizer
insertion into the latter surface structures. The insertion of
om and «är-complementizers is blocked, if the main matrix verb
is for instance PÅSTÅ 'claim' or ANSE 'consider'. These are not
syntactically marked as taking such complementizers and the
insertion of them is therefore blocked according to a general
principle of lexical insertion in generative grammar formulated
by Chomsky 1965:84.

13.5 H0RAX OCCURRING WITH NOMINALIZED COMPLEMENTS

So far we have described only sentential complements which occur


as objects after HÖRA^ 'hear auditorily', i.e. ACI-Constructions,
att and /zwr-Embeddings. There are also examples of so-called
nominalized verbs which cooccur as nominals with HÖRA^. Some
excerpts from the Gothenburg corpus will be presented, whereby
the unbalanced and balanced view of nominals will be discussed
in section 13.5.1, the problem of paraphrases in connection with
nominals is treated in section 13.5.2,and finally the status of
metaphors is touched upon in section 13.5.3.

13.5.1 THE UNBALANCED OR THE B ALANCED VIEW FOR D ESCRIBING


N O M I N A L I Z A T I ON S

In generative literature one can find two different approaches to


the problem of nominalization: 1) that of the transformationalists
who maintain the unbalanced view that all noun phrases between
whose parts there is a syntactic relationship should be derived
from underlying sentences through the use of syntactic transforma­
tions (cf. Lees 1960); and 2) that of the lexicalists who maintain
the balanced view that in the case of nomina actionis rule deriva­
tion of nomina actionis may be motivated when underlying verbs
exist, but if no authentic verbs exist from which the nouns in
question can be derived, then no transformational process of
derivation is motivated from a syntactic point of view, i.e. the
so-called balanced (lexicalist) view of nominalization as suggested
by Chomsky 1970a.60»61 Instead a process of word formation within
the lexicon is suggested which is similar to the treatment of
nominalizations as found in traditional grammars or even histori­
cal grammars. Compare for instance Chomsky's 1965 discussion of
the derivation of telegra-ph, horrify3 and frighten(p. 187) and
his treatment of author in the example the author of the book
in Chomsky 1970a:197. The nömen agentis author cannot be derived
from an underlying sentence *the one who auths the book, since
auth does not exist as an independent verb.62
128

It would likewise make little sense to derive RÖST 'voice' in


(175) from an underlying structure 'something which sounds
characteristically human' since there is no syntactic motivation
for positing such a verb as * RÖSTA in the sense of 'sounding
characteristically human':^2

(175) VILKET I NNEBÄR ATT MAN HÖR SIN EGEN R ÖST (13:79:4/112013) (GHT)
(which implies that one hears one's own voice)

Therefore, RÖST 'voice' has to be registered in the lexicon as a


noun denoting 'something which sounds characteristically human'.
A generative semanticist would claim, however, that RÖST is not
the smallest unit of meaning, but that it can be represented in a s emantic
representation in terms of smaller semantic primitives in a
semantic tree representation maintaining as for instance McCawley
does that "semantic structure is not representable by feature
representation, but rather requires a tree representation"
(cf. McCawley 1972:2).63 However, in (176) below a syntactic
process of deriving the nomina actionis SKALL ' bark',MULLER
'thunder', BULLER 'noise', and TRAMP 'clop' from underlying verbs
is motivated on syntactic grounds, since authentic verbs do exist,
i.e. SKÄLLA 'to bark', MULLRA 'to thunder', BULLRA 'to make noise',
and TRAMPA 'to clop'.64

(176) a. FÖRSTA DAGEN HÖRDE JAG BLODHUNDARNAS SKALL EFTER OSS


(933:44:4/425411) (DN)

(on the first day I heard the barks of the bloodhounds


behind us)

b. NU HÖR MAN DET EVIGA MULLRET FRÄN PAPPERSMASKINERNA


(1179:208:2/214234)(SvD)
(Now one hears the eternal thunder from the paper mills.)

C. RENSKÖTARE SOM KONTAKTATS PER RADIO ATT DE HÖRT MOTORBULLER


(127:86:3/215192) (SvD)
(reindeer herders who had been contacted by radio that
they had heard engine noise.)

d. GALOPP KAN VARA SAMLAD ELLER KORT, DÀ MAN HÖR FYRA TRAMP
(1761:4:6/526601) (SDS)
(gallop can be concentrated or short, when one hears
four clops.)

The reasons for taking a balanced view on the derivation of nomina


actionis in (176) are threefold: first, the transformational
rules can here be .claimed to constitute a productive syntactic
process of derivation which reflects the knowledge of the native
speaker to correlate the sentences of (176) with those of (177);
secondly, the ACI-Constructiòns of (177) are in some undefined
way paraphrases of the nomina actionis in (176);thirdly, a
syntactic underlying representation (179) is set up to reconstruct
the native speaker's subconscious knowledge of the paraphrase
relationship holding between the surface structure of (176) and
129

those of (177).

(177) a. första dagen hörde jag blodhundar skälla efter oss.


(on the first day I hea rd bloodhounds barking behind
us.)

b. nu hör man pappersmaskinerna mullra evigt*


(now one hears the paper mills thundering eternally.)

c. renskötare som kontaktats per radio att de hört


motorer bullra.
(reindeer herders who had been contacted by radio that
they had heard engines making noise.)

d. galopp kan vara samlad eller kort, då man hör någon häst
trampa fyra gånger i takt.
(gallop can be concentrated or short when one hears a
horse clop four times rhythmically-)

In section 13.2 examples of HÖRA occurring with object noun


phrases are presented which have to be considered as derived
occurrences where the verb denoting a sound source has been
deleted on its way to the surface structure:

(178) a. och med ens hör jag de sena syrsorna i grannskapet.


(and suddenly I hea r the late crickets in the neigh­
borhood.)

(178) b. jag hörde en klocka i rummet intill.


(I heard a clock in the room next door.)

A shallow structure like (178a1 ) and 0.7 8b1) must be posited


for the surface structure of (178a) and (178b):

(178) a'. OCH MED ENS HÖR JAG DE SENA SYRSORNA SPELA I GRANNSKAPET
(1647:139:4/322399) (ST)
(and suddenly I he ar the late crickets chirping in
the neighborhood»)

(178) b'. jag hörde en klocka slå i rummet intill


(I heard a clock chime in the room next door.)

In (176) no such deletion of a verb denoting a sound source


need be assumed, since the nomina aotionis occurring after
HÖRA are here nominalizations of verbs denoting the act or
state of making sounds. We are therefore motivated from both
semantic as well as syntactic evidence to set up the following
underlying representation of the nomina aotionis in (176), the
ACI-Constructions in (177), and the Complementizer-Embeddings
in (180).65
130

(1791 S

NP VP
V NP
I

t
+V
+Subj Rais
+att-Rgs
+hur-Rgs

X HÖRA

(a) blodhundar skäll- efter oss


(b) pappersmaskiner mullr- evigt
(c) motorer bullr­
ad) någon häst tramp- fyra gånger
i takt

Only the roots of the complement verbs have been represented in


the underl ' phrase marker (179), i.e. as the complex symbol
marked as which occurs in the untensed complement sentence,

Si. To phrase marker (179) either Subject Raising can apply


triggered by the syntactic rule feature in the matrix verb
yielding the surface structures of (177) or Nominalization may
apply triggered by the syntactic categorial symbol Nom before
the complement sentence yielding the surface structures of (176).
All the Swedish noun phrases SKALL 'bark', MULLER 'thunder,
B U L L E R 'noise', and T R A M P 'clop' can thus be paraphrased by
tenseless verbal infinitives as demonstrated by the ACI-Con-
structions of (177). For the English ENGINE NOISE in (176c) there
is no corresponding English verb, so MAKE NOISE must be inserted
as a translation of the Swedish infinitive BULLRA. It is doubtful
on these grounds that a nominalization process should be assumed
for the English translation, whereas it is motivated on syntactic
grounds for the Swedish BULLER.

The underlying subject of F Y R A T R A M P 'four clops' has been dele­


ted in (176d): DÂ MAN HÖR F YRA TRAMP 'when one hears four clops'.
Since the main sentence contains the noun GALOPP 'gallop' one can
infer that the agent of FYRA TRAMP must be a horse or several
horses. This inference motivates us to set up NÅGON HÄST'some
horse' as the complement subject (cf. Weinreich 1966 on the
semantic transfer rules which assign the correct interpretation
to discontinous syntactic elements). Despite this possibility
of reconstructing a deep complement subject for TRAMP by means
of Weinreich's transfer rules, we are sceptical about the
plausibility of such an ad-hoc underlying syntactic structure
for (176d) as given in (179d).
131

The problem with generative grammar is its excessive generative


power. No principles have so far been formulated which delimit
the range of possible representations for various surface struc­
tures. The nature of transformational rules has not yet been
clearly formulated by any generative grammarian, not even by
Chomsky. If the concept of syntactic transformation is to have
any meaning at all, some principles for the use of transforma­
tions must be formulated, otherwise it will be used to mean
practically anything as is the case in the recent writings of
generative semanticists, where phrase markers, syntactic
categorial symbols., and transformations constitute completely
different concepts than originally intended by Chomsky 1957
and 1965.

To solve the problems of tropes like MAN H ÖR F YRA TRAMP which is


also a pars pro toto expression, one could posit two different
kinds of syntactic representations, one involving a subject-
predicate relationship and the other involving a genitive-noun
relationship. Only the smaller part of the noun phrase expression
is explicitly stated; the remainder of the expression must be
implied by the listener. It is clear that horse must be inferred
somehow, since the word GALOPP occurs in the preceding sentence.
The question is whether the process of inferring should be
regarded as a structure-dependent as structure-independent
operation via a deep subject and a verb which is nominalized or
v i a t h e n o u n T RA M P w h i c h o c c u r s w i t h a d e l e t e d n o u n , i . e . HÄST
'horse'. We know of no clear answer to this choice between
underlying representations for (179d). From a psycholinguistic
point of view there is no way to tell us whether a subject-
predicate relationship or a genitive-noun relationship is primary.
From a syntactic point of view a subject-predicate relationship
is motivated within the balan ced lexicalist framework because
of the existence of an authentic verb for T R A M P , i.e. T R AM P A .

Under the assumption of a subject-predicate origin of the nomina

syntactic recategorization from resulting in


nominalized verbs , i.e. as nomina aotionis. The Nominalization
contemporaneously triggers the Genitivization of the complement
s u b j e c t i n c h a n g i n g BLODHUNDA RNA SKÄLL- i n t o BLODH UNDARNAS S KA L L
in (176a). In (176d) the same kind of Nominalization is
accompanied by an outspelling of a preposition plus the inversion
of the underlying recategorized verb and genitivized subject
noun which changes PAPP ERSMA SKINE RNA MULLR A EVI GT in to DET EVIGA
M UL LR ET FRÄN P APPERS MASKIN ERNA a l t h o u g h PA PPER SMASK INERN AS EVIG A MU LLER
would also be possible. The choice of a prepositional outspell­
ing or a genitivized subject is optional here. An elimination
of major syntactic boundaries has taken place in (176c) together
with the nominalization where MOTORER BU LLRA is transformed into
MOTORBULLER. In (176d) we can assume a deletion of the comple­
ment subject HÄST 'horse' plus the Nominalization of TRAMP- into
TR A M P A t o g e t h e r w i t h a r e c a t e g o r i z a t i o n o f t h e a d v e r b i a l FYRA
GÅNGER 'four times' into thé cardinal number FYRA and possibly
the deletion of I TAKT 'rhythmically'. However, we find such a
derivation of the methaphor in %(176d) unnatural and counter­
intuitive and suggest the lexicalist solution for the insertion
o f TRA MP.
132

13.5.2 THE PROBLEM OF PARAPHRASES

To assign identical meanings to the nomina aotionis of (176)


the ACI-Constructions of (177) and the complementizer-embedded
structures of (180) below would also be counterintuitive.66

(180) a. första dagen hörde jag blodhundarna skällde


efter oss. <•

(on the first day I hea rd Ijjowf tlie bloodJlounds

barked behind us-) * I J

b. nu hör man Jattl pappersmaskinerna mullrar evigt.


Ihurl
(now one hears Kqw ] t*ie paper mills thunder eternally.)

renskötare som kontaktats per radio att de hörtij^B


motorer bullrat. '

(reindeer herders who had been contacted by radio that


they had heard' Jjow f en8^nes made noise.)

d. galopp kan vara samlad eller kort, då man hör


någon häst trampar fyra gånger i takt. * '

\
(gallop can be concentrated or short when one hears
(&a
lhow* ra ^orse C^°PS four times rhythmically.)

In sections 5.1.1 and 13.4 it is argued that att and fcwr-Embeddings


obtain different semantic readings from ACI-Constructions, be­
cause of the fact that they contain complementizers and also
that they are tensed sentences. A holistic interpretation will
therefore result in complementizer-embedded sentences, whereas
an ACI-Construction gives a focused interpretation of the raised
noun phrase.

In the case of nomina aotionis represented in (176) a similar


focusing is involved in interpreting the genitivized complement
subject and nominalized verb in (176a), the nominalized verb
in (176b),(176c), and (176d), Considering the relatively large
difference in semantic readings between nomina aotionis, ACI-
Constructions, and two types of complementizer-embedded struc­
tures, as exemplified by (176), (177), and (180) respectively,
we cannot accept the postulate of the Standard Theory that the
underlying phrase marker (179) alone determines the semantic
interpretations of the four types of surface structures treated
in this section.
133

In the Extended Standard Theory as suggested recently by Chomsky


1970b, 1971a,and 1971b,and Katz 1970, 1971,and 1972, the
cognitive meaning is determined by a deep structure configura­
tion in the sense of the Standard Theory, whereas aspects of
focus, topicalization, emphasis, scope of negation, and quanti­
fier interpretation are determined by surface structures. This
recent proposal by interpretative semanticists is certainly an
improvement for tackling the problem of describing the semantic
relationships between paraphrases like nomina aotionis, ACI-
Constructions and att or /zwr-embedded structures.

Since this investigation is primarily syntactic, we are more


concerned with the syntactic operations which a native speaker
masters when he is able to correlate the surface structures of
(176), (177), and (180) with each other. We are convinced that
the operations involved are part of his syntactic competence.
The abstract underlying structure (179) is therefore to be
seen as a descriptive device for illustrating the syntactic
operations which we here have called Nominalization, Subject
Raising, and Complementizer Embedding.

By positing such an underlying structure we are not claiming


that (179) is a psychological reality in the sense of the
classical phoneme. We do, however, believe that the transforma­
tions involved in the generation of the surface structures of
(176), (177), and (180) must be represented in some neurologi­
cal way in the human brain (cf. Weigl & Bierwisch 1970 and
Whitaker 1971 on arguments for and against neurological corre­
lates of transformational rules).

13.5.3 METAPHORS AS ARGUMENTS F OR A S EMANTI C I ST SOLUTION OF


NOMINALIZATIONS

Metaphors like (181a) clearly demonstrate the implausibility


of the unbalanced view of deriving all noun phrases between
whose parts there is a syntactic relationship from underlying
sentences through the use of syntactic transformations, i.e.
the transformationalist position as represented by Lees 1960.

(181) a. O C KS Å H Ä R HÖR MAN ROPET I B R EV E T FRÅN SOLDATE N V ID FRO NTE N


(636:179:6/422405)(DN)
(also here one hears the call in the letter from
the soldier at the front-)

Chafe 1970 has suggested a special set of rules, the so-called


postsemantic rules, which operate between the semantic units
and the symbolization rules (=Chomsky's 1965 lexical insertion
rules) in order to handle the effects of idiomatization. Also
a semantic unit like ROPET I BREVET 'the call in the letter' can­
not have a direct symbolization in terms of the lexical items
which constitute the noun phrase but trades on the symboliza­
tion of other semantic units. Even a balanced (lexicalist)
position is inadequate for handling the semantic interpretation
134

of this noun phrase, since nobody would claim that the under­
lying structure of (181a) is something like (181b) or (181c):

(181) b,. också här hör man [ soldaten vid fronten ropar i
brevet] S-,
si

c. också här hör man [ ropet i brevet från soldaten


vid fronten] NP
NP

Syntactic underlying structures like (181b) and (181c) are not


in agreement with the condition of truth value which must be
considered as basic for setting up underlying semantic and syn­
tactic structures. Despite the fact that we can both find a
s u i t a b l e d e e p v e r b a s u n d e r l y i n g ROPET i n ( 1 8 1 a ) , i . e . ROPA
which would fulfil the lexicalist's requirement, and elaborate
plausible transformations to nominalize this underlying verb,
the postualted derivation would be ad hoc. An idiom for Chafe
1970 is a semantic unit 1) which "does not have a direct
symbolization of its own", and 2) which is characterized by its
"limitation to very narrow contexts" (cf. p. 44). As long as
syntactic theory as proposed by Chomsky and other syntacticists
cannot handle the numerous examples of idioms and metaphors in
a natural language, the level of explanatory adequacy has not
been reached. The inadequacy of the Standard Theory to solve
such problems as the semantic interpretation of idioms and
metaphors has lead Chafe and others to claim that there is
directionality in the derivational process so that semantic struc­
tures are prior to syntactic structures. By postulating post-
semantic processes along the line of Chafe 1970 and 1971 or
Weinreich 1966 a better understanding of the status of idioms and
metaphors within our linguistic knowledge can be attained.

13,6 THE VERBAL COMPLEMENTS OF FÂ HÖRA^ 'HEARJAUDITORILY' AND


FÂ HÖRA2 'HEAR2OF, BE INFORMED'
13.6.1 THE C ATEGORI AL STATUS OF SWEDISH F A I N FÂ HÖRA

In section 3.1 the verb compound FÅ HÖRAi 'hear auditorily and


suddenly' is classified as a simplex verb which expresses a
passive auditory perception on the part of the human Experiencer.
The nonperformative compound verb FÅ HÖRA2 'be told, hear of'
could be subcategorized by Rule (iv) in section 3.4 and could
also be treated as a verb of perception but of a passive mental
or cognitive character, if it took ACI-Construction.

The two verbs occur in Modern Swedish as instantaneous indeliber­


ate auditory and mental verbs of perception which take human
Experiencer NP's as deep subject. Since they are semantically
as well as syntactically two different verbs we will call them
FÄ HÖRAi 'suddenly hear' and FÂ HÖRA2 'be told, hear of'.
135

In section 5.5.1 the categorial status of Swedish FÅ before the


verb SE 'see' was discussed. The categorial status of a perfec­
tive aspectual auxiliary was assigned to FÅi in the verb compound
FÂ SE 'catch sight of' on the basis of comparative syntactic
evidence, i.e. comparisons with the perfective prefix /u-/ in
Slavic languages as presented in section 5.5.2. The perfective
aspectual auxiliary status could also be assigned to FÅi on the
basis of semantic Aktionsart blocking of the cooccurrence of
FÄ, with inherently durative verbs of active perception such as
OBSERVERA and IAKTTAGA, both meaning 'watch', because of the
instantaneous and indeliberate Aktionsart inherent in FÅi (cf.
section 5.5.4).

The function word FÂi, the perfective-aspectual auxiliary, was


also contrasted in section 5.5.2 with the four other verbs in
Swedish which are homographs or homophones of FÂ]_, that is FÄ£
'the future auxiliary', FA3'the modal auxiliary', FÂ4'the
permissive auxiliary', and finally FÅ5
'the causative verb'.
(The main verb FÄ6 which means 'obtain' was disregarded as it
will be here).

For the sake of a parallel description of FÂi as occurring be­


fore the two verbs SE 'see' and HÖRA 'hear' the occurrences of
FÂ SE 'catch sight of ' in (91) through (96) have been replaced
by FÅ HÖRA 'suddenly hear' in the following examples, which
have been slightly altered to construct appropriate linguistic
contexts for this verb of auditory perception:

FÅ^ HÖRA^ 'perfective-aspectual auxiliary'

(182) när jag kom gatan fram och på avstånd fick höra två
unga människor skälla på varandra.
(when I came up the street and at some distance I hea rd
two young people yelling at each other.)

FÅ^HÖRA^ 'future auxiliary'

(183) i morgon får du höra hur det blir i fortsättningen,


(tomorrow, you will hear what happens in the future.)

FA3 HöRA2 'modal auxiliary'

(184) varje ärende följs upp en tid efteråt så att man får
höra hur den handikappade kunnat anpassa sig.
(each case is carefully examined so that one can hear
how the handicapped person has been able to adapt himself.)

FÂ4 HÖRA^ 'permissive auxiliary'

(185) ingen under 21 år får höra prostituerade sjunga i Hamburg,


(no one under 21 may listen to prostitutes sing in Hamburg.)
136

FÅg HÖRAj. ' causative verb'

(186) jag kan kanske få dig att höra rätt


(I may be able to make you hear right.)

Of all these types of FÄ only the perfective-auxiliary interest


us here, that is FÄ^.

13.6.2 T H E S E N T E N T I A L C O M P L E M E N T S T R U C T U R E S O F FÅ H Ö R A 1
' s u d d e n l y h ear^ '

In (187) through (190) FÅ-^ is a perfective-aspectual auxiliary


expressing an inchoative or indeliberate Aktionsart under­
lying the semantic contents of these sentences. In (187a) the
reading of 'be allowed to' is also possible within the given
context, but this meaning will be disregarded here.

ACI-Construction

(187) a. Ä N I DA G FÅR M AN HÖRA ÄLDR E P E RS O NE R BERÄTTA OM OMÄN SKLI GA.. .


(1036:87:4/123084) (GHT)
(still today one hears (can hear) elderly people
speaking of inhuman )

b. domaren fick höra flickan korsförhöras på den


punkten.
(the judge heard the girl cross-examined on that
issue.)

att-Embedding

(188) a. OF TA FÅR M AN HÖR A A V FOLK, ATT DE T ÄNKER VINTERBONA ST UG OR NA


(969:179:8/515636) (SDS)
(Often one hears people say that they are going to
equip their cottages for the winter.)

b. jag fick höra att han kom i trappan.


(I suddenly heard him coming up the stairs.)

hur-Embedding

(189) när jag kom uppför trappan fick jag höra hur
grannarna slogs i våningen nedanför#
(when I came up the stairs I sud denly heard how
the neighbors were having a fight downstairs.)
Accusative-with-Past Participle
(190) ?domaren fick höra flickan korsförhörad av advokaten,
(the judge heard the girl cross-examined by the
1awye r.)
137

As (187a), (187b), (188), and (189) demonstrate, the inchoative


or indeliberate Aktionsart needs an overt aspectual marker. The
auxiliary FÅ^ in these sentences can be characterized as such
from both a comparative syntactic and semantic point of view; it
is furthermore accurate to regard FÅ^ in the verb compounds FÂ SE^
'catch sight of1, FÄ HÖRA^ 'suddenly hear', FÂ HÖRA2 'be told',
and FÅ KÄNNA 'feel tactually', as a perfective-aspectual auxiliary
which is limited as a functional word to these four main verbs
of physical and mental perception. The historical origin of this
use of FÅi in Modern Swedish (and Norwegian) remains in the
dark, since no historical handbook which describes the history
of this auxiliary gives any explicit information on the perfective-
aspectual use and its origin (cf. for instance Hellqvistl939:250
and the SAOB 1926:col. F 1936).67 We assume that the meaning of
'begin to' was prior to the inchoative-indeliberate aspectual
meaning (cf. German Er fängt an zu hören 'He begins to hear')).

However, FÄi is no longer productive as an independent perfec-


tive-aspectual auxiliary in Modern Swedish, but is limited to
the five main verbs just enumerated. (Cf. the SAOB 1928:col.F
1950 which claims that FÄi as the aspectual-perfective auxiliary
is avoided in higher literary style).

As evidence for the inherent inchoative or indeliberate Aktionsart


in FÂ'i compare sentences (187) through (189) with the following
sentences where a simplex HÖRA^ 'hear auditorily' occurs in the
past tense:

(187) a', än idag hör man äldre personer berätta om omänskliga..


(still today one hears elderly people speaking of
inhuman )

b'. domaren hörde flickan korsförhöras av advokaten på


den punkten.
(the judge heard the girl being cross-examined by the
lawyer on that issue.)

(188) a', ofta hör han av folk att de tänker vinterbona


stugorna.
(he often hears from people that they are going to
equip their cottages for the winter.)

b', jag hörde att han kom i trappan.


(I heard him coming up the stairs.)

(189') när han kom uppför trappan hörde han hur grannarna
slogs i våningen nedanför.
(when he came up the stairs he heard how the
neighbors were fighting downstairs.)

The Swedish HÖRA^ 'hear auditorily' alone in (187a') through


(188b1) does not express explicitly the inchoative or indeliberate
Aktionsart implied by the sentences (187a) through (188b) which
contain occurrences of FÄ HÖRA^. In the former sentences no
138

explicit inchoative or indeliberate readings are obtained. The


hearing of the girl by the judge in (187b*) goes on from the
beginning to the end as the lawyer examines the girl, but not
necessarily so in (187b). The current process of hearing is
expressed in the English translation of (187b') by means of
the aspectual durative marker being, which expresses the
semantic difference to (187b).

In (189') the person who is climbing the stairs from the first
step to the last step hears the neighbors fighting downstairs,
whereas no such total durative implication of hearing is
experienced in interpreting (189). To express the semantic
difference in aspectual meaning the Swedish fick is not used
in (1891 ). In the English translation of (189) the time adverb
suddenly helps to assign the intended perfective-aspectual inter­
pretation to the process of hearing, which is not possible in
the translation of (189f).

In the light of the contrasts of aspectual meaning between (187)


through (189) on the one hand, and (187') through (189*) on
the other, a different lexical status of FÅ HÖRA^ 'hear suddenly'
can be claimed than for the simplex HÖRA^ 'hear auditorily'. It
is practical to regard the verbs as two surface verbs, irrespec­
tive of the fact that the aspectual auxiliary has been described
as being spelled out by a special Aspectual Marker Outspelling
Rule in section 5.5.2 (cf. also Subcategorization Rule (iii) in
section 3.3),

13.6.3 THE SENTENTIAL COMPLEMENT STRUCTURES OF FÅ HÖRA2


• h e a ^ o f, b e t o i d 1

In the Gothenburg corpus att-Embeddings are more numerous than


any other sentential complement after FÂ HÖRA2 'hear of'. ACI-
Constructions are impossible after this verb as seen in *(192)
(cf. the non-cooccurrence of ACI-Constructions after the mental
cognitive verb SE2 'realize' in section 5.1.2). FÄ HÖRA2 'hear
of, be told' must therefore be marked as not allowing Subject
Raising, which also explains the noncooccurrence of Accusative-
with-Past Participle Constructions (cf. *(194)).

att-Construction

(191) a. J AG FI CK HÖ RA ,A TT HAN S K U L L E Â K A IN I N O RG E
(568:18:6/212198) (SvD)
(I heard that he was going into Norway.)

b. NÄR DE F I C K HÖRA A T T GREVE EHRENSVÄRD UPP EHÖ LL SI G I NK OG NI T O


(307:294:6/522590) (SDS)
(when /they heard that Count Ehrensvärd stayed in­
cognito«)
139

ACI-Constructiçn

(192) *jag fick höra honom åka in i Norge.


(I heard him going into Norway.)

hu?-Embedding

(193) när de fick höra hur han hade behandlat henne.


(when they heard (learnt) how he had treated her.)

Accusative-with-Past Participle

(194) *de fick höra av folk greven trakasserad av personalen


(•they heard from people the count molested by the staff)

13.6.4 METAPHORICAL USAGE O F FÅ HÖRA2 'hear of2,be told'

The auditory verb HÖRA normally requires a human Experiencer


as subject noun phrase. However, metaphorical usages of FÂ
HÖRA2 'be told' give evidence of acceptable violations of this
semantic well-formedness constraint, which Chomsky 1965 calls
"violation of selectional restrictions".

(195) NÄSTAN T I L L LE DA H A R DE ES TETISKA Ä M N E NA FÅTT HÖ RA A TT D E I


ME TODISKT HÄN SEE NDE S LÄ P AR LÅNG T EFTER
(936:2:3/422444)(DN)
(The esthetic subjects have heard almost ad nauseum that
they are retarded in a methodological sense.)

A Chomsky type of grammar would block metaphorical sentences like


(195), since the Aspects-theory does not allow for violations
of the human-subject constraint on the occurrence of FÄ HÖRA 2
as the main verb. A lexical insertion would not apply here. It
is clear that we want our grammar to be able to generate metaphors
like (195) as well. By assigning DE ESTETISKA ÄMNENA 'the esthetic
subjects' the feature [+Human] derived from a larger but partially
deleted noun phrase förespråkare för de estetiska ämnena
'proponents of the esthetic subjects' we believe that the problem
of the selectional violation in (195) can be adequately solved.
McCawley 1968 considers all selectional restrictions to be
semantic in nature and claims that selectional restrictions are
brought about by entire syntactic constituents and not by some
"head" contained in them as claimed by Chomsky 1965. Since the
"head" in the subject noun phrase in (195) is förespråkare
'proponent' or something similar which has been deleted in the
surface structure (195), McCawley's 1968 suggestion to assign
the [+Human] feature to the entire underlying noun phrase node
förespråkare till de estetiska ämnena constitutes a nice solu­
tion for explaining the possible violation of the human-subject
well-formedness condition for the occurrence of FÅ HÖRA2 in (195)
(cf. the discussion of selectional restrictions and presupposi­
tions in Kuroda 1969).
140

M. RESULTS OF THE LINGUISTIC PERFORMANCE STUDY OF VERBAL


COMPLEMENTS OCCURRING AFTER HÖRA AND FÅ HÖRA
In the Gothenburg corpus sentences have been retrieved which can
be shown to have undergone four different transformational rules
of subordination in our sense: 1) Subject Raising, 2) att-
Embedding, 3) Nominalization occurring with genitive as well as
prepositional noun phrases, and 4) hur-Embedding.

There are fourteen occurrences of lexicalized passive infinitives


occurring after HÖRAi 'hear auditorily' such as TALAS OM 'spoken
of' SÄGAS 'said', and RYKTAS 'rumored'» These passive infini­
tives have been treated as idioms for reasons presented in sec­
tion 13.3. However, some passive infinitives such as those
given in (151) and (161) must be considered to b e the result
of the operation of the Passive, Agent Deletion,and Subject
Raising Rules. Only two such passive ACI-Constructions have been
retrieved in the Gothenburg corpus, i.e. (161a) and (161b).

An examination of the embeddings occurring after the different


tense forms of the verbs HÖRA^ 'hear auditorily' and HÖRA2
'hear of, be told' as illustrated by Tables 10 and 11, shows
that the Subject Raising Rule is a very productive rule of
embedding as compared with the three other rules of embedding
concentrated upon here:

M.L HÖRAI 'HEARIAUDITORILY'


Table 10 THE PRODUCTIVITY OF EMBEDDING R ULES AFTER HÖRA,
1 hea r auditorily1

Type of Embedding Number Distribution of Embedding


Types after the Verb HÖRAj
'hear auditorily' in the
Gothenburg Corpus
1) ACI-Constructions(with
active or passive 25 52 %
infinitives)
2) Passivized Idioms
(HÖRA TALAS OM, 14 29.2V
HÖRA RYKTAS)
3) Nominalizations 7 14.6 %
4) /zur-Embeddings 2 4.2 %
5) att-Embeddings 0 -

6) Accusative-with-Past
Participle Constructions 0 -

Total number of embeddings: 48


Total number of examples of
the homograph HÖRA in the
Gothenburg corpus: 502
141

The statistics of Table 1. show us that more than half of all


embeddings occurring after HÖRA-^ in this corpus are carried
out by means of the Subject Raising Rule. If, contrary to our
suggestion,the passivized idioms such as HÖRA TALAS OM 'hear
of' and HÖRA RYKTAS 'hear it rumored' etc are considered not
to be idioms as we suggest, but rather structures derived
through the Subject Raising Rule an even higher percentage is
attained for this rule, i.e. almost 82 per cent.

Due to insufficient data-coverage in the Gothenburg corpus no


att-Embeddings after HÖRA^ have been found. Of the /zur-Embedding
only two embeddings have been retrieved which yield a percentage
of 4.2 %.

If nomina aotionis are treated in the balanced view of describing


nominalizations, they can be said to be relatively numerous,
i.e. more numerous than att-Embeddings, (14.6 % of all embeddings
after HÖRA.^).

w.2 höra 2 'hear 2 of, be told'


Table 11 THE PRODUCTIVITY OF EMBEDDING R ULES AFTER HÖRA2
'hear of, b e told'

Type of Embedding Number Distribution of Embedding


Types after the Verb HÖRA^
'hear of in the Gothen­
burg Corpus
1) att-Embeddings 5 100%
2) Passivized Idioms
(HÖRA TALAS OM, 0 -

HÖRA RYKTAS)
3) ACI-Constructions 0 -

4) Accusative-with-Past
Participle Constructions 0 -

5) /zwr-Embeddings 0 -

6) Nominalizations 0 -

Total number of embeddings: 5


Total number of examples
of the homograph HÖRA in
the Gothenburg corpus: 502

The statistical survey of the embedding mechanism which is


presented here for HÖRAj 'hear auditorily' and HÖRA2 'hear of,
be told' shows the productivity of transformational rules which
gives functional perspectives of syntactic rules in performance
as they can be studied on the basis of observations of a given
corpus.
142

The data-coverage in any corpus must necessarily be limited, but


this fact does not mean that any aspect of data is uninterest­
ing. Additional depth tò syntactic research is gained if
aspects of the use of transformational rules are included.

The results gained through studies in performance tell us that


the Subject Raising Rule in our sense is more frequently used
for generating complement structures after HÖRA^ 'hear auditorily1
than was assumed at the outset of the investigation of sentential
complementation in written Modern Standard Swedish.

The limitations of a finite corpus approach must be complemented


by intuitive observations of sentences to make up for the
insufficient data coverage demonstrated above in Tables 10 and
11, where for instance in Table 10 att-Embeddings after HÖRA^ do
not occur in the Gothenburg corpus, but they are known to exist
as demonstrated by examples (188a) and (188b) after FÂ HÖRA^.
Sometimes the lack of a sentential complement type in the Gothen­
burg corpus as illustrated by the zero of ACI-Constructions in
Table 11, depends upon the impossibility of this kind of sentential
complement after HORA 2 'hear of'. A mechanical enumeration of the
frequency of sentential complements of a given corpus does not
give us a realistic picture of the embedding process in a given
language, if the distribution of complement types is not comple­
mented by a syntactic description of possible and impossible
sentential complements. An intuition-based approach is indispens-
ible when the linguistic corpus is incomplete (cf. section 1.6).

The tables of distribution of embedding structures must therefore


be used together with our description of the transformational
rules, the conditions on these rules, and the semantic descrip­
tions of the verbs of perception preceding the sentential comple­
ments.
143

14,3 FÄ HÖRAX 'HEAR!SUDDENLY OR UNEXPECTEDLY'


Table 12 THE PRODUCTIVITY OF EMBEDDING R ULES AFTER FÅ Hö RA ^
1 hear sudden 1 y 1

Type of Embedding Number Distribution of Embedding


Types after the Verb FÅ
HÖRA * 'hear suddenly' in
the Gothenburg Corpus
1) ACI-Constructions 1 33.3 *
2) Passivized Idioms
(HÖRA TALAS OM) 1 33.3 %
3) at£-Embeddings 1 33.3 %
4) /zur-Embeddings 0 -

5) Nominalizations 0 -

Total number of embeddings: 3


Total number of examples
of the homograph FÄ HÖRA
in the Gothenburg corpus 26

No ftwr-Embeddings or Nominalizations have been found to occur as


verbal complements after FÅ HÖRA^ 'hear suddenly or unexpectedly1
in the Gothenburg corpus. As demonstrated by (189) above and
(196) below /zwr-Embeddings are possible as verbal complements
after FA HÖRA^. Also Nominalizations are possible as verbal
complements after this verb, although these are not found in
the Gothenburg corpus either (cf. section 14.2 on the limitation
of a finite corpus).

hur-Embedding

(196) Han fick plötsligt höra hur hon skrek högt.


(He suddenly heard how she screamed loudly)

Nominalization

(197) Han fick plötsligt höra hennes skrik på hjälp.


(He suddenly heard her scream for help.)
144

14,4 FÂ HÖRA2 'BE TOLD/ HEMHOF'


Table 13 THE PRODUCTIVITY OF EMBEDDING R ULES AFTER FÂ HÖRA2
1hear of 1

Type of Embedding Number Distribution of Embedding


Types after the Verb FÅ
HÖRA 2 ' hear of in the
Gothenburg Corpus
1) att-Embeddings 6 100 %
2) ACI-Constructions 0 -

3) Passivized Idioms
(HÖRA TALAS OM) 0 _

4) ftur-Embeddings 0 -

5) Nominalizations 0 -

Total number of embeddings: 6


Total number of examples of
the homograph FÄ HÖRA in
the Gothenburg corpus: 26

As was the case with the simplex verb HÖRA2 'be told, hear of'
in section 14.2, the compound verb FÅ HÖRA2 'be told unexpectedly1
only takes att-Embeddings as verbal complements in the Gothenburg
corpus. No complementizer embeddings with ftwr have been found,
although examples like (193) demonstrate that such verbal comple­
ments are well-formed after FÄ HÖRA2.

ACI-Constructions are on the other hand not possible as demon­


strated by (192). No Nominalizations have been retrieved which
occur after FÅ HÖRA2.
145

15, THE SUBJECT RAISING RU LE IN S WEDISH (KÄNNA)


The SAOB 1939:cols. K 3643-K 3650 gives eight classes of the
homograph KÄNNA, several of which are obsolete and will there­
fore be disregarded.

Of the remaining current meanings, KÄNNA in the sense of 'to


know somebody' cannot take sentential complements and will
therefore also be disregarded. Some classes of KÄNNA described
in the SAOB 1939 however, take such complements, i.e. KÄNNAi
'feel tactually' and KÄNNA3 'feel mentally'. Svensk handordbok
1960:414, and Illustrerad svensk ordbok 1958:807 do not explicitly
mention the fact that KÄNNA meaning 'smell', i.e. KÄNNA2, also
occurs in Swedish.68 A fourth distinction which we are motivated
to make here is to separate KÄNNA^ 'feel tactually' from KÄNNA4
(SIG) 'feel somatically' (cf.Han kände sig trött och nere 'He
felt tired and depressed').

To classify KÄNNAx 'feel tactually', KÄNNA7 'smell', and KÄNNA4


'feel somatically' into the same semantic class as implied by the
description in the Swedish dictionaries mentioned here seems
unmotivated, since the three verb classes express three completely
different processes of perception, i.e. tactual, olfactory, and
common somatic perception. All four verb classes require
Experiencer subject NP's in the Deep Structure and are therefore
passive perceptual verbs in the sense defined in section 3.3.69

Subclassification Rule (iii) in section 3.3 enumerates four


homophonous verbs which represent four different kinds of per­
ception, i.e. tactual (Class (3a)), olfactory (Class (3b)),
mental (Class (3c)), and somatic (Class (3d)), although they are
associated with one and the same phonological matrix KÄNNA. In
order to distinguish between the four classes we will call them
KANNÄI TACT , KANNA
2 O 1£ , KÄNNAj
ment> and KÄNNA,
somat respective-

ly. Such a subclassification of KÄNNA is motivated on semantic


as well as syntactic grounds.70

15,1 THE VERBAL COMPLEMENTS OF KÄNNA} 'FEEL^ TACTUALLY'


The more frequent use of KÄNNA plus an ACI-Construction occurs
when this verb denutes tactual perception, i.e. KÄNNAj tact:

(198) a. MA N KÄ NNE R GRU SET K NASTRA UNDER FO TSULO RNA


(180:8:4/226190) (SvD)
(you feel the gravel crunching under the soles of
your feet-)

b. SOM J A G SETT MÄNGA G ÅN GE R OCH KÄNT D AR RA GENOM MIG


(1305:74:6/522540) (SDS)
(which I had seen many times and felt shivering
through me.)
146

(198) C. FRI DE GÅRD H A R KÄNT SI N P E N NA L EDA S A V EN M AK T SO M ST ÅR UTANFÖR


H O NO M ( 1 2 9 9 : 1 0 1 : 2 / 5 2 6 5 3 3 ) ( S v D )
(Fridegård has felt his pen guided by a power which
is beyond him*)

att-Embedding and hur-Embedding


hur) r u s e t
(198) a1, man känner attl^ knastrar under fotsulorna,
IU tj
ai
(one feels \ th a t j t * i e g r a v e l cr unc hes u n d e r t h e s o l e s
of your feet.)

*som^ jag sett många gånger och känt den^


m^ir
darrar genom mig

(•which^ I ha d seen many times and felt (s,|»t


shivers through me.)

fhur \
Fridegård har känt|att f bans penna leds av en makt
lattl
onom-,
som står utanför honom»,
(Fridegård has f^M^hatih
power which is beyond him.)
{H. " ^ en by a

Both att and ftur-Embeddings are possible as can be seen from


(198a1) and (198c'). However, there is a constraint on the att
and /zur-Embedding Rules when the sentence undergoes the Wh-
Movement Rule in (198b') (cf. The same constraint on Wh-Embedd-
ing and complementizer embedding after SE-^ in section 5.1.2)

Nominalization

(199) H O N KAN MÅL A BJ ÖR KAR S Å ATT MAN NÄ S TA N K ÄNN ER S TA MM A RN AS SK RO VL I G­


H ET KLI STRAD P Å Ö GO NEN ( 1 8 0:5: 7 / 2 2 6 1 9 0 )( S v D )
(She can paint birches so that one almost feels the
roughness of the trunks glued on one's eyes.)

Sentence (199) contains a metaphor after K ÄNNER 'feels' which


can be syntactically explained in terms of a Nominalization
Rule changing the adjective SKROVLIG into SKROVLIGHET. However,
a c o n c r e t e i n t e r p r e t a t i o n o f t h e n o u n p h r a s e ST AMMA RNA S S K RO V LI G ­
HET 'the roughness of the trunks' together with the verb form
KLISTRAD 'glued' is not possible. It is clear that a surface
structure interpretation rule in the sense of Katz 1972 is
necessary to account for the semantic interpretation of metaphors
like (199). It is therefore unsure whether we should regard
SKROVLIGHET as a Nominalization here.
The ACI-Construction in (198c) occurs after a verb which could
also be classified as a verb of mental perception, i.e. as
KÄNNA3 'feel mentally', since it is used here in a metaphorical
sense. As both KÄNNAi 'feel tactually' and 'feel mental­ KÄNNA3
ly' take ACI-Constructions in
147

Swedish, it is irrelevant to which of the two classes of KÄNNA


the main verb in (198c) is assigned. The semantic vacillation
between two classes of verbs explains the metaphorical use,
so the ACI-Complement after KÄNNA3
'feel mentally1 is therefore
well-formed. The semantic recategorization of KÄNNA^ 'feel
tactually'into KÄNNA3
'feel mentally' does not change the
syntactic rule feature [+Subj Rais] in Swedish KÄNNA. (In
contrast to this syntactic fact in Swedish, compare the corre­
sponding verbs in English FEELi tactually and FEElementally, /vfte r
FEEL3 mentally it is never possible for ACI-Constructions to occur
without an infinitive marker (to)?1). Compare the Swedish and
English surface structures in (200) where in the English
translations no infinitival form of the complement verb is
possible. Instead the passive auxiliary (BE) is deleted obliga­
torily or a different complement structure is selected, e.g.
AS IF plus a tensed sentence.

15.2 KÄNNA3 'FEELJMENTALLY'


ACI-Construction

(200) a. G EN OM S I T T OF ÖRKL ARA DE FÖRHÅLLANDE KÄNN ER HAN S I G STÂ I SK ULD


TILL D . (1257:163:2/426428) (DN)
(Through his indefinite relation he feels indebted
to D.)

b. S O M VA NLI G FA TTIG MÄNNISKA K ÄNNER H A N SI G T I L LH Ö R A KOMMUNISTERNA


(1384:232:3/126064) (GHT)
(As an ordinary poor person he feels that he belongs
to the Communists.)

A T T H A N INTE KÄND E S I G H A N ÅG ON D EL I DEM (950:107:5/422412)


(DN)
(that he did not feel as if he had a share in them.)

However, there is an Equal Subject Constraint on the Subject


Raising Rule also when the matrix verb is KÄNNA3
'feel mentally'.
(The existence of this constraint for verba dioendi and verba
putandi is discussed in Ureland 1970:231-234. All the examples
of ACI-Constructions after KÄNNA3
which are well-formed show
coreference linkage between the underlying subjects before
Subject Raising. Only KÄNNA3
plus reflexive pronouns yield well-
formed ACI-Constructions. In other words there is a semantic
constraint on syntactic well-formedness in sentences like (201)
below, since KÄNNA3
'feel mentally' refers to the Experiencer
(in the sense of Fillmore 1971) and requires of the complement
subject that it be one and the same person. This semantic
constraint on the Subject Raising Rule is violated in (201) and
ill-formed ACI-Constructions result:
148

(201) a. *genom sitt ouppklarade förhållande känner han^ hennej


stå i skuld till D.
(through his indefinite relation he^ feels her^ indebted
to D.)

b. *som vanlig fattig människa känner han^ honomj till­


höra kommunisterna.
(as an ordinary poor person he^ feels him^ to belong
to the Communists.)

c. *att han^inte kännde hennej ha någon del i dem.


(that he^ did not feel her^ to have a share in them)

The choice of complement structure after KÄNNA3


'feel mentally' is
consequently ruled by the existence of coreference linkage between
the underlying subjects. If no such deep structure linkage is
present, then the ACI-Construction is ruled out for semantic
reasons and instead either the att or the ftur-Embedding Rule is
obligatory. See (202) and (203) where the Equal Subject Constraint
blocks the application of the Subject Raising Rule and imposes
a triggering of the att or /zwr-Embedding Rule:

att-Embedding

(202) a. KLAGSTORP, R YD SG ÅR D OC H VEMMEN HÖG KÄNNER ATT MAN Ä GE R D E T TA


M USEU M ( 6 8 2 : 8 2 : 1 / 5 1 5 6 0 6 ) (SDS)
(Klagstorp, Rydsgård, and Vemmenhög feel that they
own this museum.)

b. OCH M ER OC H ME R K ÄN DE H A N , ATT D ET FANNS NÅGO T OK RISTL IGT OCH


OR ÄT T ( 1 8 9 1 : 1 0 2 : 5 / 1 2 3 1 0 2 ) ( G H T )
(and more and more he felt that there was something
unchristian and injust.)

hur-Embedding

(203) a. HA N SOM S Å FÅ K Ä NN E R HUR S TÄMN INGEN I KOMMUNEN S VÄ N G T OM


(1455:49:8/215215) (SvD)
(He,like few people, feels how the atmosphere has
changed in the community.)

b. A T T HAN Ä N D A UPP I F0LKPENSIONSÂLDERN KÄ NNE R HU R D ET ÖMMAR


EFTE R DE HÅ RD A TÖRNAR H A N FI CK ( 1 2 9 9 : 3 7 : 1 / 5 2 6 5 3 3 ) ( S D S )
(that up to his retirement age he feels how it hurts from
the hard knocks he got.)

It is certainly not accurate to claim that the choice of embedd­


ing type is optional and that there is identity in meaning
between the complement structures occurring after 'feel KÄNNA3
149,

mentally'. There are both syntactic and semantic constraints on


the selection of complement structures which the native speaker
must master in order to produce well-formed complement structures.
Such disregard of syntactic constraints on transformational rules
can be explained by the existence of two paraphrases the choice
of which is optional under the Equal Subject Constraint as in
the following sentences: 72

(204) a. A L LA . KÄN NER A T T D E . GÖR EN I N S AT S F Ö R P A T I E N T E R N A


(1679:82:2/313303) (ST)
(Everybody^ feels that he^ makes a contribution for
the benefit of the patients.)

b. alla^ känner sig^ göra en insats för patienterna.


(everybody^ feels that he^ makes a contribution for
the benefit of the patients.)

There are examples where the choice of the ACI-Construction is


doubtful, although the requirement of coreference linkage is
met in both ?(205b) and ?(206b):

(205) a. H ÅL L NI N GE N Ä R F Ö R B U L AT O V I C E T T A L I B I OC H H A N . K ÄNNER A T T H A N .
B E HÖV ER D ET (92 1: 331: 2/422507)(DN) 1

(the behavior is to Bulatovic an alibi and he- feels


that he^ needs it.)

b. ?hållningen är för Bulatovic ett alibi och han^ känner


sig^ behöva det.
(the behavior is to Bulatovic an alibi and he^ feels
that he^ needs it.)

(206) a. ME D EN S KÄNNER M AN . AT T M A N. B L I V I T A L L D E L E S KA LL
(918:140:8/415408) (DN)
(all of a sudden one- feels that one- has become quite
cold.)

b. ?med ens känner man^sig^ (ha) blivit alldeles kall


(all of a sudden one. feels that one- has become quite
cold.) 1 1

There is a Tense Constraint on the Subject Raising Rule which


blocks the generation of ?(206b). On the necessity of including
such a constraint in sentences containing SE 'see' as the matrix
verb see section 5.1.2 above.
The adjective kail can refer both to a mental and physical state
of being. Under the former interpretation känner menas 'feel
mentally', whereas under the latter interpretation it means
'feel somatically', i.e. KÄNNA3 and KÄNNA4 respectively (cf. also
(220)).
150

15.2.1 SUBJECT RAISING AND PASSIV E AUXILIARY OR COPULA DEL ETION

The most numerous complement structure after 'feel KÄNNA3


mentally' is the surface structure where no infinitival form of
a passive auxiliary (BLI(VA)) or a copula (VARA) occurs as the
complement verb. The Accusâtive-with-Past Participle (cf. (207)),
or Accusative-with-Predicative Adjective (cf. (208)), or
Accusative-with-Adverbial (cf. (209)) complements constitute
together abcut seventyeight per cent of all complements occurring
after KÄNNA 3 in the Gothenburg corpus. Since in all instances a
passive auxiliary or a copula is optional (indicated in the
excerpted examples in parentheses) the complement structures
of (207) through (209) are here considered as being generated
by the Passive Auxiliary or Copula Deletion Rules after Subject
Raising has taken place:

Accusative-with-Past Participle Construction

(207) a. OCH S Å F I N N S BE RL IN M U REN D ÄR . L I S E KÄN NER S I G H OTAD (vara


hotad) (1257:9:5/426428) (DN)
(And then there is the Berlin Wall there. Lise feels
threatened.)

b. FÖR KL AR AR A TT H A N K ÄNNER S I G D J U P T S Å RA D (vara djupt sårad)


AV EN HÅRD OM VÄRLD (39:69:6/226253) (SvD)
(explains that he feels deeply offended by a hard
world around him.)

Accusative-with-Predicative Adjective Complement

(208) a. B E L G I E R K Ö P ER U PP H EL A K US TE N O CH S P A N JO RE R NA K ÄNNER S I G
U RA RVA (vara/bli urarva) (802:170:1/225220)(SvD)
(the Belgians buy up the whole coast and the
Spandiards feel deprived of their inheritance.)

b. DE K ÄNNER S I G O L US T I G A O CH O S ÄK R A (vara olyckliga och


osäkra) (1764:2 5:6/526655) (SDS)
(they feel uncomfortable and insecure.)

Accusative-with-Adverbial Complement

(209) a. ST U D E NT E RN A S E G E N F RÄ ML I NG S LE G I ON D I T D E S O M E J K Ä NN ER
S I G H EMMA(vara hemma) P Å N AT I O N ER NA K AN T A V ÄGEN
(394:45:5/115040) (GHT)
(the students' own foreign legion where those who
don't feel at home in the student unions can go.)

b. (när) INV AN D RARN A I N T E F Ö R ST Å R V A D K A MRA TE RN A SÄ G E R K ÄN NE R


DE S I G U T ANF ÖR ( v a r a u t a n f ö r ) K R E T SE N P Å A RB E T ET
(1589:1:8/213225) (SvD)
(when the immigrants don't understand what their
colleagues say they feel excluded from them at
work.)
151

15.2.2 NOMINALIZAT I ON

In the balanced view of treating nominals as products of a


Nominalization process where underlying verbs exist, the verb
KÄNNA? 'feel mentally' has been found to occur with the follow­
ing three Nominalizations in the Gothenburg corpus:

(210) a. VARJE M ÄN N IS K A, ÅTMIN STONE NÅ G ON GÄNG, H AR KÄNT EN ÖNSKAN AT T


BLI F RI FR ÄN D E SOCIALA KO NVENTIONERNA (318:235:2/522647)
(SDS)
(at least once, every person has felt a wish to be
free from social conventions.)

b. Ä VEN I KANSLIHUSET KÄNNER MAN NATURLIGTVIS S PLITTRING EN


(183:223:4/213268) (SvD)
(even in the Chancellery one feels of course the
disunity.)

C. SV AR E T LÄ R VARA A T T H A N K Ä NT E TT V ISS T M OT STÅ ND INOM S I NA EGNA


LED (956:100:6/533624) (SDS)
(The answer is probably that he has felt a certain
resistance among his own adherents.)

It is possible to paraphrase the nominals occurring immediately


after KÄNNA3
in (210) by both att and ftur-Embeddings as (210')
demonstrates:
att
(210) a', varje människa, åtminstone någon gång , har känt
hur
hon önskar bli fri från de sociala konventionern
(at least once, every person has felt 1 £ow 1 ke wishes
to be free from social conventions.)
f a tt i
även i Kanslihuset känner man naturligtvis
är splittrad.
that
(also in the Chancellery one feels of course
how f o ne
is disunited.)
att
c1. svaret lär vara att han känt man gör motstånd
hur
inom de egna leden.
that!
(a possible answer is that he has felt they
how
create resistance among his adherents.)

By positing three different underlying syntactic structures for


the nominals after KÄNNA3 'feel mentally' along the lines of
section 13.5, the difference in semantic readings of the three
surface structures exemplified here can be accounted for. It is
c l e a r t h a t t h e d e e p m a t r i x s u b j e c t o r E x p e r i e n c e r N P o f ÖNSKA N
152

'wish' in (210a) must be the matrix subject V A R J E M ÄN NI S K A 'every


man' of the dominating sentence, because it is at the same time the
subject of BLI FRI 'be free' in the second complement sentence
of (210a). A nominalization process together with Equi-NP Dele­
tion can adequately explain the association of the Experiencer
subject VARJE MÄNNISKA with the deleted deep subject of the
nominal ÖNSKAN 'wish' and the deep subject of the second comple­
ment sentence BLI FRI FRÄN DE SOCIALA KONVENTIONERNA, which has like­
wise been deleted in the process of derivation.

Also in (210b) the nominal S P L I T T R I N G E N 'the disunity' can be


derived from the underlying verb SPLITTRA 'disunite' which occurs
in the deep structure with an empty subject NP node. Through the
Nominalization process the paraphrase relationship to (210b') is
accounted for, which has undergone the att or /zwr-Embedding Rule
with considerable semantic consequences for the interpretation
of the sentences exemplified.

In (210c') there is no direct underlying verb to M O TST ÅN D


'resistance' as there is no corresponding verb * M0TSTÅNDA 'resist'
in^ Swedish as there is in English. Even so, the nominal is
n e v e r t h e l e s s r e g a r d e d h e r e a s d e r i v e d f r o m t h e p h r a s e GÖ RA M O T­
STÅND 'create resistance'. Such a solution is not in conflict
with our principle of treating nominals as derived products
only in those cases where there are corresponding underlying
verbs, since our intuition tells us that the underlying verb of
M OT ST ÅN D i n t h e s e n s e o f ( 2 1 0 c ) i s t h e v e r b c o m p o u n d GÖR A M OT ST ÅN D
'make resistance'. It is an accidental gap in the Swedish lexicon that
there is no direct simplex underlying verb to this nominal. How­
ever, we do not accept the development of the so-called Abstract
Syntax, when linguists posit any abstract verb as underlying a
given nominal (cf. Lakoff 1965 who for semantic reasons feels
motivated in doing so).An underlying verb like * AGRESS is
certainly not syntactically motivated as underlying the nominal
AGRESSION as argued in section 13.5.1.

15.3 KÄNNA2 'SMELL'


In section 15.1 we pointed out that the Swedish standard
dictionaries do not explicitly mention the fact that the homo­
graph KÄNNA in Swedish can also mean 'smell', i.e. KÄNNA2
olfactory. Even an additional meaning of KÄNNA is neglected in
the dictionaries mentioned in section 15, that is KÄNNA in the
sense of 'taste'. These two minor meanings of KÄNNA in Swedish
are demonstrated by the following examples:

(211) O CH K ÄN NER D EN S V AGA L U K TE N A V M ASSAVE DSL AGRE N


(1179:209:3/214234) (SvD)
(and smells the odor from the stores of pulpwood.)
153

(212) OH S A NN I N G E N S K AL L F RAM S Å K Ä NDE V I I N G E T A NN AT Ä N S MAKE N A V E T T


VÄLBRYGGT ÖL (999:56:6/515621) (SDS)
(To tell the truth, we tasted nothing but well-brewed
beer)

The problem here is how to treat the two types of KÄNNA. They
could be considered as semantically empty verbs with respect
to the modality of perception, so that not until the interpreta­
tion process is almost finished are they assigned perceptual
modality if connected by either LUKT 'smell' or SMAK 'taste'.
The solution of an empty perceptual modality notation is simple and attrac­
tive.

An alternative solution would be to treat the two examples of


KÄNNA in (211) and (212) as two separate lexical entries, i.e.
as KÄNNA2a 'smell' and KÄNNA2b 'taste', The following examples
show that Swedish KÄNNA can be used without cooccurring in the
same sentence with the nominals lukt and smak:

(213) a. I Umeå luktar det Obbola. Känner du det?


(In Umeå there is an odor of Obbola. Can you smell
it?)

b. Det smakar vitlök. Känner du det?


(There is a taste of garlic. Can you taste it?)

Furthermore, the fact that both nominals in ACI-Constructions


and complementizer embeddings can occur after Swedish KÄNNA£a
'smell' motivates us to treat KÄNNA here as an olfactory verb:

ACI-Construction

(214) a. I Ume å känner man den elaka lukten från Obbola sprida
sig utefter älven.
(In Umeå one smells the terrible odor from the Obbola
factory spreading along the river.)

att-Embedding

b. I Umeå känner man att det luktar illa från Obbola.


(In Umeå one smells that it stinks from the Obbola
factory.)

hur-Embedding

c. I Umeå känner man hur det luktar illa från Obbola.


(In Umeå one smells how it stinks from the Obbola
factory.)

By positing syntactic structures as indicated in (214') the


paraphrase relationship between the object nominal of the ACI-
Construction in (214a) and the tensed verb of the complementizer-
embeddings in (214b) and (214c) can be accounted for:
154

(214) a1. Man känner [ [ [Pro luktar illa från Obbola]


NP Sl S2 S2
sprid- sig ] ]
S1 NP

b1. Man känner [ Sub [Pro luktar illa från Obbola] ]


NP [+att] S1 S^NP

c'. Man känner [ Sub [Pro luktar illa från Obbola] ]


NP [+hur] Sx SXNP

In (214a*) a Nominalization of the complement verb after Pro,


luktar 'smells' takes place, which results in lukt 'odor'; a
recategorization of the manner adverb ilia 'badly' then occurs,
so that elak is instead inserted before lukt. A deletion of the
empty subject node, det 'it', is also a result of the nominaliza­
tion process. The new nominal is then raised into the matrix
sentence, whereby the complement verb sprid- undergoes Infini-
tivization to sprida'spread'. (We disregard the problem of
describing the generation of the reflexive pronoun sig 'itself',
which probably results after an NP Copying Rule).

At the underlying structures of (214b') and (214c') the att


and hur-Complementizer Rules are triggered by the sentence type
nodes marked as T+Subl and [+Subl after the Complementizer
L+attj \+hur\
Feature Copying and Segmentalization Rules which have been de­
scribed in section 13.4. Subject Raising seems to be correlated
with the nominalization process as shown by (214a), since a
single infinitive yields an ill-formed ACI-Construction, e.g.
*Han kände hunden bajsa. 'He smelled that the dog made a Bowel
Movement'.
15.4 KÄNNA 'FEEL SOMATICALLY'
4 4

In section 15.2 KÄNNA3 'feel mentally' was shown to take Accusa-


tive-with-Past Participle, Accusative-with-Predicative Adjective,
and Accusative-with-Adverbial Constructions. The semantic border­
line between 'feel mentally' and 'feel somatically' is often
unclear. For instance in examples like (215) it is impossible
to assign the semantic perceptual class to the verb of percep­
tion, as the limited linguistic context of (215) does not give
us a clue whether to interpret KÄNNER as 'feel mentally' or as
'feel somatically':

(215) EN A V F RÅ GOR NA H AR G Ä LL T H UR F O L K K ÄNN ER S I G N ÄR G U D ST J Ä N S TE N Ä R


SLUT (1557:71:5/212215) (SvD)
(One of the questions has been how people feel when the
church service is over.)

It may be the case that people have attended a long church service
and have been tired out from sitting hours in an over-heated
church room. If a person then asks one of the church visitors
afterwards in front of the church how he feels, then it is
155

certainly a question which concerns the physical well-being


of the church visitors, that is KÄNNER is under those
circumstances KÄNNA^ 'feel somatically'.

If on the other hand, the same people are asked about how they
feel after a relatively short church service for the funeral of
a little child killed in a car accident, then it is a question
which concerns the mental state of being, that is KÄNNER is
under such circumstances KÄNNA3 'feel mentally'.

These two situations are possible and constitute an argument


for considering KÄNNA in Swedish as a semantically unspecified
lexical entry as far as the type of perception is concerned.
Only from the linguistic context or the situational context
can the mental or physical process of perception be deduced.

As support for the hypothesis of the empty perceptual process


suggested for the Swedish lexical entry KÄNNA one could also
refer to the following example where KÄNDE 'felt' acts as a
verb of both physical and mental perception:

(216) GENOM ATT VI VUXNA SJÄLVA SÂ LÄTT KÄNNER OSS HUTTRIGA OCH MOD­
STULNA EN KULEN DAG (1743:144:1/515634) (SDS)
(Because we adults ourselves so easily feel
chilly and depressed one forbidding day.)

The predicative adjectives HUTTRIGA 'chilly' and MODSTULNA


'depressed' contain semantic properties which are of consequence
for the perceptual class assignment and identification of the
preceding verb KÄNDE 'felt'. This verb form must be claimed to
obtain the semantic reading of physical (from HUTTRI GA 'chilly')
as well as mental (from MODSTULNA 'depressed') perception.

As a rule there are more clear-cut cases of physical predicate


adjectives, past participles, or prepositional phrases which
occur after Swedish KÄNNA4 'feel somatically' as in the follow­
ing examples:

Accusative-with-Predicate Adjective Construction

(217) a. HON HADE FAKTISKT KÄNT SIG (vara) LITET KRASSLIG


(1375:69:4/115054) (GHT)
(Actually she had felt a little poorly.)

b. SAGT ATT HON KÄNDE SIG (vara) TRÖTT (1375:49:4/1150511)


(GHT)
(Said that she felt tired)

Accusative-with-Past Participle Construction


(218) a. Efter tortyren kände sig den brasilianske studenten
(vara) totalt sönderslagen,
(After the torture the Brazilian student felt totally
beaten.)
156

(218) b. gangstern kände sig (vara/bli) pressad bakåt mot


ryggstödet av en snara runt halsen, när bilen star­
tade-
(the gangster felt himself being pressed backwards
against the back support by a noose around his neck,
when the car started.)

Accusative-with-Adverbial Construction

(219) a. SOM K ÄNNE R S I G (vara) YR I MÖSSA N I N F Ö R E N S Â O S TY R I G A U K T O R


(320:237:2/522656) (SDS)
(who feels dizzy in the head before such an unruly
author.)

ACI-Construction

(219) b. Han kände sig falla i vanmakt.


(He felt that he was fainting.)

c. ?Han kände sig insjukna-


(He felt that he was getting ill.)

Under the assumption of an optional or passive auxiliary (VARA


OR BLIVA) in the shallow structures of (217) through (219)
(indicated here in parentheses), Subject Raising occurs which
causes a Reflexivization of the raised coreferent pronoun. The
deleted copula structures exemplified in (217)-(219) turn out
to be the only complements which result after Subject Raising,
since no ACI-Constructions which consist of a reflexive pro­
noun and the copula VARA 'be' have been retrieved in the
Gothenburg corpus, although they are possible as indicated by
the VARA within parentheses in(217)-(219). In the Gothenburg
corpus the copula deleted structures constitute sixty
per cent of all embeddings occurring after 'feel KÀNNA4
somatically1 (cf. Table 17).

Complementizer embeddings also occur as verbal complements after


KÄNNA4 'feel somatically'

att-Embedding

(220) MED E N S KÄN NER MAN A T T M AN B L I V I T A L L D E L E S K AL L


(918:140:8/415408) (DN)
(All of a sudden one feels that one has become completely
cold-)

hur-Emb e dding

(221) J A G K ÄN DE I N T E N S I V T H U R K VAVT R UMMET V AR


(256:130:3/426471) (DN)
(I felt intensively how close it was in the room.)

There is an Equal Subject Constraint on the occurrence of ACI-


Construction after KÄNNA4 'feel somatically', if the complement
subject is not coreferent with the matrix subject. Subject
157

Raising is then blocked as in the following sentences where the


constraint has been violated:

(222) a. *Hon kände mannen vara alldeles kall.


(She felt the man to be completely cold-)

b. *Hon kände rummet vara alldeles kallt.


(She felt the room to be completely cold.)

Instead, a different verbal complement structure is required


to avoid violating the Equal Subject Constraint as demonstrated
by the ill-formed sentences *(222a) and *(222b), that is att-
Embedding or ftur-Embedding must result (cf. also section 15.2
on the Equal Subject Constraint on Subject Raising after KÄNNA,
'feel mentally'):

(222) a'. Hon kände att mannen var alldeles kall.


(She felt that the man was completely cold.)

b. Hon kände hur rummet var alldeles kallt.


(She felt how the room was completely cold')

Sentence (220) above can undergo Subject Raising and yield a


well-formed ACI-Construction, since the matrix and complement
subjects are coreferential, i.e. both are MAN 'on e':

(220') med ens känner man sig ha blivit alldeles kall


(All of a sudden one feels oneself to have become
completely cold)

Nominalizations

To claim a derivational process for all the three nominals which


occur in (223) after KÄNNA^ 'feel somatically' does not seem
adequate, even though underlying syntactic structures could be
posited to contain deep verbs in order to explain the transforma­
tional derivation of sentences like the first two of the follow­
ing sentences:

(223) a. K ÄN N ER MAN E N K A R A K T E R I S T I S K R Ö R EL SE H OS F A RTY G E T


(23:81:1/125052) (GHT)
(if one feels a characteristic movement of the ship.)

b. P Å G R UND A V A T T D E K Ä NNER EN MÄNG D O LI K A L US TF Ö R N I MM ELS ER


(1800:61:1/522567) (SDS)
(because they perceive many different desires.)

c. A T T D E K ÄN NER L U S T OC H S M ÄR T A (486:63:7/221170) (SvD)


(that they feel pleasure and pain.)
158

It is clear that in (223a) it is the ship which moves and in


(223b) it is the same Experiencer subject for the matrix as for
the underlying complement sentence, provided the nominal FÖRNIM­
MELSE 'perception' is treated as a transformational product.

Such a transformational treatment of the nominals in (223a) and


(223b) can be motivated by the existence of paraphrases like the
following sentences, where instead of a Nominalization the att­
or ftwr-Embedding Rule has applied:
J hur
(223) a', känner man > att fartyget rör sig karakteristiskt,

(one feels fct the ship moves characteristically.)

på grund av att de känner förnimmer olika

lust i en mängd.

(because they feel J Î??W.A they perceive many different


desires.) [ tnatj

For reasons discussed in section 13.5.1 one can treat the nominals
RÖRELSE 'movement' in (223a) and FÖRNIMMELSE 'perception' in (223b)
as being transformationally derived, because there are under­
l y i n g v e r b s w h i c h c a n b e p o s i t e d i n t h e d e e p s t r u c t u r e , i . e . R ÖRA
'move' and FÖRNIMMA 'perceive', respectively.

However, such a claim seems unmotivated in the balanced view of


describing nouns such as LUST 'd esire', since no underlying deep
verb exists for this noun, there is no*LUSTA. There is an under­
lying verb for SMÄRTA 'pain', i.e. as in DET SMÄRTAR 'it hurts',
but the constellation with LUST 'desire' which cannot be trans­
formationally derived, stops us from deriving also SMÄRTA in
(223c) transformationally. (For further discussion on the
balanced view for describing nominals in Swedish see also sections
13.5.1 and 15.2.2).

15,5 FÂ KÄNNA 'FEEL AND SMELL SUDDENLY AND UNEXPECTEDLY'


The auxiliary FÄi only occurs as an aspecutal perfective morpheme
before KÄNNAi 'feel tactually' and KÄNNA2 'smell an odor' as
the following examples demonstrate:

FÅ KÄNNA-^ 'feel suddenly tactually'

(224) När kan gick på grusgången fick han (plötsligt) känna


gruset knastra under fotsulorna.
(As he walked on the gravel road he (suddenly) felt the
gravel crunching under the soles of his feet)
159

FÅ KÄNNA? 'smell the odor suddenly'

(225) När hon kom till Ålidhem fick hon (plötsligt) känna
lukten av Obbola sprida sig.
(When she came to Ålidhem she suddenly smelled the bad
odor of Obbola spreading.)

FÅ KÄNNA^ 'feel mentally and suddenly'

(226) *som vanlig fattig människa fick han känna sig tillhöra
kommunisterna.
(as an ordinary person he felt that he belonged to the
Communists.)

The occurrence of fick is ill-formed in the sense of 'suddenly


feel'. In the sense of 'have to' it may be well-formed, however.
The same refers to FÄ KÄNNA^ in the following sentence:

FÅ KÄNNA^ 'feel somatically'

(227) *De fick känna sig olustiga och osäkra.


(They suddenly felt unhappy and insecure.)
160

16, RESULTS OF THE LINGUISTIC PERFORMANCE STUDY ON VE RBAL


COMPLEMENTS OCCURRING AFTER KANNA

Of the total sum of 510 examples of the homograph KÄNNA in the


Gothenburg corpus, there were 116 cases of verbal embeddings.
The distribution of embedding types after the four classes of
KÄNNA discussed in section 15 is illustrated by the following
tables:

16.1 KfiNNA x ' FEELj TACTUALLY'

Table 1 k THE PRODUCTIVITY OF EMBEDDING R ULES AFTER K SNNAj


1 f e e 1 t a c t ua Ì 1 y 1

Type of Embedding Number Distribution of Embedding


Types after the Verb KÄNNA^
in the Gothenburg Corpus
1) ACI-Constructions 3 66.7 %
2) Nominalizations 1 33.3 %
3) att-Embeddings 0 -

4) ftur-Embeddings 0 -

5) Accusative-with-Predicate
Adjective Constructions 0 -

(VARA Deletion)
6) Accusative-with-Past
Participle Constructions 0 -
[VARA or BLIVA Deletion)
7) Accusative-with-Adverbial 0 -

(VARA Deletion)Construction s

Total number of embeddings: 4


Total number of examples of
the homograph KÄNNA in the
Gothenburg corpus: 510
161

16.2 KÄNNA2 'SMELL/


Table 15 THE PRODUCTIVITY OF EM BEDDING RULES AF TER KSNNA2
'sme111

Type of E m b e d d in g Number D i s t r i b u t i o n o f E m b e d d in g
Types after the Verb KÄNNA^
i n t h e G ot h e n b u r g C o r p u s

1) Nominalizations 1 100 %
2) att-Embeddings 0 -

3) /zur-Embeddings 0 -

4) Accusative-with-Predicate
Adjective Constructions 0 -
(VARA Deletion)
5) Accusative-with-Past
Participle Co nstructions 0 -
(VARA or BLIVA Deletion)
6) Accusative-with-Adverbial 0
Constructions(VARA De le­
tion)
7) ACI-Constructions
1 a -

Total nu mber of embeddings: 1


Total number of examples of
the homograph KÄNNA in the
Gothenburg corpus: 510

16,3 KÄNNA 3 'FEEL3 MENTALLY'


Table 16 THE PRODUCTIVITY OF EMBEDDING RULES AF TER KÄNNA^
1 f e e 1 men t a l 1 y 1

Type of Embedding Number D i s t r i b u t i o n o f E m b e dd i n g


Types after the Verb KÄNNA^
i n t h e Go t h e n b u r g C o r p u s
1) Accusative-with-Predicate
Adjective Con structions 38 37.6 %
(VARA D eletion)
2) Accusative-with-Past
Participle Co nstructions 25 24.8 %
(VARA or BLIVA Deletion)
3) Accusative-with-Adverbial
Constructions 12 11.9 %
(VARA Deletion)
4) att-Embeddings 11 10.9 %
5) Nominalizations 7 6.9 %
6) Tzur-Embeddings 4 3.9 %
7) ACI-Constructions 4 3.9 %

Total number of embeddings: 101


Total number of examples of
the homograph KÄNNA in the
Gothenburg corpus: 510
162

16,4 KÄNNAS 'feel4somatically'


Table 17 THE PRODUCTIVITY OF EMBEDDING R ULES AFTER KÄNNA^
1 f e e 1 s o m a t i c a 11 y 1

Type of Embedding Number Distribution of Embedding


Types after the Verb KÄNNA^
in the Gothenburg Corpus
1) Accusative-with-Predicate
Adjective Constructions 5 50 %
(VARA Deletion)
3) Accusative-with-Adverbial
Constructions 1 10 %
(VARA Deletion)
2) Nominalizations 2 20 %
4) att-Embeddings 1 10 %
5) 7zM2»-Embeddings 1 10 %
6) Accusative-with-Past
Participle Constructions 0 -

7) ACI-Constructions 0 -

Total number of embeddings: 10


Total number of examples of
the homograph KÄNNA in the
Gothenburg corpus: 510

As can be seen from the total number of embeddings in Tables 14


through 17, the Subject Raising Rule is a very important rule
for generating verbal complement structures after the four
classes of KÄNNA. Of the 116 embeddings retrieved in the Gothen­
burg newspaper corpus 88 embeddings are the result of Subject
Raising under the assumption that besides ACI-Constructions
(7 examples) also Accusative-with-Predicate Adjective (43
examples), Accusative-with Past Participle (25 examples), and
Accusative-with-Adverbial (13 examples) Constructions are
derived by the Subject Raising Rule via a Shallow Structure
which contains either the copula VARA or the passive auxiliary
BLIVA,.that is three fourths (75.8 %) of all embeddings occur­
ring after the homograph KÄNNA in the Gothenburg corpus are
carried out by the Subject Raising Rule.

The embedding carried out by means of the complementizers att


'that' and hur 'how' are numerically fewer, twelve and five
respectively, which yields a percentage of 10.3 % for att-
Embeddings and 4.3 % for /zwr-Embeddings.

Nominalizations are also marginal with the total of eleven


embeddings which constitutes 9.4 % of all embeddings.
163

However, the numerical distribution of embedding types given in


Tables 14 through 17 does not describe adequately the syntactic
status of the less frequent embedding types. A simple enumera­
tion of the percentage of embedding types does not give an
explanation of the reasons for the distribution of complementa­
tion structures. The att and Tzur-Embedding Rules constitute
indispensible complementation rules in instances where no core-
ference linkage exists between the matrix and complement subjects.
ACI-Constructions or copula-deleted constructions are then blocked
by the Equi Subject Constraint on Subject Raising discussed in
section 15.4.

A description of the verbal sentential complements occurring


after KÄNNA which does not include the Equi Subject Constraint
on Subject Raising does not meet the requirement of descriptive
adequacy, since such description would fail to describe the
important syntactic status of the att and ftur-Embeddings in
their relation to the ACI-Construction and the copula or BLIVA
deleted structures (cf. also the same relation of sentential
complements occurring after SE2 'realize' in section 5.1.2 and
HÖRA2 'hear of, be told' in section 13.4).
164

CONCLUSION OF THE PERFORMANCE STUDY


1, THE EVIDENCE OF TRANSFORMATIONAL RULE PRODUCTIVITY IN THE
GOTHENBURG CORPUS
An important product of using a comparatively large corpus of
written Standard Swedish is the specification of transformational
rule productivity and the distribution of embedding types with­
in a given corpus.

These results which specify the productivity of embedding rules


have been established through examining the excerpts from the
Gothenburg newspaper corpus. Of course our results on trans­
formational rule productivity refer only to this corpus.

The productivity of a number of syntactic embedding rules have


in this way been described concomitant with the intuition-
based description of the operation of such rules. Also in cases
where no actual examples of a given rule of embedding existed
in the corpus the embedding rule has been discussed.

We have been primarily involved with the following transforma­


tional rules: the Subject Raising Rule which generates ACI-
Constructions and Accusative-with-Past Participle Constructions,
the att and /iwr-Complementizer Embedding Rules, the Passive Rule,
the Copula or Passive Auxiliary Deletion Rule, the Nominaliza-
tion Rule, and the Pronominalization Rule.

By including aspects of the productivity of embedding rules in


a given corpus, we obtain an additional dimension in the de­
scription of verbal complementation, i.e. that of distribution
between the types of embedding. (See the next paragraph on the
reliability and validity of such rule distribution).

In the theory of generative grammar linguists have often claimed


that native speakers have intuitive notions about syntactic
structures (cf. for instance Chomsky 1965:21). Obviously such
a self-evident claim cannot be completely contradicted. There
is in the present investigation, however, evidence that an
exclusive use of the intuition-based knowledge of a native
speaker does not necessarily include all aspects of an analysis
which are gained through a linguistic investigation of the
present format. It can hardly be claimed that the native
speaker's knowledge about the productivity of certain trans­
formational rules is a part of a general or subconscious
linguistic intuition, because such insight does not exist for
the native speaker.

On examining for instance the statistical distribution of the


ACI-Construction in its relation to other types of embeddings
after Swedish verba sentiendi as occurring in the Gothenburg
corpus, it was found that the linguist's intuition about the
frequency of occurrence of this surface structure does not at
all correspond to the facts of distribution as presented in
for instance Table 1, section 6.1.
165

Judging from the results stated in this table and those of the
succeeding tables of Part One, we have reason to doubt whether
there is anything like a refined native speaker's intuition about
the productivity of a given transformational embedding rule in
its relation to that of other embedding rules.

This claim will be supported by the following three concrete


examples: i) If the ACI-Constructions and Accusative-with-Past
Participle Constructions are treated as being generated by the
Subject Raising Rule in sentences which contain SE, 'see visually'
as the matrix verb of perception, then the att-Emoedding Rule
can be claimed to play a numerically smaller role. After SE^
'see visually' the att-Embeddings constitute only 13.8 per cent
of all the embeddings found after SEi in the Gothenburg corpus,
whereas ACI-Constructions anu Accusative-with-Past Participle
Constructions together amount to slightly more than half of all
embeddings after SEi, i.e. 50.4 %.

We could not predict such a productive dominance of the Subject


Raising Rule over the complementizer embedding rules, the att
and /zur-Embedding Rules, at the outset of this investigation.

ii) The second example concerns the embedding structures after


l
HÖRA- 'hear auditorily' which in the Gothenburg corpus also
takes half of all occurring embeddings as ACI-Constructions,
i.e. 52 % as indicated in Table 10, section 14.1. If passivized
infinitives of the type TALAS OM 'spoken of' and RYKTAS 'rumored'
are treated not as passivized idioms (as in section 13.3),
but as derived structures,an even higher percentage is attained
for the Subject Raising Rule, i.e. 81.2 %. The att and hur-
Embedding Rules are here extremely marginal, since they carry
out only 4.2 % of all embeddings after HÖRA^ 'hear auditorily'.

iii) The third piece of evidence that the native speaker has
no intuitive knowledge about the productivity of embedding
rules is the fact that also after the homograph KÄNNA 'feel'
the Subject Raising Rule is the most frequently applied rule
of embedding, whereas the complementizer embedding rules
constitute less productive embedding rules. Tables 14 through
17 in section 16 indicate that the complement structures which
result after Subject Raising plus the additional rules of Copula
and Passive Auxiliary Deletion out-number by far the complement
structures resulting after the att and ftur-Embedding Rules. Of
the total 116 embeddings after the homograph KÄNNA, 88 examples
contain embeddings carried out by the Subject Raising Rule,
that is ACI-Constructions and Copula Deleted Structures like
Accusative-with-Past Participle Constructions, which yields
a percentage of more than 75.3 % for the Subject Raising Rule
as the rule of embedding in sentences containing KÄNNA. The
att and /zwr-Embedding Rules again play a numerically marginal
role as embedding rules, with their total of 17 embeddings,
which is 14.6 % of all embeddings after KÄNNA.

Irrespective of the lower frequency of the application of the


att-Embedding Rule as evidenced in the Gothenburg corpus, the
syntactic status of the att-Embedding Rule must not be under­
estimated. It is an important embedding rule when Subject
Raising is blocked by for instance the semantic or tense
166

constraints as demonstrated for SE-, 'see visually' in section


5.1.2 by *(56b), *(57b), and *(68bj. After cognitive verbs
like SE2 'realize' and HÖRA2 'hear of, be told' there is also
a semantic constraint which blocks the Subject Raising Rule
from applying as is demonstrated by *(59b) and*(60b) in section
5.1.2 on the one hand, and by *(167b) and *(168b) in section
13.4 on the other. The latter two verbs must consequently be
marked as not allowing Subject Raising, since embedding is
possible only in these sentences by means of the att and hur-
Embedding Rules.

Therefore, a simple enumeration of the embedding types after


the perceptual verbs in question does not give an explanatory
description of the complement structures of Swedish. It is
necessary to state the Equal Subject Constraint on the genera­
tion of ACI-Constructions and Copula Deleted Structures like
Accusative-with-Past Participle Constructions in a descrip­
t
tion of SE 'see visually', SE2 'realize', HÖRA^ 'hear audi­
torily' HÖRA2 'hear of', and KÄNNA? 'feel mentally'.
(Cf. sections 5.1.2, 13.3, and 15.2).

In order to obtain a general survey of the productivity of the


embedding rules as evidenced in the Gothenburg corpus and
treated in the present investigation the following two figures
summarize the distributional facts of Tables 1 through 17:
167

.o O
a: O
r—)

to to to
:o to to to
K to
to to
<
pH

w
CO

CO VU
3<
o
CSI (NI
o O
vO
q_
OC
o <
o
CD
2
2
O
LO
o
rH
o O
rH rH (NIO
q: :<
«
zd to CT>
cq
z
lu <
en

00

vO


rH
en en
• •
en
x 2 to
(NI rH o to vO
h- S
o
to rH
cd cm
LU < o
m
1- 2
o
2 rH
z: :<m
<c •—•
rH to
lo lu CO < vO to
o_ 2
LcüjJ >- 2 vO to
z=> h" :< «
cd
H—« CD CNJ
ll_ z 3 ex O
Q :o O
Q x rH
llj (NI CNJ vO
CQ rH
21
lu â CNJm (NIen rH
u_ :o
o X, to r-.
2 CN) O to
o W
•—I co
h-
(NI
o
LO CM vO
13 en LO 00 CO en oo to
CQ rH
t—I w
ca co to
LO
r—1
o t* o o
rH to
to
h-
co
Q o
e
tìo 1
X! <D
i
X! -f-»
, ,
_j •H •H P rH P P P Ö
< P XJ •H fX •H < •H •H bO t>0 o
3 CD O
2 3
t—1 £ -H Ö £
1 L> O 1 <1> O
Ö £ Ö
1 O
£
1
co C
•H
Ö
•H
Ö
o 2
•H
P
1—
o CO •M
Jh X) CD >H -H <D > -H
(D > P P > -H P
<D -H
> rH P > <L)
Q) •H
P
w
tN) -a
x> nd HH h
•H
P
O cd
(SI <D
<
ULO Dw h—1
<D C/> •H •H Jh O
M •H -M U
P cd 3 P Cd 3
•H cd O •H > U
P -H 3
(-H
XIe<L>
a> <D
rH
H
<
•H J
rH 3
c*
« <LJ
C
o > cd eu f -t Cd
•H 7) P </> -H -P
(J Jh «CD ,Û(h Jh J
P -H 3
P
cd t
V) tìP
H
2 CD XI
W I 1 I
2 W

W Q
<U
C tSJ
HH
Cd P
o
CD< 2
cd H HH1 V) 3 P w 313 (fl
7) U <n p! O 0) Ö
3 <D 71
O > Ö
3 37)
h
U J G
SQ
W P U P
•H D

<O DJhh UO
u C B U
cd U es O O TJ O O V O Q P 3 P 2 OP
1- 1—1
2Q
w U
»"5 <
PQ PL, < P« u < < U < CD CJ eu
PQW cd I—I O C
4S cd
S
2
S
lu Q H /—\
CO f \ /—\ t —%
o X >—\ O
O W
QU PQ
rH (SI to
,—X /—\ /—\
LO
/•—\
vO u W rH CSI to
<•—\
2 rH
a. SW
lu «; PQ ,~s
CJ
168

•—« CSI

LO LO

ct> c n

c
o
Ë
O
1
CD
1
rC T-»
, 43
p
•H •H P rH •P TJ P G
P Ti •H fit •H •< •H •H bû bO O
CJ3
2
O
3
I—1 £ *H
1 U O
tì £
1 o o
c £
i
ö
o
£
1
Ö
O OS
e
•H
C
•H 2
•H
P
i—i J-t Ti <D -H •H <D
> -H <D •H O •H w TJ T3 O cd
CO +-> <D > P •P >-H P > rH +-> > <D P Kl Ti T3 t—t M <D
tn M •H JH O •HP O •H cd O •H > O i—i <D <D H •H *H
<; tì•H P Cd 3 P (U 3 P *H 3 4-> -H E- JQ < rH 3
oc o > Cd Oh u cö
U fn cd cd 2 <-3 S B KJ cd P
u •H (/> P V) >H -P V) ÎH P 10 Ö P W 2 W W 1—1 Ö O
H 1 { n 3 P (/) 3T3 V) 3 <D t/> 3 3 w S HH 1 1 -Î •H 3
u hH m ü to C O <L> Ö O > C U ^ C W Q
Q
<P|
<P M
3
< e ^
w u cd O CÖ o O JH O O T J O O <D O 2 O P
< O-i < cu o < a*
u < C u < C3 u CU W cd| jd i—i 2 CO
PQ 2 PQ S
D /—\ f—\ ,—\ ,—\ /—\ /—\ O S /—\ /—\ O •—\
C/D rH CSI K) *3- LO vO u W rH CSI 2 rH

/•—\ /—\
PQ CJ
169

Because there are so few examples of FÄ SE^ 'catch sight of' in


the Gothenburg corpus, no reliable conclusions can be drawn
about the distribution of sentential complements after this verb
compound. However, /zwr-Embeddings (60$), att-Embeddings (20$)
as well as ACI-Constructions (20$) have been retrieved as
sentential complements after FÄ SE^, whereas no copula or BLIVA
deleted structures or Nominalizations have been found after FÅ
SEi as Figure 6A demonstrates (cf. also Table 3, section 6.4).
An Accusative-with-Past Participle Construction is also possible
after FÅ SEi as illustrated in section 5.5.3, although such a
surface structure is not frequently exemplified after FÄ SEi in
the Gothenburg corpus.

After FÅ HÖRAi 'suddenly hear' the first third of embeddings


consists of ACI-Constructions, the second third of passivized
idioms (TALAS OM 'spoken of'), and the last third of att-Embed-
dings. It is doubtful whether Accusative-with-Past Participle Con­
structions are well-formed after this verb compound as demonstrat­
ed by ? (190) in section 13.6.2.

The cognitive verb FÅ HÖRA2 'hear of, be told' cooccurs only with
att-Embeddings in the Gothenburg corpus (100$), although examples
of at least ftur-Embeddings are potential embeddings as shown in
section 13.6.3 by (193). On the otherhand, the Subject Raising
Rule cannot be applied to complement sentences after this
cognitive verb, since ACI-Constructions and Accusative-with-Past
Participle Constructions are ill-formed as sentential complements
(cf. *(192) and *(194)).

The remainder of the Swedish perceptual verbs, i.e. the six verbs
of Figure 6B have been found in the Gothenburg corpus to take
only complementizer embeddings and nominalization structures as
sentential complements. The blank space under the six verbs of
perception in Figure 6B clearly indicates that examples of
embedding by the Subject Raising Rule are nonexistent in the
Gothenburg corpus. The blank space here is not accidental, since
it reflects the low actual frequency with which Subject Raising
carries out embeddings in Swedish sentences which contain these
six verbs of perception.

The instantaneous verb of perception MÄRKA 'notice' takes mainly


att-Embeddings (59.5$), but also /zwr-Embeddings (9.5$) and
Nominalizations (30.9$) occur. The Subject Raising Rule seems
to be an infrequent rule of embedding also for MÄRKA, although
both ACI-Constructions (cf. (112)), Accusative-with-Past
Participle Constructions (cf. (113a)), and Accusative-with-
Predicate Constructions (cf. (113b)) are potential sentential
complements under coreference linkage between the matrix and
complement subjects (see *(112') and (112) in section 7.1).

A semantic distinction between the verbs UPPTÄCKA^ 'discover


visually', UPPTÄCKA2 'discover cognitively', and UPPTÄCKA3
'make a discovery' has been made in section 9. The productivity
of the att and fcur-Embedding Rules is overwhelming: of all
embeddings occurring after the homograph UPPTÄCKA 85.4 $ are
carried out by the complementizer rules. No ACI-Constructions
or copula deleted structures have been retrieved in the Gothen­
burg corpus, although under the given circumstances, ACI-
170

Constructions (cf. (119a)-(119c)), and Accusative-with-Past


Participle Constructions (cf. (120a) and (120b)) are possible
sentential complements if the Existential Verb Constraint on
Subject Raising is not violated (cf. (120a') and (120b1)).

In the Gothenburg corpus the durative verb of active visual


perception, OBSERVERA^ 'watch' has been found to take only att-
Embeddings (100%) as sentential complements (cf. section 11.1).
It is doubful whether any other embedding is appropriate here.
ACI-Constructions are potentially possible under Aktionsart
agreement between the matrix and complement verbs (cf. Part Two
section 2 ), but are hardly ever used in Swedish (cf. ?(131a)
and ?(131b)).

After OBSERVERA^ 'become aware of' both att-Embeddings (501)


and Nominalizations (50%) occur as sentential complements. ACI-
Constructions are here ruled out for semantic reasons (cf. also
SE2 'realize' in section 5.1.1 and HÖRA2 'be told, hear of' in
section 13.4). Even though no Tzwr-Embeddings were found in the
Gothenburg corpus, their well-formedness is evidenced by (137).
Also on this point the Gothenburg corpus is too small to make
further generalizations about the statistical distribution of
sentential complements after OBSERVERA^ 'become aware of'.

2, THE PROBLEM OF LEXICAL AND SYNTACTIC REPRESENTATIVITY


The 1387 newspaper articles which constitute the raw data basis
of the Gothenburg corpus have been written by 569 different
persons. The articles have been excerpted from the following
five Swedish newspapers which appear in the three major cities
of Central and South Sweden: Göteborgs Handels- och Sjöfarts-
Tidning (GHT)j Svenska Dagbladet (SvD), Stockholms Tidningen
(ST), Dagens Nyheter (DN), and Sydsvenska Dagbladet Snäll-posten
(SDS).

The 1387 artides which include more than one million running
words correspond roughly to 3000-4000 normal printed pages or
to th e vocabulary of 15 to 20 novels of average size. The body
of text of one million running words which is the total size
of the Gothenburg corpus can be said to be relatively small,
too small to describe the total lexical and syntactic sompetence
of a native speaker of Swedish.

Karlgren 1972 has recently criticized the Gothenburg corpus as


being insufficent for describing lexical competence. Obviously,
our linguistic intuition about both the lexical and syntactic
structures of Swedish is based on observations of linguistic
material which surpasses the size of the Gothenburg corpus many
times over (cf. p. 12). Such criticism is justified if it refers
to a one-sided, corpus-oriented approach to describing native
speaker's linguistic competence and thus his total lexical and
syntactic knowledge of his language.
171

Allén 1973 has countered Karlgren's 1972 criticism by claiming


that the Gothenburg corpus gives an adequate "approximation of
the general numerical relationships of lexical entries" within
the Swedish lexicon as represented by a newspaper text and an
adequate picture of the "linguistic structure of the basic
vocabulary" (cf. p. 5). Furthermore, according to Allén,
reliable lexical and morphological data on written Modern
Swedish are obtainable from the Gothenburg corpus especially
regarding the frequency distribution of words, the length of
words, the number of sublemma forms, word classes, and grammati­
cal categories,(i.e. comparative classes of adjectives, de­
clensional classes of nouns and conjugational classes of verbs
(cf. p. 6)). Finally, Allén indicates that newspaper texts
constitute more than 75 per cent of all the three billion words
printed in the Swedish language in both Sweden and Finland,
which supports the contention that newspaper texts provide a
representative corpus.

The limitation of a corpus-oriented approach illuminated by


Karlgren concerning the lexical items does not mean that all
results which are gained through a study of for instance the
syntactic rule application for embedding within a given text
are uninteresting. Quite the contrary, pertinent observations
on the productivity of embedding rules can be made on the Swedish
sentential complement system by studying a given corpus which
is representative of a certain style of writing. Since the
Gothenburg corpus is so far the largest corpus of written
Standard Swedish presently available, it has been used as a point
of departure for describing the native speaker's syntactic
competence to embed one sentence into another.

There is of course the great problem of representativity. To


what extent is the Gothenburg corpus representative as a de­
scriptive point of departure for describing syntactic competence?
Does the Gothenburg corpus reflect adequately the potential
occurrence of sentential complements after the given set of
perceptual verbs in Swedish? What criteria are to be applied
in deciding what makes a given corpus representative and to
what extent can it tell us about the native speaker's know­
ledge of a set of syntactic structures and rules? Does the
corpus contain more syntactic structures than the native speaker
knows actively? Should one not make a distinction between the
native speaker's active and passive knowledge of syntactic
structures? Such a distinction is never made in generative
literature, although it is certainly necessary that it be made.
It is a well-known fact that a native speaker has a greater
passive knowledge of syntactic structures and rules than he has
an active command of these structures and rules. So far no one
has given any useful definition of what representativity means
when applied to the syntactic facts of a linguistic corpus.
Chomsky 1957:51 has some superficial remarks about the relation­
ship between a given corpus and the construction of the best
grammar, which Chomsky does not characterize as being an active
or passive part of the native speaker's linguistic competence
but rather treats it as a single concept.
172

The present investigation is a contribution to the description


of what a native speaker knows about the embedding mechanism
and what evidence of this mechanism is offered us by studying
a given corpus. We are involved only with written language
here and will not discuss the embedding mechanism in spoken
language. We believe, however, that several surface structure
embeddings are part of written language rather than spoken
language. The need for distinguishing between different levels
of style has been stressed in Part One and will be stressed even
more in Part Two. The distinction between written and spoken
language is perhaps the most important.

3, ACTUAL AND POTENTIAL DISTRIBUTION OF EMBEDDING TYPES AFTER


VERBA SENTIENDI
In the present investigation of the Swedish verb complement
system a corpus is considered representative if it contains two
thirds or more of the potential syntactic structures under study.
That means two thirds of all possible embedding types must occur
after each of the seventeen verbs of perception investigated here

Assuming that the percentage of two thirds is fair enough to be


used as the requirement for determining that a corpus is
representative of the type of written language from which it is
taken, but not to reflect the native speaker's linguistic know­
ledge, then the missing third of the examples of surface struc­
tures must be supplemented in the competence part of the study by
linguistic intuition.

In order to provide an illustrative example of the gap between


actual and potential verb complementation after the seventeen
Swedish verba sentiendi, Figure 7 below is presented. The encir­
cled plusses denote actual embedding types retrieved from the
Gothenburg corpus, whereas the unmarked plusses denote potential
but not exemplified embedding types from the same corpus.
173

:o
as
<
P-.
©
:0
K
<
©© ©
PH

<Uh © 1 0 + ©
2
:<
«
+ 1 + © © + ©© ©
<
2
«
© 1 © © © ©© ©
<
2
2 ©
<
2
2 © 0

:o
a:
©• 1

:0
K
©© 1 •© ©
W
C/D
1 1 © ©© ©
W
CO © 1 © © ©© ©
I
rfi <D
P Hr i
•H ÇU <C 00 ö0
> *H fi
I O O 0)
fi £I I o
fi fi
•H •H
0 «H 'H >• 0 *H 0 »H ti tj cd
>Pp •H P > rH P > 0 P 1J T* N0
CQ W •H fn O P >L •H Cd U •H > O 0 0 H i-«
•—« -1 P ÖC 3 Cd 3 P «H 3 P irl 3 rH 3
o:z £3 CTJ CX, î-i o u cd ,û U CÖT3 ^ cd P
I- o Pä
CO »-•
in p
3 P t/)
•H P tfi *-i P l/> fi P
•"Ö < /) 3 0 c/> 3 3 t/>
S Ü
pq W
I
fi u
H3
— 1- U
Q DI 2
O c/) fi
Urto
0 fi O > fi UJ-.fi
* O uu o O 0 O
M
3
S >H
OP
* CQ h—1
_J —• Q
<O U h U < < u <u u ,fil 2 OC
< CC Q o^
3 -h W 2 Hr
1- CO PQ
O — S
< Qw
174

w
£
m
m
© ©
pq
o

w
>
o:
w
co
©
PQ
O

U
:<
H
Cl
,
© ©
O.
S3

CJ
:<
H

©© ©
Cli
CD
r—I
<
«
u
:<
H
a.
© ©
a.
D

5 S
©0 ©

42 <L> X! r-ì
P rH P T3 c
•H a. •H < bO b O o
£ -H Ö £ Ö Ö Ö '•H
p
I o o I <D O I
<D -H «H Q> > «H Ü -H <D »H -ö cd
> p +J > -H P > H P > <D P TJ m a>
CQ pq •H îh U •H P U •rl Oj U •H > U •H ^
— -J •P Cd 3 P dc 3 P -H 3 P »H 3 1-» 3
ÛCZ tD Cd ÇL, U Cd Ü U Cd X) M Cd tJ U e s cd P
h- O C6 V) -P tO -H P Ul 4-J 10 ö P w w ä o
tO — 3 P m ST* V) 3 (1) <0 3 3to i •H 3
-- I->= 2
Q U o to tì U <D Ö
Urto U îh O
Ü > tì O M Ö
UTS O O 0 O 3
H C e*
2OP
CQ M < Pu u <C Û4 u < < U <uu t— i 2 O C
-I —- Q S
< OC Q O'
3i- m z »
I— (/> PQ
a— s
< QW
175

As indicated at the bottom of Figures 7A and 7B the following


perceptual verbs cooccur in the Gothenburg corpus with more
than two thirds of the potential embedding types: SE^ 'see
visually1 actually occurs with six of the eight potential
embedding types (751), SE7 'realize' with four of the four poten­
tial embedding types (100%), HÖRAi 'hear auditorily' with four
of the six potential embedding types (66.71), KÄNNA3 'feel mentally'
with seven of the eight possible embedding types (87.5%), KÄNNA4
'feel somatically' with five of the eight possible embedding
types (62.5%), UPPTÄCKA2 'discover cognitively' with all three
of the three potential embedding types (100%), and UPPTÄCKA3
'make a discovery' with two actual of the three potential embed­
ding types (66.7%).

In summary, out of 40 potential embedding types after these


seven verbs of perception, 31 actual embedding types have been
retrieved in the Gothenburg corpus, i.e. 77.5%. In the light of
these distributional facts one can certainly claim the Gothen­
burg corpus to be representative as a data basis at least as
far as these seven verbs of perception are concerned. They meet
the requirement of being exemplified by two thirds or even three
fourths of all the potential embedding types.

The remaining set of perceptual verbs enumerated in Figures 7A


and 7B does not show such a high correlation between actual and
potential embedding types since the average number of examples
is less than fifty per cent, i.e. 42.5% of the potential embed­
ding types are represented by actual embedding types in the
Gothenburg corpus after these verbs.

The following perceptual verbs cooccur in the Gothenburg corpus


with less than two thirds of the potential embedding types:
HÖRA2 'hear of' with one of two potential embedding types (50%),
KÄNNA^ 'feel tactually' with two of five potential embedding
types (40%), KÄNNA2 'smell' with one of four potential embedding
types (25%), FÅ SEi 'catch sight of1 with three of eight potential
embedding types (37.5%), FÅ HÖRA2 'hear of' with one of two
potential embedding types (50%), MÄRKA 'notice' with three of
seven embedding types (42.9%), UPPTÄCKAi 'discover visually' with
two of eight potential embedding types (25%), OBSERVERAI 'watch1
with one of two potential embedding types (50%), OBSERVERA2
'become aware of' with two of three potential embedding types
(66.7%), and finally FÄ HÖRA^ 'hear suddenly' with three of the
six potential embedding types (50%).
As far as the latter set of perceptual verbs are concerned the
Gothenburg corpus is, according to the two-thirds coverage re­
quirement, not representative.
All together there are eighty-seven potential embedding types
after the seventeen verbs of perception focused upon in this
investigation. Fifty embedding types have been retrieved as
actual embeddings in the Gothenburg corpus. There is consequently
a rather weak correlation between the rate of potential and
actual embedding types which occur after this set of verbs in
the Cothenburg corpus, i.e. 57.5% of the potential embedding
176

types have actual correspondences in this corpus. From this


point of view the Gothenburg corpus can hardly be claimed to
be representative as a corpus of the sentential complementa­
tion types studied here.

Furthermore, it does not reflect correctly what a native speaker


knows about the Swedish sentential complementation system, as
it contains only slightly more than half of the total number of
embedding types. However, the Gothenburg corpus is representa­
tive only as far as the first set of perceptual verbs is con­
cerned which have been discussed above and which constitute the
major verb classes of perception, i.e. SE 'see', HÖRA 'hear',
and KÄNNA 'feel'. For these verbs a sufficiently large number
of examples have been found in the Gothenburg corpus, whereas
the minor verb classes such as FÅ SE 'catch sight of', FÄ HÖRA
'hear suddenly', MÄRKA 'notice' etc are too little represented
to be able to draw any conclusion about their distribution.

Statements about the productivity of embedding rules which


concern the major classes of verba sentiendi can therefore be
considered to be reliable and statistically valid, whereas
such statements concerning the minor verb classes of verba
sentiendi must be treated with caution due to insufficient data
coverage in the Gothenburg corpus.
177

CONCLUSION OF THE COMPETENCE STUDY


1. EARLIER AND CURRENT RESEARCH ON THE INFINITIVIZATION PROCESS
In the Introduction the background and the goal of the present
investigation are iiscussed, whereby the merits of previous
approaches to the ACI-Construction by historical, traditional,
and structuralist (glossematic) grammarians are emphasized. The
traditional grammarians' concept of the so-called "logical sub­
ject of the infinitive" and their treatment of auxiliaries as
main verbs from a historical point of view are given as examples
of insightful syntactic research of the past.

Such concepts have lately been presented in a generative-trans­


formational framework as a "new" discovery of an underlying
complement subject which is associated with the complement verb,
that is the infinitive (cf. Rosenbaum 1967a and Kiparsky &
Kiparsky 1970) or as a pseudodiscovery of auxiliaries as main
verbs (cf. Ross 1967b).

Except for some cliche references to Jespersen 1954 most American


generative linguists who treat the problem of infinitivization
in English neglect the earlier historical and descriptive studies
of the infinitivization process carried out long ago by European
philologists and historical grammarians. Only in the last few
years have some American linguists who work within a generative
framework begun to consider earlier syntactic research on the
infinitive and other types of embedding (cf. e.g. Closs 1965,
Bever & Langendoen 1971, Bresnan 1972 etc).

However, throughout the present investigation of the Germanic


complementation system the results of a long series of European
historical linguists have been considered and evaluated.
Philologists or historical grammarians who have achieved pertinent
results on the conditions of the infinitivization or gerundiviza-
tion processes mainly in ancient Germanic languages (e.g. Old
Swedish, Old Icelandic, Old English, Old High German and Gothic)
have been acknowledged.

Furthermore, the Two-Subject-Hypothesis which was suggested by


traditional grammarians at the beginning of this century (cf.
e.g. Beckman 1959:252, Alving 1918:1 etc) has been accepted as
a working hypotheses for describing the infinitivization process
in Swedish (Part One) and Infinitivization and Gerundiviza-
tion in German and English (Part Two). It is clear that modern
transformational thinking in terms of syntactic processes and
constraints has been indispensible for the present investiga­
tion of verb complementation in Germanic languages. Modern
syntactic descriptions include far more abstract concepts than
did the works of earlier philologists and historical linguists.
Their concern was mostly surface structures (cf. e.g. the
works of Grimberg 1905 on ACI-Constructions in Old Swedish,
Nygaard 1905 on Old Norse syntax, Callaway 1913 on Old English
infinitives etc), whereas the present approach to the verb
complementation systems of Germanic languages include abstract
common deep structures, transformational rules, and derivational
constraints. The latter improvements of syntactic theory are
178

bound to yield quite different results in the description of


the syntax of Germanic languages (cf. Lehmann 1970 and 1972
on the significance of transformational theories for diachronic
syntactic research).

Under the assumption o f two subject NPfs, a matrix and a comple­


ment subject NP, a useful common syntactic basis is posited
which will be the starting point of the comparative approach
developed in detail in Part Two concerning the verb syntax of
Germanic verba sentiendi. Also for the description of the verb
complement system after these verbs in Modern Swedish, the
abstract deep structures facilitate the description of a host
of surface structures.

The rules which generate the ACI-Constructions for instance


can be claimed to be rules which are common to Germanic. It has
been the task of the present investigation to describe the
nature of these rules, and to demonstrate the various level
constraints on the generation of primarily the ACI-Construction,
and also of other verbal complements which must be included in
a description of the embedding processes in Germanic languages.

2. SYNCHRONIC AND DIACHRONIC COMPARATIVE SYNTAX


The basis of a synchronic comparative syntax model for describing
sentential complements after verba sentiendi is established in
Part One and Part Two of this investigation.

To disregard the results in infinitive syntax accomplished by


the philologists mentioned in the Introduction (1.4) would
mean both a waste of previous research and a loss of insight.
Synchronic comparative syntax of the infinitive cannot possibly
reach the rigor of descriptive adequacy as found in these
diachronic comparative works on infinitive syntax, unless the
same requirement of empirical observations of linguistic
material is observed in present-day investigations.We must
draw upon the results of comparative diachronic syntax and
apply modern linguistic methods to elaborate a comparative
synchronic syntax.

Therefore, in sections 5 through 16 the intuition-based method


of describing syntactic structures in Modern Swedish is
supplemented by empirical observations of a given corpus, the
Gothenburg corpus. Such an empirical approach is nowadays fully
accepted in socio«* psych o-and neurolinguistics.

Section 1.5 discusses at some length the theoretical framework


for describing the Swedish sentential complementation system,
which is the starting point of the synchronic comparative
syntax of Part One and Part Two. A combination of the syntac-
ticist and the semanticist models was found to be the most
appropriate framework for elaborating a comparative syntactic
179

model in our description of the infinitivization process in


Germanic languages, that is the Extended Standard Theory modi­
fied by Perlmutter's 1971 concept of deep and surface structure
constraints together with some of Chafe's 1970 semantic distinc­
tions.

3. A THEORETICAL ISSUE OF WORD ORDER REPRESENTATION


Four arguments are presented in section 1.5 against treating
Swedish as a VSO language along the lines of Dahl 1971,who
applies to the Swedish language McCawley's 1970 proposal
to treat English as a VSO language. The psycholinguistic argu­
ment against an underlying VSO order in English is supported
by Bever's 1970 results on the perceptual strategies of the
listener to identify syntactic structures; the typological
argument is supported by some of Greenberg's language universals;
the historical argument is supported by Lehmann's 1970 findings
on the historical development of the Proto-Germanic SOV word
order to modern Germanic SVO order; and finally the inherent
generative semanticist argument refutes any hypothesis of a
fundamental syntactic VSO order for any of the Germanic
languages.

4. SUBCATEGORIZATION OF SWEDISH ACI-VERBS


In section 2 on Modern Swedish, ACI-Constructions are exemplified
as occurring after twenty major ACI-verbs in declarative,
subordinate, and interrogative sentences. In order to gain
perspectives of the syntactic domain on which the rule of Subject
Raising operates, the syntactic rule which generates ACI-
Constructions, the reader is presented an alphabetic list in
section 3.1 of twenty-nine verbs which can take ACI-Construc­
tions as embeddings under given circumstances.

In section 3.2 a semantic subclassification of the Swedish ACI-


verbs follows a syntactic Subcategorization Rule (i) which
assigns the syntactic rule feature [+ Subj Rais] for Subject
Raising. Detailed semantic subcategorization is found in section
3.3 for the verbs of perception (verba sentiendi), in section
3.4 for the performative and mental verbs (verba dicendi and
putandi), in section 3.5 for the instantaneous-perfective verb(s)
(e.g. FINNA 'find'), in section 3.6 for the stative-nonperfective
verb(s) (e.g. VETA 'know'), in section 3.7 for the causative
verbs, and in section 3.8 for the reflexive verbs (e.g. VISA
SIG 'turn out'). Part One and Part Two of the present investiga­
tion are devoted to verba sentiendi, (whereas Part Three will
describe verba dicendi).
180

5, the concept of syntactic rule domain and derivational


level constraints
Section 4 presents the concept of syntactic rule domain. It is
necessary to curtail the generative power of the Standard Theory.
Such a limitation is implemented by introducing various level
constraints into syntactic theory. Some of these constraints
which block ill-formed ACI-Constructions are exemplified in
section 4.2. For instance, the Equal Subject Constraint on
Subject Raising prevents the following ill-formed ACI-Construc­
tions from occuring in Swedish: (31a) *John sade flickan vara
sjuk 'John said that the girl was ill1 and (31b) *Vem sade flickan
vara sjuk 'Who said that the girl was ill'; the Aspectual
Constraint on Subject Raising blocks an ACI-Construction like
(32c) *Han märkte Peter falla från taket 'He noticed that Peter
fell from the roof'; the Coreference Linkage Constraint on the
Passive Rule blocks (34) *Sig hördes av FNL:ama tala om Viet­
nam '^Themselves were heard by the FNL adherents to speak of
Vietnam'; and finally the Indefinite Subject Constraint on the
Passive Rule blocks (35) *Palme hördes av LO tala om löneför­
höjningar och skatter 'Palme was heard by the LO speaking of
pay raises and taxes'.

These constraints are called deep or shallow structure con­


straints and are indispensible in making adequate descriptions
of the ACI-Constructions in Swedish (and other Germanic languages
for that matter).

In Swedish there is also a surface structure constraint which is


of stylistic nature and which blocks the cooccurrence of two
s-passives as exemplified by (36a) *Den turkiske revolutionären
uppges skjutas i dag 'The Turkish revolutionary is reported
to be going to be shot today'.

All these examples of violation of deep, shallow, and surface


structure constraints illustrate how constraints can be used as
devices for describing adequately the Swedish (and Germanic)
complementation system of embedding.

6. the general process of infinitivization i n germanic languages


In section 4.3 the general process of Infinitivization in
Germanic languages is discussed, whereby a transformational
syntactic rule, the Subject Raising Rule in our sense, is
formulated to carry out infinitivizations of verb phrase
sentential complements (cf. (37a)) on one hand, and on the other
hand of such sentential complements which are preceded by an
object NP which is coreferential either with the subject or
object NP of the complement sentence (cf. (37b)). By treating
the rule which infinitivizes the complement verb in verb phrase
complements as well as noun phrase complements as a general
Subject Raising Rule, a useful generalization is gained, although
in the latter case, i.e. in phrase marker (37b) an Equi NP
181

Deletion Rule also operates after Subject Raising has moved the
complement subject into the matrix sentence.

Some examples of the conditions of the Swedish Subject Raising


Rule are given which concern the generation of ACI-Constructions
after verba dioendi. The rise of the Equal Subject Constraint
in Modern Swedish is demonstrated by examples from Old Swedish
and Old Icelandic where this constraint did not exist in senter-
ceswhicn contained performative verbs like Old Swedish SW^ERIA
'swear', BINDA SIG 'pledge', SIGHIA 'say' etc (cf. Ureland 1972V
(forthcoming) on the change in syntactic domain in Swedish),

7. SPECIFIC SWEDISH SENTENTIAL COMPLEMENTATION


Sections 5 through 16 describe the occurrence of nine sentential
complements after seventeen verba sentiendi in Modern Swedish.
In paragraphs 1 and 2 of the Conclusion of the Performance Study
above,the distribution of the embedding types after these verbs
is summarized (cf. especially Figures 7A and 7B). The theoretical
points of view will therefore be summarized here.

In section 5.1 the operation of the Subject Raising Rule is


demonstrated and arguments are presented for the existence of
such a syntactic rule. In section 5.1.1 the complementizer
embedded sentences after SE^ 'see visually' are discussed.
Sections 5.1.2 and 5.1.3 give syntactic and semantic evidence
for treating complementizer embedded structures as derived from
different underlying syntactic structures than the ACI"Construc­
tions. The claim of some American generativists that comple­
mentizers are surface phenomena and therefore transformationally
derived (cf. Rosenbaum 1967a and R. Lakoff 1968) is rejected in
this investigation. (Cf. also Bresnan's 1972 criticism of the
transformational hypothesis for deriving complementizers).

Section 5.2 discusses the Neutralization of Aktionsart or


Auxiliary Deletion in connection with the Subject Raising Rule.
It is demonstrated that without a complete text it is often
impossible for a reader to assign the correct Aktionsart to
numerous aspectually ambiguous Accusative-with-Past Participle
Constructions. The following three sentences are examples of
aspectual ambiguities which are due to the Neutralization of
Aktionsart or Deletion of the Passive Auxiliary: (76a) Man hade
nog velat se dem verifierade i andra källor 'One would have
liked to have seen them verified in other sources', (76b) och
se butik efter butik fullproppade av k öpmedvetna köpstarka
tonåringar 'and see shop after shop stuffed full with eager
affluent teenagers', and (76c) Först när man ser faktorerna
kartmässigt framställda 'First when one sees the factors
plotted on the map'.
182

Therefore, in order to enable the reader to disambiguate these


Accusative-with-Past Participle Constructions, complete pertinent
contextual information is supplied from the Gothenburg Corpus
(cf. especially footnote 51).

A careful study of the linguistic contexts of (76) results in a


straight-forward stative aspectual assignment to verifierade
'verified' in (76a), but an ingressive aspectual assignment to
fullproppade 'being stuffed' in (76b) and framställda 'being
plotted' in (76c).

However, there are instances where the correct aspectual assign­


ment to a past participle may be problematic, even though the
whole linguistic context is scrutinized as in (77a) har sonen
till Österrikes kejsare förklarat att han ser sig tvingad vid­
ta speciella åtgärder 'the son of the Austrian Emperor has declared
that he sees himself forced to take special measures' and (77b)
en ung flioka som då hon ser sig övergiven låter prostituera sig
'a young girl who prostitutes herself, when she sees herself
abandoned'. In the latter two cases the stative point of view
is probably the Aktionsart which is intended by the writer.

In order to explain the reader's competence to assign the correct


aspectual reading to the complement verbs involved a rule of
Aktionsart Copying is also suggested in section 5.2 (cf. (85)).

In sections 5.3 (SEi), 7.3 (MÄRKA), 9.1 (UPPTÄCKA), 11.2


(OBSERVERA), 13.5.1 (HÖRA), and 15.2.2 (KÄNNA) nominalized
structures are discussed which occur as verbal complements after
verba sentiendi.

Under the balanced hypothesis of treating nominals, these embed-


dings are seen as transformationally derived if underlying verbs
or adjectives exist. The following examples illustrate such
nominals which can be derived transformationally: (86) en ung
kvinna som från bussens övervåning sett striden vid hållplatsen
'a young woman who from the top deck of the bus had seen the
fighting at the bus stop', (nominal after SE^ 'see visually');
(114a) man märker förändringen inte minst på den del av perso­
nalen 'one notices the change, not least from
the section of the employees' (nominal after MÄRKA 'notice');
(121a) mamma upptäcker slarvet och syndaren sätts att skriva
'Mother discovers the carelessness and the offender is put to
writing' (nominal after UPPTÄCKA^ 'discover visually'); (135b)
den som seglar båt har säker observerat skillnaden i rörelse
under segel 'Whoever sails boats has certainly noticed the
difference in movement while sailing' (nominal after OBSERVERA2
'become aware of'); (210a) Varje människa åtminstone någon gång,
har känt en önskan att bli fri från de s ociala konventionerna
'At least once, every person has felt a wish to be free from
social conventions' (nominal after KÄNNA3 'feel mentally1) and
(176a) första dagen hörde jag blodhundarnas skall efter oss
'on the first day I he ard the barks of the bloodhounds behind
us' (nominal after HÖRA-^ 'hear auditorily').
183

An extensive discussion of the syntactic status of Swedish FÂ


which occurs in the verb compounds FÅ SE. 'catch sight of' FÅ
HÖRA2 'hear of' is presented in sections 5.5.1 and 13.6.1.
Comparative syntactic arguments are forwarded which support the
view of treating FÅ in Swedish as a perfective-aspectual auxili­
ary, which is gelled out by a special morphological rule in front
of a limited number of Swedish main verbs, i.e. as in FÅ SE
'see suddenly' FÅ HÖRA 'hear of', FÅ KÄNNA 'feel suddenly', FÅ
VETA 'be told unexpectedly', and FÄ ERFARA 'be informed unexpect­
edly'.

Comparative evidence is presented which refutes Ross's 1967b


suggestion to treat an auxiliary like FÅ as a main verb. If
syntactic deep structures shall be constructed by linguists to
be as universal as possible, then aspectual overt markers like
the aspect prefixes in the Slavic languages (e.g. Russian /u/)
or imperfective or perfective auxiliaries in Germanic languages
(e.g. English ing-forms and the Swedish auxiliary FÅ respectively)
are to be treated as surface structure manifestations, because
such prefixes or auxiliaries express the semantic Aktionsarten
of the deep structure. Since Danish, Modern Icelandic, German, and
English all lack an overt perfective-aspectual marker which
corresponds to the Swedish FÅ before the verbs just mentioned
(cf. (97a)-(97f)), it is from a comparative syntactic point of
view more adequate to describe such surface structure differences
in terms of a common Germanic deep structure, and language-
specific transformational or morphological rules, than to posit
a syntactic deep structure for Danish, Icelandic, German, and
English on the one hand (cf. (99)), which is different from that
for Swedish on the other (cf. (98)).

In section 15 the sentential complements which occur


after KÄNNA 'feel' are treated. There are semantic reasons for
subclassifying the homograph KÄNNA into four subclasses which
denote four different ways of perceiving, i.e. KÄNNA^ 'feel
tactually', KÄNNA2 'smell', KÄNNA3 'feel mentally', and KÄNNA4
'feel somatically'.

An important distinction is also maintained between perceptual


verbs which denote passive or active perception on the part of
the deep subject. Most verbs treated here denote passive per­
ception. Only OBSERVERA 'watch' is an active perceptual verb
which takes an Agent as the deep structure subject. By sub-
classifying the perceptual verbs as being either [+Purpose] or
[-Purpose] verbs, Redundancy Rules (i) and (ii) in section 3.3
automatically assign them the kind of deep subject that they
can take. If they are marked as [+Purpose] they take an Agent,
if marked as [-Purpose] they take an Experiencer as the deep
subj ect.

The language-specific study on the Swedish verb complementation


system after verba sentiendi has been treated in detail in Part
One of the present investigation.

In Part Two we shall present a comparative study of verb comple­


mentation in three Germanic languages, whereby the deep struc­
tures, the general transformational rules, the language-specific
184

level constraints, etc will be treated in the descriptions of


Swedish, German, and English AC! and AWG-Constructions. Because
of space limitation it is not possible to treat all the surface
structures discussed in Part One in the comparative approach
developed in Part Two, but we shall focus on the infinitiviza-
tion and gerundivization processes in the three Germanic languages.

Furthermore, in order to demonstrate that the comparative model


of Part Two can be applied also to non-Germanic languages, the
generation of some infinitive and gerundive constructions in
North Lapp and Standard Finnish will also be demonstrated.
185

FOOTNOTES TO PART ONE

1. In th e ol dest historical documents of Swedish, the Old Sw edish laws


and c h r o n i c l e s o f t h e t h i r t e e n t h and f o u r t e e n t h c e n t u r i e s , AC I -
Constructions and .at-Embeddings ar e found as frequent verbal comple­
ments a fter verba sentie ndi and dicendi. The transformational rules
of embedding sentences into each oth er can t hus be claimed to be a t
l e a s t a s o l d a s t h e f o l l o w i n g e x a m p l e s i n ( i ) t h r o u g h ( v ) . ( We h a v e
reason to believe that Complementizer Em bedding an d Subject Raising
were synt actic rules also in Proto-GermanÎc. Historical comparative
evidence from Old English, Old Hig h Germ an, and Gothic supports such
a clairr.

SE (A) ' s e e '

AC I - C o n s t r u c t i o n

(i) a. eth korss saa man ther staa( E k 2 5 0 6 : G 3 ^ 5 )


(one sa w a cross standing there)

at-Embedding

b. iak sa at maper hafK viliœ sin mep faar{Vgl l l , R 8 : G 3 ^ 6 )


(I saw t h a t a man had sexual intercourse with sheep)

HÖ'RA ' h e a r '

ACI-Construct ion

(ii) a. Jak hörde wisa men thz swärya{Ek 3 ^ 6 9 : G 3 W


(I heard wise me n swear on it)

at-Embedding

b. nu far hin fat höra sum aerft hauœr ( Ö g L / E 7' G 3 ^ 9 )


at hon œr hauande
(Now t h e person wh o h a s r e c i e v e d t h e i n h e r i t a n c e h e a r s t h a t
she is pregnant)

K/E N NA 'smel 1 '

ACI-Construction

(iii) a. vm ena nat kœnde faseren sötasta luot vtganga af bamsens


hœrbaerghe ( C o d . B u r . 5 2 5 : A h l 1 9 )
(one night the father smelt a sweet odor comi ng out o f the
children's bedroom)

S1GHIA ' s a y '


ACI-Construct ion

(iv) a. um mafcer sigher man haua fœ œller hors{M g L I I R 8 : G 3 1 6 )


( i f somebody says t h a t somebody e l s e has sexual intercourse
with c attle or horse)
186

at-Embeddinq

(v) a. sighir at han skal gangœ til roms ( V g L I A 1 2 : G 3 1 1 )


(says t h a t he i s going t o Rome)

2. The Got henburg corpus covers a body of text of 1 000 669 running w ords
dis tribute d ov er 1387 newspaper a r t i c l es which a r e written by 569
d i f f e r e n t persons. The a r t i c l e s a r e of t h r e e major gen res: (A) general
news rep orts (42.6 %), (K) a r t i c l e s on c u l t u r a l s u b j ec t s (46 %) , and
(U) r e p o r t s w r i t t e n by correspondents on foreign p o l i t i c s (11.4 %).
The frequency dictio nary i s based on a random s ample of a r t i c l e s from
the following f ive major Swe dish mo rning papers of the year 1965:
SD
Svenska Dagblad et ( V )(29.6 %). Stockholms-Tidningen (ST) (12.1 %),
D a g e n s N y h e t e r Tdn ) ( 1 2 . 5 % ) , G ö t e b o r g s H a n d e l s - o c h S j ö f a r t s - T i d n i n g
(GHT) ( 1 7 . 6 %), and Syds venska Dagblad et S n ä l l p o s t e n (SDS) ( 2 8 . 3 %) •
For specific details concerning the Gothenburg corpus the reader is
r e f e r r e d t o t h e i n t r o d u c t i o n o f Allén 1970aXI I I-XLVI and Allén 1970b.

3. Beckman's 1959 s t a r t in g point for the treatment o f the ACI-Construction


i s the clause subordinated by the conjunction a t t . This underlying
structure for the generation of the ACI-Construction is different from
the one presented in s ection 4.3. The underlying st ructure of an A C I -
Construction i s thought t o contain only a sentence complement without
any conjuncti on, that is only an S node. The underlying s tructure of
an att-construction is thought t o contain a type node, a T , onto which
the complementizer i s inserted, a t t . The remainde r of Beckman's r ule in
§ 227 is , however, in accordance with the Sub ject Raising Rule as
formulated on independent grounds in the present investigation:
"Objekt med infinitiv och subjekt med infinitiv. Dessa ersätta vanligen
fullständiga satser med att eller huru och uppkomma genom att den under­
ordnande satsens subjekt göres till objekt i djen överordnande satsen
eller om denna är passiv, till subjekt varefter infinitiven bifogas
s å s om e n p red ika ts bes täm ni ng " (B e c k m a n 1 9 5 9 : 2 5 2 : § 2 2 7 ) .

4. In a recent c r i t i c a l study Col l i n d er 1970 a t t a c k s t h e generative concept


of deep stru ctu re and tran sformations a s a means of ex plaining the native
speaker's learning and understanding of surface structures. Without
recognizing or accepting the results of linguistic and psy cho-linguistic
research o f t h e past ten t o f i f t e e n years Col Under adheres t o t h e
empiricist and beha viorist view of language learning of the nineteen
f o r t i e s and f i f t i e s . He a s s e r t s t h a t a four-year-old c h i l d does not
know what an o b j e c t i s in t h e sentence J ohn persuaded B i l l t o leave
with the motivation:
"es (das Kind/S.U.) versteht aber mit Hilfe der Muster, die es im Gehirn
vorrätig hatj dass derjenige}der sich entfernen soll, kein anderer ist
als derjenige, der nach persuaded erwähnt wird Um zu erklären,
wie es kommt, dass das vierjährige Kind nicht glaubt3 dass John den Bill
davon überzeugte, dass er selbst (John) sich entfernen würde (oder
sollte) haben wir keine Tiefenstrukturen oder Tilgungs-Prozesse oder
We gr ad ie ru ng sp ri nz ip ie n nöti g", ( c f . 2 7 " 2 8 ) .
187

Consequently, the occurrence of a noun phrase immediately a f t e r the


matrix verb persuaded gives the listener, accordning to Collinder,
enough information about who i s the subject noun phrase o f t h e
complement verb leave. However, Collinder does not mention in h i s
c r i t i c i sm another very important example discussed by Ch omsky 1968:
48-49 which show s t he inadequacy of Col lind er1s explanation in terms
of surface structure categories: John promi sed Bil l to leave.Al though
Bill is the clos est noun phrase whic h occurs a f t er the matrix verb
i t c a n n o t p o s s i b l y be t h e s u b j e c t o f l e a v e . By n o t r e c o g n i z i n g t h e
deletion of the underlying sub ject noun phrase Collinder even con­
tradicts traditional grammarians, historical grammarians, and
s t r u c t u ra l i s t grammarians (see section 1.3).

5. The e x i s t e n c e of a common s y n t a c t i c Germanic r u l e , t h e Subject Raising


Rule, which generates ACI-Constructions in a l l Germanic languages, i s
unquestionable. One ma y, however , argue whether t he deep s t r u c t u r e s,
(37a) and (37b) a s d e s c r i b e d in s e c t i o n 4 . 3 a r e a l s o common t o Germanic
languages. In an investigation of comparative syntax such abstract
representations as (37a) and (37b ) are motivated f or comparative reasons,
although they ma y somet imes be questio nable from a language-specific
point of view ( c f . f o r instance Part Two sectio n 11, where only in
Swedish but not in English does such an underlying st ructure a s (37b)
seem motivated from a 1anugage-specifi c point of view, that is for the
generation of English senten ces l i k e He f e l t i l l .

6. In Ureland 197^-b (forthcoming) a h i s t o r i c a l s t u d y of change in s y n t a c t i c


domain i s presented which descr ibes the change of the conditions for
the applic ation of the Subject Raising Rule in Swedis h between the
time o f t h e Old Sw edish laws (13th and 14th c e n t u r y A.D.) and Mo dern
Swedish (20th century). The observation o f an Equal Subject Constraint
on Subject Raising in Mod ern Swed ish sentences containing verba dicendi
(see Ureland 1970) led to the discovery o f the lack of such a constraint
on Subject Raising in the tex ts of the Old Sw edish laws which contain
verba dicendi . The ri se of the Equal Subject Constraint was discovered
by comp aring t h e Modern Swed ish ACI-Const ructions with Old Swedi sh A C I -
Constructions. By having f u l l e r access t o t h e s y n t a c t i c processes in
Modern Swedi sh we could consequently ma ke progress on t h e problem of
historic al syntax by com paring th e conditions of a syn tactic rule
between two periods of the Swedish language. Our knowl edge of the
present Subject Raising Rule could be used to explain the Subject Raising
Rule of Old Sw edish and the syntac tic change, (cf. Labov 1972:1110-1112
On t h e Use o f t h e P r e s e n t t o E x p l a i n t h e P a s t ) .

7. I o w e t h e i n h e r e n t g e n e r a t i v e s e m a n t i c i s t a r g u m e n t a g a i n s t a V SO b a s i c
o r d e r in English t o J o e Emonds w ho p r e s e n t e d a paper a t t h e L i n g u i s t i c
I n s t i t u t e a t t h e S t a t e U n i v e r s i t y o f N ew Y o r k , B u f f a l o , i n J u l y 1 9 7 1 .
T h e i n a d e q u a c y o f t h e s e m a n t i c V SO o r d e r f o r g e n e r a t i n g E n g l i s h s e n t e n c e s
which contain i n d i r e c t o b j e c t s can be shown t o be t h e same in g e n e r a t i n g Sw.
sentences of the same st ructur e. The following de rivational history
demonstrates th e inadequacy of the generative semantic predicate-and-
argument or dering:
188

givit John Bill pengarna


given the money

Subj +T Pred
liT R a i s
(H)
Ra i s

seem

givit Bill» pengarna


have given t h e money

Il T NP
Jl Fron t

John

Bill ha givit pengarna


to have gì ven the money
189

8. Also from the stand-point of predicate calculus there is no impelling


reason to have a predicate-first notation wit h the arguments following
the predicate. Logicians wri te the predicates first to obtain a non-
ambiguous representation. This i s the practice of logicians since
th eir notations do not contain labelled brackets. In synta x, however,
the basic order of elements and the labelling of brackets are crucial.
It is this mixture of logic and sy ntactic formalism which has caused
s o much confusion among t he l i n g u i s ts o f a generati ve semanticist
o r i e n t a t i o n w h o u s e t h e s y m b o l V , n o t f o r v e r b , b u t f o r p r e d i c a t e , a nd
t h e symbol NP, not f o r noun phrase in a s y n t a c t i c s e n s e , but f o r argu­
ment in a semantic sen se. Their t r e e representations a r e not s y nt a c t i c
t r e e s , r a t h e r they a r e s o - c a l l e d semantic t r e e s ( c f . McCawley 1972:2).

9. Lakoff s Ross 1967, McCawley 1968 and 1972, Chafe 1970 e t c have d i s ­
claimed the value of a s y n t a c ti c a ll y based gra mmar whi ch contain s
s y n t a c t i c deep s t r u c t u r e s f o r d e s c r i b i n g how m an com bines semantic
contents with phonetic exp ression.

10. I n t h e e a r l y s i x t i e s t h e t r a n s f o r m a t i o n a l r u l e s o f a l i n g u i s t i c g r a mm a r
were thought t o have psychological correlates in man's ment al grammar.
The m ore co mplex a surface s tructu re could be sho wn t o be transf orma­
tionally, the more di ff icul t i t was claimed to be from a perceptual
p o i n t o f view ( c f . Chomsky & M i l l e r 1 9 6 3: 4 8 1 - 4 8 3 , M i l l e r & McKean 1 9 6 4 ,
Savin & Perchonock 1965 etc on the Correlation Hypothesis in psycho-
l i n gu i s t i c s which claimed t h a t performance complex ity mirrored
competence complexity). However, Slob in 1963, Fodor S Garett 1966,
Slobin 1968 etc soon found in psycholinguistic experiments that for
instance truncated passives (John was foun d) and comparative con­
structions with deleted verbs (John runs faster than Bill (runs))
take less time t o process than d id full passive and fu ll comparative
constructions, although they have under gone a t least one more
transformational rule than e ither the full passives or comparatives.

11. Chafe I97O launches an a ttac k on the Interpre tative Standard Theory f o r
its syntactically based theory of semantics and sugg ests instead a
semantical 1 y based theory because i t is more "nat ural" to posit a
semantic underlying structure than a syntactic one for generating
a s u r f a c e s t r u c t u r e . He a l s o claims in t h i s c o n t e x t t h a t t h e r e i s
directionality within the derivational process of a given sen tence.
The sta rting point of a derivation should be the semantic representa­
tion with the phonetic s urface s tructure as the end prod uct. The
generative part par e xcellence is, in h is view, not the syntactic
component of the grammar a s in Chomsky's 1965 theory , but the seman­
t i c component (cf. especially pp. 59~65) .

12. Examples of such competence-oriented publi cations a re Linguistic


Inqui ry, Vols. 1-1 I I (1970-1972), M.l.T. Press, Papers in Linguistics,
V o l s . I-5, p u b l i s h e d by L i n g u i s t i c R e s e a r c h , I n c . » A l b e r t a , a n d t h e
Papers from th e Regional Meetings of the Chicago Lin guistic Society
(19^8-1971). C r i t i c i s m a ga i n s t t h i s o ne-sided o r i e nt a t i o n in g enerative
grammar has been d i r e ct e d by Ne wmeyer and Em onds 1971.
190

13. In t r a d i t i o n a l and h i s t o r i c al grammars' the concept of i n t u i t i o n was


accepted a s a criterion for judging whether a sentence was accept able
or not. During th e American s tructural ist period, however, intuition-
based gra mmars were not recognized a s empirically adequate, because
other procedures were considered to be mor e reliable which did not
involve me ntalistic aspects, e.g. the substitution test in syntax as
suggested by Bloomf ield 1933 or Fries 1952, the minimal pair t e s t in
phonemics as suggested by Bloom field and othe r s t r uc t u r a li s t linguist s.
C o m m on t o b o t h t r a d i t i o n a l a n d s t r u c t u r a l i s t g r a m m a r s i s t h e f a c t t h a t
the observations presented are based on a corpus of data which gener a­
tive grammarians consider t o be too limited to allow c orrect general­
izations about the true nature of language. The grea test contribution
of generative grammar t o 1inguistic thinking con sist s in i t s extension
of the boundaries of syntactic research to encompass prog nostic a spects
a s w e l l , and not o n l y be a d a t a - c a t a l o g u e o f sentences ( c f . Chomsky
1957 and Katz 1964 on t he mentalistic approach to syntactic stru ctures).

14. Through t he kind as sistance of Professor S.Allen and hi s colleagues


a huge corpus o f 29 Modern Swed ish A C I - v e r b s has been mad e a v a i l a b l e
t o me. I t includes more than 8000 examples o f t h e d i f f e r e n t tenses
of the 29 verbs which take a ll types of verbal complements and em bed-
dings.The examples cover active finite verb forms in the present, the
past, the perfect, and th e pluperfect. Also i nfinitival forms of the
AC I - v e r b s o c c u r i n t h i s c o r p u s .

15« See t he c r i t i c i s m directed against Allén 1970a and 1971 by Karlgren


1972 concerning the s t a t i st i c a l r el i a b i l i ty and va lidity of the
Gothenburg corpus computerized by t h e Gothenb urg Resear ch Gr oup. As f a r
as the syntactic constructions are concerned, the Gothenburg corpus
contains a sufficient number of embedding types a f te r the major verba
s e n t i e n d i , i . e . S E ' s e e ' , HÖR A ' h e a r 1 , a n d KÄN NA ' f e e l 1 f o r m a k i n g
gener alizations about the produc tivity of the embedding ru les which
a r e i n t h e f o c u s o f t h e p r e s e n t i n v e s t i g a t i o n ( c f . a i s e i: h e C o n c l u s i o n
of Performance Study of Part One. Due t o t h e limited s i z e o f t h i s
corpus, however, certain marginal embedding types a f t e r the minor group
o f v e r b a s e n t i e n d i l i k e F Å S E ' c a t c h s i g h t o f 1 , F Å H ÖR A ' h e a r s u d d e n l y
a n d u n e x p e c t e d l y 1 , MÄRKA ' n o t i c e ' e t c a r e n o t i n c l u d e d a t a l l . A s a
consequence t h i s author often had t o supplement the examples from the
corpus with examples from h i s own i n t u i t i v e knowledge of the Swedish
complementation system.

16. In Ureland 1970 two Gl obal Rules are elaborated t o explain the deriva­
tion of well-formed surfa ce s tructures via ill-formed shallow st ruc­
tures (cf. also Part Three).

17. Generative gramm arians have disagreed as to where to draw the border­
line between s yntactic and seman tic f eatur es in t h e i r descriptions of
s y n t a c t i c s t r u c t u r e s . Chomsky 1965 t r e a t e d p r o p e r t i e s such a s [+human]
and [ +abstract]as syntactic selectional features, because these pro­
perties o f the nouns were claimed t o select the type of verb which
could be used in a given c o n t e x t ( c f . 153-154). McCawley I968 c l a i m s ,
on t he other hand, that such pr operties are semantic pr operties by
nature and that there is "no reason t o have syn tactic selectional
f e a t u r e s " ( c f . 1 3 3 ) . We s h a l l d i s r e g a r d t h i s t e r m i n o l o g i c a l c o n t r o v e r s y
191

a n d i n c l u d e s u c h s e m a n t i c p r o p e r t i e s i n o u r d e s c r i p t i o n o f AC I " C o n s t r u c ­
tions a s have conseq uences for the generation of these surface structures,
that is such fe atures which subclassi fy the ACI-verbs in sections 3.2
through 3.7.

18. I t i s n o t o n l y McCawley 1968 and 1972 who has claimed t h a t t h e r e i s no


difference between s y n ta c t i c and semantic fe atures of the kind dis cuss ­
ed by C homsky 1965. Recent gene rative l i t e r at u r e abounds in suggestions
for modifying th e s yntactically biased Standard Theory. The necessi ty
of including semantic fe atures for generating sy ntactic structures has
been demonstrated by a number of l i ng u i s t s , both Ame rican and Europ ean
( c f . b e s i d e s McCawley 1968 and 1972, Lakoff 197 1a. Kiparsky S Kiparsky
1970, Bierwisch 1970, Heidolph 1970, Fodor 1970 e t c ) . In th e present
d e s c r i p t i o n o f t h e AC I - C o n s t r u c t i o n i n S w e d i s h i t i s n e c e s s a r y t o
include semantic aspects. Semantic consideratio ns a re indi spensible in
describing the constraints on t he Subject Raising Rule. If the inherent
semantic pr operties of the verbs involved were disregarded in a one­
sided syntacticist description of the infinitivization process, i l l -
formed ACI-Constr uctions wou ld result.

19. In a section on the infini tivization process in English, Kiparsky &


Kiparsky 1970 were successful in demonstrating that semantic fe atures
do play a crucial role for the operation of the Subject Raising Rule,
as they c alled the syntactic rule which generated a certain type of
infinitival constructions. Their a r t i c l e , o r more accura tely the e a r l i er
version o f i t , was written in 1967*1968 a t a time wh en syntacti cism was
the prevailing approach t o l i ng u i s t i c s t r u c t u r es . However, they insisted
that the presence o r nonpresence of a given semantic feature can explain
why som e sy nta ctic constructions a r e well-formeo, while ot hers a r e i l l -
formed.

20. Kiparsky S Kiparsky 1970 suggest that on ly so-called fact i ve verbs can
take the noun phrase the fact plus an emb edded sentence as a verb comple­
ment. Nonfactive verbs cannot cooccur with the noun phrase the fact plus
a following sent ence. Compare the two following s u b t r e e s in (A) and
(B):

r+v
[+Fact i ve] 1
suppose

As evidence f o r t h e c l a i m mad e by t h e Kiparskys t h at such a cooccurrence


r e st r i c t io n e x i s t s in Swedish, Germ an, and English compare the follow­
ing f act i ve and nonfac tive s entences:

(C) 1) Han beklagade d e t faktum a t t hon hade köpt en Volvo


[+Factive] 2 ) E r b e d a u e r t e d i e T a t s a c h e , d a s s s i e e in e n V o l v o
gekauft hatte
3) He re gretted the f a c t t h a t she had bought a Volvo
192

(D) 1) *Ha n antog det faktum a t t hon hade köpt en Volvo


[-Factive] 2 ) * E r n a h m d i e T a t s a c h e a n , d a s s s i e ei n e n V o l v o
gekauft hatte
3) *He assumed the fac t that she had bough t a Volvo

The Kipars kys p osit deep str uctures that contain the noun p hrase the
fact plus the embedded sentence (Sj) (cf. p. 157) on the strength of
such cooccurrence restrictions as exemplified in (D).

21. Karttunen 1971 does not see any reason t o postulate that surface
complements o f f a c t i ve v e r b s a r e commanded in t h e deep s t r u c t u r e
by the noun phrase the f a ct a s indicated in (A) in Footnote 20.
Instead he tak es the step beyond syntactic ism t o pure semanticism by
claiming fact i ve and nonfa ctive to be pure semantic p roperties in
h i s paper on pr edicate complement cons tructions in English ( c f . op.
c i t . p. 23, Footnote 24). Although the Kiparskys introduced semantic
features for describing syntax, some synta cticism lingered in their
postulation of ad-hoc s u b t r e e s l i k e (A) above in Footnote 20. In t h e
present investigation no such ad-hoc dev ice is needed. The semantic
properties together with the syntactic rule feature [+Subj Rais] decide
what v e r b complements can o c c u r a f t e r AC I - v e r b s .

22. A semant ically-oriented linguist would probably claim that i t is the


syntactic feature [+Subj Rais] which is predi cat ibi e and not the
semantic fe ature [+Factive] or [-Factive]. (cf. the Kiparskys' claim
that o nly nonfactive p redicates allow the ACI-Construction (op.cit.
p. 146). However, there a r e many nonfactive verbs which do not allow
the ACI-Construction a s verb compl ement, e .g. the Swedish verbs
yrka på ' i n s i s t on' kräva 'demand' e t c . Therefore, we canno t di spense
with the syntactic rule feature [+Subj Rais], since nonf acti veness
is not always corre lated with Sub ject Raising in the sense of the
Kiparskys. They noticed this fact themselves (c f. p. 147) without
drawing t he conclusion t h a t a l s o in Engl i s ^ marking the s y n t a c ti c
rule feature would be imperative f or a given s et of nonfactive verbs.

23. The term performative i s used here in the sense of Ross 1970, Fra ser
I97I» and And erson 1971.

2b. T h e t e r m m o d a l i t y i s U9ed h e r e f o r d e n o t i n g t h e v i s u a l , t h e a u d i t o r y ,
t h e t a c t u a l , t h e v e r b a l , and t h e o l f a c t o r y m o d a l i t i e s o f man a s i s common
usage am ong neur olinguists ( c f. Whitaker 1971:38-92).

25. Ellegård 1971 has informally c l a s s i f i e d t h e verbs o f perception in


English, as "varseblivningsverb", without making the semantic sub-
classification in terms of auditory, visual, tactual, and ment al
fea tures. Nor does he dis tinguish between verbs that inherently denote
p u r p o s e l i k e LOOK AT a n d LIS TEN T O a s compa red t o SEE a n d H EAR,
(cf. p. 162).

26. Through information provided by Fil.kand. Jan Anwa rd, Uppsala, I have
realized that Rogers 1971 makes a s imilar distinc tion between physical
verbs of perception in terms of 1) cognitive perception on the part of
the experiencer who i s passively exposed t o some kind of perception
( S E E , H E A R , F E E L , N O TI C E e t c ) a n d 2 ) a c t i v e p e r c e p t i o n o n t h e p a r t o f
t h e a g e n t i n F i l l m o r e ' s 1 9 7 1 s e n s e . T h e a g e n t d o e s s o m e t h i n g t o e x­
p e r i e n c e s o m e k i n d o f p e r c e p t i o n ( O B S E R V E , W A T CH , L I S T E N T O e t c ) ;
193

3) descript ive perception on the part of an imaginary experiencer who


does not directly experience the action or state of perception mentioned
( L O O K L I K E , F E E L A S I F e t c ) . We a r e o n l y c o n c e r n e d w i t h t h e f i r s t t w o
classes of Rogers perceptual verbs. His c lassification is in terms of
s e m a n t i c f e a t u r e s i n h e r e n t i n t h e nou n p h r a s e s (EXPERIENCER, AG ENT, e t c ) .
To include such fe atures in the verb c lassification attempted in 3.3 is
not p o s s i bl e without devising a much mor e complicated mec hanism f o r
subclassification. The dist inction between purpose and nonpurp ose verbs
of perception i s, however, su fficient for our goal here.

27. A c c o r d i n g t o J a n A n w a r d t h e S w e d i s h v e r b s KÄN NA P Å , L UK T A P Å , a n d S MAKA


PÅ c a n b e used a s a c t i v e v e r b s o f p e r c e p t i o n r e q u i r i n g t h e f e a t u r e
[+Purpose]:
(i) Han känd e på tyget. 'He f e l t the cloth.'
(ii) Han l u k t a d e på rosen*'He smell ed t h e rose.'
(iii) Han sm akade på soppan. 'He t a s t e d the soup.'

28. Ross's I97O claim t h a t the performative se ntence i s always a top s i n g l e


sentence has been rejected by Fraser 1971 who r e f er s t o sentences l i k e
I regret that I must inform yo u of your dismissa l in which th e per­
formative verb inform i s not part of the top s entence. Two o r more per ­
formative s entences can al so occur before the embedded sentence a ccording
t o Fraser a s i n : Let me point o u t t h a t I admit t h a t you a r e r i g h t .
Both com mand and admission cooccur a s performative s entences here.

29. Fraser concludes h is examination of Ross's 1970 performative an alysis


by asking for a better "viable alternati ve for the task of pairing
sentences with th eir force(s)" than in terms of dominating performative
sentences ( c f . p. 27). He therefo re d isclaims Ross's postul ation of a top
performative se ntence for each senten ce in ord er to explain illocutionary
force as being sy ntactically unmotivated (cf. p. 28). Not a single one
of Ross's fourteen arguments fo r postulating underlying performative
sentences i s accepted by Fraser in h is detailed criticism of Ross's
arguments fo r such underlying structures (cf. pp. 8-27).

30. Only such performative verbs have been trea ted here a s a re part of
so-called overt performative c lauses, that is such performative clauses
in which t h e r e i s an e x p l i c i t performative verb: Han säge r a t t han ä r
sjuk 'He say s that he is i l l ' .

31. In Russian there is a complex sy stem of overt a spect markers for


signalling the mode of ac tion (the procedural)of the verb, e . g . :

(He heard she was playing the piano.)

(Han hörde • -hon spelade piano.)


194

(i) b. On u s l y l a l j b ^ l o n a i g r a l a na r o j a l e .
[+Perf] lCtoJ [-Perf]
/ h o w ! 5e 1 6
(He heard Uhai-r * was P^aY*n9 the piano)
I thatj "

(Han f i c k höra hon s p e l a d e piano.)

I n ( i b ) t h e v e r b o f p e r c e p t i o n SLy5at ' ' h e a r 1 m u s t o b t a i n t h e p e r ­


fective prefix /u-/ to yield a correct Russian s urface s tructure,
i f the instantaneous mode of ac tion i s the type of underlying
Aktionsart, i.e. [+1nstantaneous] in a semantic, or[ +Perfective]
in a morphosyntactic notati on. The cooccurrence of a nonperfecti ve
verb and semantically instantaneous verb is not possible a s in *( iia)
a s compared t o ( i i b ) , where th e perfective prefix / u - / occurs before
the verb of perception:

( i i ) a . *0n sly£a1 c t o ona zakryl a dvjer.


[-Perf] [+lnst]
(He heard t h a t s h e opened t h e door.)
(Han hörde a t t hon öppna de dörr en.)

b. On u s l y l a l c t o ona z a k r y l a d v j e r .
[+Perf] [+lnst]
(He heard t h a t she opened t h e door.)
(Han f i c k höra a t t hon öppnad e dörr en.)

32. For information on aspe ct in Russian I am indebted t o F i l . k a n d . L.


Lagerstedt, Uppsala, Docent C.G. Söderberg, Umeå, and A r c h i v i s t
A. Bränström, Umeå.

33. The extensive research on Aspekt and Aktionsart cannot be treated


h e r e a t a n y l e n g t h . We s h a l l l e t i t s u f f i c e t o r e f e r t o t h e l i t e r a ­
ture on aspectology a s discussed by Wellander 1965 in h is review of
various theories of Aspekt and Aktionsart as applied by Germani sts
and Scandinavi an is t s in the past and prese nt centuries. However, a
description of the Subject Raising Rule in Germanic languages i s not
possible without including some of the d i st i n c t i o ns made bet ween
Aktionsart and Aspekt by SI a vi s t s . The aspect affixes which were
discovered a s ex pressing the procedurals of the verb in Sl avic
languages were thought by a number of Germanists t o have gram matical
correspondence in Germanic languages. Streitbe rg 1890 introduced t he
terms Perfektiv and Imperfektiv. Delbrück 1897:1^ § 636, Brugmann
1904, Behage1 1924:93, Noreen 1904-1912:195, 607 and 645, Wellander
1965 e t c developed t he terms suggested by Str eitberg in t h e i r
descriptions of the verb phr ase in Germa nic languages. W e shal l draw
upon t h i s research by mak ing a d istinction between the semantic
category o f Aktionsart 'procedural' and th e morphosyntactic category
of Aspekt as proposed by Agrell 1908:78.
195

34. There is no cl ear-cut process of derivation for the reflexive pronoun


i n VISA (SIG) ' t u r n o u t 1 . I f t h e r e f l e x i v e pronou n i s c o n s i d e r e d t o be
an integral part o f the verb, then no problem will a r i se . I f , however,
the reflexive pronoun is regarded to be a transformational product,
then the question of the underlying st ructure will arise.

35. In addi tion to ordinary cyclic rules (such as Passivization, Subject


Raising, Reflexivization etc) some transformati onal rules have been
claimed to operate before the cycle (precyclical rules), after the
cycle (postcyclical rules) (cf. Ross 1967 a and Postal 1970a),and a t
any poin t in a derivation (so-called "anywhere rul es") ( c f . Ross 1967a),

36. Lakoff 1971bhas recently suggested a new device t o explain the genera­
tion of certain well-formed surfa ce structures that cannot be described
or derived by l ocally o perating transformational rules in the sense of
Chomsky 1965» h i s s o - c a l l e d " t r a n s d e r l v a t t o n a l c o n s t r a i n t s " .

37- Cf. Chomsky's 197 1b mo st recent paper on Conditions on Transformations


w h e r e h e a t t e m p t s " t o co n s t r a i n t h e f u n c t i o n i n g o f g r a m m a t i c a l r u l e s
and thereby t o limit the generative power of grammars of a given form"
(cf. p. 4).

38. Bach I97I gi ves a survey of syn tactic research since the publication of
Chomsky's 1965 Aspects.

39. The term Subject Raising was propose d by Kiparsk y S Kiparsky 1970:160
in an early version of the ir a r t ic l e which was reproduced by the
I n d i a n a L i n g u i s t i c s C l u b i n 19 6 8 . Rosenbaum 1 9 6 7 a u s e s t h e t e r m
" p r o n o u n r e p l a c e m e n t " , Lang endoen 19 6 6 " e x p l e t i v e r e p l a c e m e n t " , a n d
Ross 1967a " it-replacement".

40. Jacobs&Rosenbaum 1967:9 and 14, and Jacobs & Rosenbaum 1968:193 suggests
t h e term verb phrase complement f o r such comp lements wh ich occur
d i r e c t l y u n d e r a VP n o d e i n a t r e e d i a g r a m , w h e r e a s a n o u n p h r a s e c o m p l e ­
ment i s c l a i m e d t o o c c u r u n d e r an NP n o d e . Because o f t h e s e m a n t i c
properties discussed in Footnote 21 i t i s not necessary t o postulate a
noun phrase the fact o r i t s pronominal form as preceding the
complement sentence Si in the deep str ucture representation (37a). The
embedded sentence i s therefore a verb phrase complement in our sense.

41. The inf initiv izatio n process has been des cribed by a number of genera­
t i v e grammarians w ho t r e a t the derivation of the i n f i n i t i ve from
various points of view. Bierwisch 1963:123 derives Ger man i n f in i t i v es
through a generalized transformation operation which ap plies a t a
two-string representation of the sentences involved in the embedding.
One- o f t h e two i d e n t i c a l s u b j e c t NP's i s d e l e t e d in t h e embedding
p r o c e s s . R os s 1 9 6 7 a a n d J a c o b s s Ro senb aum 19 6 7 u s e a n I t - R e p l a c e m e n t
R u l e i n O r d e r t o g e n e r a t e E n g l i s h i n f i n i t i v e s , w h er e b y t h e dummy i t i s
r e p l a c e d by t h e complement s u b j e c t NP and t h e i n f i n i t i v e o f t h e compl e­
ment v e r b r e s u l t s . R. Lakoff 19 6 8 i n t r o d u c e s t h e c o n c e p t o f Comple­
ment i zer-P lacement Rule w hich i nserts the complementizer that into
English senten ces and then an implausible Comple mentizer Ch ange Rule
which changes that into the infinitival marker t o under deletio n of
the identical complement sub ject ( c f . pp. 26-27). The l a t t e r soluti on
can be seen as a product of extreme syntac ticism a t a time when any
196

kind of syntact ic manipulation was acceptable in one of the camps of


Americ an g e n e r a t i v e gramm ar. Co mmon f o r a l l g e n e r a t i v e a p p r o a c h e s t o
the infinitive i s , however, the Two-Subject-Hypothesis.

42. The fac tor of syntactic interference from written Latin cannot be ignored
a s f a r a s t h e law t e x t s of Old Sw edish a r e concerned. I t i s known t h a t
the compilators and e d it o r s of the Old Sw edish laws knew Latin (see
Wessen 1968:20-23 and Ståle 1967:42-47 on t he e diting of the Old
S w e d i s h p r o v i n c e l a w s ) . T h e e d i t o r s w e r e a l s o f a m i l i a r w i t h R om a n l a w .
(See H olm 1952:283 and 1969» Ahlberg 1942, and Larsson 1931 on vario us
arguments f o r and a g a i n s t t h e i n f l u e n c e o f L a t i n and Lo w G erman on
literary Old Swedis h).

43. Cf. G r i mb e r g 1 9 0 5 : 2 0 5 a n d Nyga ard 190 5 : § 2 0 6 o n t h e AC I - C o n s t r u c t i o n i n


Scandinavian l a n g u a g e s , Bock 1931:220 and V i s s e r 1956:754 on t h e AC I -
C o n s t r u c t i o n i n E n g l i s h , a n d E r d m a n n 1 8 7 4 : 2 0 5 o n t h e AC I - C o n s t r u c t i o n
in German.

44. Both Alving 1918:1 and Beckman 1959 :252 a r e aware of t h e complement
s u b j ec t s t a t u s of t h e raised NP, while t h e p h i lo l o g i s t s and h i s t o r i c a l
l i n g u i s t s menti oned in F o o t n o t e 4 3 t r e a t b o t h t h e a c c u s a t i v e NP and
the infinitiv e a s equal object complements of the matrix sentence.
They do n o t s u g g e s t a Two-Subje ct-Theory a s A l v i n g and Beckman d o .

45. In Ame rican gene rative l i t e r a t u r e the same s y n t ac t i c deep s t r u c t u r e


i s o f t e n p o s i t e d b o t h f o r AC I - C o n s t r u c t i o n s a n d t h a t - e m b e d d e d c o n ­
s t r u c t i o ns . See f or instance Jacobs & Rosenbaum 1968:195 and 219»
R. Lakoff 1968:25-29 e t c ) .

46. The exa mples in IBM-Artisan c apital l e t t e r s presented in t h i s investiga ­


tion signify that the examples are actual sentences excerpted from the
Gothenburg corpus. The f i r s t two indices in parentheses denote the
number of the a r t i c l e computerized and the l i ne in the a r t i c l e . The
index before the slash is a serial number. The nu mbers following the
slash denote t he newapaper, genre, subject, and author (for further
details see the Introduction to Allén 1970a).

47. The question whether tense should be tr eated as a category of the


Aux symbol a s suggested by Ch omsky 1965:107, o r a s a f e a t u r e o f t h e
main verb a s proposed by Jacobs & Rosenbaum 1968:114, o r a s an inde­
pendent syntactic constituent having "a predicative function" as
suggested by Kiparsk y 196$:16, or finall y as "th e pronominal form
o f a time a d v e r b " in t h e s e n s e o f McCawley 1971 : i 11, cannot be
answered here. This question is notational and i t is sufficient here
t o c o n c l u d e t h a t t e n s e i s e v i d e n t l y n o t r e p r e s e n t e d i n t h e AC I " C o n ­
s t ru c t ion. If tense needs t o be expressed in the embedded sentence ,
then an embedding with a t t must take place.

48. C h o m s k y 1 97 1 b:9 c l a i m s t h a t l e x i c a l i t e r n s l i k e e a c h c a n n o t b e e x t r a c t e d
from tensed sentences, whereas such a movement o f each i s p o s s i bl e
from the matrix sentence into untensed sentences like infinitive con­
s t r u c t i o n s . Each-Movement can apply a t ( i ) t o generate ( i i ) :
197

(i) The candidates each expected [ the other to win]


Si Si

(ii) The candidates expected each ot her to win.

However, t h e same mo vement r u l e cannot apply a t ( i i i ) to yield (iv),


because ( i i i ) contains a tensed embe dded sentenc e:

( i i i) * The candidates each expected [ that the others would win]

(iv) * The candidates expected t hat each oth er would win .

49. This is probably a result of the "overwi 11ingness of present day


linguists t o assign identical meanings t o different surface s tructur es"
(cf. Chafe 1970:87) combined with the lingu ists' primary concern with
theory construction under negligence of empirical observations of
authentic sentences in n atural speech sit uations.

50. T h e n o t i o n o f f o c u s s u p p o r t s t h e c l a i m t h a t AC I - C o n s t r u c t i o n s h a v e
different semantic readings than compleme ntizer-embeddings. The raised
NP o b t a i n s a s u r f a c e s t r u c t u r e i n t e r p r e t a t i o n by a r h e t o r i c a l s u b ­
component as recently suggested by Katz 1972:420-427.

51 Through the kind as sistance of Prof S . Allén and hi s colleagues,


Gothenburg,the following contexts of (76) through (83) have bee n
excerpted from the Gothenburg corpus:
1) The linguistic context of (76a):Stative Aktionsart Neutralization
(VARA Deletion)
Det ä r s v å r t a t t i det här a r b e te t få något r i k t i g t grepp o m va d
Jaquette Liljecrantz betytt för kvinnosaken och de radikala idéerna.
Till större delen bygger fram ställningen på hennes egna uppgi fter i
d a g b öc k e r o c h b r o s c h y r e r . Man h a d e nog v e l a t s e dem v e r i f i e r a d e i
andra k ällor. Tonvikten ä r därför a l l t lagd vid de dramatiska om­
k a s t n i n g a r n a i h e n n e s l i v ( 1 3 7 9 : 4 4 - 5 4 : 5 / 1 2 5 0 2 6 ) ( G H T)

2) The linguistic context cf (76b): Ingressive Aktionsart Neutralization


(BLIVA Deletion)
Att s i t t a framför TV :n och se butik e f t e r butik fullproppade av köp­
medvetne Köpstarka tonårin gar med an poptonern a f y l l d e rummet som
någon sla gs kultmusik, var inte enbart en h andgriplig påminnelse
om a t t de egna ungdomsåren b ö r j a r b l i en smula avlägsna. (1321:25~3Î
-.6/515683) (SDS)

3) The linguistic context of (76c): Ingressive Aktionsart Neutralization


(BLIVA Deletion)
Först när han ser faktorerna kartmässigt framställda, när han kan
jämföra o l i k a kartogram bredvid v a r a n d r a , v a r t och e t t med s i n
fakt or karterad, f ö r s t då k larna r bilden, han s e r samspel och
sammanhang och han kan bygg a hypoteser och t e o r i e r om dem.
(954:48 -55:3/521684) (SDS)
198

The linguistic context of (77a): Stative or Ingressive Aktionsart


Neutralization (VARA or BLIVA Deletion)
I e t t brev t i l l partiets ordförande och koalitions-kab'netts-
vicekansler, Dr. Bruno Pîtterman, har sonen t i l l Österrikes s is te
k e j s a r e f ö rk l a r a t a t t han s e r s i g tvingad t i l l a t t v i d t a s p e c i e ll a
åtgärder som kan möjliggöra hans återvändande. Brevet röjer inte
v i l k a å t g ä r d e r Otto har i tankarna (975:170-179:2/533629) (SDS)

The linguistic context of (77b): Stative or Ingressive Aktionsart


Neutralization (VARA or BLIVA Deletion)
Grundtemat ä r en adelsmans f ö r f ö re l s e av en ung f l i c k a , som då
hon s er si g övergiven låter prostituera s i g , medverkar o f r i v i l l ig t
i e t t rånmord deporteras t i l l Sibirien och försonas i dödsögon­
b l i c k e t under färden t i l l f ä n g e l s e t med s i n nu u p p r i k t i g t å n g e r ­
f u l l e f ö r f ö r a r e (560:39-^6:1/226174) (SvD)

The linguistic context of (77c): Ingressive Aktionsart Neutralization


(BLIVA Deletion)
Sina 55 sekunder u t n y t t j a r Ture Nerman t i l l e t t y v i g t l o v t a l ö v e r
p l a n h u s h å l l n i n g e n , m e n s e r s i g o m g åe n d e v e d e r l a g d o c h d e t p å 4 5
sekunder (1318:106-110:5/522551) (SDS)

The linguistic context of (78a): Ingressive Aktionsart Neutralization


(BLIVA Deletion)
Fältet var väl samlat när master s läppte tempot f r i t t och f jolårets
jaktvin nare Carl Malmen hade ingen chans med s i n Sonette mot den
segervisse Lars La-Ger dahl som tog revansch me d s i n Nanette. 1964
f i c k ekipaget s e s i g s l a g e t med endast en nosläng d. Ändå v a r d e t
kanske "ål dermännen" Ragn ar Linde (nyss fy llda 80) som red r i d­
skolehästen Odea. (1567:^3-53:7/215272) (SvD)

The linguistic context of (79a): Ingressive Aktionsart Neutralization


(BLIVA Deletion)
De 14 inbjudna länderna fr ån engelska samväldet upprepade på on s­
dagen a t t de önskar se konferensen uppskjuten och a t t de inte
kommer a t t i n f i n n a s i g i Alger 29 j u n i . (874:11 -16:5/333371)(ST)

The linguistic context of (80a): Stative Aktionsart Neutralization


(VARA Deletion)
M en a l d r i g h a r m a n v ä l s e t t n å g o t s å v ä l k l ä t t o c h s å c h a r m e r a n d e
som Berta Hall i Fru Ingeborgs uppenbarelse. På balen utvecklar
hon en grace i dansen som teate rns unga f lickor kunde avun das
henne. ( 577:81-87:6/3 26294) (ST)

The linguistic context of (81a): Ingressive Aktionsart Neutralization


(BLIVA Deletion)
D e v a l u t o r s o m p å o l i k a s ä t t u n d a n h å l l i t s S t o r b r i t a n ie n b ö r j a r n u
strömma t i l l b a ka . Baissespekulanterna s e r sigtvunqna at t köpa pund
f ö r a t t komma i f r å n s i n a t e r m i n a t a g a n d en . Deras s p e k u l a t i o n h a r
s l a g i t s l i n t och k o s t a t dem dryga pengar. De ko mmer i n t e i f ö r s t a
t a g e t a t t v å g a s i g på l i k n a n d e a f f ä r e r . ( 1 3 8 5 : 8 4 : 1 / 1 2 4 0 2 2 ) (GHT)
199

11) The linguistic context of (82a): Stative Aktionsart Neutralization


(VARA Deletion)
Olika intressen drar å t var s i t t h å l l , i för sta hand markexploa ­
t ö r e r och n a t u r v å r d a r e , men o c k s å sommars tugeä gare - både redan
etablerade och sådana som ho ppas och v i l l - vidare den bofas ta
b e f o l k n i n g e n , s o m b e v a k a r k o m m u n e n s i n k o m s t i n t r e s s e n , me n o c k s å
f i s k a r n a , som s e r u p p e h å l l e t h o t a t a v n ö j e e f i s k e t . Det b l i r s e x
s j u g r u p p e r , som s k a l l jämka ihop s i n a å s i k t e r (48:40:5/215208)(SvD)

12) The linguistic context of (83a): Stative Aktionsart Neutralization


( VARA Deletion)
Två omständigheter b i d r a r t i l l a t t kampen mot de c a n c e r fa r l i g a
ämnena ham nat på e f t e r k ä l k e n . 1) Den " ä r l i g a oförmågan" hos
jämbördiga fors kare a t t enas om hur for sknin gsresultat skall t o l ­
kas, påtryckningen från i n d us t r i e r som s e r s i n a ekonomiska i n t r e s ­
sen hotade och den natu rliga benägenheten hos vetenskapsmän a t t
ge e f t e r för t r y ck e t . Det har hänt a t t forskare tving ats avgå
därför att de t rätt inflytelserika industrier.för när (1780:47:1/
511621) (SDS)

D.F. Aasen, Trondheim, has provided the information tha t in Norwegian


bokmå1 the circumlocution f å #ye på 'catch s i g h t o f ' i s often preferred
in a sentence l i k e (97d), although fikk s e i s not ung rammat i c a l . The
s i t u a t i on in Norwegian i s thus much mor e complicated than claimed h ere.
As i s t h e c a s e in Swedish, t h e r e seems t o e x i s t an o v e r t aspectual
marker a l s o in Norwegian whic h i s used t o express an indeliberate and
i n s t a n t a n e o u s p r o c e s s o f p e r c e p t i o n : F Å S E , F Å H 0 R E , F Å K J E N N E, a n d
FÅ F 0LE a s i l l u s t r a t e d by t h e f o l l o w i n g e x a m p l e s :

Norwegian FÅ SE ' c a t c h s i g h t o f '

(i) a. På avstand fikk jeg se e t ungt elskende par-


(At some distance I saw a loving young c ouple-)

Norwegi an FÅ 0YE P Å ' c a t c h s i g h t o f '

b. E t s t y k e b o r t e f i k k j e g tfye på e t u n g t e l s k e n d e par.
(At some distance I saw a loving youn g cou ple.)

Norw egian FÅ H 0RE ' s u d d e n l y h e a r '

(ii) a. Han f i k k hfóre henne s e t t e i e t s k r i k .


(He heard her scream o ut.)

Norwegian H0RTE ' s u d d e n l y h e a r d '


b. Plutselig hjórte han henne s e t t e i e t skr ik.
(Suddenly he heard he r scream out.)

Norwegi an FÅ KJENNE ' s u d d e n l y feel tactual 1y1


(i i i ) a . Men s å f i k k han e n d e l i g k j e n n e p u l s e n s l å .
(But then he fin ally f e l t the pulse beat.)
200

T h e r e i s a c o n s t r a i n t on t h e o c c u r r e n c e o f FÅ i n Norwegian whic h can ^


not be tr eated here, since a special investigation would be necessary
t o describe the difference between the.Swedish and Norw egian uses of
the overt aspectual marker.

53. In a detailed investigation o f English modal s Jenkin s 1972 has rebutted


Ross's 1967a mai n verb hypothesis of the syntactic s tatus of auxiliaries
because i t "receives no independent sy ntactic support and mus t be
abandoned in favor o f a theory which does not require these additional
modifications" ( c f . p . 3*0« The additional modifications a r e a number
of output constraints which, under the main verb hypothesis, are
necessary in ord er to block fo r instance the occurrence of twp m odals
a s in *1 may may co me. The double cooccurrence o f two m oJa I s i s
automatically excluded under the the auxiliary hypothesis by a phrase
s t r u c t u r e r u l e o f t h e f o l l o w i n g f o r m : S — » NP (Mod al) VP . The p o s i ­
tion of may in I ma y com e i s a l s o accounted f o r automatically by
t h e given p h r a s e s t r u c t u r e r u l e which a s s i g n s t h e modal i t s
post i t ion be fore the main ver b. Under the main verb hypothesis a rule
i s neede d which i n v e r t s t h e o r d e r o f * 1 come m ay t o I ma y com e. These
two examples of e x t ra machinery wh ich adherents of t h e main verb hypothesis
introduce illustra tes the inadequacy of Ross's claim t hat auxiliaries
are main ve rbs.

5^. The Duden 1959:922 p o s i t s a p s e u d o s t r u c t u r e from which t h e German A C I~


C o n s t r u c t i o n s a f t e r HÖ REN ' h e a r ' a r e a s s u m e d t o b e d e r i v e d . F o r t h e
sentence Karl hört seine Schwester d i e Lieder Schuberts singen two
sentences of kernel type a re posited: i) Karl hört seine Schwester
and i i ) S e i n e S c h w e s t er s i n g t d i e L i e d e r S c h u b e r t s . On s e m a n t i c grounds
kernel sentence(i) is uniikely,since i t is impossible to hear a person
without t h e immediate a s s o c i a t i o n o f a sound sourc e. By d e r i v i n g A C I -
C o n s t r u c t i o n s a f t e r HÖ R EN f r o m d e e p s t r u c t u r e s l i k e ( 1 4 8 ) s u c h a d - h o c
s t r u c t u r e s l i k e t h e one suggested by t h e Duden Grammar can be a v o i d e d .
(Cf. a l s o Bierwisch 1963:123 who p o s i t s t h e same type of kernel sentences
f o r h i s example Mein V a t e r h ö r t s i c h e t w a s U n e r w a r t e t e s s a g e n ) .

55« Mötsch I97O showed t h a t a number o f a t t r i b u t i v e ad j e c t i v e s in Ger man


cannot p ossibly be derived transformationally from a pr edicative
p o s i t i o n by t h e A d j e c t i v e Movement Rule under R e l a t i v e Pronn oun and
Copula Deletion, e.g. ein al ter Freund 'an old Friend' cannot be derived
from ein Freund, der a l t i st ' a friend who i s o l d ' .

56. Cf. a l s o Chomsky 1965:190 who p o i n t s o u t t h a t phrases l i k e t a k e f o r g r a n t ­


ed and take of fence must be treated as "single" lexical items.

57. Whitaker 1971 convincingly argues that " i t i s not acceptable t o remove
linguist ic hypotheses to an ab stract level outside the domain of
neurological corr elation and refutation" ( c f. p. x i i ) . Thus i f ther e i s neur
logical and psychological evidence that some struct ures are more
accurately treated as idioms than as transforms, then theoretical
l i n g u is t s should ta ke such empirical f a c t s into consideration when
constructing a theory of l i n g ui s t i c competence, t h a t i s under the
assumption t hat they want t o describe what a nati ve speaker know s
about his language. The curre nt wild theories in semantics as well
as in syn tax cannot be sai d t o reflect such l ing uistic competence, if
they a r e in c onflict with the findings of psycho and ne urolinguists
201

in their empirical observations o f lingu istic behavior. Such t heories


s u f f e r from being based to o mu ch o n i n t ui t i on and too l i t t l e on
empirical socio, psycho, and neur olinguist i c observations. (Cf. Bever's
I970:3^5 stand on the qu estion of competence and performanc e:"The
r e l a ti o n s h i p between l i n g u i s t i c grammar based on i n t u i t i o n and t h a t
based on th e de scription of other kinds of explicit language p er­
formance m ay not j u st be ' a b s t r a c t 1 but may be nonexistent in some
cases").

58. A v a l u a b l e corpus o f German i n f in i t i ve c o n s t r u c t ions which o c c u r a f t e r


verba se ntiendi, verba dicend i e t c has been c ollected by Be ch 1955 and
described within a stru ctur alist-glossemat ic framework. Som e exam ples
o f h i s abundan t e x c e r p t s from G erman l i t e r a t u r e have bee n used in P a r t
One and P a r t Two t o demonstrate s e n t e n t i a l complements a f t e r German
v e r b a s e n t i e n di .

59« Mötsch S Schädlich I965 derive the Swedis h impersonal passives occurring
a f t e r ÖNSKA ' w i s h ' , A N S E . ' c o n s i d e r ' , a n d S K R I V A ' w r i t e ' i n c o n s t r u c ­
tions l i k e Det önska s 'One wishes ' , Det ansåg s 'It
was considered 1 Pet skrivs 'They wr ite ' etc from the
following underlying representation which con tains a deep str ucture
s u b j e c t w h i c h i s s p e l l e d o u t a s m a n ' o n e ' m a n a n s å g d e t . . . .( d . 2 7 7 ) .
After Passivization, Agent Pe letion obligatorily applies to the
s t r u c t u r a l index w hich no w c o n t a i n s man a s t h e a g e n t :
Pet ansågs av man Tftg,=»Pet ansags Consequently, their
Agent P eletion Rule cannot apply u n t i l t h e l e x i c a l i n s e r t i o n o f man
has o c c u r r e d . In (162) and (16 3 ) t h e Agent P e l e t i o n Rule has a l s o
operated. By assu ming an a b s t r a c t i n d e f i n i t e human agent f o r such
sentences, however, no ref erence need be mad e t o the phonological
matrix, whereas such reference is implied in the Agent Peletio n
Rule propo sed by Mötsch S Schädlich.

60. Chomsky 1965 and Ch omsky 1970adiffe r a s f a r a s t h e treatment o f t h e


d e r i v a t i o n o f n o m i n a l s a r e c o n c e r n e d . W h e r e a s Cho m sky 19 6 5 s u g g e s t s
a derivation of nominals like destruction from a n underlying verb
( d e s t r o y , +V), Chomsky 1970aderives d e s t r u c t i o n from [destroy,+N]
(cf. p. 203). The l atte r notation is in our view already the product
of a recategorization o f the underlying verb .

61. The term "balanced view" has here been used t o make a d i s t in c t i o n from
Chomsky's unclear concept "lex ical i s t view". As indicated in footnote
60, Chomsky's ow n t reatment o f the word formation process involved i s
contradictor y. For som e nominals he suggests an ac tive transformational
rule process t o generate nominals l i ke the owner as being derived
from the one who o wns t h e house ( c f . Chomsky 1970:196-197)» whereas
he does no t t r ea t for instance the noun refusal in the same way , as
he does n ot assign any categ orial feature to a lexical entry like
refuse, which instead is claimed to be "fr ee with respect to the
categorial f e a t u r e [nom] and [ver b]" ( c f . p . 190). We f i n d t h i s
stand inconsistent and suggest instead a derivational morphological
process along the l i n e s of European l i n gu i s t s who have mo re c l e a r l y
formulated notions about the interaction of the lexicon and the
syn tactic component ( c f. Wurzel 1970:82, Kiefer 1970:164 and Kiefer
I972 on the derivation of nomina a c t i on i s and nomina agent i s ) . These
European l i n g u i s t s a r e now tryi ng t o formulate rules of word-forma-
tion which ar e correlated with the base rules and the transformational
202

62. The terminolo gy o f p r e d i c at e - c a l c u l u s has beco me f a s h i o n a b l e among


some lin guists in the l a st few years. However, l i t t l e is gained as
f ar as syntactic research i s concerned by c all ing verbs predicates
and noun s arguments. A generative semanticist or a logician whose
primary i nterest is semant i es, would claim that by wo rking with th e
concept of one, two, or three-place predicates, a more dire ct seman­
t i c representation i s attained. The burd en f a l l s upon him t o prove
that such a new level of representation i s motivated fo r syntact ic
s t r u c t u r e s . Compare Chafe 1970:96 who c o n s i d e r s i t awkward t h a t t h e
terms predicat e and argument have t o change t o verb and noun a t some
unmotivated p o i n t a l o n g t h e p a t h from s e m a n t i c t o su r f a c e s t r u c t u r e s .

63. McCawley 1972 r e p r e s e n t s a deve lopm ent w i t h i n American l i n g u i s t i c s


which is the outgrowth of extreme transformat ionalism, i .e . the
abandonment of syntactically motivated stru ctures for describing
surface structures. Lees I960 initia ted transformationalism by
d e r i v i ng a 11 nominal s by t r a n s f o r m a t i o n s . Lakoff 1965 was even more
daring by postu lating his so-called abstract verbs which oblig atorily
undergo nominalization s, e.g. *agress woul d under an ab stract syntactic
hypothesis be considered to be th e underlying verb o f agression. It is
t h i s view of the so-called abs trac t syntax ( c f . especially R. Lakoff
1968) which i s the s t ar t i n g point of generative sem antics. I t i s be­
lieved here, however, t hat the balanced view of describing nominals
is empirically more motivated than the wild s peculations of abstract,
nonexistent verbs along the lines of generative semant i ci s t s . Such an
approach t o the nominalization process in Germa nic language s a lso agrees
with the results gained in the theory of derivational morphology.

6k. Several treatments of nominalizations have bee n published besides


Lees I960 and Ch omsky 1965 and 1970a, e . g . Gruber 1967, Chapin 1967,
Fraser 1970b on nominalizetions in English, Wurzel 1970 on nominaliza-
tions in German, Kiefer 1970 and 1972, Teleman 1971, Urèland 1972a and
Linell I972 on nominalization in Swedish.

65. I am gratefu l t o B. Wik, Umeå, who pointed o ut t o me in a seminar held


on nominalizatio ns in November 1972 that the Latin term nomina a c t i on i s
as used in Ureland 1972a cannot be translated into English action nominal
since the latter term stands for a specific kind of construction in
English, e . g . John's riding of h i s bicycle. However, s i nee Swed ish lac ks
such a construction, the Latin term nomina actionis can be used t o denote
nominals like skal 1 'bark', mul1er 'thunder' etc without being mis­
interpreted.

66. No s a t i s f a c t o r y s o l u t i o n f o r d e s c r i b i n g t h e embedd ing me chanism h a s s o


far been presented for any language, although sever al attempts have
been made ( c f . Bierwisch 1963, Härtung 196^, Rosenbaum 1967a , R.La koff
1968, and Bresna n 1972). In s e c t i o n 1 3 ^ a proposal i s made f o r a s s i g n ­
ing compleme ntizer feat ures to the embedding sy mbol which i s generated
by the base rules.
67. In Söderwall 1918:229 there i s one exam ple of perfective-aspectual use
o f O l d S w e d i s h F A : f i n g h o m w i bra dhelika h ö r a r ö s te r ( L g 3 : ^ 6 1 : S ö w 2 2 9 )
'we suddenl y heard voices'. Söderwall 1918 paraphrases this use of Old
Swedish F A by "komma i t i l f ä l l e a t t " 'have t h e o p p o r t u n i t y t o ' . (Cf.
also the following sentenc e from th e province law of Östergötland:
nu far hin J>œt höra sum œrfft hauœr at hun œr hauande{'6qL Œ . 7:G3^9)
(Now t h e o n e who h a s a c c e p t e d t h e i n h e r i t a n c e h e a r s t h a t s h e i s p r e g n a n t )
203

68. T h e S AO B 1 9 3 9 d i s c u s s e s KÄ NNA2 ' s m e l l 1 c o n c o m i t a n t w i t h KÄNNA-j ' f e e l


t a c t u a l 1 y 1 a n d KÄNNAi* ' f e e l p h y s i c a l l y ' . I t i n f o r m s u s t h a t KÄNNA c a n
also be used to express perceptions of smelling and fe eling (cf.
column K 3 648)

69. KÄNNA m a y i n S w e d i s h a l s o o c c u r a s a v e r b o f a c t i v e p e r c e p t i o n a s
i n d i c a t e d w i t h r e s p e c t t o KÄNNA P Å ' t o u c h ' i n f o o t n o t e 2 7 . H o w e v e r ,
t h e s i m p l e x KÄNNA a l s o e x p r e s s e s a c t i v e p e r c e p t i o n a n d t a k e s a n A g e n t
a s deep s u b j e c t in a sentence l i k e Han kän de i a l l a f i c k o r me n fann
i n t e en enda s l a n t (SAOB 193 9: K365TI 'He f e l t in a l l h i s p o c k e t s b u t
did not find a single coin'.

70. In t h e phrase Känn d i g s j ä l v ! 'Know t h y s e l f ! ' känn can be regarded


as a deleted s tructure of Lär känna dig själ v! The Ag ent r ole of
the deleted subject is clearly distinguishable here.

71. Part Two, sec tion 10.

72. Cf. also footnote 49.


PART TWO: COMPARATIVE GERMANIC SYNTAX OF VERBA SENTIENDI
204

INTRODUCTION TO PART TWO


Part Two of the present investigation of verbal complementation
in Germanic languages comprises a trilingual approach, in which
a number of verbal complements after Swedish, German, and
English verba sentiendi are described within a comparative syntactic
framework. Emphasis has thereby been put on the two sentential
complements which result after Subject Raising, that is the ACI
and AWG-Constructions.

By tr eating these two surface structures as being generated by


a single general syntactic rule which is common to the three
Germanic languages involved in this description, an important
basis for interlingual comparisons is established. The goal of
Part Two is to demonstrate the extent of the syntactic domain
of the Subject Raising Rule in Swedish, German, and English
by examining the language-specific level constraints which are
necessary to block ill-formed verbal complements which occur
after verba sentiendi.

To undertake the comparison of a specific set of surface


structures in one language with a corresponding set of surface
structures in another language is by no means novel in the
history of linguistics. Such comparisons were the concern of
comparative philologists and historical linguists of the past
century (e.g. Rask 1818a, Bopp 1816, Grimm 1819, Verner 1877 etc).
Their primary interest was historical comparative phonology or
morphology with the final goal of describing genetic relation­
ships between the languages of Europe and Asia. The reconstruc­
tion of a Proto-Germanic and a Proto-Indo European 'Ursprache'
was also a product of their comparative efforts. The term
comparative came to be synonymous with diachronic-comparative
in tnis earlier research.

However, the comparison between different languages can of


course be carried out for synchronic purposes as well. There
exists alongside the diachronic tradition in Europe a synchronic
tradition which emphasizes interlingual studies, e.g. Kröger
1914, Jespersen 1969 etc. The synchronic comparative approach
of these linguists was often overshadowed by the diachronic
approach of the proponents of the neogrammarian concept of
language, because historicism was the more spectacular and
predominant interest at the beginning of this century.

Since a multilingual approach can break down the specific


language bias which often influences both the conceptualization
and the formalization of rules in a given language, a compara­
tive approach to syntactic surface structures in more than one
language serves as a valuable empirical basis of an inter­
lingual description to the linguist.

In the last fifteen years the term oontrastive grammar has often
been used for such multilingual studies. It has served as a
cover term for a multitude of interlingual descriptions which
have been carried out within very disparate theoretical frame­
works. The term contrastive has been adopted by for instance
205

linguists who represent structuralist thinking e.g. Kufner 1962,


Moulton 1962, Dahlstedt 1966, Malmberg 1967 etc, by linguists
who represent structuralist-generative thinking e.g. Stockwell
& Bowen & Martin 1965, Nickel 1971a, Nickel 1971b, Oksaar 1972,
Hammarberg 1972 etc, by linguists who represent generative-
transformational thinking e.g. Ellegård 1971, Köning & Nickel 1970,
Köning 1971, Krzeszowski 1971, Wyatt 1971 etc, and even by
linguists who represent stratificational thinking e.g. Snook
1971.

Therefore the term contrastive does not imply any special


linguistic theory but it has been used for denoting all kinds
of interlingual descriptions. The fact that so many theoretical
frameworks have been used is the strength of contrastive grammar.
All approaches to interlingual descriptions are justified, since
there is no theoretical apparatus which can come to grips with
the complexities involved in the simultaneous description of
two, three or more language systems. At the present time we are
witnessing a world-wide development of interlingual analysis.
Interlingual approaches to the study of language is a fruitful
area of linguistic research. This research has barely been
started. The theoretical tolerance which characterizes the
majority of linguists who work within interlingual studies can
result in a stimulating exchange of ideas.

In the present investigation the term contrastive syntax has not


been used for reasons which will be discussed further in section
12.5. As d emonstrated there the term contrastive too clearly
implies an orientation to the surface structure to be appropriate
as a denotation for the generative type of interlingual descrip­
tion which is attempted here.

Instead the term comparative synchronic syntax will be used for the
description of sentential complements in Swedish, German, and
English, because there are a great number of instances where the
surface structures are identical in the three languages under
study.1 To use the term contrastive would not be meaningful
under such circumstances. Our term comparative synchronic syntax
can, however, be used with respect to similarities as well as
dissimilarities. It is just as important for a linguist to
demonstrate in his description that the domain of a given
syntactic rule is equally extensive in two different languages
as it is to demonstrate that the rule domain is unequal in
extent. A one-sided concern with surface structure dissimilar*
ities may even present a distorted picture of the way syntactic
rules operate in the languages under study, although such
concern for surface structures may be motivated from a peda­
gogical point of view. (Cf. Nickel's 1971aand James's 1972
discussion of the pedagogical advantages of contrastive
analysis).,

The generative-transformational approach applied here to the


sentential complementation systems of Swedish, German, and
English is focused more on dynamic syntactic processes and their
constraints than on surface-structure phenomena. We are striving
not only to c ompare different verbal complements in Swedish,
206

German, and English, but we also wish to compare the various


transformational rules and level constraints which give rise
or block the verbal complements under study; they are primarily
ACI and AWG-Constructions and complementizer embeddings.
Comparative synchronic syntax as conceived of here is more
abstract than the majority of analysis carried out within a
contrastive syntactic framework.

It is a natural descriptive procedure for us to posit common


Germanic deep structures as underlying a number of language-
specific verbal complements. These varying language-specific
surface structures can be generated by a set number of trans­
formational and morphological rules, if certain deep structures
are assumed to be common to the synchronic description of the
three Germanic languages discussed here. It is necessary to
formulate the abstract Germanic deep structures as generally as
possible and to find common syntactic transformations such as
Subject Raising, Complementizer Insertion etc in order to enable
us to express what constitutes the common syntactic core of the
three Germanic languages. Language-specific characteristics
such as varying lexical entries, varying constraints on syntac­
tic rule application, varying morphological and phonological
rules can then be described in fairly general terms.

In classical comparative philology the postulation of abstract


underlying proto-structures was a means to reconstruct historical
sound changes and to establish a theory of the genetic relation­
ship between the Germanic languages. The interest was then in
historical linguistics, but in modern lingustics the interest
has mostly been in typological, socio, psycho, and neuro-
linguistics. However, the recent rediscovery of abstract under­
lying structures for describing surface structures by Chomsky
and others has been important for developing the present
comparative synchronic approach to the syntax of Germanic
languages. The syntactic base rules are to be formulated as
universally as possible to be of use for interlingual studies.
The syntactic deep structures are therefore assumed to be of a
language universal character. Now and then linguists have
suggested that there exist such universal deep structures
which underlie language-specific surface structures. Chomsky
1965:27-30 discusses such deep structures and assumes that they
are of a syntactic nature. In recent theory these universal
base structures have become more and more abstract until they
now resemble more closely the formulas of predicate logic
(cf. for instance Lakoff 1971a, McCawley 1972, Keenan 1973 etc).

Since our interest is not semantics but rather the syntax of


three natural languages, we are not motivated to begin with
underlying representations of too abstract semantic nature,
but we will instead test the first assumption made by Chomsky
1965 that common underlying syntactic structures can be used
for describing the surface structure differences between
different languages in terms of varying syntactic operations
and constraints on rules.
207

0, PRESENTATION OF ACI-VERBS IN THREE GERMANIC LANGUAGES


The Subject Raising Rule which generates ACI-Constructions
exists in all Germanic languages, modern as well as a ncient.
The rule which moves the subject noun phrase from the constituent
sentence into the matrix sentence can be looked upon as an over­
all Germanic syntactic rule, the constraints of which are the
task of the present investigation to explore. We have demonstrat­
ed some of these constraints for generating well-formed ACI-
Cons tructions in Modern Swedish. In section 1. comparisons will
be made between Swedish and German ACI-Constructions on one
hand, and English ACI-Constructions on the other. As ca n be
expected the Subject Raising Rule is constrained in different
ways in the three different Germanic languages focused upon
here. The advantage of a generative transformational approach
for describing syntax will be demonstrated in the comparison
of the ACI-surface structures of Swedish, German, and English.
We will be able to show that the syntactic domain of the Subject
Raising Rule is considerably wider in Swedish than in German
or English. The constraints on Subject Raising in sentences
containing verba sentiendi w ill be discussed.

The domain of the same rule in Latin will also be demonstrated


by comparing Latin ACI-Constructions with the Germanic ACI-
Cons tructions in Swedish, German, and English. As will be shown
the domain of the Latin Subject Raising Rule is wider than that
of any Germanic language, also that of Swedish (cf. Part Three).
It is important to demonstrate such a difference in syntactic
rule domain between Latin on one hand, and Germanic language's
on the other, since throughout the history of Germanic languages
there has been a considerable syntactic rule interference from
written Latin on the written languages of Western Europe.

In order to show that the Subject Raising Rule is not only an


Indo-European rule, the corresponding structures of Germanic
ACI-Constructions will be treated in North Lapp and Standard
Finnish, two Finno-Ugric languages. Sometimes ACI-Constructions
are possible in these two Finno-Ugric languages (especially
after verba dioendi and verba causativa J , but sometimes a
Nominalization of the constituent verb and a Genitivization
of the constituent sentence subject occurs instead of Subject
Raising (that is after verba sentiendi ) (cf. section 13.2).

The subclassification rules worked out for subdividing Swedish


ACI-Verbs (cf. Part One section 3.3) can be used for sub­
dividing German and English ACI-Verbs, since the same class
notation for the categories of verba sentiendi is useful for
comparisons between the three Germanic languages.

The subclassification of the Swedish, German, and English ACI-


Verbs in Table 1 has been made on the basis of both syntactic
and semantic characteristics. The starred verbs are such verbs
that only in a literary style or under special linguistic
context conditions may take ACI-Constructions. A distinction
between different styles must be made in order to account for
the fact that in a formal style of speaking or writing ACI-
208

Constructions are acceptable, whereas in a colloquial style of


speaking and writing ACI-Constructions are less acceptable.
Especially verbs like English LISTEN TO (Class la), LOOK AT,
OBSERVE, WATCH, WITNESS (Class 2a), German BEMERKEN, MERKEN
(Class 2b), and Swedish IAKTTAGA, OBSERVERA (Class 2a), MÄRKA,
UPPTÄCKA (Class 2b) are style-sensitive as far as the ACI-
Construction is concerned. The need to make a distinction
between different styles was also felt by Nygaard 1905 who
indicated that Old Icelandic ACI-Constructions appeared more
frequently in learned style ("las rd stil") than in popular
style ("folkelig stil")(cf. p. 2).

FIGURE 1
COMPARATIVE TABLE OF VERBA SENTI END I IN SWEDISH, GERMAN, AND ENGLISH
ACI-Class (A)

[+V]
[+Subj Rais]
[-Fact]
[+Perceptive]

ACI-Subclasses
(la) (lb) (2a) (2b) (3a) (3b) (3c) (3d)
[+Process] t+Process] [+Process] [+Process] [+Process] [+Process] [+Process] [+Process]
[+Auditory][+Auditory][+Visual] [+Visual] [+Tactual] t+Olfact ] [+Mental] t+Somatic]
[•Purpose] [* Durâti ve]t+Purpose] [-Purpose] [^Durative]
t+Durative][iDurative]

FÅ HÖRA-, IAKTTAGA . FA SE KÄNNAi KÄNNA2 KÄNNA3 KÄNNA4


HÖRAi OBSERVERA* SE; 'feel 's me11' 'feel 'feel
MÄRKA tactually' mentally' somatically1
UPPTÄCKA*

BEMERKEN* FÜHLEN!
MERKEN* SPÜHRENi
SEHEN x 'feel tactually'

ENG: LISTEN TO HEARX LOOK AT ßtnuLiJ:


BEHOLD rtELi
FEEL-i FEEL3**
OBSERVE? DISO
DISCOVER* PERCEIVEi PERCEIVE3**
WATCH* NOTICE*
NOTI
WITNESS* SEEi

Starred verbs take ACI-Constructions which occur only in a certain literary style
Only Accusative-with-TO-Infinitive
209

1. MORPHO-SYNTACTIC AND SEMANTIC ASPECTUAL FEATURES IN


SWEDISH, GERMAN, AND ENGLISH
Comparisons between the ACI-Constructions of Swedish, German,
and English give perspectives of the domain of the Subject
Raising Rule in Swedish which otherwise would remain
unilluminated. It has proved practical to treat the Swedish and
German surface structures together and describe the corresponding
English structures in contrast, since English raises a special
problem with its Accusative-with-Gerundive (AWG) Construction;.
The obligatory character of the AWG-Construction considerably
limits the number of well-formed ACI-Constructions and thus
constrains the domain of the English Subject Raising Rule in
its generation of ACI-Constructions which occur after verba
sentiendi. The progressive gerundive form is not obligatory in
Swedish or German (cf. (2c) and (3c)), although it may be used
in a certain literary style (cf.(2d) and (3d)). Sometimes the
AWG-Construction may, for some unknown reason, even be ill-
formed in Swedish and German.2

It is a well-known fact that the AWG-Construction in English


expresses a continous^ noninstantaneous, and noniterative mode
of action, i.e. durative Aktionsart, on the part of the underlying
complement subject. That is,the role of the complement subject
must be of a continous character to take the ING-suffix. Compare
the following sentences;

(1) a. I saw him run to the store every day.


[+Iter] [+Iter]
b. I saw him cross the street-
[+Mom] [+Result]
*
c. I saw him run down the street.
[+Mom] [+Dur]
d. I saw him running down the street.
[+Dur ] [+Dur]

The durative, [+Dur], mode of action in the complement sentence


of (Id) must be morphologically manifested by an ING-suffix
attached to the verb, if a well-formed sentence is to result in
English. Compare the ill-formed *(lc), where this morphological
condition is not met.

In English only under special conditions can a plain ACI-


Construction result after Subject Raising, if the verb of per­
ception ( heard or saw ) cooccurs with an iterative mode of action
[+Iter], in the complement sentence. Compare (la) where the
time adverbial every day specifies the semantic mode of action of the
complement verb run to be iterative, although no surface mani­
festation is visible in the verb. The Progressive Suffix Rule
is blocked here, because of the aspectual semantic feature [+Iter],
as it is in (lb), where the contextual resultative mode of
action, [+Result], likewise blocks the outspelling of an ING-
suffix on to the complement verb. Thus the verb of perception,
sawj in (1) can in a given context be interpreted as either an
210

iterative verb (cf. (la)) or as an instantaneous verb (cf. (lb))


or as a durative verb of perception (cf.(Id)) due to the Aktionsart
of the complement sentence.

Whether semantic aspectual features are transferred from time


adverbials as in (la), or across the sentence boundaries in a
given linguistic context as in (lb), semantic features of
Aktionsart play a crucial role in the generation of well-formed
ACI-Constructions as exemplified in these sentences.

The necessity of including semantic features for describing such


aspectual cooccurrence restrictions on verba sentiendi (SEE, HEAR,
OBSERVE etc) and complement verbs has long been recognized by
European philologists and historical linguists as evidenced by
the works of Nygaard 1905, Callaway 1913, Poustma 1929:804,
Jespersen 1954b:178-214, Wood 1956 etc. The recent "discovery"
of including semantic features for describing syntax by Kiparsky
& Kiparsky 1970 within the theory of generative grammar is a
discovery only to those linguists who worked within the genera­
tive framework to the exclusion of others and whose primary
concern during the sixties was the description of syntactic
processes without semantic considerations, i.e. in the days of a
syntactically-based generative theory of grammar.

Thus, under special circumstances such as those of an iterative


or resultative mode of action implied by the complement sentence,
English ACI-Constructions can be generated by the Subject
Raising Rule. However, there is a considerable morphological
and semantic difference between the English AWG-Construction
on one hand, and the Swedish and German ACI-Constructions on
the other. Compare (1) with (2) and (3).

(2) a. Jag såg honom springa till affären varje dag


b. Jag såg honom gå över gatan.
c. Jag såg honom springa nedför gatan.
d. Jag såg honom springande nedför gatan.

(3) a. Ich sah ihn jeden Tag in den Laden rennen.


b. Ich sah ihn über die Strasse gehen.
c. Ich sah ihn die Strasse hinunterrennen.
d. Ich sah ihn die Strasse hinunterrennend.

To be able to account for the explicit durative semantic reading


of the AWG-Construction in English in (1) and the lack of such
an explicit aspectual reading in the Swedish and German ACI-
Constructions in (2) and (3) respectively, two different under­
lying cognitive structures must be posited, i.e. phrase marker
(4a) for English and phrase marker (4b) for Swedish and German:
211

FIGURE 2

(4a) (4b)

S s

In English it is crucial that the complement verb must be


marked as either durative or nondurative, i.e. with the features
or the features respectively. Such marking is
morphologically redundant in Swedish and German as represented
by the plain [+V]under Si in phrase marker (4b). The optionality
of a durative marking of the Swedish and German verb matrix as
demonstrated by (2d) and (3d) belongs to a literary style. The
feature DURATIVE may b e regarded as a semantic aspectual feature
with morphosyntactic consequences. In American generative
literature no clear distinction is made between morphosyntactic
and semantic aspectual features. Chomsky's 1965 Aspect-category
which is expanded from the Auxiliary symbol in the base rules
must be seen as a morphosyntactic category (cf. p. 107).
Since not all languages have overt markers for Aktionsart like
the English ING-suffix,4 it is more adequate to keep the base
rules as universal as possible by avoiding special categorial
symbols for surface structure phenomena like morphological
endings, and let such language-specific manifestations be the
task of the morphological component (cf. Bierwisch 1967,
Wurzel 1970, and Kiefer 1970). Following the practice of
European philologists and linguists, the semantic aspectual
feature DURATIVE is here considered to be a semantic deep
structure feature of Aktionsart, whereas the morphologically
manifested suffixes are treated as m orphological surface
structure characteristics of Aspect spelle d out by the morpho­
logical component (see the works on Aktionsart and Aspekt in
Meyer - Lübke 1899:39 , Agrell 1908:78, Pollak 1960:203,
Isaîfenkc 1960 , Forsyth 1970 etc). The existence of a morpho­
logical rule like (5) which spells out the ING-suffix onto
durative complement verbs in English gives rise to t he
considerable difference between English on the one hand,and
Swedish and German on the other (see the historical explana­
tion of the rise of the English ING-construction as described
by Curme 1912, Mustanoja 1960:572-573, and Visser 1969:1065-
1069).5
212

English Progressive Affix Rule

(5) 0 /ING/ /^ur]_

If the durative feature triggers the outspelling of the present


participle suffix in Swedish, /-NDE/, or the present participle
suffix in German, /-ND/, according to morphological rules (6)
and (7), then a surfacc structure results both in Swedish and
German which is r eminiscent of the English ING-construction.
Compare (Id) with (2d) and (3d).

Swedish Progressive Affix Rule

(.» „ _ /ml/ j [:v_r] _

German Progressive Affix Rule

(7) 0 — /ND/ j gur]-

The basic difference between English on one hand and Swedish


and German on the other is that in English the Progressive
Affix Rule is obligatory, whereas it is optional in Swedish and
German, at least in the (d) sentences of (2) and (3).

The advantage of distinguishing between semantic aspectual


features and morphological characteristics as suggested by t he
European philologists and historical linguists referred to
above, can be seen from sentences like

(8) a. I saw him crossing the street.


[+Dur] [+Dur]

b. I saw him cross the street.


[+Mom] [+Result]

c. I spotted him crossing the street.


[+Mom] [+Dur]
*
d. I s potted him cross the street.
[+Mom] [+Result]

Both the durative mode of action expressed by crossing i n (8a)


and the resultative mode of action expressed by cross in (8b)
can be combined with the instantaneous, [+Mom], mode of visual
perception in saw or with the potential durative mode of action
which can also be expressed by saw. In *(8d), where the matrix
verb spotted expresses an instantaneous mode of action, no
complement verb of resultative mode of action can occur."
213

It seems adequate to regard the deep structure semantic con­


straints exemplified in (8) and in section 2. below as being
more fundamental for the generation of well-formed ACI-
Constructions than a mechanical enumeration of abstract
syntactic verb categories in the sense of Chomsky's 1965
Aspects -theory. A pure syntactic description of ACI-Constructions
in Germanic languages without the inclusion of semantic aspectual
specification of the deep structure modes of action in the
matrix and complement sentences will never yield satisfactory
results in generating well-formed ACI-Constructions.
214

2. ASPECTUAL CONSTRAINTS AND THE PROBLEM OF LEXICAL INSERTION

Serious work on the ACI-Construction in Germanic languages can­


not exclude semantic viewpoints in describing the cooccurrence
restrictions holding between the matrix and complement verbs.
A linguist is confronted with the problem of lexical insertion
of verbs expressing varying modes of action, i.e. verbs of
resultative, instantaneous, durative, inchoative, and
terminative Aktionsart. A linguist interested in writing a grammar
of the ACI-Construction in Germanic languages may wish to write
a grammar which is not too powerful. He does not want a grammar
whose rules generate well and ill-formed surface structures
alike in the same unconstrained way. Ill-formed cooccurrences
of matrix and complement sentences nave to be avoided.
Conditions on the insertion of particular matrix and complement
verbs must therefore be formulated, so as to block the generation
of ill-formed ACI-Constructions as exemplified by the following
Swedish sentences:

(9) a. *iian observerade barnen tappa mjölkkrukan«


[+Dur] [+Mom]
(He observed the children drop the can of milk,)
*
b. Han märkte Peter falla ned fran taket.
[+Mom] [+Dur]
(He noticed that Peter fell from the roof.)
* o
c. Han iakttog henne köra pa hans bil.
[+Dur] [+Mom]
(He watched her collide with his car.)

Because of an Aspectual Constraint on the cooccurrence of


certain matrix and complement sentences, Subject Raising cannot
apply in sentences which contain OBSERVERA 'observe', MÄRKA
'notice', or IAKTTAGA 'watch' as m atrix verbs, unless the mode
of actiçjn in the complement sentence is durative or iterative.
The sentences in (9) in which this semantic aspectual constraint
is violated are therefore ill-formed, whereas in (10), where this
requirement is met, no violation results.
(10) a. Han observerade några barn leka på skolgården
[+Dur] ODur]
(He observed some children playing in the schoolyard)

b. Han märkte Peter komma emot sig med en kniv


[+Mom] [+Dur]
(He noticed Peter approaching him with a knife)

c. Han iakttog sin kollega skriva på maskin


[+Dur] [+Dur]
(He watched his colleague writing on a type writer)
215

(10) d. Han observerade skolbarn komma ut på gården.


[+Dur] [+Mom]
(He observed school children enter the school yard.)

e. Han observerade skolbarn komma ut på gården varje


I + Iterl [+ Iter]

morgon-
(He observed school children enter the school yard
every morning.)

Whether one works within a generative semanticist's framework


or within the Extended Standard Theory as sketched by Chomsky
1970a, 1970b, and 1971athe requirement of aspectual well-
formedness of the cooccurrence of matrix and complement
sentences will be imperative. Crucial for the generation of
well-formed ACI-Constructions is thus the choice of matrix and
complement verbs. In neither of the two current generative
theories is there explicitly stated in what order and under
what circumstances main verbs are inserted into the different
phrase markers of complex sentences as in the following phrase
marker:

FIGURE 3

(ii) s

It is clear from the ill-formed examples of (9) and the well-


formed examples of (10) that the inherent semantic characteristics
of both the matrix sentence verbs and the complement sentence
verbs are decisive for generating well-formed ACI-Constructions.7
By e xtending Chomsky's 1965 subcategorizational mechanism of
rules to include aspectual semantic features as well, so that
features for modes of action can govern the insertion of main
verbs, the lexical insertion problem can be solved without
radically altering the Standard Theory.
216

(12) [+Durative]

[+Mom]

[+V]—> < [+Result]

ClIngressive]
Alterative]

Such a mechanism is a n essential improvement of the Aspects-


subcategorizational mechanism, but it is still too powerful
for generating well-formed ACI-Constructions because a whole set
of modes of action can appear in the matrix and complement
sentences without any constraint whatsoever. Several deep
structure constraints of the type suggested by Perlmutter 1971
are necessary to delimit the lexical insertion of verbs which
carry varying modes of action, so that the Subject Raising Rule
will not generate ill-formed ACI-structures like (9). The concept
of the constraint is an important innovation for describing
syntactic structures. Some kind of a semantic filter is needed
to block the insertion of certain complement verbs whose mode
of action is semantically incompatible with such matrix verbs
as d emonstrated by (9). Under the assumption that the matrix
verb is inserted first, the choice of complement verb is
governed by the mode of action of the perceptual matrix verb.
Compare the ill-formed occurrence of instantaneous complement
verbs in the (b) sentences with the well-formed occurrence of
durative complement verbs in the (a) sentences in (13):
MÄRKA 'notice* .

(13) a. Han märkte sina krafter avta långsamt.


[+Mom] [+Dur]
(He noticed that his strength decreased slowly.)
*
b. Han märkte Peter klippa till honom.
[+Mom] [+Mom]
(He noticed that Peter hit him}

UPPTÄCKA 'discover'

(14) a. Han upptäckte sin vän stå ute på bryggan.


[+Mom] [+Dur]
(He discovered his driend standing on the bridge.)
*
b. Han upptäckte henne komma in i rummet.
[+Mom3 [+Mom]
(He discovered her coming into the room.)
217

For the Swedish verbs OBSERVERA 'observe', IAKTTAGA 'observe,


watch', MÄRKA 'notice', and UPPTÄCKA 'discover' it is of grea^
grammatical consequence which mode of action the complement verb
expresses. If the choice falls upon an instantaneous matrix verb
of perception such as MÄRKA 'notice' or UPPTÄCKA 'discover', the
deep structure Aspectual Constraints block the insertion of all
complement verbs that are not durative, that is no embedding by
the Subject Raising Rule can take place if there is semantic
incompatibility between the matrix and complement sentences as
far as t ne mode of action is concerned. The (a) sentences of
(13) -(14) are well-formed, because they fulfil the aspectual
requirement of well-formedness so that Subject Raising can occur
and yield well-formed ACI-Constructions. If the durative matrix
verbs OBSERVERA 'observe' or IAKTTAGA 'watch' are selected, then
only those complement verbs may be inserted which also express
a durative mode of action. Compare the well-formed (10a) and
(10c) above^ where this aspectual requirement is met, with the
ill-formed (9a) and*(9c), where the aspectual constraint is
violated.notice also that iterative complement verbs occur
(cf. (10e)).
The following filters are therefore suggested as blocking
devices in the generation of ACI-Constructions in Swedish(and
in German when Subject Raising is possible):

(15) Aspectual Constraint on the Insertion of Durative Verbs and the


Application of the Subject Raising Rule

t
S0
[+DUr| —
VP
t
Sl
EIdUr
kiTE5!vp
]
sx
]
s0

OBSERVERA
IAKTTAGA

(16) Aspectual Constraint on the Insertion of Instantaneous Verbs and,


the Application of the Subject Raising Rule

[ Fv i
[+MOM] — C r1+DURJ
+v i
] ]
S0 VP S-, VP s, s„
o 1 lo
MÄRKA
UPPTÄCKA

Filters (15) and (16) refer to inherently durative and instanta­


neous verbs respectively, as indicated under the matrix verbs.
Filter Constraints (15) and (16) do not block the insertion of
the aspectually ambiguous verbs SEj 'see visually' and HÖRA^
'hear auditorily' which are marked as +V
f+DURl
l+MOMJ
218

Perhaps it would be more adequate to have these two verbs of


perception unmarked as f ar as Aktionsart is concerned and copy-
in aspectual features from the linguistic contexts, time
adverbials or complement sentences. In the classification
table of Figure 1 two aspectual features have been written in the
verb matrices of HÖRA and SE, since these verbs are aspectually
ambiguous. In any case, they must be marked in such a way so
that they are not blocked from being inserted by aspectual
constraints (15) and (16).

Deep Structure Aspectual Constraint (15) stipulates positively


in the sense of Stanley 1967 that durative matrix verbs

EV ~T s
DurJ 1 * ^eeP
Structure Aspectual Constraint (16) stipulates that perfective,
^

nondurative matrix verbs like UPPTÄCKA 'discover' and MÄRKA


'notice' may only cooccur with durative complement verbs.
Notice that the structural description of the two positive
conditions on the lexical insertion of verbs into complement
sentences specifies the embedding through the Subject Raising
Rule. If another rule of embedding is selected, the two
constraints formulated in (15) and (16) do not block the
insertion of a nondurativé verb after OBSERVERA, or the insertion
of an instantaneous verb occurring after UPPTÄCKA or MÄRKA:

(17) a. Han observerade att barnen tappade mjölkkrukan.


(He observed that the children dropped the can of
milk.)

b. Han märkte att Peter föll ned från taket.


(He noticed that Peter fell down from the roof.)

c. Han upptäckte att hon kom in i rummet.


(He discovered that she entered the room}

The well-formedness of cooccurring matrix and complement verbs


can here be explained by the fact that a tense feature exists in
the complement sentence. There is an interesting correlation
between tense and Aktionsart, which is little known. Current
theories of tense as proposed by Hofman 1966, Kiparsky 1969,
Wunderlich 1970s and McCawley 1971 are inadequate for describing
the problem ot tense and Aktionsart. The fact that ACI-
Constructions are untensed seems to make them much more sensitive
to aspectual constraints as formulated in (15) and (16). (Cf.
also section 10. below on the role of tense in the choice of
that -Embeddings and Accusative-with-To-Infinitives after FEEL3
and PERCEIVE3 'feel mentally').

The complement verb seems to influence the Aktionsart of the


matrix verb in (17), so that the inherently durative OBSERVERA
undergoes an Aspectual Reeategorization from denoting lexically
a durative process of perception, to d enoting an immediate
process of perception in this linguistic context. The
instantaneous mode of action in tappade mjölkkrukan 'dropped the
can of milk' is copied upwards in the phrase marker by a kind
of agreement rule similar to gender and case agreement rules.
219

FIGURE 4

Han observerade

+v
+Past
_+Mom_
barnen tappade mjölkkrukan

TASPECTUAL
V COPYING

Han observerade Nsy

.9

+Pas t

barnen tappade mjölkkruka

Another possibility would be to treat observerade 'noticed' in


(17a) as belonging to a different verb cl ass, since it denotes
a different mode of perception, that is bli varse, OBSERVERA2
'become aware of' instead of vara varse, OBSERVERA, 'be aware
of' (cf. Part One, section 11,2).
220

3, ASPECTUAL COPYING RULE AND VARIABLE SYNTACTIC RULES IN


SWEDISH AND GERMAN

The domain of the Subject Raising Rule in German is very


restricted after BEMERKEN 'notice', MERKEN 'notice', and
ENTDECKEN 'discover'. German informants hesitate about the
well-formedness of both the (a) and (b) sentences in (19) through
(21), rejecting the (b) sentences completely, but finding the
(a) sentences more acceptable for some reason. The (a) sentences
in (19) and (20) are, however, excerpted from authentic literary
texts, so their grammaticality is, at least in literary style,
beyond question.°

BEMERKEN 'notice'
(19) a. Er bemerkte sich vor der Dame vom Hause einen Dank murmeln .
[+Mom] [+Dur]
(G. Freytag. Soll 154:Bech 162)
(He noticed that he murmured a thank you before
the house wife.)
*
b. Er bemerkte sich eine Verbeugung machen .
[+Mom] [+Mom]
(G. Freytag. Soll 154:Bech 162)
(He noticed that he made a bow.)

MERKEN 'notice'
(20) a. Ich merkte nicht Menschen neben mir vordrängen.
[+Mom] [+Dur]
(S. Zweig. Phantast. Nacht. 150:Bech 162)
(I did not notice that people were pushing past me.)
*
b. Ich merkte nicht andere Hände Geld hinwerfen.
[+Mom] [+Mom]
(S. Zweig. Phantast. Nacht. 150:Bech 162)
(I did not notice that other hands threw money.)

ENTDECKEN 'discover'
(21) a. Er entdeckte einen Freund auf der Brücke stehen.
[+Mom] [+Dur]
(He discovered a friend standing on the bridge.)
*
b. Er entdeckte sie gegen seinen Wagen fahren.
[+Mom] [+Mom]
(He discovered that she drove into his car.)

The German ACI-Construction of (20b) occurring after MERKEN has


been starred because it appears in too limited a linguistic
context to be felt as well-formed by native speakers of German.
The cooccurrence of the instantaneous verb hinwerfen 'throw' and
MERKEN will, however, become well-formed in a more extensive
linguistic context as in the following:
221

(22) Ich merkte nicht Menschen neben mir vordrängen, andere Hände wie
[ +Mom] [+Dur]
Fühler sich -plötzlich vorstrecken9 Geld hinwerfen oder einkarren.
C +Dur] [ +Dur ] [+Dur]
(S. Zweig. Phantast. Nacht. 150:Bech 162)
(I did not notice that people were pushing past me, that
other hands than mine were stretching out like tentacles,
throwing or gathering in money.)

The durative interpretation of vorstrecken 'stretch1forward',


hinwerfen 'throw1, and einkarren 'gather in' is due to the durative
Aktionsart of the first complement verb vordrängen 'push past',
which is signalled through the dative case after the preposition
neben 'beside'. A copying rule for changing aspectual features of
complement sentences (the lower embedded sentences) in agreement
with the aspectual feature of the topmost complement sentence
is needed in order to capture the generality of aspectual
copying as evidenced in (22). Therefore, the operation of the
Aspectual Copying Rule is assumed to have occurred between the
shallow phrase marker (23) and the surface phrase marker (24)
as indicated below:

FIGURE 5

\T+v I
|_+MomJ
merkteN

ILI
vordränge]

vorstrecken*

hinwerfet
ASPECTUAL
COPYING
RULE

einkarrei
S,

1+Diir)
vorstreckei

l+Duq
hinwerfeir
222

If a rule of embedding other than Subject Raising applies, then


the Aspectual Constraint (16), which is valid for both Swedish
and German, will not block the cooccurrence of nondurative
German verbs like BEMERKEN 'notice', MERKEN 'notice', and
ENTDECKEN 'discover' with instantaneous mode of action in the
complement sentence:

BEMERKEN 'notice'

(25) Er bemerkte, dass er eine Verbeugung machte.


(He noticed that he made a bow.)

MERKEN 'notice'
(26) Er merkte, dass andere Hände Geld hinwarfen.
(He noticed that other hands threw money.)

ENTDECKEN 'discover'
(27) Er entdeckte, dass sie gegen seinen Wagen fuhr.
(He discovered that she drove into his car.)

Two semantic Aspectual Constraints, (15) and (16), have been


formulated for restricting the cooccurrence of matrix and
complement verbs after the Subject Raising transformation in
order to block the generation of the German (b) sentences
in *(19) through*(211 and the Swedish (b) sentences in *(13)-
*(14), *(9a) and *(9c). in Swedish, the aspectual constraints
are more clear-cut, since (14a) and (13a), (10b) plus sentence
(10c) are all acceptable sentences for native speakers of
Swedish. The corresponding German sentences in (19) through
(21), however, are somewhat more controversial, as several
German native speakers do not accept them as well-formed,
although literary evidence for the existence of such ACI-
Constructions has been presented above. It does not matter for
such speakers of German whether the Aspectual Compatibility
Requirement is met or not, in their idiolects of German no
ACI-Constructions are possible at all after the verbs BEMERKEN,
MERKEN, and ENTDECKEN. In their lexicon these verbs are quite
simply not marked as being able to undergo Subject Raising.

It is a known fact that the constraints on syntactic rules vary


from one speaker to another within one and the same language.
Judging from the literary evidence that ACI-Constructions exist
after these verbs, it would be inadequate and hence wrong to
ignore the existence of such ACI-Constructions in a description
of German ACI-Constructions (cf. Bech 1955:162 who includes
these three verbs in his discussion of the verbs which take
German infinitives ).

However greatly the Aspectual Constraints (15) and (16) may


vary among different speakers of German, it is a fact that
Swedish and German share similar conditions on the operation
of the Subject Raising Rule, when German well-formed ACI-
Constructions do appear after this category of verba sentiendt.
223

To state it differently, the German Subject Raising Rule is not


constrained inliterary as it is in colloquial style in sentences
which contain BEMERKEN, MERKEN, and ENTDECKEN, as matrix verbs.
One could also claim that the German Subject Raising Rule seems
to be more generally restricted than the corresponding Swedish
rule which operates on complement sentences occurring after
MÄRKA 'notice' and UPPTÄCKA 'discover'. The domain of a syntactic
rule can thus be said to extend more widely or less widely in
one language or another. The greater the number of lexical items
(verbs) which are marked for the rule, the wider the paradigmatic
syntactic domain must be claimed to be, quantitatively.
Consequently, we would not mark BEMERKEN, MERKEN, and ENTDECKEN
for Subject Raising in colloquial German as we would do for
literary German. In Swedish such marking can be carried out for
both literary and colloquial style. The same marking is necessary
for Swedish OBSERVERA 'observe' and IAKTTAGA 'watch', whereas
German BEOBACHTEN 'watch' cannot cooccur at all with ACI-
Constructions; neither in literary nor in colloquial style:

(28) a. *Er beobachtete die Kinder auf dem Schulhof spielen.


(He watched the children playing in the school yard.)

b. *Er beobachtete die Kinder eine Milchkanne fallenlassen.


(He watched the children dropping a can of milk.)

A syntactic description of the ACI-Construction in Germanic


languages must also include various stylistic levels which
constitute important variables in the judgement of well-formedness.
Such consideration.of stylistic levels is necessary to obtain an
adequate treatment of the conditions on the syntactic rule which
generates ACI-Constructions, the Subject Raising Rule. Suggestions
for incorporating stylistic levels into linguistic description
have been made by Labov 1970, primarily for describing variables
of pronunciation. Syntactic rules which are dependent on certain
social variables such as t hose mentioned above have also been
suggested. It is in performance-oriented investigations that the
need for less categorical types of rules than those formulated
by most generative grammarians are to be found.(Cf. also Nygaard's
1905:2 clear distinction between popular style ("folkelig stil")
and learned style ("laerd stil")).
224

4, OVERT AND COVERT ASPECTUAL MARKERS I N GERMANIC LANGUAGES

A further comparison of the most frequent verba sentiendi in Germanic


languages, the two verbs of visual and auditory perception, shows
us a typological difference between the North Germanic languages
and the West Germanic languages. Some Scandinavian languages have
an overt marker for the ingressive mode of action.9 Compare
Swedish FÅ which cooccurs as an aspectual marker primarly with
verbs which denote seeing, hearing, and feeling i.e. FÄ SE 'catch
sight of', FÄ HÖRA 'hear, be told', and FÂ KÄNNA 'feel'.

As examples (29) and (30) demonstrate, no aspectual constraint


on the cooccurrence of matrix and complement verbs seems to block
the ability of the Subject Raising Rule to generate ACI-Constructions:

FÅ HÖRA 'hear'
(29) a. Han fick höra henne tala illa om sin väninna.
[+Ingr] [+Dur]
(He heard her speaking evil of her girl friend-)

b. Han fick höra (hörde) henne skrika till.


[+Ingr] [+Mom]
(He heard that she yelled out.)

FÅ SE 'see, catch sight of'


(30) a. Han fick se henne stå där nere på gatan.
[+Ingr] [+Dur]
(He saw her standing down there in the street.)

b. Han fick se (såg) henne köra på hans bil.


[+Ingr] [+Mom]
(He saw that she drove into his car.)

Both durative and riondurative complement verbs can cooccur in


well-formed Swedish ACI-Constructions with the ingressive [+Ingr]
matrix verbs, FÄ HÖRA 'hear' and FÂ SE 'catch sight of'.

Past tense forms of HÖRA and SE, hörde 'heard' and såg1s aw' are
also possible in (29b) and (30b), because the complement verbs
express the instantaneous mode of action. The latter occurrence
of HÖRA or SE is well-formed, since the two perceptual verbs
are aspectually ambiguous and obtain the aspectual reading
according to the Aktionsart of the complement sentence.

However the occurrence of FÄ as an aspectual marker seems to


be limited to the main verbs SE 'see' and HÖRA 'hear', and
KÄNNA 'feel', since the inherent durative verbs of perception
such as OBSERVERA 'observe' and IAKTTAGA 'watch' yield ill-
formed cooccurrences with the aspectual marker FÂ; probably
because there is semantic incompatability between the modes of
action inherent in the aspectual marker FÅ on one hand, i.e.
instantaneous [+Mom] Aktionsart, and the durative [ +Dur]Aktionsart
inherent in the two latter verbs of perception on the other
hand.10
225

(31) a. *Han fick observera barnen leka på skolgården-


(He observed that the school children were playing in
the school yard.)

b. *Han fick iakttaga eleverna skriva på maskin.


(He watched the pupils writing on typewriters.)

A cooccurrence of the ingressive marker FÅ and inherently


instantaneous verbs of perception such as MÄRKA 'notice' and
UPPTÄCKA 'discover' is not semantically incompatible, but is
redundant, so the following sentences are therefore ill-formed;
compare the ill-formed compound verbs with the well-formed single
verbs in the (a) sentences.

(32) a. Haii*£ick märka > sina krafter avtaga långsamt,


isärkte [
(He noticed that his strength was slowly decreasing.)

b. Han I*fick märkal henne komma in i rummet.


i märkte J
(He noticed that she entered the room}

(33) a. Han J*fick upptäcka I sin vän stå ute på bryggan.


1 upptäck te J
(He discovered his friend standing on the bridge.)

b. Han I*fick up ptäcka I henne komma in i rummet.


\ upptäckte J
(He discovered that she entered the room.)

In (32) afid (33) an overt aspectual marker meaning 'the be­


ginning of an act of perception' fick is ungrammatical, irre­
spective of the mode of action in the complement sentence.

If interpreted as a main verb in the sense of 'was allowed to'


(31a) and (31b) can be grammatical. We are here confronted with
two different categories of Swedish verbs: FÄ,, the ingressive
aspectual marker, and FÅ., the main verb of permission (cf. the
discussion of FÄ SE. and FÅ HÖRA. in section 5.4 and 13.6
respectively of Parr One).

The ingressive aspectual marker FÂ which cooccurs with HÖRA, SE,


and KÄNNA 'feel' under given semantic constraints is
characteristic of Swedish, whereas in Other Scandinavian languages
(Danish and Icelandic) except perhaps Norwegian,FÅ as overt
aspectual ingressive marker is unknown.H

German and English, on the other hand, have other means than that
of an overt aspectual marker to express 'the beginning of a
process of perception'. This ingressive aspectual meaning may in
these two languages be expressed by either the past tense form
of the verb of perception (hörte and heard, sah and saw)* or the
226

choice of another lexical item as i n the case of German


ERBLICKEN 'catch sight of'. A third possibility to express
ingressive Aktionsart is by means of a circumlocution as in
English aatoh sight of or German gewahr werden. In Swedish one
could also use få syn på and b li varse to express a similar
ingressive process of seeing something.

HÖREN/ HEAR

(34) a. Er hörte sie ihre Freundin verleumden. =Sw. (29a)


(He heard her defaming her girl friend.)

b. Er hörte sie aufschreien. = Sw.(29b)


(He heard that she yelled out.)

SEHEN/ SEE
(35) a. Er sah sie da unten auf der Strasse stehen. = Sw. (30a)
(He saw her standing down there on the street.)

b. Er sah sie gegen seinen Wagen fahren. = Sw. (30b)


(He saw that she drove into his car.)

ERBLICKEN 'catch sight of'


(36) a. Er erblickte sie auf der Strasse.
(He caught sight of her on the street.)
(Han fick syn på henne på gatan»)

As the past tense form of HÖREN/ HEAR and SEHEN/ SEE in (34b)
and (35b) indicate, the ingressive mode of actionis expressed
sufficiently by a simplex verb, whereas Swedish may have a
compound verb consisting of the ingressive marker FÅ plus either
of the two verbs of perception, HÖRA or SE.

The following German sentences, most of which have been excerpted


from literary texts, are translated into both English and
Swedish for the sake of syntactic comparison in order to
demonstrate the difference between overt and covert aspectual
markers in the three Germanic languages:

HÖREN 'hear'
(37) a. Ich hörte meine Mutter durchs Haus gehen.{E. Glaeser. Fried. 98:
[+Dur] [+Dur]Bech 139)
(I heard my m other walking through the house.)
[+Dur] [+Dur]
(Jag hörde min mor gå genom huset»)
[+Dur] [+Dur]
227

(37) b. Er hörte eine Tür in der Nähe zuschlagen.


[+Ingrl [+Mom]
(He heard a door slam close byO
[+Ingr] [+Mom]
(Han fick höra en dörr slå igen i närheten.)
[+Ingr] [+Mom]
i
hörde
[+Ingr]

Under the assumption that the hearing process occurred during


the whole course of the mother's walking from the beginning to
the end in (37a), then a translation by the Swedish simplex
verb hörde is t he only possible verb form. A translation by
fick höra here would under these circumstances yield a wrong
interpretation, since' the compound verb fick höra signals that
the hearing process starts somewhere during the mother's walking,
but not explicitly from the beginning to the end as in (37a).
The German verb form hörte and English heard must therefore be
aspectually ambiguous, as there is no overt marker to signal
whether the durative or ingressive aspectual mode of action is
implied by the matrix verbs in (37a) and (37b). However, a
more complete linguistic context will disambiguate such a two­
fold aspectual reading and will always yield the mode of action
intended by the author of the text.*2

The Aspectual Copying Rule suggested above in section 2. must


be modified so as to be suitable for formalizii ^ the description
of a correct interpretation of the matrix verb hörte' heard1 in
(37a). The aspectual characteristics of the preceding sentence(s)
in the same text will automatically yield the aspectual reading
which is intended by the writer for hörte in (37a).

The following sentences contain the linguistic information which


is necessary for the aspectual interpretation of (37a):

(37) a'. Ioh hatte das Fenster meines Zimmers geöffnet... Die Strasse war
leer. Selbst das Licht aus den anliegenden Häusern war abgedämpft..,
Ich hörte meine Mutter durchs Haus gehen; ihr Schritt war klar
und bestimmt; sie schloss die Lädent während Kathinka das Tor
abriegelte.(E. Glaeser.Fried. 98: Bech 139)
(I had opened the window to my r oom... The street was
empty. Even the lights from near-by houses were sub­
dued. I heard my m other walking through the house;
her steps were clear and distinct; she closed the
shutters while Kathinka locked the gate«)

It is clear from this linguistic context that the listener heard


how his mother walked through the whole house closing one window
after the other. The process of hearing goes on contemporaneously
with the walking through the whole house and the closing of all the
windows. To be able to transfer this durative Aktionsart into
English, the English paraphrase has the progressive suffix attached to the
228

complement verb, i.e. walking. However, if the result of the


hearing is more in the focus of attention then the English
translation by an ACI-Construction is the only possible verb
complement: I heard my mother walk through the house. In Swedish this
resultative Aktionsart can be signalled by the auxiliary FÅ:
Jag fick höra min mor gå igenom huset. The latter Swedish sentence is
aspectually ambiguous and can also mean that the listener
suddenly heard his mother walking through the house.

In (37b) no such ambiguous aspectual interpretation of hörte


is possible because the complement verb, zuschlagen 'close',
expresses an instantaneous mode of action, [+Momj. Both the matrix
verb fick höra and the simplex verb hörde are possible as
Swedish translations of German hörte here.

We see that Swedish hörde can occur both as a durative matrix


verb (cf. (37a)) and as an ingressive or resultative matrix
verb (cf. (37b)). The latter verb is often expressed by fick
höra to avoid aspectual ambiguity.

There is a parallel case in (38a) where sah 'saw' can yield


a simplex verb translation into Swedish såg 'saw' under the
assumption that visual perception is contemporaneous with the
Aktionsart of the complement sentence as follows: as the general
was watching the soldier> the soldier was standing before him.

SEHEN 'see'
(38) a. Der General sah einen einfachen Soldaten von etwa
[+Dur]
fünfundzwanzig Jahren vor sich stehen.
[+Dur]
( B.Kelleraiann Nov.l03:Bech 151)
(The general saw a private soldier about twenty-five
years old standing before him.)
(Generalen såg en menig soldat på omkring tjugofem år
stå framför honom.)

However,another reading of the German past tense sah is also


possible, given a linguistic context which implies a sudden
perception on the part of the experiencer, that is the general
in this case, in the matrix sentence of (38a). In the Swedish
translation this ingressive Aktionsart can be marked by the
overt ingressive aspectual marker FÅ(SE):

(38) b. Der General sah einen einfachen Soldaten von etwa


[+ Ingr]
fünfundzwanzig Jahren vor sich stehen.
I + Dur|
(The general saw (caught sight of) a private soldier
about twenty-five years old standing before him}
(Generalen fick se (fick syn på, blev varse) en menig
soldat på tjugofem år stå framför honom.)
229

If we refer to the sentence in Kellermann's novel Der 9.


November, p. 103, we find that the preceding paragraph gives
us a clue to the aspectual interpretation of sah in (38a).
There is no way in (38a) to decide if durative or ingressive
Aktionsart is intended here by Kellerman, if only the isolated
sentence is our basis of information. The verb of perception
sah is aspectually ambiguous. In order to d isambiguate such a
construction the relevant text is given:

(38c) (i) Ein weiter, grauer Soldatenmantel flatterte durch sie


hinduroh! (die strategische Ostlinie/ S.U.)...(ii)
Seit Wochen schon war ihm dieser Mantel aufgefallen,
und zwar nur, weil er so merkwürdig flatterte, wie kein
Mantel sonst, (iii) Obschon er immer nur einen Zipfel
dieses Mantels verschwinden sah, konnte er doch fest­
stellen, dass es der Mantel eines gemeinen Soldaten war...
(iv) In besonderen Stimmungen hatte er in dem Flattern
dieses Mantels sogar etwas Herausforderndes erblickt.

§2 (v) Diesmal aber lief ihm der Mantel direkt in die Hände. §2

§2 (vi) Der Soldat kam näher, und nun, da er den Schritt ver­
langsamte, sah der General, dass er das eine Bein etwas
nachschleppte, (vii) Der weite Mantel stand an der Wand
still, wie alles, was sich hier bewegte, wenn der General
in Sicht -kam. § ^

§ ^(vi i i)Der General sah einen einfachen Soldaten von etwa fünf­
undzwanzig Jahren vor sich stehen, mittelgross, breit­
schulterig, mit schlichten» für sein Alter auffallend
ernsten Zügen, (ix) Was dem General aber besonders an dem
Gesicht auffiel, das waren die Augen. (x) Sie waren braun
und ausserordentlich sanft. § .

The key sentences (i) through (x) of the coherent text in Keller­
mann 1921:103 have been rendered here to illustrate, on one hand,
how the author conveys meaning and the reader uses a certain kind
of interpretation stategy to assign the correct reading to der
Mantel 'the coat' in (vii), and on the other hand, how the ACI-
Constructions in (iii) and (viii) (=(38a)) are disambiguated
by the linguistic context as far as the Aktionsart is concerned.

Due to th e preceding sentence (ii) which contains a verb of


repetitive Aktionsart, i.e. flatterte 'flapped', the sa/z of (iii)
may b e interpreted as expressing iterative or ingressive
Aktionsart, although durative Aktionsart is not exclu ded: 'saw that
a piece of the coat disappeared'or 'fick se en bit av kappan
försvinna'.

To disambiguate sah in (viii) (=(38a)) it is a lso the Aktionsart


of the preceding sentences which in some way give pertinent
information on the interpretation of the verb of visual per­
ception. Under the assumption that a rule of metaphor inter­
pretation has already operated on the noun phrase der weite Mantel
'the broad coat1 in (vii), so that this pars pro toto expression
is understood as der Soldat, then it is clear in the given context,
230

that the sah in (viii) occurring in the succeeding paragraph must


be of durative Aktionsart, since the main verb phrase in (vii)
stand an der Wand is a verb phrase expressing a state of being.
An additional circumstance must be mentioned, however, which
causes a durative, stative, interpretation of sah in (viii).
That is the expression in the immediately preceding subordinate
clause of (vii): der General in Sicht kam. Since,according to the
given text, everything came to a standstill in the presence of
the general, the object of his visual perception is already
in front of the general as the general sees him.

Therefore, it is not the Aktionsart of all the verb phrases in the


preceding third paragraph, §3, kam näher 'came closer', den
Schritt verlangsamte 'slowed down', sich bewegte 'moved' etc which
can be said to yield the aspectual interpretation of sah in
(viii). What gives the direct clue to the durative aspectual
interpretation of sah in (viii) is the verb phrase in (vii)
der weite Mantel stand an der Wand still. Through the selectional
restriction feature present in stand which requires a human being
to be the subject noun phrase, plus the contextual rule of
metaphor interpretation of der weite Mantel, the reader must first
interpolate the pars pro toto relation; then the aspectual
durative Aktionsart feature can be copied into the German sah
of (38a).

In order to illustrate the Copying of Aspectual Feature in a


linguistic context together with interpretation process of
metaphors the following text marker is given in Figure 6 below.

FIGURE 6
//TEXT(J__
Paragrapn ^ §2 Paragraph §2 §3 Paragraph §3 §4 Paragraph §4

Mantel\
(i)Ein Soldatenmante (v) lief ihm in die^
(ix),'der General sah einen,
y/ flatterte ^ Hände ^
/ ' e. S. vor sich stehen

,(ii) war ihm dieser Mantel aufgefallei ^*(vii) der weite Mantel stand an der Wand stilly

(iii) er... einen Zipfel dieseìr-Mantels (viii) wenn der General in SicKt^-kam^
' verschwinden sah ^

in dem Flattern dîeSes Mantels


231

The i-index Noun phrases refer to noun phrases which have


different morphosyntactic manifestations (ein Soldatenmantel,
dieser Mantel, einen Zipfel dieses Mantels, dem Flattern dieses Mantels, der
Soldati einen einfachen Soldaten) but all of which have the same referent,
that is, der Soldat 'the soldier'. The circled NP's are combined
with each other by a track to indicate that only the connection
of the circled NP's in Si, S2, S3, S4, S5, and S7 with the
circled NP of S6 will yield the correct interpretation of these
NP's, that is both NP's preceding the NP under S(j and NP's
following the same NP under S(, must somehow be interrelated by a
rule of metaphoric interpretation.

The -aspectual copying is represented by a broken line going from


the verb phrase under S2 to S3 which copies in the ingressive
feature and from the verb phrase under S7 to S9 which copies in
the durative feature.

The reader of (38a) must read the entire page, i.e. p. 103 in
Kellermann's Der. 9, November, in order to understand the pars pro
toto relation holding between the NP's of S1 through Sg. He
must also have read sentences (vi) and (vii) in the text in
order to be able to assign the aspectual feature intended by
Kellermann.

It is evident that a linguistic theory which does not account for the
text-theoretical relations which exist between the paragraphs
and sentences for a given text will fail to d escribe what a
native speaker knows about his language, be it as a n author or
a reader of a text. Our demonstration of the reader's
competence to interpret metaphors and his strategies for
assigning Aktionsart from one sentence to a nother in (38c), are
supporting evidence for those linguists who are elaborating an
explicit text theory (cf. Isenberg 1968, Ihwe 1971, Petöfi 1971a
etc).
232

5, VERB COMPLEMENTS OCCURRING AFTER TACTUAL VERBS IN


SWEDISH AND GERMAN

The importance of making the proper choice of embedding rule


(Subject Raising or att/dass Comple mentizer Embedding),
demonstrated above in section 3. is further illustrated by-
sentences (39) through (41) which contain the Swedish and
German verbs of tactual and mental percepticm, i.e. KÄNNA-,
'feel tactually', KÄNNA3 'feel mentally.'., FÜHLEN^ 'feel
tactually', FÜHLEN3 'feel mentally', SPURENj 'feel tactually',
and SPÜREN3 'feel mentally'. These verbs cooccur with ACI-
Constructions as well as with att/dass-Embeddings acc ording
to semantic conditions to be discussed here. The index 1
denotes a verb of tactual perception, whereas the index 3
denotes a verb of mental perception. (The reason for indexing
the latter verb in this way is that, index 2 denotes 'smell' in
Swedish KÄNNA2 which also takes the ACI-Construction). A
subclassification into the two groups of verbs is motivated
on syntactic as well as semantic grounds. Only the verbs
indexed as 1 can undergo Subject Raising, cooccur with ACI-
Constructions, and obtain the semantic reading of physical,
tactual perception as illustrated by the following sentences:

KÄNNA^ 'feel tactually'


(39) a. Han kände hjärtat slå fort.
[+Dur]
(He felt his heart beating fast.)

b. *Han kände flickan vidröra honom.


[+Mom]
(He felt that the girl touched him.)

FUHLEN^ 'feel tactually'


(40) a. Er fühlte das Herz klopfen.
[+Dur]
(He felt his heart beating)

b. *Er fühlte das Mädchenlihn [berühren.


|sichj[ +Mom]
(He felt that the girl touched him.)

SPUREN1 'feel tactually'


(41) a. Er spürte das Herz schnell klopfen.
[+Dur]
(He felt his heart beating fast.)

*Er spürte das Mädchenlihn berühren.


i sich 'C +Mom]
(He felt that the girl touched him.)
233

The verbs of tactual perception in Swedish and German are


constrained in the same way as the verbs of visual and audi­
tory perception are as far as the Subject Raising Rule is con­
cerned. If Aspectual Constraint (16) is violated, then ill-
formed ACI-Constructions result as de monstrated by the
Swedish and German (b) sentences in (39)- (41). Only dura­
tive complement sentences can occur after the Swedish and
German verbs of tactual perception.

In Part One section 1 5 . 2 another semantic constraint i.e. the Equal


Subject Constraint was shown to be necessary to block ill-
formed Swedish ACI-Constructions after KÄNNA3 'feel mentally1.
If the Equal Subject Constraint on Subject Raising is violated,
ill-formed ACI-Constructions result in Swedish:

KÄNNA3 'feel mentally'

(4 2) a. GENOM S I T T O U P PK L ARA D E F ÖR HÅL L ANDE K Ä NN ER H A N S I G S T Å l SKULD


T I L L D . ( 1 2 5 7 : 1 6 3 : 2 / 4 2 6 4 28 ) ( D N )
(Through his indefinite relation he feels indebted to D,J

b. S OM V AN LI G F A T T I G MÄ N N I SK A K ÄN N ER H AN S I G T I L L H Ö R A K OM M U N IS T ERN A
(1384:232:3/126064)(GHT)
(As an ordinary poor person he feels that he belongs
to the communists)

c. A T T H AN I N T E K ÄNDE S I G H A N ÅG ON D EL I DE M ( 9 5 0 : 1 0 7 : 5/ 4 2 2 4 1 2 )
(DN)
(that he did not feel as if he had a share in them)

(43) a. * genom sitt ouppklarade förhållande känner han. henne-


stå i skuld till D. J

b. * som vanlig fattig människa känner han^ honom, tillhöra


kommunisterna. J

c. * att han^ inte kände henne^ ha någon del i dem.

If no deep structure coreference linkage is present between the


matrix and the complement subjects, then ill-formed ACI-
Constructions are generated as evidenced by (43).

In German no ACI-Construction, whatsoever can occur after


FÜHLEN3 'feel mentally', whether there is coreference linkage
between the matrix and complement subject or not. Either another
syntactic construction is selected, that is an adjectival
construction with a deleted copula {sein) in (44b), or another
matrix verb GLAUBEN 'believe', is selected with an infinitive
(zu haben) in (44 c), or a dass-Complementizer must be selected
in (44a).
234

FUHLEN? 'feel mentally'

(44) a. durch sein unerklärtes Verhältnis fühlt er,

{* sich in D:s Schuld stehen j}


r dass er D. etwas schuldig ist l

b. als gewöhnlicher Mensch fühlt er sich den Kommunisten


gehörig
•gehörig sein
< verbunden '
•verbunden sein
•gehören

dass er nicht[ glaubte daran Teil zu haben


^•fühlte sich nicht daran Teil haben
•fühlte nicht als ob er daran Teil hätte

The German sentences in (44) which are correct are no ACI-


Constructions, but syntactic constructions of different embeddings.
It is doubtful, if tnere is any reason to postulate a deleted sein
as underlying the construction with gehörig in (44b), since a
surface structure manifestation of the copula yields an incorrect
ACI-Construction (gehörig sein), the same refers to the past
participle verbunden, which cannot cooccur with the copula
(verbunden sein). A syntactic solution along the line of Bach 1967
seems tó yield-a more adequate description of the nonoccurrence
of the copula. The segmentalization of the copula cannot take
place in sentences like (44b). It is more adequate to design a
deep structure without the copula node, since it is obligatorily
deleted.here. In (44c) there is a problem of lexical insertion,
since FUHLEN cannot be inserted at all if an infinitival
complementizer is selected. A dass-Embedding must then take
place:

(45) dass er nicht fühlt, dass er daran Teil hat.

Although the Equal Subject Constraint is not violated here, no


reflexive ACI-Construction is acceptable •fühlte sich daran Teil
haben. An embedding through an als ofr-Complementizer is not
possible.13

What was said about the constraints on the Subject Raising Rule
in its generation of ACI-Constructions after FUHLEN^ is v alid also
for the constraints on the occurrence of ACI-Constructions after
SPÜREN^, and therefore need not be repeated here.

The Swedish verb of mental perception, KÄNNA3 'feel mentally', must


therefore be marked as taking the ACI-Construction, that is,
as containing the syntactic feature [+Subj Rais], and Subject
Raising can then apply if the Equal Subject Constraint is not
violated. If no coreference linkage exists between the matrix
and the complement subjects, then another embedding rule is
obligatory, that is, the a£t-Complementizer Rule (cf. (42) as
compared to (43)).
235

The German verbs of mental perception, FUHLEN3 and SPÜREN3 'feel


mentally' must be marked negatively as far as Subject Raising
is concerned, i.e. as [-Subj Rais], whether the Equal Subject
Requirement, which is so crucial for well-formed Swedish ACI-
Constructions,„is met or noj:. No ACI -Construction is possible
after German FÜHLEN3 anc* SPÙREN3 . For a general survey of the
partial constraint on Subject Raising in Swedish and the total
blocking of the same rule in German, compare the following
sentences:

KÄNNA3 'feel mentally1

(46) a. Han kände att hans plan var helt misslyckad.


(He felt that his plan was a total failure.)

b. *Han kände sin plan vara helt misslyckad.


*(He felt his plan to be a total failure.)

c. Han kände sig vara helt misslyckad som författare.


(He felt that he was a total failure as a writer.)

FÜHLENz 'feel mentally'

(47) a. Er fühlte, dass sein Plan ganz falsch war.


(He felt that his plan was a total failure.)

b. *Er fühlte seinen Plan ganz falsch sein.


*(He felt his plan to be a total failure)

c. *Er fühlte sich als Schriftsteller ganz misslungen sein.


(He felt that he was a total failure as a writer.)

SPÜREN3 'feel mentally'

(48) a. Er spürte, dass sein Plan ganz falsch war.


(He felt that his plan was a total failure.)

b. *Er spürte seinen Plan ganz falsch sein.


(He felt his plan to be a total failure.)

c. *Er spürte sich als Schriftsteller ganz misslungen sein.


(He felt that he was a total failure as a writer.)

If the copula segment is not regarded to be present in the deep


structure, then the blocking of *(47b) -*(47c) and *(48b) -*(48c)
may be carried out by a constraint on the Copula Segmentalization
Rule. The latter solution is to be preferred for reasons to be
discussed in section 9.
236

6, THE SUBJECT RAISING RULE IN ENGLISH AND THE GENERATION OF


ACI AND AWG-CONSTRUCTIONS (HEAR AND SEE)
If the domain of the English Subject Raising Rule is compared
to the domain of the same rule in Swedish and German, one will
find that in English the rule must be more narrowly constrained
in its generation of English ACI-Constructions, i.e. in
sentences containing the verbs of perception of class (A).
Some of the restrictions holding in English between the co­
occurrence of verbs of perception and the constituent sentence
verbs were mentioned in section 1. in connection with the
discussion of the aspectual restrictions in German and Swedish.
Since English poses a special problem with its Accusative-with-
Gerundive Construction, the specific constraints on the English
Subject Raising Rule require a special treatment. Swedish and
German are syntactically closer as far as the generation of the
ACI-Construction is concerned, and could therefore easily be
treated together in sections 1. through 5..

The perceptual verbs SEE and HEAR allow also in English the
greatest number of cooccurrences of verbs with varying modes
of action, Aktionsarten, expressed in the matrix and complement
verbs. The semantic incompatibility of certain matrix verbs
of perception with complement sentences will be discussed.

There is a general morphological rule which obligatorily spells


out the suffix ING and adds BE as an auxiliary verb, if the
mode of action is continuous, nonrepetitive, and nonhabitual etc
This language-specific morphological rule is a characteristic of
the English verb which neither the Swedish nor the German verbs
share. Such an obligatory rule considerably limits the number of
potential English ACI-Constructions which can occur as compared
to the number of potential Swedish and German ACI-Constructions.

To support this claim we can give the following examples:

(49) Han hörde henne spela piano.

(50) Er hörte sie Klavier spielen.

(51) *He heard her play the piano.

The Swedish and German ACI-Constructions in (50) and (51) have


already been shown to be fully grammatical in section 1.,
whereas the English ACI-Construction in (51) was found to be
ill-formed in certain contexts(to be di scussed below). The
sudden act of perception requires in English a verb form in the
complement sentence which expresses the continuous mode of
action on the part of the complement subject (her), i.e. a
gerundive ING-form.
237

However, the morphological rule which spells out the ING-


suffix (Rule(5) above), is heavily constrained in three ways
as far as the semantic characteristics of both the matrix and
complement verbs are concerned. First, the inherently non­
durative verbs of perception such as BEHOLD, DISCOVER, and
NOTICE (see subclasses (lb) and (2b) in Figure 1) and the
aspectually ambiguous HEAR and SEE cannot undergo the Progressive
Affix Rule (5):

(52) a. *1 am hearing a noise.


b. *1 am seeing a big moose.
c. *1 am noticing a beautiful girl.
d. *1 am beholding a strange thing,
e. *1 am discovering a new machine.

The semantic characteristics of immediate perception inherent in


these verbs forbid their cooccurrence with a progressive suffix
plus BE which expresses a continuous, durative process.

Secondly, inherently durative verbs of perception like LOOK AT,


OBSERVE, WATCH, and WITNESS do occur with the progressive form,
since these verbs are semantically compatible with the suffix
which expresses an action in progress (cf. Subclass (2a)):

(53) a. He was listening to the radio.


b. He was looking at the beautiful woman.
c. He was observing a man in the street.
d. He was walking in the airport watching lots of people,
e. He was witnessing a great event.

Thirdly, the restrictions on the occurrence of the ING-form after


complement verbs, if the complements follow the continuous verbs
of perception i.e. Subclass (2a) and noncontinuous verbs of
perception i.e. Subclass (2b) will be the issue of the following
sections on English ACI-verbs and the constraints on English
ACI-Constructions.

Rather than an ACI-Construction as verb complementation in


English we are confronted with a different type of verb
complementation, the AWG-Construction: Compare *(51) with (54):

(54) He heard her playing the piano.

The AWG-Construction is not, as will be shown, an ordinary


morphological variant of the English ACI-Construction, but
reflects, in most cases, a fundamental semantic difference.
The ACI-Construction and the AWG-Construction are therefore
considered to be derived from two different underlying con­
ceptual structures (cf. section 1. above). This claim has to
be weakened when certain interpretations are made of the ACI-
Cons truction- which occurs after the verbs NOTICE, FEEL^ , and
PERCEIVE^ as d emonstrated by examples (84c), (90c) and (92c).
238

A simple infinitival form of the complement verb is, however,


possible after the matrix verbs SEE and HEAR, if the mode of
action is nòndurative as in (55b) and (56b). If the mode of
action is durative the Progressive Affix Rule (5) applies, and
spells out the ING-form as in (55c) and (56c):

HEAR
(55) a. *He heard her play the piano.
[+Dur]
b. He heard her shut the door.
[+Mom]
c. He heard her singing a folktune.
[+Dur]

SEE
(56) a. *He saw her play in the yard,
[+Dur]
b. He saw the firing squad execute a man.
[+Mom]
c. He saw her playing in the yard.
[+Dur]

If the Progressive Affix Rule does not apply in complement


sentences expressing durative Aktionsart, violations of well-
formedness occur as can be seen in *(55a) and *(56a).

There are at least two ways to derive the surface structures


exemplified by (55c) and (56c). The AWG-Constructions of these
sentences are either derived through the Subject Raising Rule
in our sense (cf. Part One, section 4.3) or via the Equi NP
and the Temporal Complementizer Deletion Rules. At present
there is no empirical evidence for deciding which of the two
possible underlying structures is more adequate for generating
the AWG-Constructions under discussion. (See the discussion
of the underlying representations of the ACI and AWG-Construc­
tions in section 1.).

The ACI-Constructions are considered to be generated by the


Subject Raising Rule, whereas the derivation of the AWG-
Constructions is more problematic. Besides the underlying
structure (B) presented in section 1. one could posit the
phrase marker (C) as underlying the AWG-Constructions in
English:
239

+V
+ Sub Dui
±Timg
as
iwhen

The alternative conceptual structure of the AWG-Constructions


in (C) corresponds roughly to the following strings to which
the Equi-NP and Temporal Complementizer Deletion Rules have
not applied:

(58) He heard her- [Ja? \ she. sang a folktune]


1 Sfhenj 1 [+Dur] Sx

(59) He saw her-, [ C? I sh e played in the yard]


1 s]whenJ [+Dur] S1

Through the existence of a morphological difference between the


durative verb (i.e. the marked verb form) and the terminative
verb (i.e. the unmarked verb form) the English speaker can use
the gerundive ING-form in certain speech acts to make a clear
semantic distinction between different modes of action.

In sentence (60) the fact is stressed that the action of the


complement sentence is terminative, because the terminative
mode of action is expressed by the ACI-Construction:

(6o) a. HeJheard the runner X win the race Jon BBC's first
I saw t on TV
program.

In (61), however, it is not true that the action of the complement


verb is terminated when the subject of the matrix sentence
sees or hears the action. The AWG-Construction here expresses
the process of the complement verb, i.e. the durative Aktionsart.

(61) b. HeJheardfthe runner X winning the race Jon BBC's first


Is aw J ion TV
program.j

In the future tense, only the terminative verb form is acceptable,


since a nonterminative AWG-construction yields an ill-formed
sentence, due to the fact that the future tense form of the
matrix verb [will hear and will see) and the durative mode of
action expressed by the AWG-Construction are morphologically
unacceptable.
240

(62) a. I shall Jhearlthe runner X win the race on BBC's first


seel J on TV
program,]

b. «I shall hearl the runner X winning the racelon BBC's


see J [on TV
program.l

The semantic distinction between terminative and nonterminative


modes of action is thus crucial for the well-formedness of such
ACI-Constructions as exemplified in (60) and (61) and morpho­
logical forms of matrix and complement verbs for blocking the
AWG-Construction in *(62b). This is a constraint in English
which neither Swedish nor German shares.

Due to the existence of the AWG-Construction a further semantic


feature is required for blocking or generating other ACI-
Constructions in English which contain transitive verbs which
denote the action of the complement sentence as an achievement
or an accomplishment. Vendler 1967:113-115 speaks of
achievement and accomplishment verbs to d enote the Aktionsart in
sentences like:

(63) His wife heard him washing the dishes.


saw [+DurJ
His wife heard him wash the dishes.
saw [+Achieve]

In (63b) the action of the matrix verb is completed, whereas the


continuous mode of action in the complement sentence in (63a) is
expressed by the AWG-Construction. The choice of ACI-Construction
results in a quite different semantic interpretation from the
AWG-Construction, By positing the underlying structure (C) (=57)
which contains the temporal complementizers (as or when) a natural
explanation of the two semantic readings is obtained. The
English ACI-Construction is hence derived from an underlying
structure which contains no durative Aktionsart or any temporal
complementizer, i.e. from phrase structure (B) (=4b) described
in section 1., repeated here for the convenience of the reader
as (B'). In order to account for the well-formedness of ACI-
Constructions like (65a) and (66b) below, we include the
semantic feature [+Achievement] to express the mode of action
in the complement sentence.

(64)
(B1 )

+Achieve
241

(65) a. The husband Jheardl his wife fixing the lunch.


[saw J

b. *The husband jheardl his wife fix the lunch.


I saw J

The English ACI-Construction of (65b) is not well-formed, as it


is considered no achievement for a wife to fix lunch. If
another verb is chosen to express the Aktionsart of achievement
on the part of the complement subject, both ACI and AWG-
Constructions are acceptable:

(66) a. I <hea rd\ her concocting a fantastic meal in the


I saw [ C+Dur]
kitchen,

I J heardl her concoct a fantastic meal in the kitchen.


I saw [ [+Achieve]

Whether an AWG-Construction or an ACI-Construction is to be


chosen cannot be described without considering the linguistic
and extralinguistic contextual variables. In recent generative
grammar and philosophical writing several lingusists Chomsky
1970b, Bierwisch 1970, McCawley 1968, Lakoffl97la, Kiparsky &
Kiparsky 1970, Keenan 1971, 1972, and 1973 have indicated what
an indispensible role so-called presuppositions play for the
acceptability of sentences. They claim that there are certain
conditions which the world must meet in order that sentences
make sense and are acceptable. Some sentences are accepted only
under given presuppositions (cf. especially Keenan 1971:48).
Some examples supporting the necessity of including presuppositions
into a generative formalism have been given by the Kiparskys,
without much acclamation from other linguists. Most linguists
prefer to de scribe presuppositions verbally by giving examples
of sentences which can be grammatical under given conditions
in the outer world (cf. Keenan 1971 and 1973, and Kiefer 1972
etc). Because of the complicated mechanism for formalizing
presuppositions within generative grammar, we shall refrain
from attempting such formalism and shall here discuss only the
presuppositions which are necessary for generating well-formed
ACI-constructions in English.

Under the presupposition that the complement subject of (67)


played the piano at a special occasion, for instance at Carnegie
Hall, then the ACI-Construction which we in (55a) have starred
as v iolating the Aspectual Constraint (15) may be acceptable,
since the complement verb then expresses an achievement:

(67) iiheardl her play the piano at the Carnegie Hall.


[saw J [+Achieve]

The problem here is how to assign the aspectual feature. The


cooccurrence of play the piano with the official concert hall
as a location Carnegie Hall, yields the impression of an
accomplishment, so that the aspectual feature [+Achievement]
242

can be inserted as representing the kind of official performance


with which we are concerned here. How such an assignment of a
feature is to be formalized in an explicit theory of generative
grammar has not been discussed in any of the treatments of the
verb phrase in English. The location Carnegie Hall does in
itself not contain any inherent feature of accomplishment as
can be claimed about the time adverbial several times in (68),
which inherently contains the iterative aspectual feature which
is copied into the verb and prevents the Progressive Affix
Rule (5) from applying. For the locative noun phrase at Carnegie
üall we have no such feature to copy into the verb play. If
no such copying of an aspectual feature can be made, the block­
ing of the morphological rule which spells out the ING-suffix
must be accomplished by the whole verb phrase play the piano at
Carnegie hall. A description merely in terms of immediate
constituents or aspectual features which are copied, etc will
not suffice for describing the well-formed ACI-Construction in
(67). The concept of presupposition might enable us to explain
why (67) is grammatical in the sense that to play the piano at
Carnegie Hall is an achievement and if somebody hears a person
under such a circumstance, then it is an accomplishment on the
part of the person performing. An extralinguistic feature of
accomplishment must then be copied into the complement verb so
that it may occur in an English ACI-Construction.

It is also possible to generate ACI-Constructions under the


presupposition of a repeated series of performances :

(68) I have seen her play the piano (several times.)


[+Iter] [+Iter]

The iterative time adverbial must not necessarily be explicitly


expressed in (68) to make this sentence well-formed. AWG-
Constructions in the two latter cases are impossible under the
presuppositions of achievement and iteration as given above:
14
(69) *1 saw her playing the piano at Carnegie Hall.

(70) *1 have seen her playing the piano several times.1^


243

7. ENGLISH VERBAL COMPLEMENTS AFTER DURATIVE VERBS OF


PERCEPTION (LOOK AT, LISTEN TO, OBSERVE, WATCH, AND
WITNESS)

The semantic compatibility of the Aktionsarten in the matrix


verbs (HEAR AND SEE) and the complement verbs has been demonstrat­
ed to be crucial in generating well-formed English, German, and
Swedish constructions. Above in section 6. we were involved
with the nondurative use of HEAR and SEE (cf. (55b) and (56b)
for instance) or the durative use of HEAR and SEE (cf. for
instance (55c) and (56c)). The semantic restrictions on the use
of AWG-Constructions and ACI-Constructions were also discussed
at some length. HEAR and SEE can be used either as durative or as
nondurative verbs of perception, depending on the Aktionsart
of the following complement verb or the linguistic context,
(cf. (67)), or on a certain presupposition (cf. (67) and (68)).

If inherently durative verbs of perception such as LOOK AT,LISTEN' TO


OBSERVE, WATCH, and WITNESS enumerated in subclass (2b) in
FIGURE 1 are selected as m atrix verbs an even narrower constraint
on the Subject Raising Rule must be formulated for blocking
certain ACI-Constructions, at least in a gi ven style of
English (cf. *(71b) and *(77a) ;-s compar ed to (72b)and (79)
respectively). Neither durative nor nondurative complement
verbs will yield well-formed ACI-Constructions after Subject
Raising. Only AWG-Constructions seem to be acceptable con­
structions in a given style of English, so that for this style
of English a different underlying structure is postulated,
i.e. phrase marker (B') (=64) and instead of Subject Raising
the Equi-NP and Temporal Complementizer Deletion Rules are
assumed to generate the AWG-Construction.

LOOK AT
(71) a. *He looked at her fix lunch.

b. *He lcoked at her leave the room.

c. He looked at her fixing lunch.

There is, however, literary evidence that ACI-Constructions may


occur after the durative LOOK AT:

(72) a. Look at Glorsvina enter a room.


L-Dur]
(Thackery, Vanity Fair: Ch. XLIII)

b. Look at hi m run. CF. Norris, The Octopus)

Older grammars often quote such examples of ACI-Constructions


after LOOK AT as evidence, but a closer examination of their
stylistic value tells us that ACI-Constructions are here felt
to be artificial and of a higher literary style than are the
AWG-Constructions.16
244

The verb LISTEN TO listed under subclass (la) denotes an audi­


tory process which is characterized as [+Purpose] in contrast
to HEAR, which can occur nonintentionally. LISTEN TO is
inherently characterized by the durative Aktionsart and as a
result requires a complement sentence which is also of durative
Aktionsart. We are here dealing with the same type of semantic
compatibility or agreement between the matrix verb and
complement verb which was described as being an important
aspectual constraint also in Swedish and German (cf. Aspectual
Constraint (15) in section !.)•

LISTEN TO
(73) a. *He listened to the woman mention his name.
b. •He listened to the bomb explode.
c. He listened to a woman talking about her husband.

(74) a. He listened to the woman slander her neighbor.


b. He listened to the bomb exploding.

For some speakers of English sentences like (74a) which contain


durative complement verbs are well-formed. The Subject Raising
Rule is applicable here, since the Aktionsart of the matrix
and complement verbs match, that is they are both [+Durative].
The ACI-Construction in (74a) may have a literary flavor to most
native speakers of English.

A contextual circumstance must be mentioned here which can


change the status of the ACI-Construction after LISTEN TO from
being literary to ordinary English. If the action of the
subject of the complement sentence is performed in a special
social role, then LISTEN TO seems to be an acceptable cooccurrence
with the ACI-Construction. In that case such an ACI-Construction
does not have any special literary flavor:

(75) He listened to Brandt speak of a reunification of Germany.

(76) He listened to Nixon speak of the economic crisis in the


USA.

Compare the use of ACI-Constructions after WATCH under the same


social circumstances below. Searle 1969 and Fillmore 1972
have in their works on syntactic structures pointed out the
importance of including the role concept for describing linguistic
structures. The restrictions on the generation of the ACI-
Construction can be weakened under certain given circumstances
of social roles which the complement subjects plav in the
given linguistic context (cf. (75), (76), and (78c)).

OBSERVE
(77) a. *He observed the children play in the yard-
b. *He observed her break the vase.
c. He observed the children playing in the yard.
245

WATCH
(78) a. *He watched her fix lunch.
b. *He watched the policeman stop the car.
c. He watched Queen Elizabeth enter the limosene,
d. He watched his wife fixing the lunch.
e. *He watched the policeman stopping the car.

There is literary evidence that OBSERVE, WATCH, and WITNESS


also take the ACI-Construction.
(79) I observed her draw a handkerchief from the pocket.
(Bronte)
(80) a. He watched Cards walk slowly down the hill. (Walpole)
b. Lizzie watched the parents return.(Wells)

(81) I witnessed a large hawk-moth meet his e nd .(Hudson)

WITNESS
(82) a.? I witnessed a spider eat a fly.
b.? I witnessed a man shoot a woman.
c. I witnessed it happening.
d. I witnessed a spider eating a fly.

The Aspectual Constraint on Subject Raising, i.e. (15) formulated


for Swedish and German verbs of perception in section 2. seems
to be a common Germanic constraint on the Subject Raising Rule,
since the ill-formed starred (b) sentences in (71) through (78)
do not meet this semantic condition on Subject Raising.
Constraint (15) blocks primarily the (b) sentences here, since
the durative mode of action in LOOK AT, LISTEN TO, OBSERVE,
WATCH, and WITNESS is felt to be incompatible with the
instantaneous mode of action of the complement sentences (leave
the room in (71b), break the vase in (77b), stop the car in
(78b), and shoot a woman in (82b).

An additional constraint must, however, be added in English to


the semantic constraint (15). The (a) sentences have complement
sentences that are of durative mode of action (fix lunch in
(71a), play in the yard in (77a), fix lunch in (78a), and eat
a fly in (82a)), but they are nevertheless ill-formed, due to
the lack of the aspectual marker ING.

However, this morphological constraint which prevents durative


complement sentences from occurring without the progressive
ING-form is often violated in literary texts as evidenced by
(79) through (81). The durative process of observing somebody
in the act of drawing out a handkerchief cannot normally be
expressed by a nonprogressive complement verb form as in (79).
The same is true of the other verbs of durative perception in
(80a), (80b) and (81).
246

The only well-formed sentence in (78) which contains an ACI-


Construction after WATCH is (c) which is a cceptable because
of the official role of the complement sentence subject
(Queen Elizabeth). Under the presupposition that it was an
official person who was entering the car, then a plain ACI-
Construction is acceptable to some native speakers of English.
We do not know how to incorporate such a presupposition into
the present theory of generative syntax. As was pointed out
in section 6., generative theory does not provide us with any
formalism which can be used to express the generality of
presuppositional constraints. Chomsky himself has no concrete
suggestion as to how to formalize concepts such as focus,
presupposition, scopus etc (cf. Chomsky 1970b, 1971a, and
19 71b) By including a means to account for the role of pre­
suppositions within a grammatical theory, the Standard Theory
would increase in explanatory adequacy. At present only loose
suggestions of surface structure interpretations to be added
to t he semantic readings which operate on deep structures have
been made by Chomsky 1970b and 1971a and Katz 19 72:414-441.

8. ENGLISH VERBAL COMPLEMENTS AFTER NONDURATIVE VERBS OF


PERCEPTION (BEHOLD, DISCOVER, AND NOTICE)

There does not seem to be any general aspectual constraint


which can be set up for blocking ill-formed ACI-Constructions
of this class. The instantaneous verbs of perception, BEHOLD,
DISCOVER, and NOTICE should, for standard English, be marked
as not undergoing the Subject Raising Rule and as a result do
not take ACI-Constructions as verbal complements. They do take
AWG-Constructions, however, and in order to show the difference
from the Swedish and German verbs of perception of the same
class (2b) the ill-formed English ACI-Constructions are also
given here.

BEHOLD
(83) a. *She beheld him imitate her father,
b. ?She beheld him pass three young men.
c. She beheld him imitating her father.

NOTICE
.(84) a. *The runner noticed his strength decrease.
b. ?I noticed Peter fall off the roof.
c. The doctor noticed the patient's heart beating fast.
d. The runner noticed his strength decreasing.

DISCOVER
(85) a. *She dicovered his friend stand on the bridge,
b. *He discovered his friend jump onto the bridge,
c. He discovered his friend standing on the bridge*
247

The underlying structure of (84) must be considered to b e the


same as the one which underlies normal ACI-Constructiors* that is
no temporal complementizer is posited as being deleted in the
derivational process. The verb NOTICE takes an object sentential
complement from which the Subject Raising Rule moves the
complement subject into the object position of the matrix
sentence (i.e. phrase marker (A) in section 1.). A derivation
of the AWG-Construction from a temporal sentential complement
is i nadequate, because it would yield an inadequate reading,
since the patient's heart was beating fast during the doctor's
examination, a fact which was noticed by him. But (84c) can
also be interpreted as meaning that the doctor noticed when
the patient's heart was beating fast, if the phrase structure
contains the temporal complementizer, i.e. phrase structure (C)
in section 6,(57j.

Compare the following strings which roughly reflect the two


underlying phrase structures being discussed here:

(86) a. The doctor noticed the patient's heart,


[the patient's heart was beating]
si si

b.*The doctor was noticing the patient's heart,


[when the patient's heart was beating]
S1 S1

It makes little sense to posit (86b) as the underlying structure


of (84c), since hopefully the patient's heart continued to beat
even after the doctor had noticed it.

Such cases as (84c) motivate'us to assume an underlying structure


for the English AWG-Construction which is identical to phrase
marker (A) (-4a) in section 1. An underlying string like the
following will avoid inadequate semantic interpretations as
exemplified by the string (86b):

(87) [ [Experiencer] +V NP. [. , ,NPi ..V ] ]


S NP NP +Perception] S ì [+Dur]S S
1 o

By cl aiming that (87) underlies all AWG-Constructions a general


Subject Raising Rule further can be claimed to generate both
ACI and AWG-Constructions, according to the presence or non-
presence of the feature [+Dur].

Some literary examples of ACI-Constructions are found after


BEHOLD and NOTICE:

(88) She beheld her brothers pass three young men. CMeredithJ

(89) He noticed them pause in front of a high obstacle.


248

9, ENGLISH VERBAL COMPLEMENTS AFTER TACTUAL VERBS OF PERCEPTION


(FEELX AND PERCEIVE^
Semantic characteristics are also decisive for the choice of
English complement structures after the two verbs of tactual
and mental perception enumerated in class (3a) in FIGURE 1,
that is, FEEL-^ and PERCEIVE, 'feel tactuallyf and 'feel mentally1,
respectively. Both the ACI-uonstruction and t/zat-Embedding
are possible verbal complements after FEEL, and PERCEIVE-,.
Compare (90) through (93) as evidence of tnis claim:

FEEL^ 'feel tactually'


ACI-Construetion
(90) a. *Peter felt his pulse beat fast.
[+Dur] [+Dur]
b. Peter felt the girl make a movement.
[+Mom] [+Mom]
touch his lips.
[+Mom] ,
Peter felt his pulse^ beating fast.
[+Dur] [+Dur]

That-Embedding
(91) a. *Peter felt that his pulse beat fast.
[+Uur] [+Dur]
Peter felt that the girl made a movement.
[+Kom] [+Mom]
touched his lips
[+Mom]
Peter felt that his pulse" was beating fast,
nir] [+Dur]

PERCEIVE^ 'feel tactually'


ACI-Construetion
(92) a. *Peter perceived his pulse beat fast-
[+i:ur] [+Dur]
b. Peter perceived the girl make a movement.
[+MOIP] [+Mom]
touch his lips.
[+Mom]
Peter perceived his pulse^beating fast.
[+Dur] [+Dur]
249

That-Embedding
(93) a. *Peter perceived that his pulse beat fast.
[+Dur] [+Dur]
b. Peter perceived that the girlf made a movement.
[+Moni] l[+Mom] I
] touched his lips.f
^[+Mom] /
c. Peter perceived that his pulse was beating fast.
[+Dur] [+Dur]

The English ACI-Construction is limited to the cases where


the Aktionsart in the complement sentence is not durative but
instantaneous, [+Mom], (cf. (90b) and (91b)). When the
instantaneous matrix verb is combined with a durative complement
verb which is not of a progressive form, as in (90a) and (92a),
then the morphological Progressive Affix Rule (5) is violated.
Two different phrase markers are therefore assumed to underlie
the AWG-Constructions #in the well-formed (c) sentences and
the ACI-Constructions in the (b) sentences, i.e. phrase
structure (A) and phrase structure (B) respectively, described
in section 1. Phrase structure (C) discussed in section 6., would,
however, be inadequate for deriving (90c) and (92c), since no
Temporal Complementizer makes sense here:

(94) [PeterJfelt his pulse [Jas I his pulse was beating] ]


jperceived S-jjwhenf S-, S _

Hopefully, Peter's pulse was beating even after Peter had felt
that it was beating.

The Subject Raising Rule of both Swedish and German is more


general in the sense that it can operate on underlying structures
without the surface structure constraint of an overt durative
aspectual marker. The starred English ACI-Constructions in (90)
through (92) are therefore perfectly well-formed in Swedish
and German ACI-translations (cf. *(90a) and *(92a) with Swedish
(39a) and German (40a) plus (41a)).18
250

10, FEEL3, PERCEIVE3, AND THE ACCUSATIVE-WITH-T0-INFINITIVE

The two homonymous verbs FEEL^ 'feel mentally' and FEEL.


'feel tactually' on one hand, and the two other homonymous
verbs PERCEIVE.'feel tactually' and PERCEIVE^ 'feel mentally'
on the other, take different infinitive complements after
Subject Raising. The verbs indexed as 3 denote not physical
but mental perception and take the Aecusative-with-TO-Infinitive
as embedding construction, whereas the verbs indexed as 1
denote tactual perception and take the ACI-Construction as
embedding. This surface structure is characteristic of the
English language and is found in neither of the other two Germanic
languages treated here in sentences containing verbs of mental
perception. Since the infinitival marker TO signals another
verb of perception than the plain ACI-Construction the feature
[Tactual] is motivated on semantic as well as syntactic grounds.

In (90b), (91b), and I92b) ACI-Constructions were generated by


Subject Raising, whereas in (95a) and (96a) an additional rule,
the Infinitival Marker Placement Rule, has operated because of
the nonphysical character of the verbs FEEL^ and PERCEIVE

FEEL^ 'feel mentally'


(95) a. Peter felt the plan to be all wrong.
[-Tactual] [+Dur]
b. *Peter felt himself to cut a poor figure.
[-Tactual] [+Mom]
c. Peter felt himself to be cutting a poor figure.
[-Tactual] [+Dur]

PERCEIVE^'feel mentally'

(96) a. Peter perceived the plan to be all wrong.


[-Tactual]
b. *Peter perceived himself to cu t a poor figure.
[-Tactual] [+Mom]
c. Peter perceived himself to be cutting a poor figure.
[-Tactual] [+Dur]
Doomsday to be approaching.
[+Dur]

No complement verbs of immédiate mode of action seem to be


possible as c ooccurrences with FEEL3 and PERCEIVE3. The action
of the matrix verb and the complement verb must be different
so that the complement sentence expresses durative mode of
action, if the Accusative-with-TO-Infinitive is to be selected.
(Cf. The Aspectual Constraint on the selection of the Swedish
perfective verbs MÄRKA, UPPTÄCKA, and KÄNNAi formalized in
(16), section 2.). Notice, however, that nondurative perceptual
verbs do cooccur with instantaneous complement sentences in
251

plain ACI-Constructions as exemplified in (90b) and (92b), that


is felt the girl make a movementorperceived the girl touch his
[+Mom] [+Mom] [+Mom] [+Mom]
lips.

Most native speakers of English prefer i/zat-Embeddings to ACI-


Constructions with the infinitival marker T0 after FEEL^ and
PERCEIVEJ 'feel mentally'.

That-Embedding

FEEL^ 'feel mentally'

(9 7) a. Peter felt that the plan was all wrong.


[+Dur3
b. Peter felt that he cut a poor figure.
[+Mom]
c. Peter felt that he was cutting a poor figure.
[+Dur]

PERCEIVE, 'feel mentally'

(98) a. Peter perceived that * Doomsday approached.


Doomsday was approaching/
[+Dur] )
b. Peter perceived that he cut a poor figure.
[+Mom]
c. Peter perceived that he was cutting a poor figure,
[+Dur]

The complement sentences in (97) and (98) are tensed sentences


and can therefore be interpreted as expressing something which
occured before the instant of perception expressed in the
matrix sentence by felt and perceived. Such a difference in
Aktionsart seems impossible to express by the Accusative-with-
TO-Infinitive as de monstrated by *(95b) and *(96b). There is
an interesting interrelation between tense and Aktionsart in
the t?zat-Embedding which will not be treated here at any length
since this requires a special investigation. It is sufficient
for our purpose to point out that the tensed complement that-
sentences allow such matrix and complement verbs to cooccur.
The cooccurrence with an Accusative-with-TO-Infinitive
Construction is ill-formed, because after Subject Raising
infinitive constructions are untensed. (See section 2. the
discussion about Aspectual Constraints (15) and (16), which
block Swedish and German ACI-Constructions, but allow the
generation of t/zat-Embeddings, because the instantaneous
Aktionsart of the complement sentence is combined with tense).
252

11, REFLEXIVE COMPLEMENTS AND PREDICATE COMPLEMENTS AFTER


VERBS OF SOMATIC PERCEPTION IN GERMANIC LANGUAGES

The English verb of somatic perception, FEEL4 'feel somatically1


takes a surface structure complement which is very different
from the corresponding Swedish and German complements. Any Swedish
or German undergraduate course in English instructs the fact
that the Swedish or German reflexive verbs KÄNNA4 SIG and SICH
FÜHLEN 'feel somatically' cannot be translated into *FEEL4 ONESELF.
Any further explanation is not given, although the syntactic
difference between the constructions which follow the Swedish
or German reflexive verb and those which follow the English non-
reflexive surface verb is considerable.

The derivation of the surface structure verb complements in (96)


through (101) is illustrated by the derivational steps of
(102a) - (102c).
ENGLISH
Predicaté Complement
(96) a. He felt ill. (well),
b. He felt bad (good).

Reflexive Complement and Predicate Complement


(97) a. *He felt himself ill (well).
b. *He felt himself bad (good).

SWEDISH
Predicate Complement
(98) a. *Han kände sjuk (frisk).
b. •Han kände dålig(bra).

Reflexive Complement and Predicate Complement


(99) a. Han kände sig sjuk (frisk).
b. Han kände sig dålig (bra).

GERMAN
Predicate Complement
(100) a. *Er fühlte krank (gesund).
b. *Er fühlte schlecht (wohl).

Reflexive Complement and Predicate Complement


(101) a. Er fühlte sich krank (gesund).
b. Er fühlte sich schlecht (wohl).

An English native speaker has no intuitive notion that a


potential complement subject in (96), i.e. a reflexive pronoun
is absent. The introduction of such a reflexive pronoun makes
the sentences utterly ill-formed (cf. *(97a) and •(97b)). In
order to describe the surface structure of the well-formed
English sentences with predicate complements in (96a) and (96b),
we may claim either that FEEL4 takes only predicate complements
253

but no reflexive pronoun as an object (this claim is false in


the case of FEEL3, since reflexive pronouns do cooccur with
FEEL3, cf. (95c) above), or that the predicate complement ill
and weil in (96a) and bad and good in (96b) are derived from
underlying structures which contain a complement subject NP
an possibly a copula. There is a comparative syntactic advantage
with the latter claim. By considering the predicate complements
of (96) as being derived from a phrase marker which contains
a complement subject node, syntactic comparisons with Swedish
and German are possible in terms of the number and ty pes of
syntactic transformations which are common to the three Germanic
languages involved here. By as suming such a common Germanic
phrase marker and postulating common syntactic rules, the genera­
tion of well-formed surface structures in (96) through (110)
can be adequately described in a comparative syntactic descrip­
tion. In this way we can establish a common basis for a syntactic
comparison between the surface structures of the three Germanic
languages and, what is more, we can describe the surface structure
differences in terms of syntactic constraints on rules.

By postulating (102a) as the common Germanic phrase marker for


deriving the surface structures of (96) through (101), we are
not claiming that the underlying phrase marker is a psychological
reality in any of the three languages. A pure comparative
approach is intended here for describing surface structure
differences. The positive results of such a comparative approach
have been demonstrated in historical linguistics, where abstract
phonological and morphological base forms have resulted in
efficient devices for reconstructing Proto-Germanic and Proto-
Indo-European forms and for describing the genealogical and
typological similarities between the languages of Europe
(cf. Bopp 1816, Grimm 1819 etc).

Peter.felt S
1.
Pete^kände
Peterifühlte

[+ADJ]

Peter.^ ill
Peteri sjuk
Peter^ krank
254

After Subject Raising the following intermediate structure


will be generated:

Peter^ felt Peter^ ill


Peter^ kände Peter^ sjuk
Peter^ fühlte Peter^ krank

The Reflexive Rule now applies in Swedish and German, whereas


in English the object NP (the second Peter) is deleted for
some reason.

(102) c.
,VP
NP, NP, [+ADJ]

Peter^ felt 0 ill


Peter^ kände sig sjuk
Peter^ fühlte sich krank

InSwedish, the copula VARA can also be spelled out, so that


a plain ACI-Construction results after Subject Raising and
Reflexivization:

(103) Peter kände sig vara sjuk (frisk).


(Peter felt ill).

In English and German the Copula Segmentalization Rule yields


ill-formed verbal complements:

(104) a. *Peter felt be ill (well).

b.*Peter fühlte sich krank (gesund) sein.

Again the Swedish Subject Raising Rule in combination with the


Copula Segmentalization Rule generates an ACI-Construction which
is not found in either German or English. The surface structure
of German is syntactically closer to that of Swedish in that
a reflexive object noun phrase is allowed which is obligatorily
deleted in English (cf. (101a) and (96a)).

There is a problem with postulating an adequate common deep


structure for the verb complements of the three Germanic
languages exemplified in (96) through (101). One may either
posit an underlying copula node along the lines of early
transformational grammars (e.g. Chomsky 1957:39 and 1965:107 or
255

regard the copula as a result of a language-specific segment-


alization rule which spells out the copula under given
conditions (cf. Bach 1967 and 1968:118, Jacobs & Rosenbaum 1968:
101 etc). In Germanic languages this outspelling of a copula
segment is mostly optional as exemplified by the Swedish ACI-
Construction in (103). Sometimes, however, there are heavy
constraints on Copula Outspelling, especially in embedding
processes as exemplified by English (104a) and German (104b).
(Cf. also the constraint on VARA Segmentalization in Swedish
Accusative-with-Past Participle Constructions discussed in
Part One, Section 5.2, examples (76a)-(76c)).

In some languages like Russian the copula does not exist at


all as verb between the subject noun phrase and the predicate
adjective.

(105) a. On éyvstvoval sebja bolnim.


He felt (himself) ill.
b. On cyvstvoval sebja njezdorovim.
He felt (himself) bad.
c. Anna boïna-
Ann (is) ill.

Deep Structure (102a) can therefore be used as an underlying


syntactic structure common both to the three Germanic languages
involved and to the corresponding Russian surface structure of
(105b). (See also the discussion of common Germanic deep
structures for deriving surface structures in the succeeding
section).
256

12. TOWARDS A THEORY OF COMPARATIVE SYNCHRONIC SYNTAX


12.1 COMMON GERMANIC DEEP STRUCTURES FOR DERIVING SURFACE
STRUCTURES

The present investigation of the verbal complement constructions


in Swedish, German, and English has been undertaken to serve
as a basis for further research to find a mechanism for
comparing and deriving surface structures from abstract deep
structures which are syntactically motivated. If possible we
would like to find a mechanism for reversing the operation of
transformations in order to reconstruct the steps of derivation
from the surface backwards to the original deep structures.
The operation of the Subject Raising Rule must be assumed to
be somehow mastered also in reverse by a native speaker or
"native listener". Although such a claim is controversial from
a psycholinguistic point of view,(see Watt 1970, Bever 1970,
and Slobin 1968), syntactic surface phenomena can be better
described and explained within a comparative framework if an
abstract level of syntactic representation is taken as the
starting point for generating a multitude of surface structures
in various Germanic or even non-Germanic languages.

A long series of yet unanswered questions have»however, arisen


in the course of this comparative undertaking. The first
important one is the following:

(106) Can the deep structures (37a) and (37b) in Part One
section 4.3 be considered as Common Germanic underlying
representation for deriving ACI-Constructions in Swedish,
German, and English?

The second, equally important, question is:

(107) Is the Subject Raising Rule as defined in section 4.3 a


formal universal in the sense of Chomsky 1965 or is this
syntactic rule specific to only Germanic and Indo-European
languages?

Some details in the approach to the ACI-syntax of Germanic


languages may yield an affirmative reply to question (106) and
an affirmative reply to question (107) with some reservation.

One example to elaborate question (106): The convention of de­


scribing German sentences in terms of a verb final order in
the deep structure is feasible in language-specific descriptions
of German (cf. Bach 1962, Bierwisch 1963, Härtung 1964, Mötsch
1964, Steinitz 1969, Wurzel 1970 etc), but certainly not for
a comparative or typological study. Both German and Swedish are
verb^econd languages, that is main clauses require that the
tensed verb appears in the second position of the sentence, a
requirement which could be described as a surface constraint
(cf. for instance Haiman 1971). English, on the other hand,
does not maintain such a constraint any longer, since after
topicalization of the object or the adverbial, the subject noun
phrase follows:
257
e

(108) a. Yesterday I saw the man.


b. Igår såg jag mannen-
c. Gestern habe ich den Mann gesehen.

German seems to have developed a different surface structure


constraint in the course of its history as far as the order
of the verbal elements in the subordinate clause is concerned.
Compare the well-known syntactic constellations of verbal
elements in the following English, Swedish, and German clauses:

(109) a. because I saw the man yesterday-


fa. eftersom jag såg mannen igår-
c. weil ich den Mann gestern gesehen habe.

Lehmann 19 70 and Ross 1970a have maintained that the final


order of verbal elements in German is not basic for German
syntax. Their arguments are based on typological evidence from
languages which are true verb-final languages, so-called SOV
languages, like Japanese and Turkish. If German were a true SOV
language as claimed indirectly by Bach 1962 and Bierwisch 1963
from a language-specific point of view, then it should share
characteristics of SOV languages other than only the verb-final
order, that is:

1) the relative clause should precede the NP modified


2) the genitive attribute should always précédé the NP modified
3) postpositions should occur after NP's
4) comparatives should have the pattern: Standard/Pivot/Adjective

The four points of typological evidence for classifying a


language as an SOV language are of course not found in German.
Lehmann 19 70:6 therefore classifies Modern German as a typical
SVO language like Spanish, although he indicates some remainders
of SOV typology which may still be found from an earlier Proto-
Indo-European epoch (cf. for instance the post-positional use of
gemäss, nach, gegenüber). Ross also identifies German as an
SVO language on the basis of the operation of the Equi VP
Deletion Rule in coordinate sentences. In this case German
implements the deletion rule common for all SVO languages, so
that it deletes all but the first identical verb in conjunctive
constructions:

(no) sv i o,sv i o, sv i o —*SV-0,S0,S0

(111) Peter liebt Kuchen und Maria (liebt) Bonbons.

(112) weil Peter Kuchen liebt und Maria Bonbons (liebt).

Japanese, an SOV language, deletes all verbs but the last:

(113) sov i ,sov i ,sov i —* so,so,sov i


258

The comparative and typological criteria presented here for


classifying languages as SOV, SVO, or VSO languages (cf. also
Greenberg 196 3) have been found to be more convincing and more
adequate for the purpose of conducting a comparative study of
Germanic infinitival constructions than have such language-
specific arguments for descriptive economy as suggested by Bach
1962 and Bierwisch 1963.

The abstract verb-final order for German base structures is


inappropriate for a typological and comparative approach. If a
linguist who is interested in the comparative aspect in Germanic
languages wishes to use the SOV order proposed by Bach and
Bierwisch, a difficult dilemma will arise as soon as he wishes
to compare the deep structure configuration of German and those
of Swedish and English. He will have to set up a completely
different deep structure order for German and dispense with a
deep structure order common to Germanic. For the sake of
comparison, language typology, language teaching,and machine
translation, a common Germanic deep structure is m ore adequate
than would be a language-specific deep structure for each
language. By postulating a common Germanic base order for
verbal elements and a given deep structure for the derivation
of ACI-Constructions in Germanic languages, we do not claim
that the proposed order of elements or deep structure is
universal in the sense understood by those linguists who have
suggested a Universal Base Hypothesis (see recent criticism of
this hypothesis by Peters & Richie 1969, 1971 and Peters 1970).

Few generative grammarians have been confronted with the di­


lemma of comparing the surface structures of various langugaes,
since most generative studies are primarily involved with
intralinguistic aspects rathe,r than interlinguistic or compara­
tive aspects of the language(s) under study.

By introducing the concept of the constraint (see Part One


section 4.1) we can explain the SOV order in German subordinate
clauses as being a surface structure idiosyncracy which results
in the final position of the tensed verb. A surface structure
constraint which acts as a filter for German subordinate
clauses (cf. Surface Structure Filter III in Step XI of
FIGURE lib) would enable us t o remove language-specific traits
from the more generally applicable base rules, and instead
account for such traits in the transformational rules and the
morphological rules. In this way the similarity of German deep
structure to that of the other Germanic languages can be stressed.
At the same time the criteria of language typology and the
needs of a comparative syntax between the Germanic languages
can be met.
259

12,2 THE SYNTHESIS AND ANALYSIS OF ACI"CONSTRUCTIONS

In the preceding section typological and comparative arguments


have been presented for setting up a common Germanic synchronic
deep structure. The derivation of surface structures from
underlying syntactic base structures (deep structures) is a
dynamic process. By emphasizing the dynamics in the derivational
process a more operational view and definition of the ACI-
Construction is accomplished than a mere pattern and slot-
oriented description along the lines of traditional grammar or
structuralist grammar. In order to illustrate the relation
between the deep syntactic trees generated by the base rules
and the surface structure trees generated by the transforma­
tional rules, FIGURE 7 is presented. It represents a given set
of derivational steps which include some language-specific and
some common Germanic constraints on the deep ,as well as the
surface structure levels.

FIGURE 7

BASE RULES

Aspectual Constraints
on lexical insertion

DEEP
STRUCTURE TREES

FILTER II:
i—r
Equal Subject Constraint

TRANSFORMATIONAL RULES

J^HALLOW^V^
1TRUCTURE TREES

Shallow Structure Constraint


on Indefinite Subjects for the
Passive Rule after Subject Raising

MORPHOLOGICAL RULES

^J^YNTACT
SURFACE STRUCTURt:

PHONOLOGICAL RULES

1 r
Surface Structure Constraints on
the Cooccurrence of Double s-Passive
T and Double BLIVA-passive
260

FIGURE 7 is drawn on the basis of different suggestions for


outlining a syntactically motivated theory. It includes the
fundamental concepts of the Standard Theory of 1965 by positing
a number of syntactic base rules which generate an indefinite
number of deep structure trees or phrase marker if the
requirement of the (Aspectual) Constraints in Filter I are met
(cf. section 2.). The derivational process thus starts in the
base component. The deep structure phrase markers or trees
are then mapped into shallow structures by the transformational
rules (T^^, xPass, -pReflex etcj provided that the given Deep
Structure Constraints on the transformational rules in question
(e.g. the Equal Subject Constraint, the Aspectual Constraint
on Durative Verbs, the Aspectual Constraint on Perfective Verbs
etc (see sections 2. and 3.) are not violated in Filter II.
Examples in which the Equal Subject Constraint has been violated
were given in Part One in (201), and in which the Aspectual
Constraints on Durative and Perfective Verbs have been violated
were given in (9) and (14b) respectively.

Sometimes certain syntactic configurations are blocked on the


Shallow Structure Level by Filter III, as for instance by the
Indefinite Subject Noun Phrase Constraint on the Swedish Passive
Rule after Subject Raising demonstrated by *(35) in Part One.
The operation of morphological rules is important for the out-
spelling of the infinitival endings in Swedish and German (-a
and ~(e)n respectively) and for the assignment of correct case
suffixes where these occur in the Germanic languages under
study (i.e. in the pronouns of all modern Germanic languages,
the articles of Modern German, certain inflectional forms of
German masculine nouns). The terminal strings of the syntactic
surface structure trees which are the input to the phonological
component undergo phonological modifications in accordance with
the language-specific phonology of each Germanic language.

The output from the phonological component, the phonological


surface structure tree, then passes Filter IV and obtains a
phonetic representation. There are some trees in which the
double s-passive occurs in Swedish generated by the morpho­
logical Passive Suffix Rule. The double s-passive must be
blocked sometimes for articulatory and sometimes for
stylistic reasons. As discussed earlier in Part One, section
4.2 a surface or shallow structure constraint is easier to
elaborate than a complicated morphological rule within the
derivational process itself would be:

(114) a. *De turkiska revolutionärerna sågs skjutas på TV idag


•mn . , . fi?to
?t be sho t)
(The Turkish revolutionaries were seen<
seenj'j. ueina shoti
today on TV.) I 8 J
De turkiska revolutionärerna sågs bli skjutna på
TV idag. .
(The Turkish revolutionaries were seenl'heing shoti
today on TV.) [ g J
261

It is not the passive as such which is ill-formed in *(114a),


but it is the morphological double occurrence of s-forms which
is disturbing. If the alternative passive with BLIVA 'be' is
selected, then a well-formed double passive may occur, if the
first passive form is an s-passive. Compare (114b) and *(114c):

(114) c. *De turkiska revolutionärerna blev sedda skjutas


idag på TV.
(The Turkish revolutionaries were seenj:to be Ishot
today on TV.) [being J

Only (114b) which contains the passive sequence [s-Passive...


BLIVA Passive] is well-formed. It seems as if a morphological
asymmetry between BLIVA Passive and s-Passive is obligatory in
Swedish when Subject Raising occurs. If the matrix verb contains
an s-Passive, then a BLIVA Passive is obligatory in the
complement sentence (cf. *(114c)).

However,there are additional constraints on the Swedish Passive


Rule to be discovered. Also the cooccurrence of two BLIVA
Passives in connection with Subject Raising is ill-formed:

(114) af. *De turkiska revolutionärerna blev sedda bli skjutna


idag på TV.
(The Turkish revolutionaries were seenrto bel shot
today on TV.) ] being [

Such constraints on the application of the Swedish Passive Rule


as demonstrated here are of great consequence for the theory of
syntax, since the constraints involved cannot be formulated
one-sidedly as deep or surface structure constraints. The choice
of BLIVA as the passive auxiliary is dependent upon the Aktions­
art of the matrix verb, i.e. SE 'see visually' and is thus of
deep structure nature. The choice of the s-Passive is, on the
other hand, also dependent upon whether there is another s-
Passive within the same sentence or not and is thus applicable
with respect to the phonological structure of the second verb
within the same sentence.

It is important here to stress the fact that it is only SEi


'see visually' and HÖRA^ 'hear auditorily', not SE2 'realize'
and HÖRA2 'be told' which can take ACI-Constructions, e.g. (61b)
*Jag kan inte se lagen ha visat henne mera aktning 'I cannot
see that the law has (*the law have) shown her more respect'
(Part One, section 5.) and (167b) * När ägaren hörde mig h eta
Ragnar 'When the proprietor heard that I was called Ragnar'
(Part One, section 13.). As pointed out in Part One the Subject
Raising Rule is blocked in sentences which contain SE2
'realize' and HÖRA2 'be told' as m atrix verbs. Consequently,
there are two constraints which block the occurrence of a
double s-Passive. The double s-Passive would automatically be
blocked in a sentence like *Lagerkrantz har setts tröstats
av D antes verk 'Lagerkrantz has been seen to be comforted by
the works of Dante'; first, because there is no object NP which
262

can be moved into subject position by the Passive Rule, since


Subject Raising cannot apply (there is no well-formed shallow
structure like *Någon har sett Lagerkrantz tröstats av Dantes
verk 'Somebody has seen Lagerkrantz being comforted by the
work of Dante'), and secondly, because there is in Swedish a
Surface Structure Constraint on the Occurrence of a Double
s-Passive under the same sentence node, e.g. *Ute på stan har
han hörts baktalas 'Downtown he has been heard being slandered'.

In light of these constraints it seems to be more adequate to


work out such morphological rules for the Swedish passive
which are sensitive to semantic as well as phonological pro­
perties. By drawing upon the theory of morphology as suggested
recently by some European linguists (cf. Bierwisch 1967,
Wurzel 1970, and Kiefer 1970) the constraints on the occurrence
of the Swedish passive forms can be elaborated as shallow or
surface structure constraints. At present no single generative
treatment exists which can adequately account for the complex
set of derivational constraints to generate well-formed
Swedish passives.

In a generative theory which includes morphological rules,


constraints can be formulated as reflecting semantic deep
structure properties like Aktionsarten and surface structure
properties like the phonological form of inflectional endings,
i.e. the s-Passive form.

What we need in Filter IV of FIGURE 7 is a cooccurrence


restriction with the following formulation:

SURFACE STRUCTURE CONSTRAINT ON THE SWEDISH PASSIVE RULE

If SEi 'see visually', HÖRA^ 'hear auditorily', and a number


of verba dioendi and putandi (SÄGA 'say', PÅSTÅ 'claim', TRO
'believe', ANSE 'consider'etc) are the matrix verbs, then the
two possible passive constructions in connection with the
Swedish Subject Raising Rule are [...s-Passive...BLIVA Passive...]
So S

and the passives must occur in that order only.


263

12,3 A MODEL FOR DESCRIBING FORWARD AND BACKWARD OPERATING


TRANSFORMATIONS

In this investigation of Germanic ACI-Constructions the deriva­


tional steps from base structures to phonological surface struc­
tures in FIGURE 7 have been presented and described without
detailed formalism as far as t he base and transformational
rules are concerned. The forward path of derivation has been in
the focus of the generative description, i.e. synthesis, whereas
the reverse path of derivation from the ACI-Constructions to
their underlying representations i.e. analysis, has been
neglected. No attempts have been made to write rules for the
analytic process of derivation which the listener must be
assumed to master somehow. That is, he must be able to re­
construct a cognitive deep structure configuration of elements
from various surface structure manifestations (ACI-Constructions,
passivized ACI-Constructions, nominalized verb constructions,
complementizer embeddings etc). Some computational linguists ,
have claimed that so-called reverse transformational rules (T s)
can be allowed to operate on surface structure strings with
retained deep structure node labels in order to be able to retrace
mechanically the destroyed deep structure configurations in a
reverse path of derivation.19 There remains»however, the problem
of whether the dominating nodes of the syntactic configurations
involved should be kept after various transformational rules
have destroyed the preconditions for labelling the nodes in that
manner (cf. for instance the complement sentence symbol Si which
is deleted after Subject Raising or the object noun phrase of
the complement sentence which becomes the subject of the matrix
sentence after Subject Raising and Passivization:PaZme ansågs
göra en fjäskresa till USA 'Palme was considered to make a trip
of courtship to the US'.

Ross's 1969 tree pruning convention destroys the possibility of


reversing the path of derivation, that is of analysis,since
there is no way to reconstruct mechanically or formally the
nodes that have been pruned once they have been deleted.2^ For
analysis the origin of various surface elements, e.g. the origin
of the NP's, are to be retained through some kind of listener's
strategy which must reconstruct the deletions of the tranforma-
tional rules.21 it is not our intention to elaborate the

Principles of such strategies which are assumed to be at the


istener's disposal. Petöfi 1971:114-118 suggested that two
completely different models of derivation are needed for
synthesis and analysis. A generative semanticist model was
claimed to be more appropriate for synthesis, whereas a syntactic
model along the lines of the Standard Theory was considered to
be more adequate for analysis. Arguments against the latter
claim have been forwarded here in the discussion of the problem
of deleted nodes which are not formally recoverable, The whole
issue of which model is most appropriate for analysis or
synthesis cannot be solved by theoretical linguists alone, but
has to be developed in cooperation with language engineers who
construct machines for speech production or speech perception.22
264

What is desired by a linguist who has comparative linguistic


interests is a formal mechanism for comparing the surface
structures of different languages in terms of a given set of
syntactic transformations and (if possible) a common deep
structure which is syntactically motivated. It is practical
from a comparative point of view, if syntactic deep structures
could be elaborated which are common to the Germanic languages
under study so that a common syntactic frame of reference for
describing the complement structures could be obtained. Such
a comparative procedure has been adopted by historical
linguists in the reconstruction of Proto-Germanic and Proto-
Indo-European phonology and morphology (cf. Grimm 1819,
Rask 1818a, Bopp 1816 etc). We are convinced that such a
comparative procedure can be used in a synchronic syntactic
approach as well. If it is possible to m ake a synchronic
description within synchronic comparative syntax, then the
results and experience can also be used for diachronic compara­
tive syntax (cf. Ureland 1972b(forthcoming)).

For the purpose of comparative syntax (and not of psycho-


linguistics) the following diagram can be drawn to illustrate
the idea presented so far:

FIGURE 8
COMMON GERMANIC
BASE RULES

FILTER I: Aspectual Constraints on Lexical


Insertion

COMMON
ERMANIC DEEP
STRUCTURES

LANGUAGE^SPECIFIC DS LANGUAGE SPECIFIC DS LANGUAGE SPECIFIC DS


FILTER X: FILTER Z:

SS FILTER X: SS FILTER Y: SS FILTER Z:


SWEDISH^ GERMAN I ENGLISH^
ACI-Cons truction ACI-Construetion
(AWG-Construction) (AWG-Cons truetion) AWG-Construetion
Nominalization Nominalization
Complementizer Complementizer Complementizer
Embedding Embedding Embedding
265

FIGURE 8 reconstructs the derivational process of four Germanic


verbal complements which have been selected as examples of
comparative syntax: the ACI-Construction, the AWG-Construction,
the nomina aotionis, and the Complementizer Embeddings. If
given conditions of aspectual constraints (assumed to be common
Germanic) are fulfilled, the common Germanic phrase structures
are enumerated by the common base rules. Various language-
specific deep structure and surface structure constraints
indicated under the common deep structure tree are posited to
describe the difference in the occurrences of Germanic verbal
complements. In the sections on verba sentiendi it was
demonstrated that Accusative-with-Infinitive Constructions were
for instance more frequent in Swedish and German than in English.
English sometimes lacked this surface structure due to the
necessity of generating Accusative-with-Gerundive constructions.
This surface structure was, on the other hand, found to be only
a peripheral or even an ill-formed surface structure in Swedish
and German. The parentheses around the AWG-Constructions under
Swedish and German in FIGURE 8 symbolize this fact.

Nomina actionis are very idiosyncratic embeddings. Sometimes


embeddings through the nominalization of a complement verb can
take place in language X, whereas in language Y the semantically
related verb cannot undergo such a syntactic recategorization.
Compare the discussion of the Swedish nominal buller derived
from [bullr-] and the lack of a corresponding underlying verb
V V
for the English noun noise in (176c) in section 13.5. The empty
space in the column of English represents such a syntactic
idiosyncrasy.^

In section 1 through 11 of Part Two the syntactic and semantic


constraints on the generation of ACI-Constructions and AWG-
Constructions have been discussed for Swedish and German on
one hand, and English on the other. FIGURE 8 is an abstraction
of the detailed discussion found in Part Two. (For further
details see the treatment of surface structures in sections 1
through 11 of Part Two and the specific Swedish constraints in
section 5 through 15 of Part One).

Each surface structure presented here can thus be described in


terms of a common syntactic deep structure, a set of constraints
on all levels .(prelexical, pretransformational, posttrans-
formational, prephonological and postphonological etc), and
a given set of transformational rules. Instead of elaborating
surface structure differences of morphology, phonology, and syntactic
configuration, emphasis has been placed on what is common to
Germanic from a synchronic syntactic point of view. FIGURE 8
constitutes an attempt to abstract from all the language-
specific idiosyncrasies and stresses the common syntactic
characteristics.

The forward-operating rules can be thought to apply in a


reverse manner, although the formalism of generative grammar
does not allow a backward derivational process for analysis.
Such reverse derivation of surface structures remains to be
carried out by such linguists who are more concerned with
analysis than with synthesis.
266

12.1 COMPARATIVE DERIVATION OF VERBAL COMPLEMENT STRUCTURES

In order to describe how comparative (Germanic) syntax works,


a German idiosyncratic surface structure has been chosen for
the sake of demonstration. It is a well-known fact that
certain pseudoactive infinitives occur after German HÖREN
'hear' and SEHEN 'see1 which must be interpreted in a passive
manner. Compare the infinitives of the following sentences,
where passive infinitives yield well-formed complement structures:

HÖREN 'hear'
(115) a. plötzlich hörte sie ihren Namen rufen CR.Huch,
Triumph. 120: Bech 141)
(plötsligt hörde hon sitt namn ropas)
(suddenly she heard her name being called)

b. *plötzlich hörte sie ihren Namen gerufen werden

(116) a. man... hört einen Stuhl rüaken(E. Dwinger, Weiss.


93: Bech 141)
(man hör en stol flyttas)
(a chair was heardjto draw up)
Idrawing up

b. *man... hört einen Stuhl gerückt werden

(117) a. aber als er dann im Gerichtssaal sitzt und das


Mort Tod aussprechen hört (S. Zweig, Fouché 39:
Bech 141)
(men när han sedan sitter i rättssalen och hör ordet
död uttalas)
(but later as he is sitting in the court-room and
hears the word death pronunced)

b, *aber als er dann im Gerichtssaal sitzt und das


Wort Tod ausgesprochen werden hört)

SEHEN 'see'
(118) a. viele Gräber hat er graben sehen (E. Wiechert. Maj.
173: Bech 142)
(många gravar har han sett grävas)
(he has seen many graves dug)

b. *viele Gräber hat er gegraben werden sehen


267

(119) a. einst* als er spät nach Hause kam, sah er am Nach­


barhause Fässer3 Kisten, und Bündel in eine kleine
Britschka laden,(G. Freytag, Soll. 97: Bech 142)
(en gång när han kom sent hem, såg han vid grannhuset
fat, kistor, och knippen lastas på en vagn.)
(once when he came home late, he saw near the neighbor's
house barrels, boxes,and bundles being loaded onto the
chariot.)

b. *einst, als er spät nach Hause kam, sah er am Nach­


barhause Fässer, Kisten und Bündel in eine kleine
Britschka geladen werden.

A comparison between the German verb complement in (115) through


(119) on one hand, and the verb complements of the Swedish and
English translations on the other, shows that German lacks the
overt passive markers which are found in Swedish and English, and
which are obligatory in the latter languages. What is remarkable
is the occurrence of a pseudoactive infinitive in each of the
German (a) sentences which calls for an explanation. All of the
examples in (115) - (119) have been excerpted from literary texts
so it is highly likely that the pseudoactive infinitive is
confined to literary usage.

However, grammaticality tests of sentences (115) - (119) among


five educated native speakers of Standard German gave the
following results: All five informants accepted (115) as fully
grammatical (as grammatisch); three informants accepted (117)
as grammatical, one informant claimed it to be ungrammatical
(ungrammatisch), and one informant thought it was questionable
(zweifelhaft) three informants accepted (118) as f ully grammati­
cal, whereas two rejected it as ungrammatical; only one informant
accepted (119) as well-formed, whereas two informants found it
to be ill-formed, and the other two informants were uncertain
about its acceptability.

This test gives us an average of 3 for grammaticality of 5


possible responses to this pseudoactive German verb complement
structure among educated native speakers of German who hold at
least a maturation degree, i.e. Abitur. (The informants were
not informed that the test sentences were excerpted from
literary German texts, but they were asked to judge the
structures on the basis of their colloquial usage). The relative­
ly high rate of acceptance among educated native speakers of
Standard German illustrates the spread of the use of the Psych
Movement Rule. In this sense Psych Movement may be not only
literary.

In colloquial, unlearned style of speech the active construc­


tion exemplified by Sie hörte jemand ihren Namen rufen 'She
heard somebody calling her name' is the normal verb complement
structure, however.
268

The use of the pseudoactive verb complement structure exempli­


fied in (115) through (119) is some kind of a stylistic paradox.
By de leting the indefinite human subject noun phrase (jemand
'someone') which must be assumed to be the deep subject of
the complement verb, but without releasing the Passive Rule,
the remarkable German pseudoactive infinitive results after
Subject Raising. Instead some kind of Psych Movement takes
place which changes the status of the two objects after
Subject Raising and Indefinite Deep Subject Deletion (cf. a
different use of the term Psych Movement in Postal 1970 and
1971).

(120) a. Sie hörte [ [jemand] ruf- [ ihren Namen.J... 3


SlNPx NP. NPy NPySi
TSub j
4 Rais

b. Sie hörte C jemand] j [ihren Namen] prüfen


NPX * NP? . NPy NPy i
Indef +^,Psych
i NP-del Move

c. Sie hörte [ihren Namen] rufen


NPy NPy

The person who calls the name is so insignificant for the


message that any reference to him is deleted, and as a result
the focus is concentrated on the name which now becomes the
surface structure object not of rufen 'call', but of the matrix
verb hörte 'heard'. The obligatory final position of the
complement verb rufen which occurs after Subject Raising takes
place causes confusion as t o the syntactic status of the two
objects which occur after the matrix verb hörte in (120b), so
that a Psych Movement is necessary to d elete the less significant
indefinite NP. From being the object NP of rufen in (120a),
ihren Namen becomes the direct object of hörte in (120c). It
is unsure whether shallow structure (120b) needs to be postulated.

The significance of perceptual strategies for the envolvement of


new surface structures has recently been emphasized by
Bever & Langendoen 19 71. The present analysis reveals that there
is an interesting interdependence between the final order of
the complement verb in German and the object Psych Movement
Rule exemplified above in (115) through (119). If object Psych
Movement does not directly depend on verb-final order in German,
why is it that a similar Psych Movement Rule does not exist in
Swedish or English?

If the Passive Rule does apply in German and introduces a


passive auxiliary plus the past participle form of the comple­
ment verb, then an ill-formed sentence results. Compare the
German (b) sentences in (115) - (119). If an Aspects-model is
269

used for describing the non-cooccurrence of passive infinitives


after the matrix verbs HÖREN and SEHEN, then some kind of a
constraint has to be formulated which blocks the generation of
the (b) sentences above. We can see from the well-formed passive
Swedish and English translations of the pseudoactive German
infinitives that the constraint on the Passive Rule as it
applies to German complement sentences following these two
verba dicendi must be a language-specific constraint. Compare
Germ, rufen with Sw. ropas, Germ, rücken with Sw. flyttas,
Germ, aussprechen with Sw. uttalas, Germ, graben with Sw.
grävas, and Germ, laden with Sw. lastas.

The English translations do not contain passive infinitives


as the Swedish ones do. After Subject Raising and Passivization
a different surface structure occurs which we shall call the
Accusative-with-Past Participle construction. (Compare Germ.
aussprechen with Engl, pronounced (117a), Germ, graben with
Engl.dug (118a). In these sentences the passive auxiliary has
obligatorily been deleted (cf. the existence of the same surface
structure in Swedish in Part One, sections 5.2 and 15.2.1).
Another surface structure which occurs after Subject Raising
in English is the Accus ative-with-Gerundive construction which
contains a passive auxiliary in the form of BEING (Compare
Germ, rufen with Engl, being called (115a), Germ, laden with
Engl, being loaded (119a)). In (116a) the Passive Rule applies
to the matrix sentence so that after Agent Deletion an in­
transitive infinitive (draw up) is chosen instead of a transitive
(passive) infinitive as in Swedish.

Examples (115) through (119) demonstrate clearly that language-


specific constraints have to be formulated in order to account
for the considerable surface structure differences among German,
Swedish, and English. The differences in morphology reflect
basic syntactic constraints on all levels. Suppose no common deep
structure and no common transformational rules should be posited
for the verbal complements illustrated by the German sentences
and their translations into Swedish and English. Then one has
to give up the idea of describing common syntactic processes
as well. Since it is felt intuitively that a certain common
basis for the closely-related Germanic languages involved in
the present description can be established, and since certain
transformational rules used for deriving surface structures
can be said to be identical in all three languages (cf. the
existence of the Subject Raising, the Passive, and Passive
Auxiliary Deletion Rules) and since surface structure differences
often can be demonstrated to be determined by morphological
criteria, then a common derivational history of the German,
Swedish, and English surface structures in (115) through (119)
is motivated for comparative syntactic reasons. Such a
derivational history as demonstrated in FIGURE 11 A illustrates
partly what is common Germanic ( the syntactic deep structure,
the transformational rules: Subject Raising, Passivization,
Indefinite Agent NP Deletion etc) and partly what is language-
specific for each Germanic language (the Object Psych Movement
in German, the resulting constraint on the Passive Rule in
German, the obligatory character of the Auxiliary Deletion
Rule in English if an AWG-Construction does not occur.as
270

verbal complement etc). What is most striking when one compares


the Germanic surface structures exemplified in (115) through
(119) is the unconstrained nature of the Swedish Subject
Raising Rule in combination with the Passive Rule. Both German
and English show more constraints on the generation of infini­
tives here: German does not allow a passive infinitive at all,
and English does not allow infinitives of the passive auxiliary
BE. One can also say that the domain of the Swedish Subject
Raising Rule in combination with the Passive Rule is less
constrained in its application to complement sentences occurring
after verba sentiendi (HÖRA 'hear' and SEHEN 'see') as compared
to the equivalent rules in German and English.

The problem is how to derive the German infinitives in (115)-


(119). There is no motivation, from a morphological point of
view, for deriving these sentences in the rule sequence:
Subject Raising, Passivization and Indefinite Agent Deletion,
since no passive auxiliary nor any agent noun phrase are visible.
There are some semantic clues for giving these sentences a
passive interpretation irrespective of the lack of morpho­
logical evidence for the Passive Rule. Only human beings can
call out a name (115a), only human beings can pronounce a
word (117a), graves are dug by human beings (118a) and human
beings load chariots (119a). We stress the fact that the
sentences in (115) through (119) are assigned a passive
interpretation without being passive constructions. The ill-
formed (b) sentences which contain passive auxiliaries (werden)
verify the claim that the (a) sentences are not syntactic
passives. By postulating an Object Psych Movement Rule which
makes the object of the complement verb (the infinitive) the
object of the matrix verb (the verbum sentiendi) a surface
structure interpretation rule can assign the correct inter­
pretation to the mysterious German pseudoactive infinitives
in (115) through (119). The selectional restriction features óf
rufen, aussprechen, graben, and laden help the listener to
realize that the author of these sentences have applied the
Object Psych Movement Rule to generate the German active
infinitives.

However, if a human noun phrase is moved by the Psych Movement


Rule to a position immediately following the verb of perception
as in the following:

(121) Welcher Frau geht das Herz nicht auf, wenn sie ihr Kind
loben hört. (A. Droste-Hülshoff, Jud. 18: Bech 141)
(Vilken kvinna blir väl inte stolt, när hon hör sitt
barn berömmas.)
(What woman is not proud when she hears her child being
praised.)

It is not self-evident from the selectional restriction features


what syntactic status the NP ihr Kind-has. One does not know
whether it is the matrix object of loben or the complement
subject of loben. Of course this is only a hypothetical problem
since the linguistic situational context normally determines the
271

interpretation of such an ambiguous sentence so that the inten­


tion of communication on the part of thé author is correctly
conveyed. Text-theoretical considerations are indispensible for
assigning ('121) the interpretation intended by the author of
the text in which (121) occurs. An interpretative semantic theory
which includes only those structures between given sentence
bcmndaries is bound to fail to describe ambiguous sentences
like (121) adequately. (Cf. the criticism of interpretative
semantics as formulated in the Standard Theory by K atz 1964
and Chomsky 1965 by the text-theoretical linguists mentioned
in footnote 12).

In sentences (115) through (119) an active interpretation of


the occurring infinitives was excluded for semantic reasons:
a name cannot call, a chair cannot draw up, the word death
cannot pronounce, boxes and barrels cannot load etc. However,
in (121) a child can praise and be praised, so both an active
and a passive reading are possible here. We claim for these
cases that there is no passive construction which underlies
the passive reading.

The selectional restriction features which are inherent in


for instance the verb rufen 'call1 allow a speaker or a writer
to apply the Psych Movement Rule without a resulting triggering
of the Passive Rule. The human agent in Fillmore1s sense and
the object NP which denotes something that can be called e.g.
a name, give him this stylistic freedom, although the complement
verb, i.e. rufen happens to be homonymous with the active
infinitive (rufen).

However, since both the human agent and the human object are
covertly signalled by the selectional features of loben 'praise'
in (121) there exists a possibility for the speaker to carry
out a two-fold and therefore ambiguous embedding. With the
active reading of (121) only Subject Raising has taken place,
but with the passive reading both Subject Raising and Psych
Movement in our sense has occurred.

A derivation in terms of Subject Raising and Psych Movement is


more adequate than to have an ad-hoc shallow structure like
Sie hörte [Ihr Name wurde gerufen] which undergoes Subject
o o
Raising and Passive Auxiliary Deletion pljis Past Participle
Affix Deletion of /ge- en/. The latter deletion rules are not
needed in the Psych Movement solution suggested above.
272

12.5 COMPARATIVE VERSUS CONTRASTIVE SYNTAX

There are still further examples of sentences in Germanic


languages which obtain passive readings without being passive
constructions. In the following English and Swedish sentences,
pseudoactive verbs occur (sells and säljer) which cannot
be paired to a subject NP in an actor-and-action relation,
since a book cannot sell, only human beings can sell some­
thing: 24

(122) a. This book sells well.


b. Den här boken säljer bra.
c. Dieses Buch verkauft sich gut.

In spite of the active verb forms in (a) and (b), the sentences
must necessarily obtain a passive interpretation. The German
example in (c) also obtains such a passive interpretation because
of the reflexive pronoun.

The same syntactic paradox, the one that the NP which is expected
as a natural object category occurs as a subject category in the
surface structure without a passive form of the verb, is also
found in the English sentence (123a) below. Here it is also self-
evident that the play cannot read itself, or act itself, because
this action is performed by a human being only. Some remarkable
reeategorization of the transitive verbs has occurred in English,
so that they have obtained an intransitive character. In Swedish
and German it is absolutely impossible to use such a pseudoactive
form of the transitive verbs uppföra and aufführen and also läsa
and lesen respectively. A reeategorization is not possible in
the latter two languages.

(123) a. This play reads better than it acts.


b. *Detta skådespel läser bättre än det uppför.
c. *Dieses Schauspiel liest besser als es aufführt.

In (122a)-(122c) and (123a) the well-formed sentences can hardly


be explained from the standpoint of the surface structure
configuration, but the interpretation of the function of the
deep structure subject NP (the Actor in Fillmore's sense) and
the character of the verbal action must be explained in a
different, more abstract manner. With a generative approach to
the problem at hand one would claim that transformational
rules have changed the deep structure of (122a)-(122c) and
(123a) so that the syntactic paradox arises from having a
nonanimate and nonhuman NP as the surface subject-NP, although
no passive form of the verb is visible or audible. It is clear
that an abstract structure which is not visible or audible must
be posited to d escribe the cognitive process of interpretation.

The following underlying structures are postulated as being


closer to the semantic representation than the surface structures
exemplified in (122):
273

(122) a1. Pro sell [this book] well


[+Hum] NP NP

b'. Pro sälj- [den här boken] bra


[+Hum] NP NP

c1. Pro verkauf- [dieses Buch] gut


[+Hum] NP NP

(123) a1. Pro read- [this play] easier than Pro act
[+Hum] NP NP [+Hum]
[this play]
NP NP

The noun phrase this book, den här boken, dieses Buch have
a natural, syntactically motivated function as object NP's
in these postulated deep structures. In order to get to the
surface structures of (122a)-(122c), a major syntactic trans­
formation is carried out which changes the syntactic object
status of the NP's this book, den här boken, and dieses Buch,
so that they are moved into the subject position of the surface
structures. This movement is carried out under deletion of the
indefinite human subject of the deep structure. A considerable
stylistic effect is obtained by this movement. Let us call it
a Psych Movement Rule, since the psychological effect is so
great. What is remarkable here is the fact that the Passive
Rule is not triggered in connection with the Psych Movement
of the object NP. The active form of the main verb is maintained.
In contrast to the true Passive Rule only the deep structure
object-NP is moved. The deep structure subject-NP remains in its
original position but is deleted when the Psych Movement Rule
permutes the object-NP's this book, den här bokent and dieses
Buch to sentence-initial position. The semantically empty dummy
symbol for deep subject, Pro is thereby deleted.
[+Hum]

This psychological movement of the object-NP also operates in


German, with the difference that contemporaneously with the
movement of the deep object-NP dieses Buoh a trace of the NP remains
that is the coreference index which is then reflexivized
yielding sich. In Swedish no such copying of the moved object-
NP is possible, so no reflexive construction is well-formed.

The Psych Movement of the deep structure object the book into
the subject position can be said to be some kind of a
topicalization rule which gives rise to a metaphor. In the
deletion process of the human indefinite deep subject, the
feature [+Human] remains as a trace in all three languages and
it is this feature which then obtains a partial interpretation
as an agent, although from a semantic point of view no such
human interpretation would seem possible. The book will in this
way obtain a selling role and in a book seller's jargon such a
stylistic role is very expressive for his purposes.
274

A paraphrase of the Swedish sentence (122a) would bean s-form


of the verb, säljs, as in Den här boken säljs bra 'This book
is frequently sold1 or possibly 'This book sells well'. The
s-suffix after the main verb normally obtains a passive reading,
although in some instances it may also obtain the original
reflexive reading (e.g. Filmen upplöstes på grund av bristfällig
lagring 'The film dissolved because of careless storage') The
adequate German translation of the Swedish s-form is in terms of
a reflexive pronoun: Dieses Buch verkauft sich gut. We know that
in Old Swedish the reflexive readings of the s-suffis of main verbs
were extremely common, e.g. Nu lösis ben ur skenu. fylghir öris
bot (ögL VaJ>19: ö 86) 'If a bone loosens from the wound, then
a one-öre fine is paid'. Because of space limitation we cannot
go into the complex difference between passive and reflexive
readings of the Swedish s-suffix after main verbs, but refer
to a special investigation of the problem (cf. Ureland 1973
forthcoming, on the rise of the s-passive in Swedish verba
dioendi).

By positing a syntactic underlying structure, a cognitive


structure, it is possible to create a frame of reference for
explaining how the native speaker must match constituents with
each other. For the purpose of comparative syntax the postulation
of such underlying structures is practical. Using the under­
lying abstract structures as a point of departure a linguist
can then work out such syntactic rules as Psych Movement,
Copying, and Reflexivization Rules which are valuable for
describing surface structure similarities and dissimilarities
in an interlingual approach to Germanic syntax.

The comparative approach discussed so far can be illustrated


by the following figure:

FIGURE 9
DEEP STRUCTURE

Pro V
[+Hum]

sälj- bok- bra


sell- book well
verkauf- Buch gut

-pPsych
Move

Reflex
,T

SURFACE STRUCTURE

This book sells well


Den här boken säljer bra
Dieses Buch verkauft sich gut
275

By po stulating that the deep structure is common to the three


Germanic languages under study we can generate different surface
structures by applying three different transformational rules
under varying conditions. In order to illustrate the common and
language-specific processes at work in FIGURE 9, let us make the
derivational processes more explicit. FIGURE 10 illustrates how
the different rules of derivation are interrelated.

FIGURE 10
GERMANIC UNDERLYING STRUCTURE
S STEP I

TRANSFORMATIONAL RULES STEP II

English Swedish German


TPsych TPsych TPsych
Move Move Move
TObj-NP
Copy
^Reflex

i
MORPHOLOGICAL RULES STEP III
i
English
Present Suffix Present Suffix Present Suffix
Rule —* -s/
/ Rule —• /-r/ Rule—* /-t/

Dem. Pron. Dem. Pron. Dem.Pron.


Rule —•/this/ Rule —/den här/ Rule —• /dies/

Definite Art.
Suffix Rule—•/-n/

Pronoun Suffix
Rule—• /-es/

PHONOLOGICAL RULES STEP IV

Language-specific Surface Structures

English Swedish German STEP V

This book sells Den här boken säljer Dieses Buch ve rkauft
276

If such rules which are common to the three Germanic languages


are elaborated within a comparative syntactic model as sketched
in FIGURE 10, one can describe general syntactic and morpho­
logical phenomena in terms of common processes. That morpho­
logical rules generate different suffixes in different languages
is not surprising, considering the idiosyncrasies of morpho­
logical structures in each language. The syntactic rules are
elaborated so that they describe common processes, whereas the
morphological rules and the lexical rules are worked out accord­
ing to t heir language-specific nature.

In comparative syntax one can thus describe general syntactic


processes in order to capture what the Germanic languages have
in common under the assumption of a syntactic deep structure
which is pretty much identical.

In contrastive syntax it is not possible to by-pass Derivational


Step III in FIGURE 10, the morphological component, since the
concrete manifestation of the surface structures ceases at this
point in the derivational history.

Comparative syntax, on the other hand, is more abstract in its


generation and description of surface structures. It starts with
the posited deep structure and describes the similarities and
dissimilarities in terms of a number of transformational rules
and constraints on transformational rules. The advantage of
working with such a comparative model as sketched here is that it
also describes those structures which are similar on the surface-
structure level, even though no contrast is visible or audible.
For a contrastive analysis, however, only constrasting structures
are in the focus of interest.25,26

Contras tive syntax is important for describing such surface


structures which are different from each other in different
languages. For pedagogical purposes it is relevant that only
English uses a pseudoactive form of the verb, whereas in Swedish
and German other surface structures are necessary:

(124) a. This play reads better than it acts.


b. Detta skådespel är bättre som läsdrama än som teater­
drama.
c. Dieses Schauspiel liest sich besser als es sich auf­
führen lässt.

A comparative syntactic approach to such surface structures


includes a derivational history of the sentences involved. By
studying the derivational history one can more explicitly captu re
the stages at which the three Germanic languages are different.

The Psych Movement Rule just discussed is blocked in Swedish at


every stage of the derivation, whereas in German the Psych
Movement Rule is possible in the first clause which contains
lesen as t he main verb, since the clause Dieses Schauspiel liest
sich gut is well-formed after Psych Movement, Copying and
Reflexivization. The second clause of the German sentence which
contains aufführen as t he main verb cannot undergo these rules
Ill

to result in als sich aufführt, probably because it would


produce unacceptable homonymy with the phrase sich aufführen
which denotes another verb, i.e. 'to behave oneself'. This is
an interesting constraint on the selection of transformational
rules which has barely been treated in generative literature.
We are involved with the paradigmatic association of syntactic
and lexical units in the sense of Saussure 1916:173-175.
Instead of the Psych Movement Rule, another transformational
process must be chosen for the second German clause of (124c),
that is the causative construction by means of LASSEN: als es
sich aufführen lässt.

The existence of a literary Psych Movement Rule is unquestionable


and the active verb forms resulting from its operation obtain
a passive reading on the stength of the selectional restriction
features inherent in the verbs. As can be seen from (123) there
are language-specific constraints on its application, a
phenomenon which is to be expected from such a stylistically
conditioned rule as Psych Movement.
278

12,6 AN ILLUSTRATIVE EXAMPLE OF COMPARATIVE SYNCHRONIC SYNTAX

In order to further illustrate concretely how the model of


comparative Germanic syntax sketched in FIGURE 10 works, a
German sentence containing a pseudoactive infinitive after
SEHEN in an ACI-Construction will be given and compared to the
correspoiding Swedish and English structures.

(125) a. weil man Hormone hat herstellen sehen.


b. emedan man har sett hormoner framställas,
c. as t hey have seen hormones (being) produced.

Only in Swedish is a passive infinitive possible. In English


either an AWG-Construction or an Accusative-with-Past Participle
Construction is a correct translation, since a plain ACI-
Construction is ill-formed:

(125) c?. *as they have seen hormones be produced,

Furthermore, a passive ACI-Construction is also ill-formed in


German;

(125) a'. *weil man Hormone hat hergestellt werden sehen.

A deep structure constraint must be formulated to block passive


infinitives like (125a1) in German which occur after the two
verbs of perception, SEHEN and HÖREN. In order to be able to
generate passives after these verbs a German speaker has to
choose a different rule of embedding than the Subject Raising
Rule, that is the wie-Complementizer Rule:

(125) a", weil man gesehen hat, wie Hormone hergestellt werden,
b", emedan man sett hur hormoner framställes.
cM. as t hey have seen how hormones are (being) produced.

As c an be seen from the Swedish and English translations of


(12 5aM)all three Germanic surface structures are identical,
except for the German order of verbal elements. It therefore
remains to discuss the surface structures of (125) which will
serve as an illustration of a comparative syntactic derivation,
the surface structure differences being so great that a
structuralist or traditional grammarian can hardly see any
general similarities between the three surface structures under
discussion. Comparative synchronic syntax can demonstrate such
general similarity of derivation by claiming and positing that
a common underlying cognitive structure exists, a certain set
of universal constraints, a certain set of language-specific
constraints, a number of common transformational rules, and
idiosyncratic morphological rules for each Germanic language.
Of course the surface structures are different, but by over­
emphasizing the differences on the surface it is easy to d is­
regard the common syntactic and morphological processes which
279

are reconstructed in the table of derivation in FIGURE IIA and


IIB.

The derivational steps leading from the base rules to the


phonetic surface structures in (125) will be discussed step by
step in order to capture the general outline of the comparative
syntactic model presented so far.

A set of base rul<js common to Germanic is postulated in Step I


which generates the complex syntactic deep structures in Step III
which underlies each of the sentences in (125), provided that
in Step II the requirement of aspectual compat bility between
the matrix and complement sentences is met (cf. section 2.).

In Step IV t he language-specific lexical insertion rules insert


the language-specific lexical items into the deep structure
trees. Language-specific deep structure filters in Step V
constrain the general transformational rules of Step VI, so that
ill-formed syntactic structures are blocked at an early stage
of the derivational process. For instance in this step the
Passive Rule would be blocked from applying to German complement
sentences dominated by the verbs SEHEN and HÖREN (cf. *(125a')).
An aspectual constraint would also block the Subject Raising
Rule from generating (passive) ACI-Constructions in English
which are dominated by SEE and HEAR (cf. *(125c')), and instead
permit the correct AWG-Constructions (cf. (125c)). Also the
base rule generation of the passive auxiliary BE (cf. also
*(125c')) would be constrained in Step V.

In the boxes which symbolize the transformational rule component


(Step VI) the German Passive Rule and Agent Deletion Rule do not
apply. The corresponding spaces for the potential application
of these rules are therefore left empty in the derivational
table. After Subject Raising the shallow structure undergoes
the Psych Movement and Complement Subject Deletion Rules (cf.
(125a)), whereby the human pro-form is deleted and the object-
NP, Hormone 'hormones', occurs after the perceptual verb (SEHEN
or HÖREN). It is motivated to set up a constraint on the Passive
Rule in Step V for German complement sentences which are
dominated by SEHEN and HÖREN, because the insertion of the
passive auxiliary (werden) produces an ACI-Construction which
is unquestionably ill-formed.

(125) a'. *weil man Hormone hat hergestellt werden sehen.

In the Swedish and English derivational columns of FIGURE IIA


the application of the three transformational rules in Step VI,
that is the Passive Rule, the Agent Deletion Rule, and the
Subject Raising Rule are identical and need not be distinguished
here. There is no Psych Movement in these two languages which
generates the surface structures which match the German surface
structures.

In Step VII shallow structures appear which have not yet under­
gone various morphological rules in order to add suffixes of
various kinds to the verb steins acc ording to the morphosyntactic
280

FIGURE 11 A
BASE RULES STEP I

FILTER I: ASPECTUAL CONSTRAINTS ON LEXICAL INSERTION STEP II

COMMON GERMANIC DEEP STRUCTURE STEP III


LANGUAGE-SPECIFIC LEXICAL INSERTION RULES STEP IV

German: [weil man hab- seh- NP.[ NP. herstell- Hormone] ]


So I Si SlSc
Swedish: [emedan man ha- se- NP.[ NP. framställ-hormoner] ]
S0 I Si SxSq
English [as they hav- see- NP.[ NP. produce- hormones] ]
SlS0
1 Sl
FILTER HiLANGUAGE-SPECIFIC DEEP STRUCTURE CONSTRAINTS STEP V
German Swedish ^ English
1) Constraint on 1) 1) Aspectual
the Passive Rule Constraint on
in complement S Subj.Rais.
after SEHEN and
HÖREN
2) 2). 2)Constraint on
Pass.Aux.Segment.

TRANSFORMATIONAL RULES: STEP VI


German Swedish English
*
.Pass .Pass

,Agent .Agent
Del Del
,Sub j ,Sub j ,Sub j
Rais Rais Rais
.Object NP
Psych Move
,Compl Subi NP
Del

SHALLOW STRUCTURES: STEP VII


German: [weil man hab- seh- Hormone herstell-]
Sq So
Swedish:[emedan man ha- se- hormoner framställ-]
So So
English:[as they hav- see- hormones produce-]

L
So I So
j
MORPHOLOGICAL RULES: STEP VIII
German I Swedish English
(CONTINUED ON PAGE 28!
281

! FIGURE 11 B

MORPHOLOGICAL RULES: j | STEP VIII


I
German Swedish English
Tense Suffix Tense Suffix
Rule —r/-t/ Rule —•/-r/
Inf. Suffix Inf. Suffix
Rule —•/-n/ Rule —*/-a/
Passive Suffix Passive Auxiliary
Rule —f /-s/ Rule —•/BE/
Past Part.Suffix Past Part.Suffix Past Part.Suffix
Rule —•/-n/ Rule —>/-t/ Rule —*/-d/
Progressive Suffix
Rule */-ing/

SHALLOW STRUCTURES: 1 STEP IX



German: [weil man hab+t seh+n Hormone herstell+n]
S0 S0
Swedish: [emedan man ha+r se+t hormoner framställ+a+s]
So ]
English: [as they have see+n hormones be+ing produce+d
.s° i S0

POST-TRANSFORMATIONAL RULES: STEP X


German 4 Swedish J, English

Movement of verb
nodes to final
position in Ger­
man subordinate
clauses
THaben
Hopping
,Pass
Aux Del

FILTER III: SURFACE STRUCTURE CONSTRAINTS STEP XI


German Swedish English
Verb-Last
Constraint in
sub.clauses
Constraint on
final position of
tensed Aux in
sub.clauses I
+ *
SYNTACTIC SURFACE STRUCTURES STEP XII
PHONOLOGICAL RULES STEP XIII
PHONOLOGICAL SURFACE STRUCTURES
i STEP XIV
German: weil man Hormone hat herstellen sehen (125a) STEP XV
Swedish: emedan man har sett hormoner framställas (125b)
English: as th ey have seen hormones (being) produced (125c)
282

features of each Germanic language. A detailed presentation of


morphological features and rules has been omitted here for the
sake of brevity, but has been treated extensively in Bierwisch
1967 and Wurzel 1970 on German morphology, in Kiefer 1970 and
Linell 1972 on Swedish morphology. It is in Step VIII, the
component where the morphological rules are involved, that the
extreme idiosyncratic traits of each Germanic language occur.

The order of the morphological rules which operate in German


is such that the tense suffix /-t/is added to the auxiliary
h ab-: hab +t3 the infinitival suffix /-n/ is then added to the
verb of the complement sentence: herstell+n, whereby only the
past participle suffix is added to the matrix verb seh+n. As
a result the outspelling of the prefix /ge-/ is blocked, be­
cause there already exists an infinitive within the same
sentence, i.e. herstell+n. The Past Participle Suffix Rule is
crucially ordered after the Infinitive Suffix Rule.27

The Swedish morphological rules are straightforward and need


no additional explanation. They spell out the tense suffix
/-r/ onto the tense auxiliary: ha+r; the infinitival suffix
/-a/ is then inserted onto the complement verb framställ+a;
the Passive Suffix Rule then operates, adding the suffix /-s/
onto the infinitival form framställa+s.28 it is clear that
the Passive Suffix Rule must be ordered after the Infinitive
Suffix Rule. Otherwise the ill-formed sequence *framställ+s+a
would be produced. The Past Participle Suffix Rule adds the
suffix /-t/to the verbum dicendi: se+t.

The English morphological rules are fewer in number than are


the Swedish rules. No tense suffix need be spelled out for
the plural present indicative of /have/. No infinitive ending
exists in English, so empty space is to be seen here in the
English column of Step VIII. Instead of a passive suffix rule,
the Passive Auxiliary Rule spells out BE which is modified
to being by the Progressive Suffix Rule which is obligatory if
the mode of action is durative. The Past Participle Suffix Rule
spells out the suffix rule onto produce+d. By t reating the
occurrence of BE in connection with the Progressive Suffix
Rule and not generating BE as a basic auxiliary as d oes
Chomsky 1965, ill-formed structures like *(125c') are avoided.
Subject Raising can occur after Passivization only if the
Aspectual Constraint on this rule is not violated in FILTER II.
That is, only the durative mode of action allows the Subject
Raising Rule to operate. The same blocking of ill-formed
English ACI-Constructions is implemented by the second constraint
in FILTER II. No passive auxiliary can be transformationally
inserted into the deep structure if the mode of action is not
durative. Two transformational rules are blocked in this way
from generating passive ACI-Constructions in English (cf.*(125cf)

A number of post-transformational rules must be posited in order


to explain adequately the language-specific order of verbal
elements in German (cf. (125a)) and the optional deletion of
the passive progressive auxiliary in English being after the
Progressive Suffix Rule operates (cf. (125c)).

In derivational Step X the Verb Node Movement Rule and the


283

HABEN HOPPING Rule are posited to account for the final position
of German verbal elements in subordinate clauses whereby the
tensed verb occurs last. Especially HABEN HOPPING demonstrates
that this rule must be crucially ordered as a very late movement
rule, since it requires the presence of two infinitival suffixes
within the same sentence. That is, the two verbs with infinitival
suffixes must be dominated by the same sentence node.. Since
infinitival suffixes are spelled out in Step VIII in FIGURE IIB
after the transformational rules, HABEN HOPPING must be ordered
after their outspelling in Step VIII, i.e. HABEN HOPPING operates
in Step X.

The same crucial ordering refers to the deletion of the passive


progressive auxiliary kbeing in English (cf. (125c)), since a
passive auxiliary in the form of be must be blocked (cf *(125c')).
The passive auxiliary is generated in the morphological component
and is thus a post-transformational syntactic entity which
consequently may undergo a post-transformational deletion rule.

There are in Swedish and English no relevant surface structure


constraints in Step XI which correspond to the two German word-
order constraints indicated in FILTER III. The first word-order
constraint (language-specific to German) blocks any occurrence
of verbal elements in subordinate clauses which is not final,
The second word-order constraint (also language-specific to
German subordinate clauses) blocks the occurrence of the tensed
auxiliary hat or wird in a clause-final position, if two
infinitival forms are dominated by the same sentence node. The
latter word-order constraint on verbal elements in German
subordinate clauses can only be formulated with respect to
two infinitival morphemes, i.e. two occurrences of the infini­
tival suffix /-n/ within the same verb ph ra se.

Earlier transformational analyses of the order of verbal


elements in German starts out from another deep structure order
of verbal elements (cf. section 12.1), but by not including
the concept of word-order constraints, linguists like Bierwisch
1963 and Bach 1962 fail to give an adequate description of
the reasons for the two German word-order constraints exemplified
in (125a). It is impossible to g enerate a correct word-order
by only including deep structure aspects. A closer study of
their rules reveals that surface structure considerations are
indirectly included. However, the theory of transformational
grammar at that time did not allow for surface structure
constraints or morphological rules, so the two linguists
mentioned here cannot be blamed for the type of solution they
have proposed. The two word-order constraints just discussed
can be claimed to be surface structure constraints.30

The syntactic strings which result after the post-transformational


rules in Step XII now constitute the input to the phonological
rules of Step XIII, the operation of which is beyond the scope
of the present investigation.

The operation of the phonological rules produces the phono­


logical surface structures of Step XIV which in their turn
obtain a phonetic representation by a set of systematic phonetic
284

rules, if the generated sentence is to be spoken, but if it


is to be written the phonological surface structure obtains
a graphemic representation in the graphemic component as
suggested by Allén & Hellberg 1971 and Teleman 1972. It is the
graphemic representation which is visible in Step XV.

It is only the Verb-Last-Constraint in German subordinate clauses


which has been discussed here. It is clear that the Verb-Last-
Constraint on Tenseless Infinite Verbs such as infinitives and
past participles in German main clauses should also be
mentioned, since the end position of tenseless infinitives and
past participles is a language-specific characteristic of
German main clauses. In a main clause the verb order of the
German subordinate clause exemplified in (125a) would be the
following:

(125)a:1 Man hat Hormone herstellen sehen.


(They have seen hormones (being) produced.)
(De har sett hormoner framställas.)

As the English and Swedish translations indicate, the German


word order is specific to the German surface structure. From a
comparative syntactic point of view it is more appropriate to
regard the German position of the two tenseless verbs after the
object NP Hormone 'hormones' as a result of a late movement
rule than to regard the end position as the basic order of
verbal elements as in Bach 1962 and Bierwisch 1963.

By positing an underlying structure which reflects the co-


constituency of the auxiliary hat 'have'with the matrix verb
sehen 'see', the German deep structure can be kept the
same as t he English and Swedish deep structures. Furthermore,
it is much easier to formulate selectional restriction rules,
if a basic SVO order is chosen rather than ai SOV order as
suggested by Bierwisch and Bach. By introducing the concept
of a surface structure constraint on word order, the Tenseless
Verb-Last-Constraint, which acts as a filter in German main
clauses, a language-specific deep structure order in terms
of SOV as implied earlier in the treatments of German verb
syntax by Bierwisch 1963 and Ureland 1965 is no longer
necessary. A basic SOV order for German is not adequate for a
comparative approach to German, English, and Swedish surface
structures as exemplified in (125a'). (Cf. also arguments
against an SOV order in German in section 12.1 above).
285

13, THE SUBJECT RAISING RULE AS A FORMAL UNIVERSAL


The Accusative-with-Infinitive Construction which occurs after
verba sentiendi9 verba dicendi, and verba causativa is a
common surface structure not only in Germanic languages, but
also in most Indo-European languages. Comparative philologists
(e.g. Bopp 1816:38-87, Grimm 1837:105 ), Indo-Europeanists
(e.g. Jolly 1873:243-270, Delbrück 1897:450-471, Brugmann 1904:
§807, Wackernagel 1920: 257-265, Hirt 19 34:191-193), and
classical philologists (e.g. Woodcock 1958:14-35, Leumann-
Hofmann-Szantyr 1965:353-363 etc) have described this surface
structure as occurring in the major Indo-European language
families, that is in the Ayrian, the Greek, Italic,and Germanic
languages.

In order to gain a wider perspective of the domain of the general


syntactic rule which generates ACI as well as AWG-Constructions,
we will discuss the evidence of Subject Raising in modern and
ancient Indo-European languages and also in two non-Indo-European
languages, North Lapp and Finnish.

13,1 TYPOLOGICAL EVIDENCE


The Subject Raising Rule which generates ACI and AWG-Constructions
in a great number of Indo-European languages can be regarded as
a widely spread syntactic rule. The arguments for treating it
as a formal universal in the sense of Chomsky 1965 will be
presented here.

From a typological point of view it is motivated to regard the


ACI and AWG-Constructions as t wo surface structures which are
characteristic of the Indo-European langugage family. These
constructions occur after several classes of verbs, of which the
verba sentiendi, verba dioendi3 and verba causativa are the
most important. The conditions on the syntactic rule which
generates these constructions vary from one language to
another, so that in Russian for instance no ACI-Constructions
are possible after verba sentiendi. Complementizer embeddings
or AWG-Constructions must be selected as embedding structures
to yield well-formed Russian embeddings after these verbs.

13,2 HISTORICAL EVIDENCE


From a historical point of view the ACI and AWG-Constructions
can also be claimed to be general surface structures which are
characteristic of the Indo-European languages, since they
have been found in classical Greek, Latin and ancient Germanic
texts. (Of the Germanic texts the Gothic, Old Icelandic and
Old English texts constitute the most interesting evidence).
Unfortunately it cannot be proved that the ACI-Construction
was a verbal complement also in Classical Sanskrit, since no
clear-cut cases of this construction have been retrieved in
the Rìg Veda or other historical documents of Sanskrit.^
286

13.3 SUBJECT RAISING IN FINNO-UGRIC LANGUAGES


If we study in Finno-Ugric languages the occurrence of verbal
complements after verba dicendi and verba sentiendi, the
evidence of the Subject Raising Rule is more problematic,
since the verbal complements in North Lapp and Finnish present
different surface structures than do I ndo-European languages.
Instead of a clear-cut accusative form of the complement
subject, North Lapp and Finnish have a case form which is the
genitive case form. The two cases are often homophonous in
these two languages. Therefore, rather than an ACI or AWG-
Construction one may speak of a Genitive-with-Infinitive (GWI)
Construction after North Lapp verba dicendi (cf.footnote 32)
or Genitive-with Gerundive (GWG) Construction after North Lapp
and Standard Finnish verba sentiendi.

The Indo-European or Germanic ACI-Constructions which occur


after verba sentiendi must be transferred into syntactic
constructions in North Lapp and Finnish which are reminiscent
of the AWG-Construction in English, provided the Aktionsart
expressed in the complement and matrix sentences is c on­
temporaneous (cf. Nielsen 1926:384 on the durative-aspectual
constraint on GWG-Constructions in North Lapp). (This claim
will be weakened in the case of Finnish GWG-Constructions
as d emonstrated by (130) below).

13.3.1 VERBAL COMPLEMENTS AFTER NORTH LAPP VERBA SENTIENDI


The North Lapp examples of verbal complements in (126)-(127)
contain genitive or accusative forms of the noun phrase
following the verbs of perception, that is bâppa 'priest'
and a$ci 'father'. There is a smaller degree of consonant
strength in these oblique forms than in the nominative forms
bap'pa and ac'ci (gradation stage two is here unmarked for the
genitive and accusative cases, whereas gradation stage three
is marked for the nominative case by the apostrophe between
the obstruents). (Cf.Ruong 1970:21).

North Lapp
OAIDNIT 'see visually'

(126) a. Ài'dnen bâppa boattimin

res Parti
(I saw the priest coming)

b. *Åi'dnen bappa boattit

(I saw the priest come)


287

North Lapp
GULLAT 'hear auditorily'

(12 7) a. Mån gullen i dan mui'talae men.


r+N i
L+GenJ F+v
L+Pres PartJi
(I heard my father telling it.)

b. *Mån gullen â£ci dan mui'talit.

K,»] CLJ
The complement verbs occur in present participle forms boattimin
'coming' in (126a) and mui'talee men in (127a). However, the
infinitival forms of the complement verbs in *(126b) and *(127b)
are ill-formed in this position, that is boattit 'come' and
mui'talit respectively.

There seems to be a durative aspectual constraint on the Subject


Raising Rule in North Lapp which is similar to the durative
constraint demonstrated for English in section 2. Under the
assumption of a contemporaneous mode of action in both the
matrix and the complement sentences, ACI-Constructions are ill-
formed after the North Lapp verba sentiendi OAIDNIT 'see
visually' and GULLAT 'hear auditorily'.32» 33

13,3.2 VERBAL COMPLEMENTS AFTER FINNISH VERBA SENTIENDI


The Finnish complements in (128a) contain genitive forms of the
noun phrases, i.e. äidin 'the mother'* minun 'me', sinun 'you',
hänen 'him' or 'her' which occur after the two verba sentiendi
näki 'saw' and kuuli 'heard'. The genitive form is obligatory
when the noun cooccurs with the gerundive form of the complement
verb, here soittavan 'playing'. If the plain infinitive soittaa
'play' occurs as in *(128b) and *(128c) an ill-formed embedding
results. There is also a durative-aspectual constraint which
blocks the occurrence of the Finnish infinitive.3^

Finnish
NÄHDÄ 'see visually' and KUULLA 'hear auditorily'

äidin]
(128) a. Poika sinun soittavan pianoa.
hänen [+Pres Part]
[+Gen]

. x fthe mother
(The boyjheard! |you / playing the piano.)
' her/him

f äidin
b. »Poika 1 minun
soittaa pianoa.
I sinun [+ Inf]
Jhänen
I +Gen]
288

äidin
/näki I minut
(128) c.*Poika > soit taa pianoa,
Ikuuli sinut
hänet
[+Acc]

the m other
me
boy{^rd you play the piano.)
her/him

That the noun phrases after the verbs of perception in (128a)


are true genitive forms can be seen from the direct object
complements of (129a) which contain no embeddings and from the
plural forms of (129b) which contain embeddings. In (129a)
distinct accusative forms of the three pronouns occur after
näki 'saw' and kuuli 'heard', that is minut 'me', sinut 'you',
and hänet 'him' or 'her', whereas in (129b) distinct genitive
case forms occur, that is meidän 'our', teidän 'your', and
heidän 'their'occur. If the morphological form of the Finnish
word for 'mother' is compared in (128a) and (129a) we see that
the inflection is the same, although it occurs under different
cooccurrence conditions. In the plural, however, the full noun
for 'mother' obtains a special genitive case form, äitien 'the
mother's' in cooccurrence with a gerundive construction, as the
accusative plural case form is äidit 'the mothers'.

äidin
(129) a. Poika näki<®j™*
[hänet
[+Acc3

the mother
me
(The boy saw<
you
him/hei

äitien
meidän
b. Poika kuuli huutavan.
teidän
heidän
[+Gen]
r+Pl 3
the mothers
us
(The boy heard shouting)
you
them

It is appropriate to call the case form of äidin also genitive


in Finnish irrespective of the homonymity of the genitive and
accusative case forms of the singular noun, since a genitive
289

form appears after Pronominalization, e.g. we would get hanen


huutavan 'her shouting' and heidän huutavan 'them shouting'
from äidin huutavan 'the mother shouting' and äitien huutavan
'the mothers shouting' respectively. The accusative plural form
äidit huutavan would be ill-formed here.

What is important for our further discussion is t he fact that


the Finnish language requires an extra ordinary genitive case-
form of the pronouns and the plural nouns, if they happen to co-
occur with a durative embedding which in its turn is dominated
by the two verbs of perception NÄHDÄ 'see' and KUULLA 'hear'.

At first glance the Finnish Subject Raising Rule seems to be


constrained by the same aspectual durative requirement as the
corresponding rule in English in that Subject Raising must be
correlated with the Progressive Aspectual Affix Rule. However,
additional data on Finnish verb complements indicate that the
outspelling of the t?a-affix is not necessarily correlated with
the durative Aktionsart of the complement sentence, since the
same affix is spelled out also when the complement sentence
expresses instantaneous or iterative Aktionsart. As e vidence
for this claim we can compare the following Finnish embeddings:

(130) tönäisevän
Näin fMatinI [+Pres Part] ranskalais ta.
\hänen[ *tonäistä
[+Gen] [+Inf 1]
(I saw Matti shove the Frenchman)

lyövän
Näin fMatin [+Pres Part] oven kiinni.
1hänen ^*lyödä
[+Gen] C+Inf 1]
(I saw Matti close the door)

syovàn
Näin f Matinl ,[+Pres Part] makkaraa joka ilta.
\ hänenf *syödä
[+Gen] [+Inf 1]

(I saw Matti eat sausage every evening.)

In English it is impossible to spell out ing-suffixes in all


three examples of (130),under ai instantaneous or iterative
aspectual interpretation or the complement sentences. In
Finnish no such constraint on the outspelling of the va-affix
seems to exist. Its function seems therefore to be only to
signal the embedded status of the complement sentence. It
may be regarded as an accidental idiosyncrasy that the verb
form of the first infinitive may not occur in (130), since
all three ä-suffix examples of infinitive one are ill-formed
(the ä-suffix is due to vowel harmony between the stem v owel
and the suffix vowel whose base form is /-a/).
290

There is a close syntactic relationship between the gerundive


constructions after Finnish verba sentiendi and nominalized
structures after the same verbs. Compare the following sentences
in (131a) through (131c) as example of the close paraphrase
relationship:

(131) a. Poika. kuuli hänen. huutavan.


t+Gen]-' [+Pres Part]

(The boy-j^ hearcijher-'j shouting.)

b. Poika- kuuli hänen. huutonsa..


1 t + GenP [+N] [+Pois Suff]
(The boyi heardjjj^jl shouts.)

c. Poika. kuuli huutonsa.-


1 [+N] [+Poss Suff]

(The boy heard his own shouts)

The present participle affix /-va/ takes the accusative case


suffix /-n/ if the embedded gerundivized complement verb occurs
in the surface structure under the VP node after the verb of
perception {näki 'saw') as demonstrated in (128a). For reasons
discussed in connection with (130), it will be adequate to
regard the embedded verb after Finnish verba sentiendi as being
recategorized to a syntactic status which is contemporaneously
verbal and nominal, since the gerundivized verb undergoes the
Accusative Case Suffix Rule described in footnote 34 like a
nominal, i.e. a noun, pronoun or an adjective.

The object noun phrase hänen. huutonsa. in (131b) 'her shouts'


consists of the genitive fori of the personal pronoun hänen.
'his' or 'her' and the nominal huuto 'shout' plus a copy of the
possessive pronoun in the form of a possessive suffix /-nsa/
attached to the nominalized verb.(The accusative case suffix /-n/
is deleted here).
In (131c) the only remainder of the deep structure complement
subject is the possessive suffix /-nsa/, which is also attached
to the nominalized verb huuto 'shout'. The complement subject NP
has been deleted and transformed into a possessive clitical
suffix.

In the balanced view of treating nominals suggested in Part One,


section 13.5, a derivational process is motivated for the nominal
huuto 'shout', since an underlying verb exists, that is huutaa
'to shout'. After nominalization fromf+Vl to f-Vl the nominal
j_-NJ L+Nj
huuto can attract the complement subject after recategorization
as the possessive suffix /-nsa/.
291

The syntactic status of the gerundive huutavan 'shouting' whose


«-suffix theoretically could be accusative as well as g enitive
is more problematic, because it is both a verb and a noun. Since
it lacks a tense suffix and does not agree in person with the
complement subject and since it occurs after a genitive noun
phrase, that is hänen 'her', it may be appropriate to assign
it a positive value for both the syntactic features Such
a classification indicates that during the embedding operation
the verb is recategorized from being only a verb to being a
syntactic category which functions simultaneously as both a
verb and a noun. An analogous simultaneous positive classifica­
tion already exists in generative phonology, where the phonological
status of liquids is considered to be both consonantal and vocalic
at the same time, i.e. p-Vocalic 1
L+ConsonantalJ

13,4 THE TWO-SUBJECT HYPOTHESIS AND SUBJECT RAISING


By as suming a basic syntactic deep structure which contains two
coreferent subject NP's, the gerundive constructions in North
Lapp and Standard Finnish on one hand, and on the other, the
ACI-Constructions in Swedish and other Germanic languages, can
be derived by general operations. The surface structure
differences can then be described in terms of varying morpho­
logical and transformational rules.

For the Germanic languages it is adequate to describe the


infinitivization process in terms of a generally operating
Subject Raising Rule which has been discussed in Part One,
section 4.3. Such a description of the embedding process does
not always seem motivated in North Lapp and Finnish with respect
to the surface structure trees which result after the operation
of the transformational rules in question. Compare deep structure
tree (132a), the constituent structure of which can be motivated
from a comparative point of view, with the surface structure
tree (132b):
292

(132a)

N. Lapp: Mån gullen aS'ci mui'tal- dat-

S. Finnish: Poika kuuli äiti soitta- piano


Poika kuuli hänj huuta-
Poika- kuuli poika^ huuta-

t ,Nom TGen
1
r TCop TPoss t
11
NP .
Suff Del J
(13?b)

N. Lapp: (127a) Mån gullen inui'talae men dan

S. Finnish:(128a) Poika kuuli äidin soittavan pianoa


(131a) Poika^ kuuli hänen^ huutavan
(131b) Pcikai kuuli hänenj huutonsa^
(131c) Poika^ kuuli 0 huutonsa-
293

In North Lapp there is an additional surface structure con­


straint on the order of the verbal elements which is reminiscent
of the German Verb-Last-Constraint on complement verbs after
Infinitivization. In both North Lapp and Standard German the
tenseless infinite verb must occupy the last position, that is
the present participle and the infinitive respectively. Compare
the following North Lapp surface structure which results after
the Verb Last Movement Rule has applied to (127a) with its
Standard German surface structure translation:

(127) a'. Mån gullen atti dan mui'talae men.


f+Prol r+v 1
[+Acc] 1+Pres PartJ
(Ich hörte meinen Vater das erzählen)
(I heard my father telling it.)

Both the North Lapp personal pronoun dan 'it' (here in the
accusative case) and the German third personal pronoun das 'it'
occur before the infinite verb forms, the present participle
mui'talee men 'telling' and the infinitive erzählen 'tell1.
It must be considered a mere accident that the Verb-Last-
Surface Structure Constraint occurs in two completely unrelated
languages under complement sentence embedding. (It is not known
whether there has occurred word-order interference from Germanic
languages on North Lapp in the past when Germanic languages were
primarily SOV language?).

If the genitive case ending is not considerered to carry a


semantic interpretation, that is if the case ending is only a
surface structure phenomenon here, one might be tempted to
claim that Subject Raising also operates in North Lapp and Finnish.
The Subject Raising Rule may under these circumstances be
considered a formal universal which applies to deep structure
(132a).

The manifestation of accusative or genitive endings after the


raised NP's is due to arbitrary requirements within each language.
The occurrence of a gerundive form instead of an infinitival
form of the complement verb is also language specific for the
two Finno-Ugric languages in question. Whereas the case difference
can be said to be superficial, the latter aspectual difference is
more basically semantic and not only due to the occurrence of
the durative morphemes. The durative morphemes mirror fundamental
semantic differences in the manner of expressing the underlying
Aktionsart inherent in the complement sentences.

If, on the other hand, such a treatment of the North Lapp and
Standard Finnish gerundive constructions is rejected because
one prefers to claim that the genitive noun phrase together
with the gerundive forms of the complement verbs convey a
semantic reading other than that of the corresponding ^ACI-
Constructions in Germanic languages, one may indicate the close
syntactic relationship between gerundive constructions and
nominalized constructions which has been discussed above as
exemplified by (131a) on one hand, and (131b)-(131c) on the
other.
294

An alternative solution to that of Subject Raising just proposed


would involve claiming that the Genitive-with-Gerundive Construc­
tions undergo transformations which are similar to the rules that
generate the Genitive-and-Nominalization Constructions, that
is Genitivization of the complement subject NP and Nominalization
of the complement verb from "to F+n| * genitive case form
of the complement subject can now occur before the gerundive,
because the latter is partly a noun, and the gerundive ending can
be spelled out, because it is a lso partly a verb.

In the latter solution of the embedding process in Finno-Ugric


languages the Subject Raising Rule cannot be regarded as t he
embedding rule, rather it is t he Nominalization Rule applied
to the deep structure of (131a) which carries out the embedding
in a manner which is r eminiscent of the operation applied
to the deep structures of (131b) and (131c), i.e. Deep Structure
(132a).

There are three arguments from Finnish syntax which support the
Nominalization Hypothesis of the embedding process in Finnish.
The first argument for a recategorization of the complement
verb into a nominal concerns the embedding operation after the
verbum sentiendi NÄHDÄ 'see visually' which cooccurs with an
accusative form of the complement subject hänet 'her' and an
inessive case-form of the recategorized complement verb;, that
is the so-called 'Infinitive Three' in Finnish traditional
grammars the infinitive infix /-ma-/ plus the inessive case
suffix /-s.sa/.

(133) a.'\ "Näin hänet istumassa.


[+Acc] r+V"] [+ Inessive]
L+NJ
(I saw her (in the^ sitting.)

The infix /-ma-/ attached to the complement verb is traditionally


treated as an infinitive ending, but the syntactic status of
the recategorized complement verb is changed, since the inessive
suffix /-ssa/ may be spelled out. Only nominals such as nouns,
adjectives and pronouns take case endings. The complement verb
has therefore the character of a nominal in (133a).

The second argument supporting the hypothesis of a syntactic re-


categorization of the complement verb is the fact that Finnish
verba causativa such as PYYTÄÄ 'invite, ask' can also cooccur
with recategorized complement verbs which take the illative
suffix /-an/ as i n the following sentence, the so-called
'Infinitive Three' in classical grammars of Finnish; which
constitutes here the verb stem tuie-plus the infinitive infix
/-ma-/ plus the case ending (cf. Ikola 1971:63-73 on the four
Finnish infinitives).
295

(133) b. Hän pyysi miestä tulemaan.


[+Part][+V] [ + Illative]
M
(He invited the manfcton"mingi.)
v [to come J

The third argument concerns the embedding process in sentences


which contain the verbum dicendi SANCA 'say'. The embedding
process between (134a) and (134b) is more adequately explained
by Nominalization, Copying, and Complement Subject Deletion than
in terms of Subject Raising.

(134) a. Hän^ sanol [ hän. tuie- huomenna]


si J si

(He , said he come- tomorrow.)

b. Häni sanoi 0 tulevansa^ huomenna.

-, I coming his tomorrowl


(He
\ that he would come tomorrow/

Instead of genitivizing the complement subject in connection with


the Nominalization of the complement verb, the complement subject
is copied and deleted, so that a possessive suffix results after
the complement verb, i.e. /-nsa/. The complement verb occurs also
here in the gerundive form.

It would be ad hoc to assume that Subject Raising has operated


here instead of Recategorization of the complement verb followed
by Copying and Deletion of the complement subject.

In Finno-Ugric languages as well as in English there seem to


be morpho-semantic characteristics in combination which promote
Gerundivization, provided the complement sentence expresses
durative Aktionsart and the surface structure verb complement
appears with an overt progressive suffix, that is in English
/-ing/, in North Lapp /-imin/, and in Standard Finnish /-va-/.
If these deep structure and surface structure requirements are
not met, -violations of the durative aspectual constraint will
occur and ill-formed complement structures will be generated.

Although the phonological manifestations of the progressive


aspectual suffixes in English are extremely different from
those of the Finno-Ugric languages, an integrated syntactic
description has nevertheless been possible in the present
comparative approach to verb complementation in historically
unrelated and typologically different languages by postulating
common syntactic deep structures and language-specific syntactic
and morphological rules. ? ?
296

CONCLUSION TO PART TWO

1. SYNTACTIC RULES AS FORMAL UNIVERSALS


The comparative synchronic approach to verbal complementation
in Germanic and Finno-Ugric languages developed in Part Two
can be applied to languages irrespective of their typological
or historical relationship. The advantages of working within a
transformational-generative framework have been demonstrated
in the preceding sections. We have shown that although there
are considerable surface-structure differences between the
complement structures of for instance the Germanic languages on
one hand, and the Finno-Ugric languages on the other, one can
nevertheless describe the syntactic operations involved in the
generation of surface structure differences in terms of general
syntactic rules, reversing the non-similar characteristics to
be accounted for by language-specific constraints and language-
specific morphological rules. It is an open question whether
the syntactic rules focused upon here can be called 'formal
universais' in the sense of Chomsky 1965 or whether they
should be treated as language-specific rules. For the general
theory of comparative syntax it is more interesting to assume
that there exists such a set of universal syntactic operations
which are common to the languages under study than to claim
the opposite, i.e. that there are no such common syntactic
operations.

We have assumed that such a common set of syntactic operations


exists in Germanic languages and that they operate on deep
structures which are of a syntactic nature. The syntactic
operations with which we are dealing here are Subject Raising,
Complementizer Embedding, Nominalization, Genitivization, and
Copying. It is the task of a universal grammar to describe
what constitutes such common syntactic operations. Since no
such universal grammar exists at present, it is possible only
to speculate about universal syntactic rules on the strength of
the empirical evidence gained from studying the syntax of a
small number of languages. Then it is the task of the language-
specific investigations such as this one to describe the
constraints which are specific for the languages under study.
In the present investigation we have examined the constraints
on the Subject Raising Rule in three Germanic languages.

Throughout the present investigation the Two-Subject Hypothesis


has been assumed as a working hypothesis for describing and
explaining the infinitivization and gerundivization processes
in Germanic and Finno-Ugric languages. We have in this respect
followed the practice of generative grammarians who generate
infinitives by deleting or raising the complement subject noun
phrase (cf. for instance Bierwisch 1963:122-126, Chomsky 1968:
48, R. Lakoff 1968:29-32, Kiparsky& Kiparsky 1970:159-160 etc).

The Two-Subject Hypothesis is useful for comparative purposes


because by constructing syntactic deep structures which contain
both a matrix and a complement subject a common syntactic basis
297

is obtained for describing verbal complements in Germanic


languages. The deep structures posited in (37a) and (37b) in
Part One can be regarded as a kind of comparative universal
basis in the sense that they contain two subject NP's one of
which is either raised or deleted in the infinitivization and
gerundivization processes of all Germanic languages, and
perhaps even in all Indo-European languages. In Finno-Ugric
languages the second subject NP is assumed to be either genitiv-
ized, copied or deleted whereas the complement verb is nominal-
ized.
As^ far as the embedding rules are concerned which operate on
these deep structures there must be something fundamentally
universal about their operation, since we recurrently find
evidence of their application in extremely different languages.
Both Chomsky 1965:27-30 and Chomsky&Halle 1968:4 give vague
definitions of the term 'formal universal'. It seems that they
are more greatly concerned with a far more abstract grammatical
theory than with a comparative grammar of actual languages as
intended here, because they take an interest in the characteristics
"that determine the structure of grammars and the form and
organization of rules'' (cf. Chomsky & Halle 1968:4).

We are interested here more in the operations of embedding


rules and in the conditions unuer which they apply in Germanic
languages than in abstract speculations about language universals
from the standpoint of one single language, i.e. the English
language.

It must be regarded to be a mere accident that this or that


universal embedding operation has become a syntactic rule in one
language after a certain type of verb but not in another after
the same type of verb. For instance, why does Subject Raising
not apply in English to yield ACI-Constructions after verba
dioendi as is the case in Swedish. Compare the following
acceptable Swedish sentence with the unacceptable English direct
translation: Han sade sig förstå allt som sagts '»He said him­
self to understand all that had been said'. Or to take another
example, it must be an idiosyncratic characteristic of Russian
and Standard Finnish that no ACI-Constructions are possible as
verbal complements after verba sentiendi. In Russian complement­
izer embeddings by kak 'how' and &to 'that' are obligatory,35
whereas in Standard Finnish a Genitive-with-Gerundive Construction
is imperative (cf. section 13.3.2).

Even though Subject Raising does not apply to generate ACI-


Constructions in one Indo-European language and one Finno-Ugric
language, its universal character is indisputable. Subject
Raising is needed in Russian to generate AWG-Constructions.36
There are also abundant examples of its existence in Romance
languages, in Latin, in ancient Greek,and even in North Lapp,
a Finno-Ugric language, where ACI-Constructions occur after
verba dioendi and verba causativa (cf. footnote 32). Such
comparative evidence of the occurrence of the Subject Raising
Rule motivates us to regard Subject Raising as an extremely
frequent syntactic operation for embedding sentences into each
other in such disparate language families as the Indo-European
languages on one hand, and the Finno-Ugrian languages on the
other.
298

2, THE UNIVERSAL NOTION OF FOCUS


As a rgued in Part One, section 5.1.2 and 5.1.3 there are both
syntactic and semantic reasons in Swedish for treating ACI-
Constructions as being derived from a different underlying
syntactic structure than complementizer embeddings. The
syntactic and semantic arguments used in the discussion of
Swedish ACI-Constructions are language-independent, because
the same arguments are valid for German ACI-Constructions and
the limited number of English ACI-Constructions as compared
to German and Swedish complementizer embeddings. The same
arguments hold also for the English AWG-Constructions as
compared to their complementizer paraphrases. (Although North
Lapp and Standard Finnish demonstrate Genitive-with-Gerundive
Constructions, their complementizer paraphrases can certainly
also be claimed to express something different).

What is at stake in Germanic and Romance languages is the focus


of the speaker which is expressed by a raised NP as co mpared
to the same NP in a complementizer embedding. The need to focus
upon a given complement subject NP is common to all speakers
irrespective of their native language. Such an emotively based
need can be fulfilled by the surface structures which the ACI
and AWG-Constructions constitute after verba sentiendi in
Germanic languages and perhaps all Indo-European languages.
Complementizer embeddings do not seem to be able to fulfil the
need to express emotive focusing.

By ra ising the complement subject NP into the matrix sentence


a considerable stylistic effect is obtained which results in a
semantically ambivalent status of the raised NP as it con­
temporaneously performs the role of being the object of the
matrix sentence as well as th e subject of the complement sentence
in a semantic sense.

The following four Swedish sentences given in Part One and their
English translations present evidence of the emotive focusing
in the ACI-Constructions. They express a clear semantic difference
to the complementizer embeddings by att 'that'.
[70b) Plötsligen såg jag tre tyska heinkelplan dyka upp vid
horisonten 'Suddenly I saw three German Heinkel planes
appear over the horizon' and
(70a) Plötsligen såg jag att tre tyska heinkelplan dök upp vid
horisonten 'Suddenly I saw that three German Heinkel
planes appeared over the horizon';
(71b) tills jag såg rader av bomber börja falla 'until I saw
series of bombs starting to fall' and
(71a) tills jag såg att rader av bomber började falla 'until I
saw that series of bombs started to fall'.

In Germanic languages such examples provide adequate evidence for


claiming that Subject Raising has the semantic effect of focusing
upon the NP which is raised into the matrix from the complement
sentence. In Part One, section 5.1.3 three arguments are presented
which reject any hypothesis which assumes that semantic readings
of ACI-Constructions and att-Embeddings are identical.
299

First, the tense is missing in the ACI-Constructions; second,


the syntactic coconstituency of the complement subject and the
complement verb is destroyed so that the number agreement is lost
between these elements and as a result the complement verb is
infinitivized or gerundivized; third, the focus of the perceptual
process is concentrated on the noun phrase three German Heinkel
•planes in the ACI-Construction of (70b), whereas no such
specific focusing on the German Heinkel planes is obtained in
the att-Embedding of (70a), where the entire fact is in focus,
i.e. that the German Heinkel planes appeared over the horizon.

It is this possibility of focusing upon a special NP which is


universal in connection with the Subject Raising Rule in Germanic
languages. It is a characteristic of the languages studied here,
i.e. Swedish, German,and English. It is either the infinitives
which constitute the complement verb forms as in most Germanic
languages after verba sentiendi, or the gerundives as in Finno-
Ugric languages (and often in English among the Germanic
languages).

The surface case forms of the NP focused upon may vary from one
language to another; and even the syntactic rules which give
rise to the focusing effect may vary. Compare for instance the
Germanic Subject Raising Rule with the Finnish Genitivization
or Copying Rules in combination with the Nominalization Rule
discussed in section 13.3.2.

The semantic effect of focus can be explained within the


Extended Standard Theory, which has in the past few years been
constructed by Chomsky 1970b, 1971a, and 1971b, and Katz 1972.
In the latter version of generative grammar, semantic inter­
pretation rules are allowed to apply to deep as well as to
surface structures.

3. THE CORRELATION BETWEEN CASE SYSTEMS AND THE DOMAIN OF


SUBJECT RAISING
If one considers the great differences in phonology and
morphology between the various Germanic and Indo-European
languages, it is to be expected that a syntactic rule like
Subject Raising is bound to be subjected to a number of varying
language-specific constraints on all derivational levels.

However, it is not known to what extent specific morphologies


may promote or block the application of the Subject Raising
Rule in Germanic languages. It seems as i f an extensive case
system supports the use of the Subject Raising Rule for
generating ACI-Constructions, whereas concomitant with the loss
of an earlier case system the syntactic domain of this rule
is constrained (cf. Ureland 1972b(forthcoming)). Evidence from
the extensive use of ACI-Constructions in literary Latin, the
legal texts of Old Swedish, as well as spoken Modern Icelandic
supports the correlation hypothesis between a developed case
system and a more extensive domain of Subject Raising.36
300

In this context it is necessary to stress the importance of


literary influence from a language which is rich in cases upon
a language which is less rich in cases, e.g. Latin upon written
Old Swedish.

ti, COMPARATIVE AND CONTRASTI VE SYNTAX


The introduction to Part Two discussed the background of the
present comparative approach to the ACI and AWG-Constructions
in three Germanic languages (Swedish, German, and English) and
two Finno-Ugric languages (North Lapp and Standard Finnish).

We have acknowledged our debt to the abstract thinking in


historical and comparative linguistics of the past century and
the theoretical speculations about formal universals in genera­
tive linguistics of the past few years (cf. Chomsky 1965 and
Bach & Harms 196 8).

For arguments also presented in the Introduction to Part Two and


in section 12.5 we have preferred the term comparative synchronie
syntax to the term contrastive syntax9 because the latter term
too clearly implies an orientation to surface structures.
Comparisons are made in the present work more between the
syntactic operations and constraints on the syntactic operations
in Germanic ànd Finno-Ugric languages than between the surface
structures of these languages. Of course comparisons of surface
structures are also involved in our comparative-synchronic
compared but even those surface structures which are identical
need to be analysed. What we have strived for here has been
to establish the syntactic basis of similarity between the
languages involved. The term contrastive would refute such an
attempt, since it is founded on a structuralist concept of
languages as being totally different subsystems.

Comparative synchronic syntax can reveal what the nature of


syntactic operations is by describing the rules and constraints
of natural languages within historically related or historically
unrelated languages. By f ocusing more on the operation of general
syntactic rules than on the specific phonological and morpho­
logical phenomena, comparative syntax is capable of capturing
observable syntactic processes. The goal of the present study of
verb complementation in Germanic languages has been to disclose
the general nature of syntactic rules and general constraints
on these. Such derivational constraints were demonstrated to
be language specific, but some were also common to Germanic
languages (cf. for example the aspectual deep structure
constraint on lexical insertion of verbs described in section
2 of Part Two).
301

5. LEVEL CONSTRAINTS
Not only have syntactic operations like Subject Raising,
Complementizer Embedding, and Nominalization been elaborated
in Part Two, but also deep and surface structure constraints.
An example of a semantic deep structure constraint is the
condition on verb insertion and Subject Raising.

In section 2 the problem of lexical insertion is discussed,


whereby the durative-aspectual constraint on verb selection
is demonstrated to be a language-independent constraint, since
it is common to all Germanic and probably all Indo-European
languages. Such constraints which are identical can be compared
with each other in the languages under study, but not contrasted.
Therefore, also on the constraint level in syntax, it makes
more sense to use the term 'comparative' than 'contrastive'.
For instance, the English sentence*(77b)and its Swedish
translation are ruled out because of aspectual incompatibility
between the Aktionsart of the matrix and the complement
sentences: * {lib) He observed her break a vase *'Han observerade
henne slå sönder en vas'.

In English the generation of ACI-Constructions is blocked by a


similar semantic deep structure constraint, that is if the
Aktionsart of the complement sentence is durative, then Subject
Raising can only apply in connection with the Progressive Affix
Rule, a morphological rule, as in (54) He heard her playing the
piano but not in (51) * He heard her play t he piano. If the
hearing and the playing take place at the same time, then AWG-
Constructions are obligatory.

If however, the Aktionsart is instantaneous in the complement


sentence as in (55) He heard her shut the door or if the
Aktionsart implies an achievement on the part of the complement
subject as in (66) I saw her ooncoot a fantastic meal, then
English ACI-Constructions are possible, that is Subject Raising
must not be coordinated with the Progressive Affix Rule.

Thus it has not been possible to describe Infinitivization or


Gerundivization after verba sentiendi without considering
semantic characteristics. If such semantic characteristics as
duration, achievement, instantaneousness etc are not observed
there is no way to avoid violations of well-formedness.

Part Two also contains examples of another type of constraint


besides the deep structure constraint involved in Aktionsart,
that is a surface structure or shallow structure constraint
on the cooccurrence of a double s-passive within the same
sentence in Swedish. This constraint is discussed in section
12.2 and exemplified by the following sentence: (114a) *De
turkiska revolutionärerna sågs skjutas på TV idag 'The
Turkish revolutionaries were seen[to beìshot today on TV' and
Ibeingi
(114b) De turkiska revolutionärerna sågs bli skjutna på TV idag
'The Turkish revolutionaries were seen[being]shot today on TV'.
I to be!
302

Only the s-passive followed by a BLIVA-passive is well-formed.

It is not the passive as such which is ill-formed in *(114),


but the double occurrence of s-forms. In order to avoid such
an ill-formed double occurrence of s-passives, a surface or
shallow structure constraint is established and presented as
a filter (FILTER IV) in FIGURE 7. This filter blocks the
cooccurrence of Swedish double s-passives or the double
occurrence of two BLIVA-passives which is also ill-formed as
in (114 a') *De turkiska revolutionärerna blev sedda bli
skjutna på TV idag 'The Turkish revolutionaries were seen to
be shot on TV today'. Even the cooccurrence of a BLIVA-passive
plus an s-passive is ill-formed as in (114c) *De turkiska
revolutionärerna blev sedda skjutas på TV idag 'The Turkish
revolutionaries were seen to be shot on TV today'. The
sequential order, s-passive succeded by BLIVA-passive, is there­
fore crucial.

To enable us to explain such idiosyncratic surface structure


constraints on the Swedish Passive Rule in combination with
Subject Raising the filter concept is used, which allows only
those Swedish passives to pass which comply with the morpho­
logical requirements of not being two s-passives or two BLIVA-
passives within the same sentence after Subject Raising. The
burden now falls upon the shoulders of those who reject this
filter concept to show how without such a mechanism as proposed
here, the correct and only the correct Swedish passives can
be generated in the sentences exemplified here and in section
12.2.

6. ÀCI-VERBS, ASPECTUAL FEATURES,AND MORPHOLOGICAL RULES


FIGURE 1 presents a comparative Table of verba sentiendi in
Swedish, German, and English which take the ACI-Construction.
Some verbs are marked as taking this surface structure only
in a certain literary style (e.g. MÄRKA 'notice' in Swedish,
WITNESS in English, and MERKEN 'notice' in German).

Section 1. discusses the two underlying syntactic structures


for ACI and AWG-Constructions in Germanic languages. The
obligatory English morphological Progressive Affix Rule (cf.
(5)) is elaborated, which generates the ING-suffix onto
English complement verbs of a durative Aktionsart. The
corresponding Swedish and German morphological rules (cf. (6)
and (7) respectively) are also presented. The latter rules
are, however, optional in Swedish and German, except when
transitive verbs occur as complement verbs (cf. footnote 2).

The postulation of the semantic feature [+Durative] was crucial


for explaining how the English Progressive Affix Rule is
triggered.

In section 2. the base rules are formulated in such a manner as


to allow aspectual semantic features to be included in the
303

matrices. Such inclusion is also crucial to ensure that


appropriate verb classes are inserted into the phrase markers
and that the Subject Raising Rule is correctly applied, so that
ill-formed ACI or AWG-Constructions are blocked. A slight
reformulation of the Standard Theory base rules is all that is
necessary to accomplish the assignment of semantic aspectual
features to the matrix and complement verbs (cf. Base Rule
(12) in section 2.).

7. ASPECTUAL COPYING
Section 3. treats some German and Swedish cases where an
Aspectual Copying Rule is necessary to explain that some co­
occurrences of matrix and complement verbs are possible,
although they should have been blocked because of aspectual
incompatibility, e.g. the following German sentence:
(22) Ich merkte nicht Menschen neben mir vordrängen, andere
Hände wie Fühler sich plötzlich vorstrecken 3 Geld
hinwerfen oder einkarren (S. Zweig, Phantast. Nacht 150:
Bech 162)
(I did not notice that people were pushing past me, that
other hands than mine were stretching out like tentacles,
throwing or gathering in money)
On the basis of the hypothesis presented in section 3. we assign
aspectual features from one complement verb (e.g.vorstrecken
'stretch out' in (22) to other complement verbs dominated by
the same sentence symbol after conjunction reduction, i.e.
hinwerfen 'throw' and einkarren 'gather in' (cf. FIGURE 5 in
section 3.).

8, TEXT-THEORETICAL VIEWPOINTS FOR DESCRIBING ASPECTUAL


INTERPRETATION
Section 4. gives examples of an overt aspectual marker in
Swedish, i.e. FÄ, which can cooccur with a set number of
Swedish verbs (mostly verba sentiendi) as in sentence (29)
Han fick höra henne tala illa om sin väninna 'He heard her
speaking evil of her girl friend'.

The instantaneous, indeliberate Aktionsart expressed by the


Swedish auxiliary FÄ was demonstrated by the ill-formed
cooccurrence of an inherently durative main verb as i n (31)
*Han fick observera barnen leka på skolgården 'He observed
that the children were playing in the school yard'.

In German and English there are no such overt aspectual markers


which correspond to the Swedish FÅ, so the past tense forms
hörte, sah and heard, saw are used ambiguously with respect
to aspect. The aspect characteristics of these verbs can be
interpreted as c onveying a meaning of either instantaneous or
durative perception.
304

The German example (37) Ich hörte meine Mutter durchs Haue gehen
(p.Glaeser. Fri eden 98:Bech 139) 'I heard my mother walking through
the house' was found to contain such an ambiguous perceptual verb
form of HÖREN 'hear1.

In order to disambiguate the German ACI-Construction the entire


paragraph in which (37) occurs must be quoted so that we know
whether the hearing occurred the ^hole time that the mother
was walking through the house or whether the subject ich fI'
heard his mother walking through only a part of the house, i.e.
'fick höra henne gå genom huset'.

A lehgthy discussion then followed on how to assign durative


or ingressive Aktionsart to German examples of sah 'saw' which
are aspectually ambiguous in a limited linguistic context,
that is within a single sentence. Another example of such
aspectual Ambiguity was the following German sentence: (38)
Der General sah einen einfachen Soldaten von etwa fünfundzwanzig
Jahren vor sich stehen 'The general saw a private soldier about
twenty-five years old standing before him'. This sentence could
also be aspectually disambiguated by placing it into the
linguistic context in Kellermann's novel Der 9. November(p.103).
By doing so we could infer durative Aktionsart in the sense that
the visual perception was contemporaneous with the entire mode
of action of the complement verb stehen. The matrix verb sah does
not overtly express durative Aktionsart here. Reading the whole
paragraph in which the past tense form of sah occurs was
necessary in order to ascertain the aspectual interpretation
intended by Kellerman. Was a translation into såg or fick se the
correct translation into Swedish of the German past tense form
sah? Only the linguistic context could give us a clue to answer
this question.

Any description of verb complements like ACI-Constructions after


verba sentiendi which is one-sidedly confined within sentence
boundaries and which consequently does not include text-
theoretical viewpoints in the description will fail to explain
how aspect is assigned by the speaker or perceiver to the verba
sentiendi in sentences like the German sentence (38).

FIGURE 6 is a reconstruction of interpretive strategies which


the reader has to apply in order to assign correct aspectual
interpretation within a given text.

9. ASPECTUAL CONSTRAINTS AND PRESUPPOSITIONS IN ENGLISH


Sections 1 through 5 treat the ACI-Construction mainly in
Swedish and German, the generation of which seem to be con­
strained by similar conditions on the Subject Raising Rule.
The complex morphology in German and Swedish is assumed to
give rise to the syntactic similarity as far as ACI-Con-
structions are concerned. Due to the lack of such extensive
morphology, in this case an overt marker for the infinitive
ending, English has a whole series of language-specific
constraints on the generation of ACI-Constructions. To this
305

lack of an overt infinitival marker one must also add the


existence of the progressive gerundive construction, which
motivates a special treatment of the English ACI-Construction
together with its AWG-Construction which is rather unique
in Germanic languages. To what extent the AWG-Construction
and with it the Subject Raising Rule have been borrowed from
Latin or French into English cannot be answered without a
special investigation of the history of English. Many-
philologists are of the opinion that the AWG-Construction in
English is an inherent development (cf. Wood 1956, Visser 1966,
Mustanoja 1960 etc.). The comparative evidence presented here
about the ACI and AWG-Constructions in three Germanic languages
reveals that English deviates from the common Germanic verb
complementation system, and the possibility of syntactic borrowing
from Romance languages cannot be excluded as an explanation.(Cf.
Jespersen 1954b:178-214).
The domain of the Subject Raising Rule must be considerably
more constrained in English than in Swedish and German and
correlated with the Progressive Affix Rule which spells out the
ING-form onto the verb stems.

Section 6. discusses first the general constraint on the


Progressive Affix Rule which cannot apply to some verbs of
perception, although they are of durative Aktionsart ,38 e.g
HEAR and SEE in (52a)*I am hearing a noise and (52b) *1 am
seeing a big moose versus (53a) He was listening to the radio
and (53b) He was looking at the beautiful woman.

There is also a constraint on the occurrence of English ACI-


Constructions which is discussed in section 6. as for instance
in (55a) *He heard her play the piano which is ill-formed
if the hearing and the playing occur contemporaneously. The
English Subject Raising Rule must here be coordinated with
the Progressive Affix Rule (5). The outspelling of the pro­
gressive suffix occurs later than Subject Raising and there­
fore cannot possibly constrain ill-formed sentences like*(5£a)
in the underlying structure, rather we must posit some kind of a
shallow or surface structure filter wh ich blocks the occurrence
of ACI-Constructions and accepts the occurrence of ING-forms.
This implies that Subject Raising is a syntactic rule which
is sensitive to the morphological and phonological form of
formatives such as suffixes.

If, however, ingressive Aktionsart-is contained within the


complement sentence, an ACI-Construction is well-formed as
in (55b) He heard her shut the door, where the instantaneous
action allows Subject Raising.

Under a certain presupposition sentence (55a) may be well-


formed, that is the playing of the piano occurs at Carnegie
Hall for instance»so that a big performance is involved,
then the ACI-Construction is also possible as in (67) I heard
her play t he piano at the Carnegie Hall. The semantic
characteristic [+Achievement] is then a useful feature to
explain why (67) is well-formed but not*(55a).
306

For similar reasons the two sentences (65b) *The husband saw
his wife fix the lunch and (66b) The husband saw his wife con-
coot a fantastic meal are ill and well-formed respectively. It
is considered to be no accomplishment to fi x lunch, if a
woman does it (perhaps if a man does it), but concoct implies
an achievement and can therefore take an ACI-Construction as
verb complement.

Such restrictions on the generation of English ACI-Construction


have hardly been discussed in extenso in generative literature.
A one-sided enumeration of morpho-syntactic categories along
the lines of Chomsky 1965 is insufficient to explain the
complexities of the English Subject Raising Rule in its
correlation to semantic characteristics on one hand, and the
outspelling of the suffix ING on the other.

10, A MODEL OF COMPARATIVE SYNCHRONIC SYNTAX


Section 12. is the most theoretical part of the present in­
vestigation. Common Germanic deep structures of a syntactic
nature are posited for deriving ACI and AWG-Constructions in
Swedish, German, and English (cf. 12.1). The order of syntactic
elements in the Germanic deep structures is also discussed,
whereby the order of verbal elements in German is posited to
be of an SVO order for typological, historical, and comparative
reasons. Therefore, Bach's 1962 and Bierwisch's 1963 suggestion
to describe German verb syntax in terms of a basic SOV order is
rejected, since it conflicts with our comparative goal of
describing varying surface' structures in terms of a common
syntactic core, the Germanic deep structures, but language-
specific constraints and transformations. The proposal of an SOV
order of verbal elements is completely ad hoc for Swedish and
English and cannot be used as a basic order. In all likelihood
German verb syntax can also be described from the same basic
order as Swedish and English verb syntax.

Section 12.2 presents the model for the synthesis and the ana­
lysis of ACI and AWG-Constructions plus the Nominalization
structures. Four syntactic filters for blocking Swedish infinitival
constructions are exemplified, i.e. FILTER I which contains
the Aspectual Constraint on lexical insertion, (a language-
universal constraint), FILTER II which contains the Swedish
Equal Subject Constraint on Subject Raising (a language-specific
constraint), FILTER III which contains a shallow structure
constraint on the occurrence of definite agents in connection
with the Passive Rule and Subject Raising after Swedish HÖRA
'hear' (cf. *Byggnadsarbetare hördes av Palme ropa slagord mot
Sverige 'Construction workers were heard by Palme shouting
slogans against Sweden'), and finally FILTER IV which contains
a surface structure constraint on the cooccurrence of a double
s-passive in Swedish (also a language-specific constraint
due to language-specific morphology).

In section 12.3 a model for describing forward and backward


operating transformations is presented, whereby we reject the
suggestion that it is possible to allow transformational rules
to be reversed to retrace a backward path of derivation, if the
307

earlier deep structure node is pruned in the way which has


been suggested by Ross 1969. If the tree-pruning convention
is accepted, how can the nodes which have been deleted be
retrieved in analysis ? Somehow the deep structure nodes must be
maintained to allow for a reconstruction of the deep structure
configurations. The claim which has sometimes been made in
generative literature that the derivational history of a string
of syntactic formatives is somehow "memorized11 is unacceptable
for either a computational linguistic approach or a psycho-
linguistic approach.

If such"memorizing" were possible to program, why have no


tangible results been gained in the mechanical translation
from one language to another? It is clear that a generative
model of analysis which does not include deep structure nodes
cannot be used to reconstruct deep structure configurations
mechanically. Such a model is to be viewed as a mere artifact.

The model must include such deep structure information to be


able to reverse t he transformations. However, no detailed de­
scription of analysis has been intended here, since synthesis
was the primary aspect of our comparative approach.

11. LITERARY PSYCH MOVEMENT AS A RULE OF FOCUSING


Sections 12.4 and 12.5 give a number of examples of literary
structures which occur as pseudo-active structures, although
they obtain passive readings. Such surface structures are
exemplified by the German pseudo-active infinitive as i n (115a)
Plötzlich hörte sie ihren Namen rufen (R. Huch, Triumph.120 :
Bech 141) 'Suddenly she heard her name being called' and the
English pseudo-active main verb in (123a) This play reads
better than it acts.

To be able to explain how (115a) is generated,three trans­


formations are assumed which apply to an underlying syntactic
structure of this sentence, in which a human indefinite agent
is the deep subject of the complement verb rufen 'call', since
a name cannot call itself. After Subject Raising and Agent
Deletion in connection with a so-called Psych Movement Rule,
the remarkable pseudo-active German infinitive rufen results.
The Psych Movement Rule is necessary to explain why no passive
form of the complement verb is triggered in German as it is in
both Swedish and English.

It is clear that greater focus is centered upon the object


ihren Namen 'her name' if the complement subject NP is deleted,
that is the indefinite human NP. By the use of an active instead
of a passive infinitive a syntactic paradox is obtained which
has a great stylistic effect. It is thanks to th e selectional
restriction features that a correct semantic interpretation is
possible of (115a) and similar surface structures.
308

For the rise of this pseudo-active infinitive in German there


must be an interdependence between the final order of the comple­
ment verb in German and the existence of the Psych Movement Rule.
Perceptual strategies may have caused the rise of the German
Psych Movement Rule exemplified in (115a). (Cf. Bever and
Langendoen 19 71 on the interdependence of perceptual strategies
and the historical change of syntactic structures).

However, we can demonstrate that English and Swedish have a


different type of Psych Movement which also implies more
heavy focusing upon the moved object NP as in (122a) This book
sells well and in (122b) Den här boken säljer bra.

It is also evident that these sentences obtain passive readings,


as no books can sell themselves. Only human beings can sell some­
thing. The German equivalent requires that a reflexive pronoun
occur after the psych Movement has moved the object NP into the
subject position of the surface structure, e.g. (122c) Dieses
Buch verkauft sich gut.

A detailed derivation of the English, Swedish, and German then


follows in FIGURE 10 of section 12.5, which demonstrates how
comparative syntax works step by step from the deep structure
to the surface structure level.

A common syntactic deep structure is posited for the three


Germanic languages which contains a human agent as t he deep
structure subject. This deep subject is deleted by the Psych
Movement Rule in all three languages. In German an extra object-
NP Copying Rule is necessary to explain the occurrence of the
reflexive pronoun sich 'itself' in the verb phrase verkauft
sich gut.

In Step III language-specific morphological rules are demonstrat­


ed which spell out idiosyncratic endings in the language-
specific surface structures of Step V.

The literary flavor to the pseudo-active structures discussed


here is unquestionable. The Psych Movement Rule which generates
the syntactic paradox is to be seen as a rule of focusing in all
three Germanic languages.

Section 12.6 discusses a much more complicated problem of


comparative synchronic syntax, whereby the starting point of
syntactic comparisons is the deep structure underlying the
German sentence (125a)weiZ man Hormone hat herstellen sehen
'as they have seen hormones (being) produced'.

A language-specific constraint on the generation of German ACI-


Constructions is illustrated by the ill-formed sentence (125a')
*weil man Hormone hat hergestellt werden sehen 'as they have
seen hormones (being) produced'. German requires a complementizer
embedding here as in (125a") weil man gesehen hatSwie 1Hormone
fh 1 I das4
hergestellt werden 'as they have seenj J oh rmones are(being)
produced'. ' J
309

Comparative synchronic syntax can describe what constitutes the


general similarity of derivational processes. A linguist who
works within a comparative framework wishes to claim that a
common underlying syntactic structure exists which reflects the
cognitive, semantically relevant, structure. A number of common
transformational rules can be posited, and in certain cases,
language-specific transformational rules can also be posited.
There are also a number of language-specific derivational
constraints as for instance the constraint on the occurrence
of a passive auxiliary after German SEHEN 'see' exemplified in
(1253'). Besides these differences there are of course also
language-specific morphological rules which create surface
structure differences (the difference between the phonological
matrices of the three Germanic languages is disregarded here).

FIGURE 11 in section 12.6 gives us a reconstruction of the


numerous derivational steps involved in the generation of the
German sentence (125a) and its corresponding Swedish and
English structures. We refer to the details outlined in FIGURE
11 and the succeeding discussion for a general idea of how
comparative synchronic syntax can be carried out.

At the end of section 12.6 the HABEN HOPPING Rule in German is


treated as a post-transformational rule (cf. Step X). It cannot
operate before the outspelling of the two infinitival suffixes
which are dominated by the same sentence node.

Furthermore, the final position of German verbal elements is


considered to be due to a surface structure constraint which
is treated as being language-specific to German subordinate
clauses and which has no correspondence in other modern Germanic
languages. The movement of German verbal elements to sentence-
final position is therefore accomplished by a post-trans formation
rule (cf. Step .X) after Case, Gender and Number Agreement Rules
have operated and after various morphological rules have spelled
out varying verb and noun suffixes.

Rather than positing a basic SOV order for Modern Germanic as i n


former proposals (cf. Bach 1962 and Bierwisch 1963), for
comparative reasons we have posited an SVO order as the basic
order of verbal elements in German. In Part One, section 1.5 and
in Part Two, section 12.1 heavy typological and historical
arguments have been presented against a basic SOV order in
German. To these arguments we have added the comparative
synchronic argument, which emphasizes the common syntactic
core of Germanic languages rather than claiming a language-
specific deep structure order for German along the lines of
Bach and Bierwisch. Their arguments for an SOV order in German
are based on the criterion of descriptive economy within the
description of a single Germanic language. For a multilingual
comparative approach to t he syntax of several Germanic
languages such monolingual arguments of economy weigh light,
however, since the proposal of a German SOV order conflicts
with the goal of describing Germanic languages in terms of
common underlying deep structures which are syntactically
motivated.4 ü
310

12. INFINITIVIZATION AND GERUNDIVIZATION IN FINNO-UGRIC AND


GERMANIC LANGUAGES
Section 13. discusses the infinitive and gerundive constructions
in North Lapp and Standard Finnish. The surface structure
differences between Finno-Ugric languages and Germanic languages
are shown to be considerable with respect to the verbal comple­
ments after verba sentiendi. It is primarily the nominal character
of the Standard Finnish complement verb after embedding which
causes the major surface structure difference. In Modern
Standard Finnish as well as in Modern North Lapp the embedding
process is more adequately described in terms of Genitivization
of the complement subject and Recategorization of the complement
verb into a noun which takes case endings rather than in terms
of Subject Raising as described for the embedding process in
Germanic languages. Compare examples of singular object NP in
(128a) Poika näki (kuuli) äidin (minun, sinun, hänen) soittavan
ipianoa 'The boy saw (heard) the mother (me, you, him/her)
playing the piano' with those of plural NP in (129b) Poika
kuuli äitien (meidän^ teidän, heidän) huutavan 'The boy heard
the mother (us, you, them) shouting'.

The synchronic Standard Finnish embedding process can hardly be


said to have been carried out by the Subject Rule in (128a) and
(129b) although in an earlier stage of Finnish such a derivation
may be appropriate. The genitive case form of the complement
subject NP is known to have been an accusative m-suffix which
agreed with the same accusative case suffix of the complement
verb, e.g. näki poja+m soittava+m.41 Due to a historical phono-
[+Acc] [+Pres Part]
logical rule in Standard Finnish, however, the accusative case
suffix /-m/ was changed into the suffix /-n/ which became homo-
phonous with the genitive case suffix /-n/. Then through analogy
the singular genitive case form spread also to the plural (cf.
(129b). In some Finnish dialects this phonological rule has not
operated so that the original accusative case suffixes are still
maintained, e.g. in spoken Finnish the accusative plural suffix
/-t/ is acceptable as in Näki pojat soittavan 'He saw the boys
playing. [+Acc]

Also in Standard Finnish there are some traces left of the accusa­
tive case forms on NP occurring before embedded structures. It
is the accusative case form which occurs, not the genitive case
form, before Infinitive Three as demonstrated by (133a) Näin
hänet istumassa 'I saw her (in the) sitting'. For some reasons
[+Acc][+Inf3]
the analogical change from the accusative to the genitive case form
has not occured in the complement subject NP, if the Infinitive
Three /-ma-/ affix plus the inessive case suffix /-ssa/ follow
the complement verb stem. As a contrast to this state of being
compare the distinct genitive case forms of the complement subject
NP's minun 'my', sinun 'your' hänen 'his or her' in (128a) and
äitien 'the mothers*', meidän 'our', teidän 'your' and heidän
'their' in (129b).
311

It is possible that the embedding process in the earlier stage


of Standard Finnish can be described in terms of Subject Raising,
since accusative case forms were the overt case manifestations.
However, the synchronic Standard Finnish process of embedding
differs from the earlier embedding rule in that now the genitive
case form is obligatory in the Standard Finnish dialect, as
argued in section 13.4. The modern process of embedding is more
adequately described by Genitivization, Copying and Recategoriza-
tion rather than in terms of Subject Raising.

The rise of new syntactic rules of embedding can thus be claimed


to be due to the interaction of phonological rules (e.g. the
change of /-m/ to /-n/ word-finally) and the mechanisms of
analogical change. Vennemann, Antilla and others have recently
revived the old neogrammarian concept of analogical change. Some
linguisitc changes can affect one subsystem but not another.
Modern synchronic syntactic research can certainly profit from
the insights gained in historical syntax and vice versa. The
apparent irregular verb complement system after Modern Standard
Finnish verba sentiendi can be discussed with much more explana­
tory rigor in the light of historical development. Synchronic
rules of embedding can be better understood if the historical
background is also taken into consideration.
312

FOOTNOTES TO PART TWO

1. S. Eliass on o f Uppsala has kindly informed me t hat Zabrocki 1970 has


suggested the term confrontati ve linguistics for a non-applied abstract,
typological field of linguistics which deals with linguistic systems
(cf. a l s o Eliasson 1972a:9)

2. A p r e s e n t p a r t i c i p l e form o f a t r a n s i t i v e verb may b e i l l - f o rm e d a s


a verbal complement structure in Swedish and German . Compare t he
following AWG-Constructions:

(i) *Han såg henne klädande på s i g

(ii) *Er sah s i e sich anziehend

( i i i) He saw h er g e t t in g dressed

3. As indicated in Part One, section 5.^. 2 Chomsky's 1965 category Aspect


is based on a language-specific bias about the morpho-syntacti c structure
in English. It i s inadequate to treat the morpho-syntacti c category,
the ING-form in En gli sh,as being generated by the base r u l e s . For a
comparative approach t o verbal complementation in Germanic lang uages i t
is more app ropriate t o keep the base rules as general as possible and
work ou t specific morphological rules to describe the difference in
sur face s t r u c t u r e s , s i n ce German and Swedish lack an o v e rt aspectual
marker l i k e t h e d u r a t i v e s u f f i x ING ( c f . f o o t n o t e 2 ) .

k. In E ngli sh i t i s the imperfective-durati ve Akti onsart that must be


marked a s the aspectual category, whereas in Slavic languages l i k e
Russian and Polish i t is the perfective-instantaneous Aktionsart that is
themarked as pectual category. The imperfective-durati ve Aktionsart i s
thus the unmarked as pectual category in S l a vi c languages ( c f . a l s o
Part One. section 5. ^.2).

5. For information on the history o f English syntax I thank B. Wik o f Umeå,


who has provided me with valuable bibliograp hical ref erence s.

6. Vendier 1967 has clarified the difference between a ccomplishment verbs


(e.g. reach) activity verbs (e.g. run), and s tate term verbs (e.g. love),
but without drawing the conclusion that a d i s t i nc t i o n must a l s o be made
between these semantic aspectual features and surface structure
charac teristics such as inflectional endings like the ING-form.

7. Some n a t i v e s p e a k e r s o f Swedish d o n o t a c c e p t AC I - C o n s t r u c t i o ns a f t e r
OBSERVERA, MÄ RKA, a n d IAKTTAGA. T h e A s p e c t u a l C o n s t r a i n t o n S u b j e c t
Raising (cf. (15)) can therefore be said to be a totally blocking
factor in these speakers.(Cf. the literary evidence of ACI-Constructions
a f t e r MÄRKA i n P a r t O n e . s e c t i o n 7 - 1 • ) .

8. The primary and secondary ßources of the literary evidence presented


here are given after the examples within parentheses. The f i r st abrévia­
t i o n re fer s t o the author, e . g . G. Freytag f or Gustav Freytag; the second
abbreviation refers t o the novel, e . g . Sol 1 for Sol 1 und Haben, and the
third abbreviation refers t o the secondary source, e . g . Bech for Bech 1955
313

9. Cf. Part One, section 5.4. 2 on the occurrence of aspectual auxiliaries


for ingressive Aktionsart in Germanic lan guages.

10. Other examples of overt aspectual markers in Scandinavian languages are


f o u n d i n Modern I c e l a n d i c , w h e r e t h e p e r i p h r a s t i c c o n s t r u c t i o n v e r a adf
expresses durative mode o f action: é g ver aà l es a '1 was reading1;
f a r a adf e x p r e s s e s i n c h o a t i v e mod e o f a c t i o n é g f e r a à s o f a ' I am
going t o s l e e p 1 , and f i na i l y vera buin a é expresses terminative mode
o f a ct i o n é g e r bu inn aë borda* 11 am through e a t i n g 1 . In Mode rn
Swedish the periphrastic construction hålla på a tt expresses a dura­
t i v e mode o f a c t i o n which corresponds t o t h e E n g l i s h BE + ING-
C o n s tr u c t i o n and t h e I c e l a n d i c v e r a ad': han h ö l l på a t t koka k a f f e ,
när j a g kom fra m 'he was making c o f f e e , when I e n t e re d ' . (See Einarsson
1967:153-1^6 on overt aspectual makers in Modern Ice landic).

1 1 . C f . P a r t One, f o o t n o t e 5 2 o n t h e s y n t a c t i c s t a t u s o f FÂ i n Norwegian.

12. Sentence-oriented descriptions o f syntax fail to gi ve adequate explana­


ti ons of a whole s er i e s o f grammatical phenomena. For a discussion o f
the importance o f including text-theoretical viewpoints in the de­
s cr i p t i on o f a r t i c l e s el e c ti o n , topic-and-comment phenomena, pro­
nominal ization, coreference index assignment e t c see Heidolph 1966,
Isenberg 1968, Petöfi 1971a and 1971b, Dìjk 1972 and Ihwe 1971.

13. There is an interesting problem o f presupposition involved in the


s e l e c t i o n o f t h e t w o c o m p l e m e n t i z e r s t h a t o r a s i f o c c u r r i n g a f t e r FEEL
in a sentence l i k e (42b). If a person re ally f e e l s as a communist does,
he will use the complementizer that. If, however, he fe els like a
communist o n l y under c e r ta i n circumstances , he w i l l use a s i f . In
Swedish t h e AC I - c o n s t r u c t i o n i n ( 4 2 a ) o r t h e p o t e n t i a l aitt-Embedding
e x p r e s s e s t h e f e e l i n g o f being a communist. An embedding by means o f
som o m would n ot infer such a presupposition, however.

1 4 . U n d e r g i v e n c i r c u m s t a n c e s t h e s e n t e n c e (69) I s a w h e r p l a y i n g t h e p i a n o
a t Carnegie Hal l may be w e l l - f o r m e d , provided t h e o b s e r v e r s e e s t h e
woman a s she i s p ra c t i ci n g a t Carnegie H a l l .

15. A l s o (70) c a n b e w e l l - f o r m e d u n d e r t h e p r e s u p p o s i t i o n t h a t t h e o b s e r v e r
d i d n o t pay much a t t e n t i o n when t h e woman was p l a y i n g , and t h a t h e l e f t
before she was through playing.

1 6 . AC I - C o n s t r u c t i o n s seem t o b e a c c e p t a b le a f t e r an i m p e r a t i v e form o f
LOOK A T : L o o k a t h i m g o , j u m p e t c .

1 7 . The o c c u r r e n c e o f t h e s u b j e c t NP Queen E l i za b e t h i n t h e s e n t e n c e a u t o ­
ma tically e l e v a te s the s t y l e . This may be the reason that a pla in ACI-
C o n s t r u c t i o n i s t o l e r a t e d a s a v e r b c o m p l e m e n t a f t e r WATCH i n ( 7 0 c )

1 8 . R o s s I 9 7 2 s u g g e s t s a so - c a l l e d s u r f a c e s t r u c t u r e c o n s t r a i n t i n E n g l i s h
on the repetition of present participles (fNG-forms) without con­
sidering the durative Aktionsart inherent in the verbs in question
(cf. p. 61). Contrary to his claim the constraint on the surface
structure configuration * l t i s continuing raining is blocked not by
the cooccurrence of two ING-forms but by the fact that the durative
Aktionsart inherent in continue blocks the outspelling of the pro­
gressive suffix, since the single verb phrase * l t is continuing is
also M 1-formed.(Cf. our discussion o f the constraint on the English
morphological rule which spel ls out the durative ING-suffix in section
6.).
314

19. Such a backward op erating mechanism was dis cussed by S . Kuno in a


lecture given a t the Department o f Linguistics, University of Texas
a t A u st i n , i n May 1968.

20. Generative grammarians o f the MIT-school would probably claim that


t h e d e r i v a t i o n a l h i s t o r y o f a g i v e n AC I-Const r u c t i o n c o n t a i n s - a — ^
"memory o f d e l e t e d nodes". I t i s c l e a r t h a t such a c l a i m has l i t t l e
or no psycholinguistic value. Also in computational linguistics no such
memory d e v i c e can be assumed t o be p a r t o f t h e p r o c e s s i n g o f t h e
computer. One cannot infer the mind t o be l i k e a computer o r v ic e versa.
Without psycholinguisti c testing no hypothesis of the relationship
between synthesis and a nalysis can be reliable. Results from recent
psycholinguist i c research demonstrate that a diff erent model i s
needed for speech perception than for speech production (cf. Bever
I97O and foo tnote lo» Part One). I t i s nai ve t o b e l i e v e t h a t there
is a single given linguistic theory which is capable of explaining
a l l f ac e t s o f language. Chomsky 1965:177 discu sses the 1 recoverabi1 i t y
condition1 for deletions without drawing the consequences for analytic
processes from su rface t o deep structures. His formalism does not
allow retrieval of deep structures from su rface structures.

21. The p sycholinguist i c strategies in processing a given surface


structure are so complex that they cannot be dealt with here.
Suffice it therefore to stress the extremely complicated set of
strategies which are applied in analysing a given surface structure.
I t may ev en be t h e c a s e t h a t t h er e i s a d i s p a r i t y between r u l e and
strategy-directed behavior, so that i t makes l i t t l e sense t o speak o f
forward-operating rules in synthesis which operate backwards in
analysis. (Cf. Whitaker 1971:21).

22. A ge nerative grammar o f e i t h e r the Standard Theory format o r o f the


Generative Semanticist format is inappropriate for any application
t o m a c h i n e t r a n s l a ti o n o r m e c h a n i c a l s p e e c h p r o d u c t i o n o r p e r c e p t i o n .
Also for psycholinguistic or neurolingui st i c purposes o ther types
of grammatical models are presently being developed for describing
the strategies of the speaker o r listener, (Cf. the articles in
Hayes (ed.) 1970 and S lobin (ed.) 1971)»

23. Of c ourse the s i t u a t i on can be q u i t e the contrary s o that co lloquial


English has a nominal which can be derived from a verb, whereas in
German o r Swedish no underlying verb e x i s t s ( e . g . English t o tap e-
record and tape recorder but Swedish *bandspe1a and bandspelare and
German * zu tonbandgeräten and Tonbandgerät).

2k. The Swedis h use of a pseudo-active verb form lik e sä lje r in (122a) i s
probably due to syntactic borrowing from E nglish commercial jargon,
i . e . the English Psych Movement Rule has been borrowed int o Swedish
in a certain commercial style.

25. Examples of such contrastive works are Fries 19^5, Lado 1957» Kufner
1962, Stockw ell & Bowen & Martin 1965, König 1971» Nickel 1972a,1972b
et c. For further references on contrastive linguist ic works see
Hammer & Rice 1965, E l i a s s o n 1972band E l e r t 1967« Most o f t h e works
mentioned here are written within a structuralist framework, a lthough
some of the contrastive analyses include transformational concepts
a s well ( e . g . Stockwell & Bowen & Martin 1965, König 1971 e t c ) .
315

26. For information on contrastive linguistic literature I thank Stig


Eliasson, Uppsala.

27. Bierwisch I963 fi rst generates the past participle suffix symbols
(=IN2) and then a special transformation (T20) changes the past
participle symbols of gehört 'heard* or gesehen 'seen1 to the infinitive
s u f f i x s y m b o l ( = I N -j ) , p r o v i d e d a n i n f i n i t i v a l f o r m o f a n a u x i l i a r y i s
contained w ithin the same se ntence. It i s clear that the formalism
used by Bierwisch in his T20 rule involves surface structure con­
siderations like the morphological form o f the suffixes. Through the
introduction of level constraints and mor phological rules into syn­
tactic theory such ad-hoc transformational rules as T20 ca n be avoided,
however.

28. All generative descriptions of the Swedish s-passive are superficial and
s o far unsatisfactory from a syntact i c point of view. Linguists who have
worked on the Swedish passive are concerned mostly with the formalism
of the morphological rule which spells out the s-suffix (cf. for instance
Teleman 1969» Kiefer 1970» Linei 1 1972 e t c ). The c onstraints on the
use of the s - or BLIVA-passi ve remain t o be discovered (c f. section 12.2 ).
(Cf. historical treatments of the Scandinavian passive in Nygaard 1905»
öhlin 1918, Holm 1952, Wessen 1965, and Markey 1969).

29. Cf. foothote 27.

30. B i e V w i s c h 1 963 d i s c u s s e s t h e c o m p l e x i t i e s o f t h e H A B E N - U m s t e l 1 u n g in
G e r m a n ( c f . p . I0 8). H e r e f e r s t o s u c h s u r f a c e c a t e g o r i e s a s t h e s u f f i x
symbols IN] and IN2 fo r the i n f in i t i v e and p ast p ar t i ci p l e endings
respectively in his formalism. However, a s pointed out in footnote 27
these symbols stand for morphological properties which cannot be referred
to unless surface structure aspects are included in the description.
The t heoretical framework wi thin which Bierwisch worked in 1963 did not
allow for such surface structure considerations in elaborating the
transformational rules. The c oncept o f surface structure constraints
had n ot yet been discovered. Therefore the Bierwisch 1963 formalism
cannot possibly work, as only deep structure aspects could be included
in the transformational rule mechanism. It does not help i f one then
c a l l s the morphological a f f i x es IN] o r IN2 in stead o f / - n / o r / g e - / + / - n / .
In the present work suc h consid erations o f surface structure phenomena
c a n b e i n c l u d e d , s i n c e a p o s t - t r a n s f o r m a t i o n a l r u l e , t h e HABEN H o p p i n g
Rule o perates after the indefinite affixes are spelled out.

31. J ol l y 1873:253-256 claims that ACI-Constructions are not known t o have


existed in Sanskrit as verbal complements. The ev idence from Sa nskrit
is,however, contradictory and seems t o depend on how cer tain passages
from t he Rig Veda a re tra nslated i n t o Modern German ( c f . e s p e c i a ll y
Wolff 1906:591-^93)• A more ca reful and systematic investigation of
the Sanskrit te xts i s necessary in order t o make more r eliable claims
a b o u t t h e h i s t o r y o f t h e Indo-European AC I - C o n s t r u c t i o n.

32. According t o Nielsen 1926:384-385 North Lapp verba dicendi and verba
putandi may take the ACI-Constructions a s i l l u st r a te d by the fol lowing
examples excerpted from N ielssen and n ormalized t o Standard North Lapp
1 orthography:
316

LÂKKAT ' s a y 1

(i) Aei låkkan båppa gal boatti t


[+N] [+V]
[+Ack] [+lnf]

(They s aid that the priest would n ot come)

MU1 ' T A U T ' t e l i 1

(ii) Mui'talit bappa b oattit


[+N] [+V]
[+Ack}[+lnf]

(They t e l l that the priest wil l come)

DOAIVOT ' b e l i e v e , t h i n k '

(iii) Doaivot gå iezat boattit


[+Pro][+V]
C+Ack][+lnf]

(Do you think that you w ill come)

It is unsure whether these embeddings belong to a literary style or whether


they should be tre ate d a s normal embeddings. Some North Lapp n a t i v e speakers
d o n o t a c c e p t AC I - C o n s t r u c t i o n s a s v e r b complements a f t e r t h e s e v e r b s ,
rather they prefer complementizer embeddings.

33. 0 . f o r h o n e n o f J o k k m o k k h a s h e l p e d m e t o n o r m a l i z e N i e l s s e n ' s 1 9 2 6 N o r t h L a p p
s e nt e nc e s i n t o Standard North Lapp ortho graphy a s used in Sweden and Norway.

3^. According t o Standard Finnish grammars th ere are at leas t four verb-forms
which are called infinitives in Standard Finnish.
In Ikola 1971:65-73 the four Finnish i nf in it i v es and two present pa rticiples
are exemplified and their morphological suffixes are discussed from a
taxonomic point o f view. In ord er t o provide a generative viewpoint fo r the
treatment of Finnish verb complementation the following morphological rules
are necessary to generate the two infinitive affixes and on e present parti­
ciple affix exemplified in the Finnish sentences, i.e. rules ( i ) , ( i i i ) ,
and (v).

(') Finnish Infinitive 1 Suffix Rule

0-W -a /
[+1nf 1]

Example: tul1+a
(t o come)

(i•) Finnish Infinitive 2 Infix Rule

+DUR

Example: Aika kului laulaessa ja leikkiessä


[+ l n e s s i v e ]
(Time pass ed in singing and p laying)
(Tiden gick i sjungandet och spelandet)
317

( i l i ) Finnish Infinitive 3 Infix Rule


c
/ +V3
0 */-ma-/ /
C
+Inf 3] / +N
/ _+DUR_
VP

Example: Poika näki hänet istumassa


[+Acc] [+ 1 n e ss i v e ]

(The boy saw|j^™j in the sitting)

(Pojken såg|hgnneJ ' sittandet)

(i v) Finnish Infinitive k Infix Rule

SJ ,
1
0 —> / - m i s -/
[ +lnf k] +N

+ ITERI
VP

Example: Sina vain lihot lihomistasi


[+V] [+V] [+POSS S u f f ]
C+N]
(You only ge t fatt er and fa tter)

(Du b l i r bara fetare och fetare)

(v) F i n n i s h P r e s e n t P a r t i c i p l e S u f f i x Ru?1e

+V
/-va/ /
[+Pres Part]/ +N

i + m v?

Example: Poika kuuli hanen huutavan


C+Acc Suffix ]

(The boy heard!J^l shouting)

(Pojken ^r^ e|hennej s^r'ka)


318

Further rules are needed to assign correct case markers to the


infinitive affixes which have been spelled out by the morphological
rules (i) through (v) just formalized:

(vi) Finnish inessive Case Suffix Rule

+V
• /-ssa/
[+lnessi ve] +N

+DUR VP

Example: istu+ma+ssa 'in the sitt ing '

(vii) Finnish Accusative Case S uffix Rule

+V
0 —* / - n /
[+Acc] +N
VP

Example: huuta+va+n 'shouting1

( v i i i ) Finnish Illative Case S uffix Rule

SJ ,
+V 1
0 -•/-an/ +ma -
[+11 lat i ve] +N

+INCH0I
VP

Example: Hän pyysi miestä tulema+an


(He i n v i t ed t h e man i n t o coming)
(Han inbjöd mannen t i l l a t t komma)

The infi nitive and present participle rules (i) through (v) must be
crucially ordered before the case suffix rules, otherwise the wrong
sequence of affixes will result, e.g. *huutanva 'shouting',
• i s t u s s a m a ' s i t t i n g ' e t c . The i n d i c e s VP and S1 below and abo ve t h e
verb matrices denote that the morphological rules are triggered i f
t h e verbs in q u e s t i o n a r e dominated by t h e symbol VP and t h e complement
sentence symbol Si in the deep structur e.

In Russian a gerundive c o ns t r u c t i on i s a l s o p o s s i b l e , whereas an AC I -


Construction i s ill-formed . Compare t he three diff erent types o f verbal
complements in the following Russian sentences, i . e . Complementizer
Embedding i n ( i ) , AC I - C o n s t r u c t i o n i n ( i i ) , and AWG-Co nstruction i n
(iii):

CO On »!y*a,lf!«k*ona igrala na rojale


j ^v i d e l j[ £ t o
[+Past)
/ mJh e a r d ì f h o w
she was p laying the piano)
(saw jjthatj
319

35. (ii) *0n S! r je jo igrat' na rojale


•v,de J [+lnf]

(He
saw*" f ^ er t^ie P'ano^

slySall . . igraju££uju l n . ,
(i i i) On videi r JeJ° • • y * . . r a r oJj a l e
jgrajuscij J
[+Pres Part]
heard|
(He her playing the piano

The Russian AWG-Construction i s not ambiguous a s i s i t s English


t r a ns l a t i o n , because t h e gender s u f f i x /— i J / and t h e gender s u f f i x
/-uju/ are overt gender markers for masculine and feminine respectively.
After Subject Raising a gender copying rule assigns the correct features
t o the gerundivized complement verb. This gender suffix will then
automatically disambiguate the Russian AWG-Construction.

In English, Russian, and the two Finno-Ugric languages studied


here, North Lapp and Standard Finnish, Infin i t v
i ization must be blocked
after verba sentiendi and instead coordinated with the Progressive
Suffix Rule so that Subject Raising applies in these four languages
only with Gerundivizati on of the complement verb.

It is in this sense that one can treat the Subject Raising Rule a s
generating infinitive or gerundive constructions in the Indo-European
and Finno-Ugric languages. The S ubject Raising Rule can therefore be
called a formal universal for describing both infini t v
i ization and
gerundivization processes in any language.

36. The e xtensive use of the ACI-Construction in classical literary Latin


was possible due to a complex m orphological system with overt case
makers and overt infinitival endings for both passive and a ctive
infinitives. Not o nly after verba sentiendi, but also after verba
d i c e n d i , p u t a n d i , a f f e c t u s , c a u s a t i va e t c c o u l d A C I - C o n s t r u c t i o n s o c c u r .
The domain o f the Subject Raising Rule seems t o have been diminished
by the loss of overt case and infinitive markers in the transition
from l i t e r a r y Latin t o Modern Romance languages, s i n c e comparatively
f e w v e r b c a t e g o r i e s c a n s t i l l t a k e AC I - C o n s t r u c t i o n s a s v e r b c o m p l e m e n t s
in Spanish, French, Italian e t c . A sp ecial invest igation o f Romance
syntax is necessary to be able to generalize on loss of a complex c ase
system and the rise of syntactic constraints on Subject Raising. Such
a development has been observed in Germanic languages as for instance
in the t r a n si t i o n from Old t o Modern Sw edish ( c f . Ureland 1972' ( f o r t h ­
coming)).

37« Some lin guists avoid the term 'comparative grammar1, because o f i t s
associa tion with the hi st ori cal-comparati ve grammar o f the past
century (c f. for instance Nickel I97I0& 1972b ).However, by m odifying
the term 'comparative1 with 'synchronic' a confusion of terms can be
avoided. Also, we use the term 'comparative synchronic' in order to
emphasize the similarity between the abstract and interlingual approach
t o l i n g u i s t i c str uct ures o f the present comparative-synchroni c grammar
t o that o f the comparative-di ach ron i c grammar o f the pas t century.
320

The interest of historical linguists was to reconstruct proto-Germani c


base forms, whereas our interest i s t o reconstruct present Germanic
s y n t a c t i c a l l y common b a s e forms.

38. In E nglish the verb SEE can take the progressive s u f f i x , i f i t means
'intend t o se e1 o r 'intend t o meet' a s in I am se eing a fri end t h i s
evening a t the c ir c u s . In su ch a sentence the verb SEE i s simply
another l e xi c al entry and belongs t o a d i f f e r e n t verb c l a s s than SEE
in sentence (52b).

39. Chomsky's 1965 Standard Theory o f syntax can be said to be primarily


a grammatical model fo r synt hesis and n ot a model for analys is.
Chomsky h imself has denied t h i s and claimed that h i s model i s neutral
with respect to production and perception. However, the fact that i t
is impossible to retrieve deep structure configurations from su rface
structures after tree-pruning supports the claim that the Aspects -
T h e o r y i s n o t a t h e o r y f o r a n a l y s i s . We t h e r e f o r e c o n c l u d e t h a t i t i s
primarily fo r synt he sis . (Cf. Chomsky 19&5:139~140).

40. For information on Finnish syntax I thank M. Rahkonen, Jyväsky lä,


S . K a r l s s o n , Umeå, a n d M. R e u t e r , H e l s i n k i .

41. At a r e l a t i v e l y l a t e s t a g e o f my work on Finnish verb complementation


a f t e r v e r b a s e n t i e n d i (NÄHDÄ ' s e e ' a n d KUULLA ' h e a r ' ) I r e c e i v e d i n ­
formation on the historical background of the verb complements which
o c c u r a f t e r t h e s e v e r b s from S . K a r l s s o n , Urne! and from a paper by
R. An tilla, Helsinki. (Cf. his forthcoming contribution t o the Pre­
prints of the Second International Conference o f Nordic and General
L i n g u i s t i c s , Umeå June 14-19» 1973:"Was There A Gen erative H i s t o ri c a l
Linguistics?" ) . The conclusions presented in paragraph 12. are
based partly on this historical information, and partly on the results
of the present interlingual study on infinitivization and gerundiviza-
tion processes in Germanic and Finno-Ugric languages.
321

APPENDIX
A. GLOSSARY TO PART ONE
SWEDISH ENGLISH

ANSE consider
FINNAi findi, discover
finna2 find2, think
FÂ! perfective-aspectual auxiliary
fâ2 future auxiliary (will)
FA 3 modal auxiliary (can, must)
FÄ4 permissive auxiliary (be allowed
to)
FÂ5..att causative verb (cause)
FÅ(j main verb (obtain)
FÖRKLARA (SIG) declare, say
HJÄLPA help
HÖRAi hear auditorily
höra2 hear of, be told
HÖRA3 listen to
HÖRA4 interrogate, examine
HÖRA5 belong to
FÂ HÖRAi hear suddenly and unexpectedly
få höra 2 be told suddenly and unexpectedly
KÄNNA! feel! tactually
KÄNNA2 smell
KÄNNA3CSIG) feel3 mentally
KÄNNA4(SIG) feel4 somatically
LATAX allow
LÄTA2 cause
LÄRAt teach
LÄRA2(SIG) learn
MÄRKA notice
OBSERVERA! watch
OBSERVERA2 become aware of, notice
PÅSTÅ claim
SE! see! visually
SE2 realize
SE3 HELST prefer to see
FÅ SEi catch sight of
SÄGA say
TRO believe
TYTlf A be of the opinion
TÄNKAi(SIG) imagine
TÄNKA2 future auxiliary (be going to)
UPPGIVA (SIG) state, say
UPPTÄCKAi discover! visually
upptäcka2 discover2 cognitively
UPPTÄCKA3 make the discovery
UPPTÄCKA4 discover a talent
VETA know
VISA (SIG) turn out
VÄNTA (SIG) expect
322

B, GLOSSARY TO PART TWO


SWEDISH ENGLISH GERMAN

FÄ HÖRA^ fHEARi JHÖRENI


(HEARI SUDDENLY (PLÖTZLICH HÖREN!

HÖRAi
(hearx FHÖRENI
[HEARI SUDDENLY [PLÖTZLICH HÖRENI

IAKTTAGA WATCH BEOBACHTEN


IFEELI TACTUALLY FFÜHLENI
KÄNNAX
[PERCEIVE! {SPÜRENL

KÄNNA2 SMELL RIECHEN


IFEEL3 MENTALLY F(SICH) FÜHLEN3
KÄNNA3 (SIG)
[PERCEIVE3 [(SICH) SPÜREN3

KÄNNA4 (SIG) FEEL4 SOMATICALLY F(SICH) FÜHLEN4


{(SICH) SPÜREN4

MÄRKA NOTICE BEMERKEN, MERKEN

OBSERVERAI WATCH, OBSERVE BEOBACHTEN

OBSERVERA2 BECOME AWARE OF GEWAHR WERDEN


/SEE! (SEHEN!
FÄ SEI [SEEL SUDDENLY JPLÖTZLICH SEHENI
/SEEi /SEHEN!
SEi JSEEI SUDDENLY [PLÖTZLICH SEHEN!

SE2 REALIZE EINSEHEN, SEHEN2

UPPTÄCKAI DISCOVER! VISUALLY ENTDECKEN


323

REFERENCES
AGRELL,S. 1908. Aspektänderung und Aktionsartbildung, Lunds
universitets Årsskrift, N.F., AFD.l. Bd.4, Nr.2. (1908).
AHLBERG, M. 1942. Presensparticip et i fornsvenskan. En syntaktisk
studie. Lund.
ÅHLÉN, K. 1833. "Om verbets syntax i den äldre fornsvenskan."
Örebro.
ALLÉN, S. 1970a. Nusvensk frekvensordbok baserad på tidningstext.
(Frequency Dictionary of Swedish based on Newspaper Material)
1. Stockholm: Almqvist & Wiksell
1970b. "Vocabulary Data Processing," in H. Benediktsson 1970
(ed.). The Nordic Languages and Modern Linguistics. Proceed­
ings of the International Conference of Nordic and General
Linguistics, 235-261, Reykjavik: Visindafélag.
1971.Nusvensk frekvensordbok baserad på tidningstext.
(Frequency Dictionary of Swedish based on Newspaper Material)
2. Lemman (Lemmas). Stockholm: Almqvist & Wiksell.
1972 (forthcoming). "Om fraser i svenskan", paper presented
at the Symposium on Svenskans beskrivning 6, held in Turku,
April 7-8^1972 (To be published in the Proceedings of the
Turku Symposium).
1973. "Om frekvensordböcker," Språkvård 1973: 9-12.
ALLEN, S. & S. Hellberg 1971. Introduktion i grafonomi. Det
lingvistiska skriftstudiet. Stockholm: Almqvist & Wiksell.
ALVING, Hj. 1918. "Objekt-subjektets plats i ackusativ med
infinitiv. En morfemhistorisk undersökning," in Språk och
stil 18, 1-37.
ANDERSON, S.R. 1971. "On the Linguistic Status of the Performa­
tive/Constative Distinction," reproduced by the Indiana
University Linguistics Club.

ANWARD, J. 1972 (forthcoming). "Om semantisk snarlikhet mellan


syntaktiska släktningar," paper given at the KVAL sea-borne
Fall Seminar, October 23-24,1972 on "Homojosämi." Stockholm-
Turku.
APELT, 0. 1874. "Über den Accusativus cum Infinitivo im Gothischen,"
Germania 19 (1874), 280-297.
1875. "Über den Accusativus cum Infinitivo im Althoch­
deutschen und Mittelhochdeutschen," in Jahresbericht über
das Wilhelm Ernstische Gymnasium zu Weimar, 1875, 1-23.
AUSTIN, J. 1962. How to do things with words. Cambridge, Mass:
Harvard Univ. Press.
324

BACH, E. 1962. "The Order of Elements in a Transformational


Grammar of German," Language 38, (1962), 263-269.
1967. "Have and be i n English syntax," Language 43,(1967),
462-485.
1971. "Syntax since Aspects," in R.J.O. Brien (ed.).
Report of the Twenty-Second Annual Round Table Meeting on
Linguistics and Language Studies»1-17, Washington D.C.:
Georgetown University Press.
BACH, E. & R.T. HARMS (eds.) 1968. Universals in Linguistic
Theory, New York: Holt, Rinehart & Winston.
BAILEY, C-J. N. 1970. "Using Data Variation to Confirm rather
than Undermine the Validity of Abstract Syntactic Structures,"
in Working Papers in Linguistics 2/8, 77-85, University of
Hawaii, Honolulu.
1972. "The Intergration of Linguistic Theory: Internal Re­
construction and the Comparative Method in Descriptive Ana­
lysis," in R.P. Stockwell & R.K.S. McCawley (eds.) 1972,
Linguistic Change and Generative Theory, 22-31, Indiana
Univ. Press.
BECH, G. 1955. Studium über das deutsche Verbum infinitum, in
Det Kongelige danske Videnskabernes Selskab, Historisk­
filologiske Meddelelser, Bind 35, No. 2 (1955) Ktfbenhavn.
BECKMAN, N. 1959 (1916). Svensk språklära, 4th ed., Stockholm:
Bonniers.
BEHAGEL, 0. 1924. Deutsche Syntax, Band II, Die Wortklassen und
Wortformen, Heidelberg: Carl Winter.
BENEDIKTSSON, H. (ed.) 1970. The Nordic Languages and Modern
Linguistics. Proceedings of the International Conference
of Nordic and General Linguistics, Univ. of Iceland,
Reykjavik, July 6-11, 1969, Reykjavik: Visindafélag.
BEVER, T. G. 1970. "The Cognitive Basis for Linguistic Structures,"
in J. Hayes (ed.), Cognition and the Development of Language,
279-362, New York: Wiley & Sons.
BEVER, T. G. & D. T. LANGENDOEN 1971. "A Dynamic Model of the
Evolution of Language," Linguistic Inquiry 43 433-463.
BIERWISCH, M. 1963. Grammatik des deutschen Verbs in Studia
Grammatica II, Berlin: Akademie Verlag.
1967. "Syntactic Features in Morphology: General Problems
of So-called Pronominal Inflection in German," in To Honor
Roman Jakobson, Essays on the Occasion of His Seventieth
Birthday, 239-270, The Hague: Mouton.
1970. "On classifying Semantic Properties," in Bierwisch & Heidolph
1970: 27-50.
BIERWISCH, M. & K.E. HEIDOLPH (eds.) 1970. Progress in Linguistics.
Janua Linguarum. Series maior, 43, The Hague: Mouton.
BLOOMFIELD, L. 19 33. Language,New York: Holt and Company.
BOCK, H. 1931. "Studien zum Präpositionalen Infinitiv und Akkusativ
mit dem To-Infinitiv," Anglia LV> 114-249.
325

BOPP, F. 1816. Uber das Conjugationssystem der Sanskritsprache


in Vergleiohung mit jenem der griechischen, lateinischen,
persischen und germanischen Sprache, K.G. Windischnann (ed.)
Frankfurt am Main: Andrea.
BRESNAN, J.W. 1972. Theory of Complementation in English Syntax
unpublished MIT dissertation.
BRUGMANN, K. 1904. Kurze Vergleichende Grammatik der indogerm.
Sprachen, Strassburg: Trübner.
BRUNNER, K. 1962. Die E nglische Sprache. (Sammlung kurzer Gramma­
tiken germanischer Dialekte. H. de Boor (ed.),.Tübingen:
Max Niemeyer.
CALLAWAY, M. 1913. The Infinitive in Anglo-Saxon, Washington D.C.:
Carnegie Institution of Washington.
CHAFE, W. 1970. Meaning and the Structure of Language,Chicago:
The University of Chicago Press.
1971. "Directionality and Paraphrase," Language 47 (1971)
1-26.
CHAPIN, P. 1967. "The Syntax of Word Derivation in English,"
MITRE Technical Paper No. 6 8, Bedford, Mass.
CHOMSKY, N. 1957. Syntactic Structures, The Hague: Mouton.
1965. Aspects of the Theory of Syntax* Cambridge, Mass:
MIT Press.
1966. Cartesian Linguistics, New York: Harper & Row.
1968. Language and Mind, New York: Harcourt, Brace & World.
1969. Topics in the Theory of Generative Grammar, The Hague
Mouton.
1970a. "Remarks on Nominalization," in R.A.Jacobs & P.S.
Rosenbaum (eds.) Readings in English Transformational Grammar,
184-221, Waltham, Mass: Ginn and Company.
1970b. "Deep Structure, Surface Structure, and Semantic
Interpretation," in R. Jakobson & S. Kawamoto (eds.)
Studies in General and Oriental Linguistics, 52-91, TEC.Co.
Tokyo.
1971a. "Some Empirical Issues in the Theory of Transforma­
tional grammar," reproduced by the Indiana University
Linguistics Club.
1971b. "Conditions on Transformations," reproduced by the
Indiana University Linguistics Club.
CHOMSKY, N. & G.A. MILLER, 1963. "Introduction to t he Formal
Analysis of Natural Languages," in R.D. Luce, R.R. Bush,
and E. Galanter (eds.) Handbook of Mathematical Psychology,
Vol II, Ch. 11, 269-321, New York: Wiley.
CHOMSKY, N. & M. HALLE, Ì968. The Sound Pattern of English,
New York: Harper & Row.
CLOSS, E. 1965. "Diachronic Syntax and Generative Grammar,"
Language 41, 402-415.
326

COLLINDER, B. 1970. "Noam Chomsky und die Generative Grammatik.


Eine kritische Betrachtung," Acta Universitatis Upsaliensis,
Acta Societatis Linguistica^ Upsaliensis. nova Series. 2:2.
CURME, G.O. 1912. "History of the English Gerund," Englische
Studien XLV (1912), 349-380.
DAHL, ö. 1971. Generativ Grammatik på svenska,Lund: Student­
litteratur.
DAHLSTEDT, K.H. 1966. Svårigheter i svenskans uttal.(Skrifter
utgivna av Modersmålslärarnas förening nr 107). Lund.
DAL, I. 1952. Kurze deutsche Syntax, auf historischer Grundlage
Tübingen: Max Niemeyer.
DAVIDSON, D. & G. HARMAN (eds.) 1972). Semantics of Natural
Language* Dordrecht: Reidei.
DELBRÜCK, B. 1897. Vergleichende Syntax der indogermanischen
Sprachen, Zweiter Theil, Strassburg: Trübner.

VAN DIJK, T.A. 1972. Some Aspects of Textgrammars.A Study in


Theoretical Linguistics and Poetics, Phil.Diss. Univ. of
Amsterdam (Janua Linguarum) The Hague: Mouton.
DUDEN, 1959. Grammatik der deutschen Gegenwartssprache, Mannheim:
Bibliographisches Institut.
EINARSSON, S. 1967. Icelandic Grammar, Texts, Clossary, Baltimore:
Johns Hopkins.
ELIASSON, S. 1972a. Review of R. Filipovic (ed.) Zagreb Conference
on English Contrastive Projects, unpublished manuscript (TO
appear in Studia Linguistica 27:1 (1973)).
1972b. Litteraturanvisningar i kontrastiv lingvistik,
duplicated stencil at the Department of Linguistics, Uppsala
University.
ELERT, C.C. 1967. "Descriptive Contrastive Work on Swedish. A
Survey of Recent Work and Work in Progress." University of
Stockholm Institute of Linguistics: Reports and Studies No 3.
Stockholm.
ELLEGÅRD, A. 1971. Transformationen svensk-engelsk satslära,
Lund: Gleerup.
ERDMANN, 0. 1874. "Untersuchungen über die Syntax der Sprache
Otfrids," Halle: Verlag der Buchhandlung des Waisenhauses.
FALK, Hj. & A. Torp. 1900. Dansk-Norskens syntax i historisk
fremstilling, Kristiania:Aschehong & Co.
FILLMORE, Ch. 1968. "The Case for Case," in Bach & Harms (eds.)
1968, 1-88.
1971. "Some Problems for Case Grammar," in R.J.O. Brien (ed.)
1971, 35-56.
1972. "Subjects, Speakers, and Roles Semantic of Natural Language,"
in Davidson & Harman (eds.) 1972, 1-24.
327

FILLMORE, Ch. & D.T. LANGENDOEN (eds.) 1971. Studies in Linguistic


Semantics, New York: Holt, Rinehart & Winston.
FODOR, J.D. 1970. "Formal Linguistics and Formal Logic," in J.
Lyons (ed.) New H orizons in Linguistics, 198-214, Harmonds-
worth: Penguin
FODOR, J.A. & M. Garett 1966. "Some Syntactic Determinants of
Sentential Complexity," in J. Lyons & R.J. Wales (eds.)
Psycho linguistic Papers: The Proceedings of the 1966
Edinburgh Conference* Edinburgh: Edinburgh Univ. Press.
FORSYTH, J. 1970. A Gra mmar of Aspect: Usage and Meaning in the
Russian Verb, Cambridge: Univ. Press.
FRASER, B. 1970a. "Idioms within a Transformational Grammar,"
Foundations of Language 6, (1970), 22-42.
1970b. "Some Remarks on Action Nominalization in English"
in Jacobs & Rosenbaum (eds.) 1970, 83-98.
1971. "An Examination of the Performative Analysis," re­
produced by the Indiana University Linguisitcs Club.
FRIES, Ch.C. 1945. Teaching and Learning English as a Foreign
Language, Ann Arbor.
1952. The Structure of English, New York: Harcourt, Brace
& Co.
Gazzaniga, M.S. 1970. The Bisected Brain, New York: Appleton-
Century-Crofts.
GLÄüSER , E. 1930. Frieden, Berlin: Kiepenheuer.
GREENBERG, J.H. 1963. "Some Universals of Grammar with Particular
Reference to the Order of Meaningful Elements," in Greenberg,
J.H. (ed.) Universals of Language, Cambridge, Mass: MIT
Press.
GRIMBERG, C. 1905."Undersökningar om konstruktionen accusativ med
infinitiv i den äldre fornsvenskan," Arkiv för nordisk
filologi 21, 205-235, 311-357.
GRIMM, J. 1819-1837. Deutsche Grammatik, Vol I-VII, Göttingen:
Dietrich.
GRUBER, J. 1967. Function of the Lexicon in Formal Descriptive
Grammars, System Development Corporation, Santa Monica, Cal.
HAIMAN, J. 1971."Targets and Paradigmatic Borrowing in Romantsch"
Language 47, (1971), 797-809.
HAMMARBERG, B. 1972. "The Insufficiency of Error Analysis," paper
presented at the Symposium on Error Analysis" Lund 26-27 Sep­
tember 1972.
HAMMER, J.H. & F.A. RICE (eds.) 1965. A B ibliography of Contras­
tive Linguistics, Washington. D.C.: Center for Applied
Linguistics.
HÄRTUNG, W.D. 1964. Die zusammengesetzten Sätze des Deutschen, in
Studia Grammatica IV, Berlin: Akademie Verlag
HAYES, J. (ed.) 1970. Cognition and the Development of Language,
New York: Wiley & Sons.
328

HEIEOLPH, K.E. 1966. "Kontextbeziehnungen zwischen Sätzen in


einer generativen Grammatik," Kybernetika 2, (1966),
274-281.
1970. "Zur Bedeutung negativer Sätze," in Bierwisch &
Heidolph 1970, 86-101.
HELLQUIST, E. 1939. Svensk etymologiskt lexikon, Lund: Gleerups.
HEUSLER, A. 1921. Altis ländis che s Elementarbuch, Heidelberg:
Carl Winter.
HIRT, H. 1934. Indogermanische Grammatik 6. Die Hauptprobleme
der indogermanischen Syntax3 Heidelberg: Carl Winter.
HOFMAN, T.R. 1966. "Past Tense Replacement and the Modal System,"
in Harvard Computation Laboratory Report to the National
Science Foundation on Mathematical Linguistics and Automatic
Translation. Number NSF-17, Cambridge, Mass.
HOLM, G. 1952. Om s-passivum i svenskan, Lund: Gleerup.
1969. Medeltida stadgespråk9 Lund: Studentlitteratur.
IHWE, J. H. 1971 (forthcoming). Literaturwissenschaft und
Linguistik. Ergebnisse und Perspektiven. I-III. Frankfurt
am Main: Athenäum.
IKOLA, 0. (ed.) 1971) Nykysuomen Käsikirja, Helsinki: Weilin & Göös.

ILLUSTRERAD SVENSK ORDBOK 1958. Stockholm: Natur och kultur.

ISA£ENK0, A.V. 1960. "La structure sémantique des temps en russe,"


Bulletin de La Société de Linguistique de Paris 55, 74-88.
ISENBERG, H. 1968. "Überlegungen zur Textteorie," paper given at
the Arbeitsstelle Strukturelle Grammatik, Deutsche Akademie
der Wissenschaften zu Berlin.
JACKENDOFF, R.S. 196 8. "An Interpretative Theory of Pronouns and
Reflexives," reproduced by the Indiana University Linguisitcs
Club.
1970. "On Some Questionable Arguments about Quantifiers and
Negation," reproduced by the Indiana University Linguistics
Club.
JACOBS, R.A. & P.S. Rosenbaum 1967. Grammar 2a Boston: Ginn & Co.
1968. English Transformational Grammar, Waltham, Mass:
Blaisdell.
(eds.) 1970. Readings in English Transformational Grammar,
Waltham Mass: Ginn & Co.
JAMES, C. 1972. "Zur Rechtfertigung der kontrastiven Linguistik,"
in Nickel 1972 (ed.) Reader zur kontrastiven Linguistik,
21-38, Frankfurt am Main: Athenäum.
JENKINS, L. 1972. Modality in English Syntax, MIT doctoral disserta­
tion, reproduced by the Indiana Universit Linguistics Club.
JESPERSEN, 0. 1954a. A Modern English Grammar on Historical
Principles* I-VII, London: Allen & Unwin.
• 1954b. A M odern English Grammar on Historical Principles t
Part IV, London: Allen & Unwin.
1969. Analytic Syntax, New York: Holt, Rinehart & Winston.
329

JOLLY, J. 1873. Geschichte des Infinitivs im Indogermanischen*


München: Ackermann.
KARLGREN, H. 1972. "Ord, massor av ord. En databas i bok- och
bandform,11 Språkvård No. 3, (1972), 8-17.
KARTTUNEN, L. 1971. "The Logic of English Predicate Complement
Constructions," reproduced by the Indiana University
Linguistics Club.
KATZ, J.J. 1964. "Mentalism in Linguistics," Language 40, 124-
137.
1970. "Interpretative Semantics vs. Generative Semantics,"
Foundations of Language 6, (1970), 220-259.
1971. The Underlying Reality of Language and Its Philosophical
Importj New York: Harper & Row.
1972. Semantic Theory> New York: Harper & Row.
KEENAN, E. 1971. "Two Kinds of Presuppostion in Natural Language,"
in Fillmore & Langendoen (eds.) 1971, 45-52.
1972. "The Logical Status of Deep Structures (Logical Constraints
on Syntactic Processes)," in the Preprints of the Proceedings
of the Xlth International Congress of Linguists> Bologna 1972,
870-883.
1973. "On Semantically Based Grammar," Linguistic Inquiry,
Vol III (1973), 413-461.
KELLERMANN, B. 1921. Der 9. November, Berlin: Fischer.
KIEFER, F. (ed.) 1969. Studies in Syntax and Semanticss Founda­
tions of Language Supplementary Series Vol.10, Dordrecht:
Reidel.
1970. Swedish Morphology, Stockholm: Skriptor.
1972. "Apropos Derivational Morphology," in Derivational
Processes, 42-59, Proceedings of the KVAL Sea-borne Spring
Seminar held on board m/s Bore, April 9-10, 1972, Stockholm-
Turku.
1973. Generative Morphologie des Neufranzösischen,Romanistische
Arbeitshefte 2. G. Ineichen & Ch. Rohrer (eds.), Tübingen:
Max Niemeyer.
KIEFER, F. & N. Ruwet (eds.) 1973. Generative Grammar in Europe.
Foundations of Language Supplementary Series. The Hague:
Mouton.
KIPARSKY, P. 1969. "Tense and Mood in Indo-European Syntax," re­
produced by Indiana University Linguistics Club.
KIPARSKY, P. & C. KIPARSKY, 1970. "Fact", in Bierwisch & Heidolph
1970, 114-173.
KLEIN, E. 1971. A Co mprehensive Etymological Dictionary of the
English Language, Amsterdam: Elesevier.
KRZESZOWSKI, T.P. 1971. "Equivalence, Congruence, and Deep
Structure," in G. Nickel (ed.) 1971b Papers in Contrastive
Linguistics9 37-48, Cambridge.
330

KRÜGER, G. 1914. Syntax der englischen Spraohe vom englischen


und deutschen Standpunkte nebst Beiträgen zu Wortbildung 9
Wortkunde und Wortgebrauch, Vol II, Dresden und Leipzig:
Kochs Verlagsbuchhandlung.
KUFNER, H.L. 1962. The Grammatical Structures of English and
German, Chicago: The Univ of Chicago Press.
KURODA, S.-Y. 1969. "Remarks on Selectional Restrictions and
Presuppositins," inf F. Kiefer (ed.) Studies in Syntax and
Semantics, 138-167, Foundations of Language, Supplementary
Series, Vol. 10., Dordrecht: Reidel.
KÖHLER, A. 1867. "Der syntaktische Gebrauch des Infinitivs in
Gothischen," Germania 12, (1867), 435-449.
KÖNING,E. & H. Nickel. 1970. "Transformationelle Restriktionen
in der Verbalsyntax des Englischen und Deutschen," in
Probleme der Kontrastiven Sprachwissenschaft. Jahrbuch 1969
des Instituts für Deutsche Sprache, Düsseldorf.
KÖNING, E. 1971. Adjectival Constructions in English and German:
A Co ntrastive Analysis, Heidelberg: Groos.
LABOV, W. 1966. The Social Stratification of English in New York
City, Washington D.C.: Center for Applied Linguistics.
1970. "The Study of Language in its Social Context," Studium
Generale 23, 30-87.
1972. "On the Use of the Present to Explain the Past," in
the Preprints of the Proceeding of the Xlth International
Congress of Linguists, 1110-1135, Bologna.
LADO, R. 1957. Linguistics across Cultures. Ann Arbor: Univ. of
Michigan Press.
LAKOFF, G. 1965. On the Nature of Syntactic Irregularity,
Harvard University Computation Laboratory Report No. NSF-16,
Cambrdige, Mass: Reprinted as Irregularity in Syntax, ^
Chicago 1970: Holt, Rinehart & Winston.
1970a. "Global Rules," Language 40,627-639.
1971a. "On Generative Semantics," in D.D. Steinberg & L.A.
Jakobovits (eds.) 1971, Semantics. An Interdisciplinary
Reader in Philosophy, Linguistics^ and Psychology, 232-296,
Cambridge: Cambridge Univ. Press.
1971b. "Some Thoughts on Transderivational Constraints,"
unpublished paper.
1972. "The Arbitrary Basis of Transformational Grammar,"
Language 48, 76-87.
LAKOFF, G. & J. ROSS, 1967. "Is Deep Structure Necessary,"
reproduced by the Indiana Linguistics Club.
LAKOFF, R. 1968. Abstract Syntax and Latin Complementation,
Cambridge,Mass.: MIT Press.
331

LANGENDOEN, D.T. 1966. "The Syntax of the English Expletive IT,"


Georgetown University Monograph Series on Languages and
Linguistics, No, 19, 207-216.
LARSSON, C.G. 1931. Ordföljdsstudier över det finita verbet i
de nordiska språken. 1. Uppsala: Uppsala Univ. årsskrift.
Filosofi. 1931:1
LEES, R:B: 1960. The Grammar of English Nominalizations, The
Hague: Mouton.
LEHMANN, W.P. 1970. "Proto-Germanie Syntax. Preliminaries to
Further Study," in Hollander Festschrift, Austin, Texas:
Pemberton Press.
1972. "Converging Theories in Linguistics," Language 48,
(1972), 266-275.
LEUMANN, M. & J. HOFMANN & A. SZANTYR .1965. Latinische Grammatik,
zweiter Band, München: Becksche Verlagsbuchhandlung.
LINELL, P. 1972. Remarks on Swedish Morphology. Ruul 1, Reports
from Uppsala University Department of Linguistics No. 1.
LURIA, A.R. 1966. Human Brain and Psychological Processes,
translated by B. Haigh, New York: Harper & Row.
LURIA, A.R. & L.S. Tsvetkova 1968. "The Mechanism of Dynamic
Aphasia," Foundations of Language 4, 296-307.
LYONS, J. 1968. Introduction to Theoretical Linguistics.
Cambridge: Cambridge Univ. Press.
1970. New Horizons in Linguistics, Harmondsworth: Penguin.
MALMBERG, B. 1967. Uttalsundervisning, Stockholm: Almqvist &
Wikseil.
MARKEY, Th. 1968. The Verbs Varda and Bliva in Scandinavian with
Special Emphasis on Swedish, (Acta Universitatis Upsaliensis,
Studia Philogogiae Upsaliensia 7) Uppsala: Almqvist & Wiksell.
McCAWLEY, J. 1968. "The Role of Semantics in ß. Grammar," in Bach &
Harms (eds.) 1968, 124-169.
1970. "English as a VSO Language," Language 46, 286-299.
1971. "Tense and Time Reference," in Fillmore & Langendoen
1971, 97-113.
1972. "Syntactic and Logical Arguments for Semantic Structures,"
reproduced by the Indiana- University Linguistics Club.
McNEIL, D. 1966. "Developmental Psycholinguistics," in Smith, F &
G.A. Miller(eds.) The Genesis of Language, 15-84, Cambridge,
Mass: MIT Press.
1970. The Acquistion of Language, New York: Harper & Row.
v.d.MEER, M. J. 1914. "Der Gotische akk.c.inf. in Subjektssätzen
und nach swaei und swe," PBB XXXIX (1914), 201-209.
MEYER-LÜBKE, W. 1899. Grammatik der romanischen Sprachen, Bd, III,
Leipzig: O.R. Reistand.
MILLER, G.A. & K.A. McKean 1964. "A Chronometrie Study of Some
Relations between Sentences," Quarterly Journal of Experi­
mental Psychology 26, (1964), 297-308*
332

MONSTERBERG-MÜNCKENAU, S. 1885. "Der Infinitiv in den Epen Hart­


manns von Aue, (Germ. Abhandlungen No.5, V.K. Weinhold (ed.))
Breslau: W. Rœbner .
MOSER, H. 1970. Probleme der kontrastiven Grammatik. Jahrbuch
1969 des Instituts für Deutsohe Sprache in Mannheim
Düsseldorf: Pedagogischer Verlag Schwann.
MÖTSCH, W. 1964. Syntax des deutschen Adjektivs 9 in Studia
Grammatica III, Berlin: Akademie Verlag.
1970. "Können attributive Adjektive durch Transformationen
erklärt werden," Folia Linguistica. Acta Societatis Linguistica
Europœae , Tomus I, 1/2 (1970), 23-48.
MÖTSCH, W. & R. Schädlich 1965. "Die Grundstruktur schwedischer
Verbalphrasen," in Steinitz-Festschrift*Veröffentlichungen
der sprachwissenschaftlichen Kommision der DAdW. Nr. 5,
Berlin.
MOULTON, W.G. 1962. The Sounds of English and German, Chicago:
Univ. of Chicago Press.
MUSTANOJA, T.F. 1960. A Mi ddle English Syntax, Part I, Helsinki:
Société Néophilologique.
NEWMEYER, F.J. & J. EMONDS, 1971. "The Linguist in American
Society," in Papers from the Seventh Regional Meeting of
the Chicago Linguistic Society, April 16-18, 1971, 285-303.
NICKEL, G. 1971a. "Contrastive Linguistics and Foreign-Language
Teaching," in Nickel (ed.) 1971b, 1-16.
(ed.) 1971b. Papers in Contrastive Linguistics, Cambridge:
Cambridge Univ. Press.
1972a. Fehlerkunde. Beiträge zur Fehleranalyse, Fehler­
bewertung und Fehlertherapie y Berlin: Cornelsen-Velhagen &
Klasing
(ed.) 1972b. Reader zur kontrastiven Linguistik, Frankfurt
am Main: Fischer Athenäum.
NIELSEN, K. 1926. Lœrebok i lappisk* Vol 1, Oslo: Bröggers.
NOREEN, A. 1904-1912. Vårt språk I-IF, Lund: Gleerup.
NYGAARD, M. 1905. Norren Syntax, Kristiania: Aschehoug & Co.
O'BRIEN, R.J. (ed.) 1971. Monograph Series on Languages and
Linguistics* Number 24, 1971, 22nd Annual Round Table.
Linguistics: Developments of the Sixties - Viewpoints for
the Seventies. Washington D.C.: Georgetown Univ. Press.
OKSAAR, E. 1972. "Zum Passiv im Deutschen und Schwedischen,"
in Nickel (ed.) 1972b, 85-105.
PAUL, H. 1920. Deutsche Grammatik, Band 17, Halle: Max Niemeyer.
PERLMUTTER, D. 1971. Deep and Surface Structure Constraints in
Syntax, New York: Holt, Rinehart & Winston.
PETERS, S.P. & R.W.RITCHIE 1969. "A Note on the Universal Base
Hypothesis," Journal of Linguistics 5, (1969), 150-152.
PETERS, S. 1970. "On the Generative Power of Transformational
Grammars," (To appear in Information Sciences),
333

PETERS, S.P. & R.W. RITCHIE 1971. "On Restricting the Base Compo­
nent of Transformational Grammars," in Information and
Control 18, 483-501.
PETÖFI, J.S. 1971a. Beiträge zu einer Ko-Textuellen Texttheorie,
Surte: Gotab.
1971b. Transformationsgrammatiken und eine ko-textuelle
Texttheorie3 Grundfragen und Konzeptionen. Linguistische
Forschungen 1. 301. Frankfurt: Athenäum.
POLLAK, W. 1960. "Studien zum verbalen Aspekt in Französichen,"
Stitzung sberichte der österreichischen Akademie der Wissen­
schaften 233* (1960), 30-47. Wien.
POSTAL,P.1970a. "On Coreferential Complement Subject Deletion,"
Linguistic Inquiry, Vol I, (1970), 439-500.
1970b. "On the Surface Verb Remind," Linguistic Inquiry
Vol 1. (1970), 37-120.
1971. Cross-Over Phenomena, New York: Holt, Rinehart &
Winston.
POUSTMA, H. 1916. A Gr ammar of Late Modern English, Groningen
Nordhoff.
RASK, R. 1818a. "Unders^gelse om det gamle Nordiske eller
Isländske Sprogs Oprindelse," in Ausgewählte Abhandlungen,
Vol I. 1-329, Copenhagen.
1818b. Anvisning till Isländskan eller Nordiska Fornspråket,
Stockholm.
RIBBECK, A.F. 1836. "Syntax des Ulfila," Neues Jahrbuch der
Berlinerischen Gesellschaft für Deutsche Sprache und Alter­
tumskunde9 39-56, Berlin.
ROGERS, A. 1971. "Three kinds of Physical Perception Verbs," in
Papers from the Seventh Regional Meeting of the Chicago
Linguistic Society,April 16-18, 1971, 206-222.
ROSENBAUM, P.S. 1967a. The Grammar of English Predicate Comple­
ment Constructions, Cambridge, Mass: MIT Press.
1967b. IBM English Grammer II, in Specification and Utiliza­
tion of Transformational Grammar, Scientific Report No 2.,
Thomas J. Watson Research Center, Yorktown Heights, New York.
ROSS, J. 1967a. Constraints on Variables in Syntax* unpublished
MIT dissertation, reproduced by the Indiana University
Linguistics Club.
1967b. "Auxiliaries as Main Verbs," reproduced by the Indiana
University Linguistics Club.
1969. "A Proposed Rule of Tree Pruning," in Reibel & Schane,
1969, 288-299.
1970a. "Gapping and the Order of Constituents," in Bierwisch &
Heidolph (eds.) 1970, 249-259.
1970b. "On Declarative Sentences," in Jacobs & Rosenbaum 1970,
222-272.
1972. "Doubl-ing," Linguistic Inquiry9 Vol III, (1972)^61-99.
334

RUONG, J. 1970. Min sàmigiella9 Uppsala: Utbildningsförlaget.

de SAUSSURE, F. 1960. (1916). Cours de Linguistique Générale„


Ch.Bally & A. Sechehaye (eds.), 5th edition, Paris: Payot.
SAVIN, H. & E. Perchonock 1965. "Grammatical Structure and the
Imediate Recall of English Sentences," Journal of Verbal
Learning and Verbal Behavior 4, (1965), 348-353.
SCHLEICHER, A. 1848-1850. Sprachvergleichende Untersuchungen9
Vols 1-2* Bonn: König.
SCHMIDT, J. 1872. Die Verwandtschaftsverhältnisse der i£g• Sprachen,
Weimar: Böhlau.
SEARLE, J. 1969. Speech Acts, New York: Cambridge Univ.Press.
SLOBIN, D.I. 1963. Grammatical Transformations in Childhood and
Adulthood. Unpublished doctoral dissertation, Harvard
University.
1968. "Recall of Full and Truncated Passive Sentences in
Connected Discourse," Journal of Verbal Learning and Verbal
Behavior 73 876-881.
(ed.)197L2,/ze Ontogenesis of Grammar. A Theoretical Symposium,
New York and London: Academic Press.
SNOOK, R.I. 1971. "A Stratificational Approach to Contrastive
Analysis," in Nickel 1971b, 17-36.
SMITH, F. & G.A. MILLER (eds.) 1966. The Genesis of Language.
A Psy cholinguistic Approach* Cambridge, Mass: MIT Press.
STANLEY, R. 1967. "Redundancy Rules in Phonology," Language 43*
(1967).
STEIG, R. 1884. "Über den Gebrauch des Infinitivs im Altnieder­
deutschen," Zeitschrift für deutsche Philologie 16
(1884), 307-345.
STEINBERG, D.D. & L.A. Jakobovits (eds.) 1971. An Interdisciplinary
Reader in Philosophy, Linguistics, Anthropology, and Psychology,
Cambridge: Cambridge Univ. Press.
STEINITZ, R. 1969. Adverbial Syntax* in Studia Grammatica X*
Berlin: Akademie Verlag.
STOCKWELL, R.P. & J.D. BOWEN & J.W. MARTIN 1965. The Grammatical
Structures of English and Spanish* Chicago: The Univ. of
Chicago Press.
STOCKWELL, R.P. & R.K.S. McCAULAY (eds.) 1972. Linguistic Change
and Generative Theory. Essays from the UCLA Conference on
Historical Linguistics in the Perspective of Transformational
Theory 1969, Bloomington: Indiana Univ. Press.
STREITBERG, W. 1890. "Perfective und imperfective Aktionsart im
Germanischen," PBB 15* (1890), 70-177.
STÅLE, C.I. 1967. "Medeltidens profana litteratur. Lagarna," in
E.N. Tigerstedt (ed.) 1967 Ny i llustrerad svensk littera­
turhistoria* Vol 1, 37-54, Stockholm: Natur och kultur.
335

SVENSKA AKADEMIENS ORDBOK (SAOB) 1898-1970. Ordbok över svenska


språketj Lund: Gleerup & Lindstedts.
SVENSK HANDORDBOK. Konstruktioner och fraseologia 1960. Svenska
bokförlaget.
SVENSK ILLUSTRERAD ORDBOK. 1958. Stockholm: Natur och kultur.
SÖDERWALL, K.F. 1884-1918. Ordbok öfver svenska medeltidsspråket*
Lund: Berlingska.
TELEMAN, U. 1969. "Böjningssuffixens roll i nusvenskan," Arkiv
för nordisk filologi 84, (1969), 163-208.
1971. Om svenska ord* Lund: Gleerup.
1972. "Stavning och fonologi," in Teleman, U. 1972, Tre
uppsatser om grammatik9 7-32, Lund: Studentlitteratur.
TIGERSTEDT, E.N. (ed.) 1967. Ny illustrerad svensk litteratur­
historia* Vol 1* 37-54, Stockholm: Natur och kultur.
URELAND, P.S. 1965. Zur Rahmenkonstruktion im deutschen Satz*
unpublished fil.lic. thesis at Uppsala University.
1968. "A Transformational Description of the Accusative-
with-Infinitive Construction in German," report at the
Linguistics Research Center, Univ. of Texas at Austin. Also
Thesis at the Dep. of Linguistics, Stockholm
1970. "A Transformational Description of the Accusative-
with-Infinitive Construction in Swedish," in C.C. Elert
& S. Fries & Å Jonsson (eds.) Svenskans beskrivning 6*
220-252, Umeå 1972.
1972a. "Nominalized Complements Occurring after Swedish
HÖRA 'hear' in Derivational Processes* 74-92, Proceedings
of the KVAL Sea-born Spring Seminar held on m/s Bore,
April 9-lo, 1972, Stockholm-Turku.
1972b (forthcoming). "Historical Change of Syntactic Domain
in Swedish," in Svenskans beskrivning 7* paper given at
the Symposium on Swedish held at Turku, April 6-7, 1972.
1973 (forthcoming). "The Rise of the Swedish s-Passive in
verba dicendi and verba putandi* " in K.H. Dahlstedt (ed.)
Preprints of the Second International Conference on Nordic
and General Linguistics* Umeå University, Umeå, Sweden.
VENDLER, Z. 1967. Linguistics in Philosophy* Ithaca: Cornell
Univ. Press.
VERNER, K. 1877. "Eine Ausnahme der esten Lautverschiebung."
KZ 23* 97 -130.
VISSER, Th. 1956. Materials for the Study of the Old English
Drama* Vol 26. A Syn tax of the English Language of St.
Thomas Moore* Part III* Louvain: Librarie universitaire.
1969. An Historical Syntax of the English Language. First
Half. Syntactical Units with Two Verbs, Leiden: E.J. Brill
WACKERNAGEL, J. 1920. Vorlesungen über Syntax mit besonderer
Berücksichtigung von Griechisch, Lateinisch und Deutsch.
Basel.
336

WAGNER, J. 1972. "Eine kontrastive Analyse der verben:


Deutsch: dürfen* mögen* müssen* können* bekommen
Schwedisch: få* torde* mäste* kunna*"
(mimeographed paper, Universität Tübingen).
WATT, W. C. 1970. "On Two Hypotheses Concerning Psycholiguistics,"
in Hayes (ed.) 1970, 137-220.
WEIGL, E. & M. Bierwisch, 1970. "Neuropsychology and Linguistics.
Topics in Common Research," Foundations of Language 6* 1-18.
WEINREICH, U. 1966. "Explorations in Semantic Theory," in T.A.
Sebeok (ed.) Current Trends in Linguistics* Vol 3. 395-
477, The Hague: Mouton.
WELLANDER, E. 1965. "Aktionsart och aspekt," Nysvenska studier
44 (1965) Tidskrift för svensk stil- och språkforskning,
184-252.
WESSEN, E. 1965. Svensk språkhistoria III. Grundlinje till en
historisk syntax* Stockholm: Almqvist & Wiksell.
1968. Svenskt lagspråk* Lund: (Skrifter utgivna av moders­
målslärarnas förening 101).
WHITAKER, H.A. 1970. "Linguistic Competence. Evidence from Aphasia,"
Glossa 4.1.
1971. On the Representation of Language in the Human Brain*
Edmonton: Linguistic Research Inc.
WOLFF, F. 1906. "Zur Frage des Accusativus mit dem Infinitiv,"
Zeitschrift für vergleichende Sprachforschung XIX (1906),
490-550.
WOLFRAM, W. 1969. Linguistic Correlates of Social Stratifica­
tion in the Speech of Detroit Negroes* Hartford Seminary
Foundation Thesis.
WOOD. F.T. 1956. "Gerund versus Infinitive," English Language
Teaching 11 (1956) 11-16.
WOODCOCK, E.C. 1959. A New Latin Syntax* London: Methuen and Co.
WUNDERLICH, D. 1970. Tempus und Zeitferenz im Deutschen* in
Linuistische Reihe 5* K. Baumgärtner, P.V.Polenz and
H. Steger (eds.), München: Hueber.
WURZEL, W.U. 1970. Studien zur deutschen Lautstruktur* in
Studia Grammatica VII* Berlin: Akademie Verlag.
WYA.TT, J.L. 1971. "Deep Structure in a Contrastive Transformational
Grammar," in Nickel 1971b, 75-82.
ZABROCKI, L. 1970. "Grundfragen der konfrontativen Grammatik,"
in Moser (ed.) 1970, 31-52.
ÖHLIN, P. 1918. Studier över de passiva konstruktionerna i forn-
svenskan* Lund: Håkan Ohlssons.
ÖSTERGREN, O. 1919-1968. Nusvensk ordbok* Stockholm: Wahlström &
Widstrand.
VERB COMPLEMENTATION
in Swedish and Other Germanic Languages
This book is a contribution to the theory of synchronic

comparative syntax, whereby the embedding processes in


three Germanic languages are focused upon, i.e. those of

Swedish, German, and English. Part One describes the generation


of verb complements which occur after S wedish verba sentiendi.
A number of deep, shallow, and surface structure constraints are

elaborated which are imperative to limit the generative power of the


Swedish rules of embedding in focus here, that is th e tf££-Embedding, the
/^r-Embedding, the Subject Raising, and the Nominalization Rules. The
productivity of these rules is demonstrated by the results from empirical
observations of verb complementation in the newspaper corpus of t
he

Gothenburg Research Group. Part Two presents an interlingual


approach to verb complementation after Swedish, German, a
nd

English verba sentiendi. The synchronic comparative model

outlined here liberates the linguist from any monolingual


bias concerning the verb complementation of one given
Germanic language. The interlingual approach yields

broader perspectives on general embedding pro­


cesses and their constraints in G ermanic lang.

Distributors
SPRÄKFÖRLAGET SKRIPTOR
Fack S-104 65 Stockholm 15

Cover designed by Avinash Talwar, Stockho

You might also like