Diachronic Grammar: History of Old and Middle English Subjectless Constructions (Linguistische Arbeiten) [Reprint 2011 ed.] 3484300973, 9783484300972

Over the past few decades, the book series Linguistische Arbeiten [Linguistic Studies], comprising over 500 volumes, has

229 75 27MB

English Pages 188 [192] Year 1981

Report DMCA / Copyright

DOWNLOAD PDF FILE

Table of contents :
Abbreviations
PART ONE. The grammar of Old English subjectless predicates
Introduction
1. Aim
2. Subjectless constructions
2.1 Previous research
2.2 The data base for the present study
2.3 The theoretical status of the data base
3. The methodological and theoretical framework
3.1 A first grammatical characterization
3.2 Goals, methods, and data
3.3 An outline of the descriptive framework
3.4 List of texts
CHAPTER I. Subjectless constructions with sentential complements
1. The OE paradigm
2. The basic form of the construction
2.1 Word order
2.2 Constituent structure
3. Grammatical relations and semantic roles
3.1 The syntactic functions of the animate-NP
3.2 Semantic role properties of the animate-NP
4. The OE RUE class
4.1 Semantic NP roles associated with RUE verbs
4.2 Syntactic occurrence properties of RUE verbs
4.3 Operational tests and paraphrase evidence for semantic roles
5. The OE PLEASE/DESIRE class
5.1 Semantic NP roles associated with PLEASE/DESIRE verbs
5.2 Syntactic occurrence properties of PLEASE/DESIRE verbs
6. The OE BEHOVE class
6.1 Semantic NP roles associated with BEHOVE verbs
6.2 Syntactic occurrence properties of BEHOVE verbs
7. The OE HAPPEN class
7.1 Semantic NP roles associated with HAPPEN verbs
7.2 Syntactic occurrence properties of HAPPEN verbs
8. The OE SEEM class
8.1 Semantic NP roles associated with þyncan
8.2 Syntactic occurrence properties of þyncan
9. Summary
9.1 The complete occurrence paradigm
9.2 The functional aspect
10. Towards an integrated basic structure
11. A dynamic view of OE subjectless grammar
CHAPTER II. Subjectless constructions with nominal complements
1. Preliminary remarks
2. Type N
2.1 The semantic framework
2.2 The syntax of type N
2.3 The occurrence of subjectless predicates in type N
2.4 A synopsis of type N
3. Two variant constructions entered by subjectless predicates
3.1 Variant type I
3.2 Variant type II
4. Summary
PART TWO. The development of subjectless grammar
Introduction
1. Aim
2. Theoretical aspects
CHAPTER III. The history of subjectless predicates until the Early Modern English period
1. The RUE class
1.1 RUE verbs in construction with sentential complements
1.2 RUE verbs in construction with nominal complements
2. The PLEASE/DESIRE class
2.1 liken
2.2 listen (OE lystan)
2.3 longen (OE langian)
3. The BEHOVE class
3.1 BEHOVE verbs in construction with sentential complements
3.2 BEHOVE verbs in construction with nominal complements
4. The HAPPEN class
4.1 HAPPEN verbs in construction with sentential complements
4.2 HAPPEN verbs in construction with nominal complements
5. The SEEM class
6. Summary
CHAPTER IV. Linguistic change within a restrictive framework
1. Introduction
2. The 12th century
3. The 13th century
4. The 14th century
5. The 15th century
6. The 16th century
Appendix: Translation of OE examples
References
Index
Recommend Papers

Diachronic Grammar: History of Old and Middle English Subjectless Constructions (Linguistische Arbeiten) [Reprint 2011 ed.]
 3484300973, 9783484300972

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

Linguistische Arbeiten

97

Herausgegeben von Herbert E. Brekle, Hans Jürgen Heringer, Christian Rohrer, Heinz Vater und Otmar Werner

Willy Elmer

Diachronie Grammar The history of Old and Middle English subjectless constructions

Max Niemeyer Verlag Tübingen 1981

To my mother and the memory of my father

CIP-Kurztitelaufnahme der Deutschen Bibliothek Elmer, Willy: Diachronie grammar : the history of Old and Middle English subjectless constructions / Willy Elmer. — Tübingen : Niemeyer, 1981. (Linguistische Arbeiten ; 97) ISBN 3-484-30097-3 NE:GT

ISBN 3-484-30097-3

ISSN 0344-6727

Max Niemeyer Verlag Tübingen 1981 Alle Rechte vorbehalten. Ohne ausdrückliche Genehmigung des Verlages ist es auch nicht gestattet, dieses Buch oder Teile daraus auf photomechanischem Wege zu vervielfältigen. Printed in Germany. Druck: Sulzberg-Druck GmbH, Sulzberg/Allgäu.

Acknowledgments

I would especially like to thank three of my teachers who tried to further my understanding of linguistics: Eduard Kolb (University of Basel) and Oskar Bandle (Universities of Basel and Zurich), who first introduced me to historical linguistics, and Neil V. Smith, who made my stay at University College London a memorable synchronic experience. I have also benefited greatly from criticism and suggestions from H. Brekle (Regensburg), H. Burger (Zürich), X. Dekeyser (Antwerpen), H. Käsmann (Heidelberg), Ed. Kolb and H. Löffler (Basel), all of whom have read parts of an earlier version of this study. Special thanks go to Geoff Aspin and Joyce Bachmann-Clarke (Basel), who read part of the typescript and made valuable comments. Most important help — in more ways than one — I received from my friend Mike R. Rylance (Wakefield); this book would have profited had I accepted more of his northern (and therefore 'cool') suggestions. I would also like to thank my fellow-students Ingrid Bell-Krannhals, Joerg Berger and Anna Stocker-Edel for assistance in reading proof. Needless to say, the responsibility for remaining errors is my own. Finally, I owe a great debt of gratitude to Mrs. M. Wasser (Zürich), who did splendid work in typing a difficult manuscript on her own IBM-composer.

The research on which this book is based was made possible by a substantial grant from the Schweizerischer Nationalfonds zur Förderung der wissenschaftlichen Forschung.

Contents Abbreviations

XI

PART ONE The grammar of Old English subjectless predicates Introduction

3

1.

Aim

3

2. 2.1 2.2 2.3

Subjectless constructions Previous research The data base for the present study The theoretical status of the data base

3 3 5 7

3. 3.1 3.2 3.3 3.4

The methodological and theoretical framework A first grammatical characterization Goals, methods, and data An outline of the descriptive framework List of texts

8 8 10 12 15

CHAPTER I Subjectless constructions with sentential complements 1.

The OE paradigm

21

2. 2.1 2.2

The basic form of the construction Word order Constituent structure

23 23 26

3. 3.1 3.2

Grammatical relations and semantic roles The syntactic functions of the animate-NP Semantic role properties of the animate-NP

27 27 28

4. 4.1 4.2 4.3

The OE RUE class Semantic NP roles associated with RUE verbs Syntactic occurrence properties of RUE verbs Operational tests and paraphrase evidence for semantic roles

29 30 32 35

5. 5.1 5.2

The OE PLEASE/DESIRE class Semantic NP roles associated with PLEASE/DESIRE verbs Syntactic occurrence properties of PLEASE/DESIRE verbs

37 37 38

6. 6.1 6.2

The OE BEHOVE class Semantic NP roles associated with BEHOVE verbs Syntactic occurrence properties of BEHOVE verbs

40 41 41

VII

7. 7.1 7.2

The OE HAPPEN class Semantic NP roles associated with HAPPEN verbs Syntactic occurrence properties of HAPPEN verbs

43 43 43

8. 8.1 8.2

The OE SEEM class Semantic NP roles associated with pyncan Syntactic occurrence properties offryncan

44 44 44

9. 9.1 9.2

Summary The complete occurrence paradigm The functional aspect

45 45 48

10.

Towards an integrated basic structure

52

11.

A dynamic view of OE subjectless grammar

54

CHAPTER II Subjectless constructions with nominal complements 1.

Preliminary remarks

59

2. 2.1 2.2 2.3 2.3.1 2.3.2 2.3.3 2.3.4 2.4

Type N The semantic framework The syntax of type N The occurrence of subjectless predicates in type N The OE RUE class The OE PLEASE/DESIRE class The OE BEHOVE class The OE HAPPEN and SEEM classes A synopsis of type N

60 60 61 63 64 64 65 66 66

3. 3.1 3.1.1 3.1.2 3.2 3.2.1

Two variant constructions entered by subjectless predicates Variant type I Subjectless predicates in type I The type it-V-NPa Variant type II Subjectless predicates in type II

67 67 69 71 71 72

4.

Summary

75

VIII

PART TWO The development of subjectless grammar Introduction 1.

Aim

81

2.

Theoretical aspects

81

CHAPTER III The history of subjectless predicates until the Early Modern English period 1. 1.1 1.1.1 1.1.2 1.1.3 1.2 1.2.1 1.2.2 1.2.3 1.2.4 1.2.5

The RUE class RUE verbs in construction with sentential complements TypeS The /f-construction The syntactic paradigm of ME RUE verbs RUE verbs in construction with nominal complements Type N Type I Type II The transitive construction The development of the ME paradigm with nominal complements

85 85 85 86 88 92 92 94 98 101 103

2. 2.1 2.1.1 2.1.2 2.2 2.2.1 2.2.2 2.3 2.3.1 2.3.2

The PLEASE/DESIRE class liken In construction with sentential complements In construction with nominal complements listen (OE lystan) In construction with sentential complements In construction with nominal complements langen (OE langian) In construction with sentential complements In construction with nominal complements

107 107 107 110 114 114 116 118 118 119

3. 3.1 3.2

The BEHOVE class BEHOVE verbs in construction with sentential complements BEHOVE verbs in construction with nominal complements

121 121 125

4. 4.1 4.2

The HAPPEN class HAPPEN verbs in construction with sentential complements HAPPEN verbs in construction with nominal complements

129 129 132

5.

The SEEM class

133

6.

Summary

135

IX

CHAPTER IV Linguistic change within a restrictive framework 1.

Introduction

141

2.

The 12th century

142

3.

The 13th century

146

4.

The 14th century

149

5.

The 15th century

154

6.

The 16th century

156

Appendix: Translation of OE examples

163

References

167

Index

175

Abbreviations IE OE ME ON OS OHG MHG ASPR BT BTs DWB EETS MED OED vdG Vi. Wai. c.

Indo-European Old English Middle English Old Norse Old Saxon Old High German Middle High German The Anglo-Saxon Poetic Records Bosworth-Toller, An Anglo-Saxon Dictionary An Anglo-Saxon Dictionary, Supplement Deutsches Wörterbuch (J. & W. Grimm) Early English Text Society Michigan Middle English Dictionary Oxford English Dictionary van der Gaaf (1904) Visser (1963-73) Wahl6n(1925) Century (in chronological statements)

XI

PART ONE THE GRAMMAR OF OLD ENGLISH SUBJECTLESS PREDICATES

INTRODUCTION

Aim

1.

The aim of this study is twofold: first (in part I), to characterize in a comprehensive way the grammar of the class of OE predicates which generally occur with a pseudo-subject in dative/accusative case instead of a nominative subject noun phrase (henceforth NP). Predicates occurring with non-nominative noun-phrases in place of nominative subjects will be called 'subjectless'.1 In OE they are typically found in structures like (l)-(3), henceforth termed subjectless constructions: (1) (2) (3)

me hreowef) Ί rue' me hreowep paes Ί rue (of) this' me hreowef) paet... Ί rue that...'

Second (in part II), to describe the development through and beyond ME of what will thus be defined as 'the grammar of OE subjectless predicates'.

2. Subjectless constructions Before turning to the discussion of the methodological and theoretical issues arising under this general aim, we will first try to become more familiar with the nature of our data by looking at the aims, methods and results of previous research.

2.1

Previous research

Three works dealing with subjectless predicates (as defined above) have to be mentioned. The earliest systematic treatment of the class of predicates represented by hreowan is van der Gaaf s The transition from the impersonal to the personal construction in Middle English (1904).2 Van der Gaaf s pioneer study is still the most complete account available. His formulation of the problem is the traditional one: he is interested in docu-

Thus the term is not used to denote genuine subjectless constructions like weather expressions, for example (rinp 'it rains'). The class of verbs dealt with here differs from the so-called Impersonalia simply in that they do occur with a pseudo-subject. Cf. also Sunden (1918) for an investigation intolystan/hungran type predicates from a 'semological' and genetic point of view. For a historical account of the above and other kinds of impersonal construction see Brugmann, Delbr ck, Behaghel and Held (cf. the bibliography).

menting the ME change from the 'impersonal' (me hreowefj ...) to the 'personal' (I rue ...) type. On the basis of 14 OE predicates1 he illustrates the impersonal constructions of the kind me hreowep with optional nominal or sentential complementation and then goes on to describe their syntactic and semantic development in ME. On the basis of his data (mainly from 14c. and 15c. texts) he suggests that the change in question should be described as an instance of case shifting, i.e. a type of reanalysis interpreting the OE dative/accusative pseudo-subject as a nominative subject. The main causes for this reanalysis are the ME loss of case distinctions and the gradual establishment of subject—verb—object word order. With rare exceptions the resulting personal construction was in general use by the first half of the 14c.( while in most cases the old type continued until about 1500. This account is quite clearly correct and has not been challenged even by the most recent generative work on the subject.2 It is not surprising that van der Gaaf s study carries some of the marks of a pioneer work. We are not given, for example, the whole range of OE and ME predicates which occur in impersonal construction. The selection of predicates is not discussed and thus no systematic lexical or semantically-based classification is possible. Furthermore, the account of the development in the (crucial) 16c. seems rather to be dictated by the limited amount of data considered than to reflect the actual situation. There is no discussion of methodological issues, the main arrangement being along the parameter 'verbs which have/have not preserved the impersonal construction'. However, the importance of this work for any subsequent study lies in its direct and 'natural' (if largely impressionistic) presentation of the typical properties of subjectless verbs.The value of such an informal account should not be underrated. By freely referring to the different aspects involved (syntax, semantics, morphology and grammatical relations) such work can show the complex nature of a grammatical phenomenon not previously realized. While not systematically presented, the insights achieved in this way allow more methodologically-oriented approaches to devise adequate frameworks for more systematic analysis. Wahlen's study The Old English Impersonalia (1925) is a descriptive and genetic work about 'verbs of material import' occurring in impersonal construction (both genuine and quasi-impersonal). The study rests on a broad lexical basis, including verbs denoting natural phenomena (hit rinf) 'it rains'), time (hit aefenlaecefr 'it grows evening'), 'physical and mental affections' (hine caelfj 'he is cold', hine hyngrep 'he is hungry'), the state of things, course of events (gebyrian, limpian 'happen') etc. In OE the occurrence properties of these predicates do not involve the issue of 'personalization'. However, both the wealth of material and the etymological second part render this work a valuable help for any historical study concerned with the impersonal use of verbs.

1 gebyrian, eglian, hreowan, langian, labian, lap beon, lician, lystan, maetan, byncan, wa beon, wel beon, god beon, leofbeon. 2 Lightfoot (1979).

Visser's monumental An historical syntax of the English language (1963—73), also documents the impersonal to personal change. He accepts van der Gaaf s conclusions (modifying them insofar as he finds Α-type (impersonal) constructions in More, thus showing that the change was not yet completed before the mid-16c.). The lexical basis has been extended and the important aspect of possible case neutralization extensively illustrated. Apart from the necessarily fragmentary representation in Visser, however, the topic of impersonal constructions of the type dealt with here has not received any detailed attention since Gaaf s study, although there has recently been renewed interest in this kind of linguistic change on the part of theory-oriented generative grammarians. In the work done by Lightfoot1 the development of the impersonal subpart of OE grammar is given a wholly new rationale within the latest framework of generative theory. The present study is not conducted within an a priori given methodological framework, for reasons mentioned below (see section 3). Thus our results will have no direct bearing on Lightfoot's claims, because his theory has been immunized (by incorporating trace theory and the logic of markedness) to such a degree that it cannot be refuted by primary data of the form presented here. It has to be mentioned, though, that the model of generative grammar cannot accomodate the range of data which form the object of the present investigation, largely because semantics and grammatical relations play a crucial part. Rather than seeking to extend the theory of grammar to make it responsive to the current set of data, our purpose in this study is to establish and systematize as far as possible what seem to be the essential properties of the 'subjectless subpart' of OE and ME grammar. This task has not been attempted so far. It is, however, a prerequisite for a discussion of model-dependent approaches and of relevant theory-construction in general.

2.2

The data base for the present study

Lexically, the subpart of OE studied here is characterized by the fact that it can be exhaustively illustrated on the basis of a small number of verbs 2 . The data on which the present study is based form the following five semantic verb classes typically occurring in the syntactic construction types (2) and (3) above, i.e. with two semantic arguments, usually a 'person' and a 'thing', the latter in the form of a nominal or sentential complement:

1 Cf. in particular (1976b) and (1979). 2 There are about 40 simple predicates in OE taking non-nominative noun phrases as pseudo-subjects as in (l)-(3) on page 1. Of these only about 20 occur in construction with sentential and nominal complementation, for example (2) and (3). Only verbs whose attested occurrence allow valid generalizations have been included. See Visser (1963, 20-35) for a comprehensive list of subjectless predicates.

I. II. III. IV. V.

The class RUE1: (-)hreowan, (-)sceamian, eglian, ofbyncan The class PLEASE/DESIRE: (ge)lician, (ge)lystan, langian, (ge)lustfullian The class BEHOVE2: (ge)byrian, gerisan, gedafenian, behofian The class HAPPEN3: (ge)limpan, (ge)\veorban The class SEEM: byncan

Classes I—V do not form a uniform semantic framework. They are, however, restrictive in relational terms, viz. by the fact that they are typically — some of them obligatorily associated with an animate (generally human) noun phrase in non-nominative case and pseudo-subject function.4 The present data base was established by reading the major part of OE prose literature and selected texts (of both prose or poetry) for each century from the 12th to the 17th, with the aim of collecting all occurrences of the verbs mentioned above in the two construction types with no'minal (him hreoweb baes) and sentential (him hreoweb baet...) complements. The result is a representative corpus of some 3500 sentences.5 (The 'absolutive' type without complement, me shameb, is hardly ever met in narrative context. It has also been included, however, for reasons of completeness and will be used where it can throw light on problems such as the pseudo-subject status of the animate noun phrase or the semantic use of predicates).

1 Verb meanings are given at the beginning of sections 4-8. Forms with the prefixes of-, ge-, forhave been included in the data base and will be used freely for illustration, although reference will be generally made to the simplex. A systematic distinction of such variants can be made neither in semantic nor in stylistic terms, as Baron(1974,90) has shown. 2 The variant byrian is not frequently met with, while the use of dafenian is extremely rare. 3 To this class is added gebyrian when used to mean 'happen'. See p. 22. 4 They are crucially distinct from genuine impersonals (such as weather verbs) by this focus on the associated noun phrase. The term 'impersonal' has generally not been defined more specifically and remains vague, especially with regard to syntax. Complaints about the terminological inadequacy of 'impersonal' or 'subjectless' constructions go back to Paul and Behaghel. Van der Gaaf (1904) calls the type involved here 'quasi-impersonal' due to the presence of an animate noun phrase, reserving 'impersonal' for the type (it) gedafenafr pact.... The terminological confusion directly reflects the difficulty of distinguishing consistently between syntactic, semantic ('logical') and psychological subject properties. 5 Cf. the list of texts, pp. 15 ff. Evidence from poetry and drama has been mainly obtained from the concordances (Bessinger, Gburek, Preston, Kottler & Markman, Crawford, Donow, Tatlock & Kennedy, Spevack, see p. 18f.). The selection of OE texts was guided only by the principle of an even representation of the whole period, with a particular preference for non-translated prose. Within these limits, any choice will always appear arbitrary to the critic for whom completeness (in terms of 'all the texts available') is decisive. It will become quite evident, however, that when we are dealing with syntactic construction types, (relative) completeness - as in the case of OE - cannot establish any new structures. We know the set of verbs entering subjectless types and the range of subjectless constructions. All a 'complete' OE data base could add in the present case would be greater quantity (more examples per type). See the introduction to part II for a discussion of the selection of ME texts.

2.3

The theoretical status of the data base

Before dealing with the internal structure of the data base, a word has to be said concerning its theoretical status. The fact that the set of OE sentences will serve as a synchronic basis, representing the subjectless aspect of the grammar of OE implies that these data belong 'to the same grammar'. This is not only a practical problem: the fact that we are dealing with texts from the 9c.-llc. would seem to speak against such an assumption. In addition to the time span, various other factors contribute to the relatively heterogeneous form of OE data in general (for example scribes and their schools, regional dialect differences, degrees of contact with Norse or Latin, etc.). The effects of these external factors on the language are described in detail in the literature. They are, however, limited to the phonology and the lexicon and with regard to syntax do not warrant an assumption of 'different grammars' in any consistent way. Our attitude towards the notion 'synchronic OE' is that adopted implicitly by the authors of OE grammars, who never doubted its unity (in a systematic1 as well as practical, communicative sense, incorporating dialectal variation). There is no evidence of any systematic syntactic change between the language of Alfred and Wulfstan, so that for our present purpose the syntax can clearly be treated as one coherent object during the whole period. It follows from the external situation with regard to our material that the analysis of OE subjectless constructions will involve a certain degree of extrapolation with respect to single texts. It will be seen, however, that considerations of absolute chronology or geographical location, which in phonology would be decisive criteria for the classification and interpretation of data, do not carry nearly the same weight in OE syntax. Syntactic structure is obviously not subject to the kind of dialectal variation characteristic of phonology.2 As has often been stated, there are methodological limits to the ways in which the present kind of historical data can be used in linguistic analysis. This is a finite set of sentences which cannot be augmented at will. To conclude that this situation a priori does not allow any judgements as to the grammaticality of potential variants and thus excludes relevant evaluation, would be to underestimate the systematic strength of generalizations based on these data, as well as the validity of comparative evidence — historical and modern — within the present semantic framework. It is naturally true that historical data of the kind used here can fulfil only to a degree the demands of a grammatical model which inherently presupposes intuitive knowledge of the language in question and the capability to produce new correct sentences. It should

1 I.e. assuming relatedness between, for example, phonological variants in terms of correspondences. 2 The assumption of OE subjectless syntax' as a coherent phenomenon (and not the result of an underdetermined data base) receives support from the respective variability of phonology and syntax in modern dialects.

be stressed, however, that our aims are explicitly not those of generative grammar, viz. to specify the set of grammatical and ungrammatical sentences of a given language. We are without doubt dealing with fully grammatical sentences only. Moreover, the distributtional facts concerning structural types emerging from our data base are such that they allow a definite statement about the range of subjectless constructions possible in OE, as well as a strongly motivated statement about the nature of actually used (as against potentially available) syntactic variants. These are exactly the aspects of grammar which promise a valid foundation on which to evaluate diachronic change.

3.

The methodological and theoretical framework

3.1

A first grammatical characterization

As mentioned briefly at the beginning, we will be dealing with structures of the following kind:1 (4) (5) (6) (7)

Me ί>ίηφ nu paet... (B. 27.15) & hine ne lyst his willan wyrcean (BH 51,16) oft pone gepyldegestan scamaf) freies siges (CP 227,19) hine fraes langode (BH 227,1)

These constructions differ from 'regular' OE sentences by not having a subject in the nominative case; instead there is an oblique noun or pronoun. In this situation the verb remains in the 'neutral' third person singular even where the noun-phrase is in the plural (see p. 21 f.). As will be seen in chapter I, the positional syntax of this noun-phrase is exactly the same as that of 'real' subject noun-phrases. In 'personal' OE syntax the nominative NP is normally associated with grammatical subject function. The question is, however, whether the inverse can also be generalized. Syntactically, sentences like (4)—(7) have no subject. The fact that there is no overt syntactic subject does not automatically imply that no element in the sentences above stands in a functional subject relation to the predicate, i.e. has some properties of what has traditionally (in notional terms) been called a grammatical subject. It appears that in the present case the functional status of the animate-NP should not be syntactically defined but is most adequately characterized in semantic terms, viz. those determined by the specific semantic situation associated with syntactically subjectless constructions.

1 Fot text references see pp. 15ff. For technical reasons, p and d are not distinguished. 8

Notice that (4)-(7) are intransitive constructions whose exact meaning and functional interpretation is not easy to assess and paraphrase for the contemporary linguist. In particular, we note that whereas the syntactic order of elements in the main clause is XVO — corresponding to the SVO order of unmarked personal syntax — from a semantic point of view (4) and (5) at least are OVS structures.1 It is the full content (the semantic and syntactic aspects) of the grammatical relation between the verb and the animate-NP which characterizes the subpart of Old and Middle English grammar investigated by the present study. Note that not all subjectless predicates are restricted to occur in sentences like (4)—(7). We will see that some of them also appear with nominative subjects, thus allowing a variation pattern of the kind me rues ... 11 rue ...In such pairs of subjectless vs. personal construction the question arises as to what the morphosyntactic difference means for the semantic interpretation. The status and occurrence properties in OE grammar of this type of potential variation will be dealt with individually for each subclass of predicates. In OE, grammatical relations are signalled both by position and case inflection (respectively prepositional marking)2. In cases of variation as mentioned above it thus happens that subjectless constructions are only distinguished from a 'corresponding' personal variant by the case marking on the initial NP. In such a case the subjectless type can formally be characterized by reference to the criteria of case inflection and the lack of verb agreement in the plural. However, this is clearly not sufficient to determine the nature of the grammatical variation involved. A correct description of the content of this variation - and of the pseudo-subject relation in general - has to take into account semantic information. The traditional way of doing this is by reference to the use of the verbs as transitive, intransitive or quasi-transitive predicates. Another possibility is in terms of the semantic roles3 taken by the NP. It appears that a relevant characterization of the grammar of OE subjectless predicates has to try to bring this type of semantic information in line with the formal syntactic structure. It is to be noted that we are not concerned with the proper semantic structure of the present set of verbs as such, but with their use with regard to the role properties of the associated animate-NP. For the modern linguist, comparative access to the content of the verb-NPa relation is provided by those Germanic dialects (such as Swiss-German) which have preserved it in their grammar. In OE and ME, however, such evidence can only be inferred on the basis of verb meaning and evidence from the type of syntactic variation mentioned above.4 The importance 1 X stands for the pseudo-subject, V for verb, Ο for object, S for subject. See section 9.2 on the functional aspect. 2 The OE case system is generally still sufficient to allow the relevant distinctions to be expressed, although not with proper nouns. 3 I.e. the number and kind of 'deep cases' associated with a predicate. See Fillmore (1968). 4 Cf. sections 1-4 of chapter I for an illustration of how semantic and syntactic information is used in this study.

attributed to grammatical variation demands that the subjectless subpart of grammar should not be analysed in isolation from 'personal' grammar. As far as subjectless syntax is concerned, reference should be established to the properties of OE 'personal' syntax such as generalizations about word order and syntactic paraphrase possibilities. Likewise a comparative discussion of semantic NP-roles (in subjectless or personal use) has to be made possible, whatever methodological reservations might arise in so doing. The consequences of these observations for our descriptive framework are dealt with in the next section.

3.2

Goals, methods, and data

The preceding discussion of the grammatical properties of subjectless predicates allows us to return to our original formulation of the problem and to the question of method in a more informed way. Let us first try to get a clearer idea about the interrelationship between our goal and the range of data involved on the one hand and currently available methods of analysis on the other. Our general aim was formulated on the basis of the fragmentary (but clearly relevant) results of the only full-scale study so far, van der Gaaf (1904). The description attempts to 'characterize' (rather than 'explain') OE subjectless grammar and its development. This goal is well motivated for the problem before us, as will have become clear. There are no definite analytic methods associated with our formulation of the present goals. The investigation could be performed within a traditional, structuralist or generative framework. However, if we look for an adequate descriptive framework for historical and diachronic syntax we are disappointed. Far from offering any useful models, the present research situation is characterized both by a lack of common (traditional and structuralist) heritage and by completely conflicting views on the nature of linguistic change, which leave no hope of an integrated model in the near future. As far as syntax is concerned this situation is due to the fact that — unlike phonological research — the study of historical syntax cannot build on the neogrammarian heritage of well workedout theoretical concepts. It is true that general theories of natural language have been steadily forthcoming in the last decade. They have, however, almost exclusively dealt with synchronic grammar, and sometimes even excluded historical study in principle. There is hardly an outline of a theory of diachronic change which would offer the explanatory framework often discussed with regard to synchronic grammar.1

1 Outstanding genezative accounts based on single-factor concepts of change include Klima (1964), Tiaugott [Closs] (1965), Lakoff (1969), Bevei & Langendoen (1972) and Lightfoot (1979). For further discussion see the introduction to part two.

10

With respect to linguistic change most recent observers agree that no such framework is available today.1 Two remarks by well-informed linguists illustrate the point: David Lightfoot notes that 'the recent attention paid by generativists to diachronic syntax has led to no significant implications for a general theory of grammar and, from a purely historical viewpoint, has failed to focus on any well-defined concepts of historical change'.2 In his introduction to the Chicago Linguistic Society paravolume on diachronic syntax (1976), Ebert writes: 'In historical syntax there is as yet no integrated model which encompasses recent approaches'. He goes on to explain why the promise of early transformational generative grammar was not fulfilled (p. viii) and points out the recent return to explanations of new syntactic construction types in terms of surface reanalysis — i.e. in traditional terms. Ebert adequately qualifies the present research situation when he notes that 'there is a certain tension between attempts to achieve explanatory adequacy and the empirical requirements of a descriptively adequate account' (p. xiv). The syntax is naturally a fundamental part of 'the grammar of OE subjectless predicates', but we have seen that it has to be complemented by additional aspects of grammatical structure largely disregarded by previous and current generative work and thus not accommodated by the most advanced formal models used in present-day analysis.3 With regard to the present study then, the conclusion is that no descriptive framework currently available could adequately accommodate the different grammatical aspects which appear to be crucial features of the grammar of subjectless predicates. A comprehensive analysis has to take into account the fact that in the constructions we are concerned with semantic, relational4 and morpho-syntactic properties are interdependent and have to be described as such. This conclusion will have important consequences for the descriptive method, because it implies statements and assumptions not amenable to the kind of rigid formalization possible in syntax.5 Possibly the strongest model of linguistic change suggested so far is by Andersen (1973). Although it has only been applied to phonology, it could also serve as a fruitful framework for the discussion of one central aspect involved in the history of subjectless grammar, viz. the process of (semantic and functional) re-interpretation of syntactic surface structure.

1 One of the explicit alternatives to a generative account, Samuels (1972), does not include syntax. 2 (1976b,129) 3 Cf. Lightfoot (1976, 1979), Ebert (1976), both within the framework of the 'extended standard theory', or revisions of it. 4 Depending on the context, the term 'grammatical relation' refers to the (configurationally or morphologically defined) function of the animate-NP as subject, indirect object or pseudo-subject or to the relation between verb and animate-NP in terms of semantic roles. 5 Cf. Li (1975a, preface), expressing a similar idea with regard to research on subject and topic. 11

At present, however, it is hard to see in what way Andersen's model of abductive change could be extended to accommodate the present range of data and in particular in what terms classes of abstract lexical entries should be formalized to make them fitted for this descriptive framework. The assumption that, faced with indeterminacy in language, the speaker/hearer uses the reasoning of abduction to infer a new correct interpretation — in the present case, to bring syntax and grammatical relations in line — is of such general validity that it can serve as a rationale even to an informal account. To repeat, then, our aim is to describe as comprehensively as possible the various factors involved in shaping the subjectless syntax of Old and Middle English, to determine the general grammatical conditions which will lead to a situation in which reanalysis becomes possible or necessary and to follow the directions of development actually taken by the language. In a research situation which does not allow this goal to be formulated internal to some well-defined framework of assumptions concerning historical and diachronic grammar, one possible path to take is to establish a firm data base and to proceed by means of statistical generalizations allowing distinctions in terms of norm and variation where possible (i.e. in syntax) and abductive hypotheses where appropriate (i.e. in questions of grammatical relations and semantic use). The present approach will be along this path.

3.3

An outline of the descriptive framework

The present descriptive framework will be directly (though not exclusively) suggested and motivated by the structure of the primary data. The exposition of the material and the discussion of method will go hand in hand. The first methodological decision is to separate, for the description, the 'synchronic' OE and the diachronic ME data. This is largely for practical reasons, but it also reflects our conviction that linguistic change has to be described on the basis of (and preferably in terms of) the system that gives rise to it. That the study of word order is crucially relevant for an account of syntactic development has been confirmed by recent work.1 In OE it is possible to establish well-motivated order types and thus to describe sentences as regular representatives of a type or as order variants. In this way synchronic variation is brought to bear on diachronic change. The individual status of syntactic variants is not defined for us by any model, but has to be established and justified on the basis of the primary data. (Compare, for example, the question as to whether some variation is purely stylistic or represents a different construction type).

1 Cf. Lehmann (1974) and Lightfoot (1979), whose main hypothesis is based on an early English word order change from SOV to SVO. See also Li (1975a). 12

What we need, then, is a descriptive framework which allows first of all a systematic representation and discussion of the syntactic properties of subjectless constructions in OE. Justified directly by surface generalizations, it should represent the unmarked types of structure in terms of order and constituency. The syntactic behaviour of the major elements — in particular of the pseudo-subject — can then be evaluated with regard to regular occurrence and possible variation, (word order variants within a single type). In this way, the upper limits of possible variation can be stated, an aspect crucially important for the analysis of the focal element, the animateNP. Complementary to the investigation of the syntactic structure of the construction types represented by (4)—(7), however, the grammar of subjectless predicates demands a description of their syntactic behaviour in terms of lexical valency1 and their occurrence properties in different syntactic types. Evidence from the complete occurrence paradigm of subjectless verbs (for example their possible use also in personal construction) is particularly important in determining the nature of the grammatical relation to the animateNP regularly associated with them. Our first task, then, will be to establish the place of categorial elements in the two construction types (sentential complements: chapter I, nominal complements: chapter II). Justifying the abstract form of syntactic types involves two kinds of abstraction. The first is based on generalizations of the linear order of main elements. The present concept of typological structure should be regarded as one phrase marker for the class of sentences illustrated by (4)—(5) and (6)—(7) respectively.2 The form of the basic structure will not receive deductive justification, however. We are not concerned here with the rules which generate the abstract structures of OE sentences, nor with the problem of a basic distinction between grammatical and ungrammatical sentences, but in the regularities and variation possibilities within the class of attested (i.e. grammatical) sentences. Under the present approach, the form of abstract structure will be defined in terms of distributional generalizations and a straightforward notion of constituent structure (which adds the hierarchical structure of syntactic categories in different construction types). Such an abstract structure is thus explicitly a surface construct. That the linear order of syntactic categories receives priority (especially in an historical analysis) over proper constituent structure and the semantic intuition needed to argue for 'deep' structure, is the lesson that traditional research and the recent development of linguistic method have taught us.

1 Here used in a purely syntactic sense: the full range of potential occurrence properties of verbs. 2 Note that in generative grammar phrase markers are associated with the derivation of single sentences only.

13

Adopting such a concept of syntactic type — henceforth basic structure - as our descriptive framework gives priority (for initial purposes of the analysis) to the syntactic level of structure. It does not, however, exclude a relevant treatment of grammatical relations and the semantics of individual predicates. In the present context the basic structure fulfils two main purposes: (1) it will serve as a structural framework for the discussion of data from internal variation within a type in a systematic way. It will also open an approach to the issue of relatedness between different types, both in terms of general syntactic structure and the individual lexical valency of verbs. The concept of basic structure should allow reference from, for example, a subjectless type to syntactic paraphrase constructions. It should allow us, in other words, to describe the difference involved in a potential variation pattern of the kind him rues that.../ it rues him that.../ he rues that..., relative to the syntactic properties of the construction types which these examples represent. In this way, adopting the basic structure will allow a systematic treatment of subjectless constructions vis-ä-vis OE personal syntax, in which the same class of predicates occurs with syntactic nominative subjects and in subject or indirect object relation to the animate-NP. (2) Besides the account in proper syntactic terms, however, the descriptive framework of basic structure should also be compatible with a coherent discussion of the grammatical relations present within a type, both in terms of the semantic roles of the two arguments and their function as subject, indirect object or pseudo-subject. Needless to say, this aspect cannot be rigidly formalized.1 The effect of syntactic variants on the relational situation within the present semantic context has not yet been studied.2 It is clearly relevant, however, in determining the relative or absolute status of constraints on syntactic variation allowed with individual predicates. Whereas the occurrence of the animate-NP in the three functions mentioned above — especially in the pseudo-subject function — is dependent on general syntactic possibili-

1 The degree and exact nature of interrelation between the syntactic and relational aspects is not given to us by any theory. It is only on the basis of an extensive and balanced account of the factual data that questions of this kind can be fruitfully dealt with at the present stage of research. Whatever theoretical framework and methods are applied to the problem outlined above, their success will depend on the degree of principled relationship they can establish between these two levels of analysis. 2 The extent to which such an integrated view of syntactic variation is justified depends of course on the amount of evidence for the semantic paraphrase nature of variants. Such evidence might be text-internal or from the same historical period or even based on the lexical valency of modern predicates of the same semantic classes. Given the present semantic framework, we will have to decide on the kind and degree of relationship between syntactic variants for each verb class individually. 14

ties, the semantic roles which the animate-NP can take are determined by the individual semantic structure of predicates. Adopting the construct of basic structure as outlined above allows a description of OE subjectless grammar along the parameters of (1) and (2) and thus promises to serve as a valid foundation for the analysis of diachronic development. This line of approach promises to be tight enough to allow systematic — restrictive — statements about syntactic structure and at the same time flexible enough to allow reference to semantically-based aspects such as the role properties of NPs. The loss of rigid formalization is outweighed by the consideration given to the occurrence in the same grammar of syntactic variants for the expression of a specific semantic situation associated with a particular set of grammatical relations. Contrary to much current practice in synchronic syntactic work, where generalizations are often based on small sets of data, it is essential for a historical syntactic phenomenon to be amply documented. Within the limits of possible variation, syntactic variants in OE reflect choice in terms of language use, and a sufficient data base therefore allows us to assess the productivity of some types compared to others, an indispensable condition for evaluating both the status of synchronic variant types and subsequent pathways of development. The rich material at our disposal promises significant surface generalizations of the relevant aspects to be made. Although corpus-based, our reasoning is not a priori limited to inductive inference. Thanks to traditional research we are quite well informed about the general properties of OE grammar (such as phrase structure or the syntactic behaviour of predicates). The methodological framework outlined above is in line with our belief that it is necessary, before relevant theory-construction can proceed, to obtain much further insight into surface typology, in order to have a clear idea of the range and structure of phenomena to which an eventual theory is to be made responsive.

3.4

List of texts

The following OE texts have been examined with regard to the occurrence of syntactically subjectless constructions:1 The Oldest English Texts. EETS 83. (OE Texts, plus short title) Die angelsächsischen Prosabearbeitungen der Benediktinerregel. Ed. A. Schröder, 1964.

(BR) l

The abbreviations given in brackets are those used in the quotations. In prose texts, references are to pages and lines, in poetry to lines. 15

The BlicklingHomilies. EETS 58, 63, 73. (BH) The Old English Version of the Heptateuch. EETS 160. (GEN, EX) Alfred Corpus: King Alfred's West-Saxon Version of Gregory's Pastoral Care. EETS 45, 50 (HattonMS)1 (CP) King Alfred's Orosius. EETS 79. (Or.) The Old English Version of Bede's Ecclesiastical History of the English People. EETS 95, 96,110,111. (HE) King Alfred's Old English Version of Boethius. Ed. WJ. Sedgefield (1899), 1968. (B) King Alfred's Version of St. Augustine's Soliloquies. Ed. Th.A. Carnicelli, 1969. (SOL) Aelfric Corpus: Aelfric's Lives of Saints. EETS 76,82,94,114. (Lives I, II) Homilies of Aelfric, A Supplementary Collection. EETS 259,260. (Aelf.Hom. Supp. I, II) Homilies of Aelfric. Ed. B. Thorpe, 1844 (2 vols). (Aelf.Hom. I, II) Wulfstan, Sermo Lupi ad Anglos. Ed. D. Whitelock (1939), 19633 (revised). (W. Sermo) Die Hirtenbriefe Aelfrics. Ed. B. Fehr, 1966. (H) The Homilies of Wulfstan. Ed. D. Bethurum, 1957. (W) Two of the Saxon Chronicles Parallel. Ed. Ch. Plummer, (1892), 1972 (2 vols.) (Chr.) The Middle English data have been collected on the basis of the following texts:

12c. Old English Homilies. First Series. EETS 29,34. (Lamb.) Old English Homilies. Second Series. EETS 53. (Trin.) Twelfth-Century Homilies. EETS 137. (Bod. Horn.)

1 Unless otherwise indicated, references are to the manuscript (MS) mentioned after the EETS number. The choice of MS variants does not follow any linguistic criteria, because with respect to syntax there are no systematic differences.

16

Early English Homilies from the Twelfth Century MS Vesp.D.XIV. EETS 152. (E.E.Hom.) Die alt- und mittelenglischen Apolloniusbruchstücke. Ed. J. Raith, 1956. (Apol.ofT.)

13c. The Orrmulum. Ed. R.M. White and R. Holt, 1878 (2 vols.). (Orm) Vices and Virtues. EETS 89.

(W) The English Text of the Ancrene Riwle. (Titus MS). EETS 252. (AR) The English Text of the Ancrene Riwle: Ancrene Wisse. EETS 249 (Corpus Christi MS). (AW) HaliMeidenhad. EETS 18 (Titus MS). (HM) pe Liflade of St. Juliana. EETS 51 (Royal MS). (Jul.) Seinte Marherete. EETS 193 (Bod.MS): (Marh.) The Life of Saint Kathenne. EETS 80 (Royal MS). (Kath.) Layamon's Brut. Vol. I. EETS 250 (Caligula MS). (Lay.) The South English Legendary. EETS 235,236, 244. (SEL I, II) An Old English Miscellany. EETS 49. (Best., O.KentSer.) Havelok. Ed. F. Holthausen, 19283. (Hav.)

14c. Cursor Mundi. (Lines 1-12558). EETS 51, 59 (Trinity MS). (CM) English Prose Treatises of Richard Rolle de Hampole. EETS 20. (RR) Dan Michel's Ayenbite oflnwyt. EETS 23. (A) The Book of Vices and Virtues. EETS 217.

(Bk.ofW) The English Works of Wyclif. EETS 74. (Wy.) The Cloud of Unknowing and The Book of Privy Counselling. EETS 218. (Cloud) Gower, Confessio Amantis. Ed. G.C. Macanlay, 1901. (Gower, CA)

17

15c. Mandeville's Travels. Ed. M.C. Seymour, 1967. (Mand.) Three Middle-English VersionsoftheRuleofSt. Benet. EETS 120. (N.Ben.) Merlin. EETS 10,112; 21,36. (Merlin I, II) The Book of Margery Kempe. EETS 212. (MK) Malory. Works, Ed. E. Vinaver, (1947) 19482 (3 vols.) (Mai.) Caxton's Mirrour of the World. EETS (Extra Series) 110. (Cax.) The Cely Letters, 1472-1488. EETS 273. (Cely) ThePaston Letters, 1422-1509. Ed. J. Gairdner 1872-5 (3 vols.) (Supplement 1901) (PL I-III) 16c. John Fisher, The English Works. EETS (Extra Series) 27. (Fisher) St. Thomas More: A syntax of the English language of St. Thomas More. Part I. F.Th. Visser, 1946. (More) The New Testament octapla. Ed. L.A. Weigle, 1962. (NT) Spenser's Faerie Queene. Ed. J.C. Smith (Poetical Works of Edmund Spenser, vol. I, II), 1909. (Spenser) The Bibliotheca Historica ofDiodorus translated by John Skelton. EETS 233, 239. (Sk.) 17c. William Warner's Syrinx. Ed. W.A. Bacon, 1950. (Warner) The Lyfe of Syr Thomas More. EETS 222. (Life of M.) Concordances Bessinger, J.B., A concordance to Beowulf. New York: Cornell University Press, 1969. (Beow.) Bessinger, J.B., A concordance to the Anglo-Saxon Poetic Records.1 New York: Cornell University Press, 1978. Tatlock, J.S.P. and A.G. Kennedy, A concordance to the complete works of Geoffrey Chaucer. Gloucester, Mass.: Peter Smith, 1963. (Ch.)2 1 Titles of poems are given in full, followed by ASPR volume and line. 2 The abbreviations are those used in the concordance. 18

Gburek, H., Der Wortschatz des Robert Mannyng of Brunne in Handlyng Synne. Diss. Erlangen-N rnberg, 1977. (HS) Bitterling, K., Der Wortschatz von Harbours 'Bruce'. Diss. Berlin 1970. (Bruce) Preston, Μ.Ι,,Λ concordance to the Middle English shorter poem. Leeds, 1975 (2 parts). Kottler, B. and A.M. Markman,^! concordance to five Middle English poems. University of Pittsburgh Press, 1966. Donow, H.S., A concordance to the sonnet sequences of Daniel, Drayton, Shakespeare, Sidney and Spenser. London and Amsterdam: Southern Illinois University Press, 1969. Crawford, Ch., A concordance to the -works of Thomas Kyd. Louvain, 1906—10 (Kraus reprint 1963). (Kyd)1 Crawford, Ch., The Marlowe concordance. Louvain, 1911-32 (Kraus reprint 1963). (Marlowe)1 Spevack, R., A complete and systematic concordance to the works of Shakespeare. Hildesheim: Olms, 1968-75. (Sh.)1 Let us now turn to our OE material, on the basis of which the descriptive framework will be considerably elaborated and modified. In part one (chapters I and II) we will try to establish the main properties of OE subjectless grammar in terms of syntax, grammatical relations and the semantic role features of the animate-NP. Part two (chapters III and IV) will then document the historical development from the 12th to the 17th century (and beyond where possible) and deal with the methodological problems arising under the goals outlined there.

1 The abbreviations are those used in the concordance.

19

CHAPTER I SUBJECTLESS CONSTRUCTIONS WITH SENTENTIAL COMPLEMENTS It seems best for our purpose to divide the data into two classes, according to the sentential or nominal form of the complement. Chapter I will deal with the former type of construction, chapter II with the latter. The aim of this first chapter is twofold: 1) to characterize the syntax of subjectless constructions with sentential complementation in terms of word order and constituent structure; 2), building on (1), to determine the grammatical relation between the verbs and their associated noun phrases in terms of syntactic functions (as subject, indirect object, pseudo-subject); 3) to try to assess the semantic role (as recipient, goal, experiencer, agent, etc.) played by the animate-NP. As shown in the introduction, combined information from these three levels of structure is an essential condition for a comprehensive characterization of the OE system on the basis of which ME development will be described. The construction with sentential complements lends itself well for the investigation of 'subjectless grammar' as outlined above (see 1—3), because it is highly representative in terms of occurrence and shows significant syntactic patterning. As this construction, which involves only one noun phrase - the animate (regularly human) NP, henceforth NPa - occurs with great regularity in one syntactic form, we will first deal with the distributional aspects (in terms of norm and variation) and then turn to grammatical relations and the role properties associated with the NPa as they appear within this particular syntactic framework. On the basis of the analysis of the construction with sentential complements we will then turn, in chapter II, to a number of other OE syntactic constructions^ all with nominal complementation, into which the present class of verbs enters and which crucially contribute to a correct account both of OE subjectless grammar and subsequent development in ME. The following OE paradigm illustrates the type with sentential complementation, in which the complement can occur in clausal or infinitival form:

1.

The OE paradigm

RUE (8) fjaet ... me on minwn hyge hreowefj fraet hie heofonrice agan (Genesis ASPR I, 426) (9) ... fregnas \>aem ofyyncefr lit hie synd ludea folces (BH 175, 18) (10) ne sceamodepe to ceorfanne paet paetpu sylfsuce (Lives 202, 125) (11) Him eglde pact he waes betra fronne he (CP 235, 8)

21

PLEASE / DESIRE (12) pa licade hire... pact heo wolde pa baan up adon (HE 292, 5) (13) Mine lyste eac geseon hu ... (B. 39, 21) (14) Haeleb langode... hwonne hie of nearwe s taeppan mosten (Genesis, ASPR I, 1431) (15) Us gelustfullap gyt furpur to sprecenne (Aelf.Hom.I, 360, 29) BEHOVE (16) nu him behofed paet he crape in his mycele codde (Chr. 262, 5) (17) ac hym gebyrab to be standenne pa men (H 132, 23) (18) Biscepe gedafenab paet he sie taelleas (CP 53,9) (19) baem weorce nanum men aer negerise bet to fandianne bonne... (Or. 54, 30) HAPPEN (20) baem forhaebbendum hwilum gebyrede paet hie ... (CP 316, 24) (21) him gewearb pa eallum paet man funde niwe swurd (Lives 278, 233) (22) pa gelamp him semninga midgife... gehaeledne beon (HE 382, 11) SEEM1 (23) Nu bincb me... paet... (GEN 76,6) Formally these constructions are characterized by a lack of a nominative subject noun phrase. Instead there is a noun or pronoun in the dative case2. The sentences further show the verb constantly in the third person singular, i.e. there is no number agreement with the oblique NPa in the plural.3 The broadest generalization with regard to the semantic relation between the verb and the NPa is that the latter is 'affected' in some way by what is predicated in the complement, i.e. remains inherently non-agentive. As mentioned in the introduction, however, within this overall framework the semantic structure of the predicates involved here allows them to stand in different kinds of relation to the NPa, a fundamental fact of grammatical variation whose consequences on syntax and semantics we will try to determine. This latter issue relates to the important aspect of'potential' semantic structure and actual syntactic behaviour, a crucial factor in accounting for the development of subjectless constructions. If a verb can enter intransitive constructions with the NPa in nominative case (he hreoweb paet... 'he rues that...') or (quasi-)transitive ones with nominal complements (he hreoweb his synna, 'he rues his sins'), the question arises as to what this means for its proper semantic structure as well as for its capacity to take pseudo-subjects of the

1 Evidence from fryncan is only included with the verb used as a simple predicate, i.e. not in the type me fryncf) god paet... fryncan will thus only be dealt with in construction with sentential complements. 2 sceamian, lystan and langian also take the NPa in the accusative. 3 Cf. examples (9,15, 19).

22

kind met in (1)—(23) above. If the grammatical features of the animate-NP are a function of general syntactic possibilities and the specific use of a small set of verbs, our analysis has to be made responsive to this situation.

2.

The basic form of the construction

Our first task is now to establish the form of the assumed common syntactic structure of (8)-(23). The present verb classes occur most regularly in this type, whose syntactic pattern can thus serve as the general descriptive framework for the discussion of extrasyntactic aspects as well. It seems best, then, to establish the generalized unmarked form of the OE type represented by (8)-(23) and, with reference to these terms, to discuss the relational1 and semantic aspects. Moreover, evidence from such a uniform syntactic structure will allow us to assess the status of this construction type within general OE syntax. Our concept of basic structure is thus not only a systematic representation of the formal aspect of (8)—(23) above, but of the whole class of sentences which can occur in this form.

2.1

Word order

In establishing the basic syntactic form of the construction with sentential complements the first step is concerned with the linear order of the main elements, in the present case with generalizations about the positional properties of the verb, the NPa and the complement. The most regular aspect of main and subordinate clause order is the final position of the complement. With regard to this element the evidence of (8)-(23) can be generalized and subsumed under two versions of the same constraint on the position of sentential complements in this type. With clausal complements the restriction holds absolutely: structures of the form pact... gedafenap him do not occur in our texts.2 The reason for this appears to be rhythmical and thematic: the 'weight' of the complement as well as its regular nontopical function. The same situation holds with infinitival complements, although preverbal position in OE does not appear to be constrained in the same way; it is only met, however, in close translation from Latin, such as in to sittanne on mine swyfjran healfe...

1 Cf. footnote 4, p. 11. 2 In view of the great number of examples showing rAef-complements, it is unlikely that OE knew this variant at all. Visser notes that the clausal complement 'is never placed before the verb' (1963, 25). 23

nys me inc to syllanne1. The NPa regularly occupies first (or, in the presence of other initial elements, preverbal) position. This corresponds to the general - tough by no means absolute — SVO order holding for sentences with clausal and infinitival complements in OE 'personal' syntax.2 With regard to the verb position two facts are to be noted: firstly, it always precedes its complement. Its occurrence in initial position, however, seems to be highly constrained, again by general factors not limited to subjectless syntax. In schematic categorial terms, then, the generalized unmarked order of noun phrase, verb and complement in the subjectless type with sentential complementation is NPa-V-S.3 This surface pattern is not given any definite status as an abstract structure, but represents the norm of occurrence, irrespective of any external factors affecting the order of elements. (24)-(29) represent linear NPa-V-S order in the sentence type with no introducer element:4 (24) (25) (26) (27) (28) (29)

Us sceamap to secgenne ealla pa sceandlican wiglunga (Aelf.Hom. I, 370,100) hie forscamige pact hie eft sua don (CP 151,17) hi ne lyst spyrian... (B. 121, 20) him gedafenap paet he gepence... (CP 75, 4) him gebyrede paet he nyste selfhwaeper he monn waes (CP 41,1) Mepincp paet... (Lives 126,163)

It is a general fact about OE word order that a subset of predominantly initial conjunctions (CONJ) and adverbs (ADV) can occur freely in different positions in the sentence, governed, it is assumed, by general rhythmical factors.5 It is to be noted, however, that also in this situation, as illustrated by (30)—(33), the unmarked order ... NPa-V ... is preserved: (30) (31) (32) (33)

ac aelcne man lyst... paet... (SOL 48, 5) soplice paem lareowegedafenap paet... (BR21,21) ac him gedafenap paet hi hyra bisceope beon underpeodde (H 112, 1) oft swipe manegum men faerlice gelimpep pt he hine... gedaelep (BH 125,10)

1 Quoted by B. Mitchell, A guide to Old English (1975, 112), without reference. Note that with ίΛαί-clauses the absolute restriction on initial position also holds in construction without the NPa (f)a gelamp paet...). It also appears to hold for infinitive complements in non-translated prose. Apart from a whole section in Leechdoms (Cockayne, III, 199ff.) of the type 'In dreams to catch fowls betokens profit' (p. 199), initial position of the infinitive is extremely rare, and typically found in interlinear glosses following the Latin original, such as the following examples quoted by Visser (II, 948f.): feower kynna beon sutol is, lufigean his nehstan swa hine sylfne paet is mare... 2 Deviations from this rule will be discussed below, in the context of preverbal position. 3 S is the symbol for sentential clause, whether in infinitival or clausal form. 4 Such as conjunctions, adverbs or sentence connectives. 5 The most complete study on this aspect is still Bacquet (1962). 24

Possible deviation from this order is significantly restricted. There is a small class of CONJ/ADV, including pa and ponne, which does occur with systematic inversion of verb andNPa: (34) (35) (36) (37) (38)

ponne hreowep hire paet heo hire gehat ne gefylde (BT) sippan gelicade eallum folcum paet hie ... (Or. 106, 22) pa licade hire..., paet heo wolde... (HE 292, 5) pa gelamp him paet his lif wearp geendod (BH 113,7) pa gebyrede hire paet heo hie forlaeg (Or. 162, 32)

The facts are clearest with pa, where the word order variants pa-NPa- V ... and pa-V-NPa ... are fully functional in distinguishing main and subordinate clause order. Exactly the same situation holds in constructions with nominative subjects.1 This observation, then, cannot be used to as an argument for special positional behaviour of the oblique NPa in the syntactic type discussed here. On the contrary, the result is again that the NPa behaves in exactly the same way as the nominative NP in parallel sentences. Another possibility of verb-first order (again seldom met with) arises in co-ordinated structures: (39)

... and Heap him paet hie paet unaliefede dop unaliefedlice (CP 145, 10)

Absolute initial position of the verb does occur, but is extremely rare: (40) (41)

gelomp him in geoguphade his paet he... gefremede (HE 350, 12) Hreaw hine swipe paet... (Genesis, ASPR I, 1276)

As far as the NPa is concerned, the postverbal position in (40) is perfectly in line with an interpretation as indirect object, as comparison with syntactically unmarked paraphrases shows ('it happened to him that...'). However, postverbal position of the NPa is not distinctive for this construction type; the same deviation from regular word order is possible with nominative subjects. Postverbal position of the NPa is thus not a crucial indirect object property (cf. the behaviour of pronominal NPa, which tends to occur preverbally, below). Grammatical relations cannot be said to determine syntactic order in any sense, and the fact that except in the presence of some introducers postverbal position of the NPa is very rare clearly speaks against the assumption of any tendency towards a correspondence between the syntactic and functional levels of sentences in OE. It should be noted, however, that if the initial position is occupied by a place-holder element like it, a functional indirect object interpretation naturally conforms to the syntactic form. In any type of affirmative structure in OE the scarcity of verb-first order is the result of a general syntactic tendency which demands that the verb be in second place in unmarked 1 Cf. Andrews (1940, chapter II).

25

structured order of main declarative clauses. Although its general force must be equally assumed in the present context, this word order constraint is weaker with subjectless predicates than with other verb classes, a fact possibly due to the lack of a syntactic subject in the construction type dealt with here. Although this first sketch will be modified in detail, it represents the typical word order variants of type NPa-V-S. Thus there seems to be a strong tendency for the NPa to occur in initial position, although it is always marked for object case and does not govern verb agreement, i.e. has none of the typical subject features. We have seen that the main word order variant is of the fomX-V-NPa .... where X is an element of the class of CONJ/ADV. Here the postverbal occurrence of the NPa is not due to its inflection or possible nonsubject reading, but follows from systematic order variation in main and subordinate clauses, i.e. from a general aspect of OE syntax. Our next task is to establish the constituent form of the schematic linear pattern NPa- V-S.

2.2

Constituent structure

The main constituents of the unmarked type of the OE sentences met so far are noun phrase (NP) and verb phrase (VP), and the constituents of VP verb and complement. Under such a binary analysis the internal structure of the syntactic type is, in hierarchical categorial terms (represented in tree form), as follows:1

VP

Basic structure of (8)-(41) This is a direct translation of the regular schematic form NPa-V-S into constituent structure. Although there is obvious syntactic motivation for this basic structure (viz. word order), it is only of limited value as a characterization of the present range of sentences, because it lacks any link to morphology, which is what formally marks the type through case inflection on the NPa and absence of verb agreement. Neither of these aspects has been associated with our concept of basic structure. Thus, whereas word order is determined by this abstract form, morphology is not. As it stands, there are no statements concerning the subcategorization of verbs or a derivational concept of sentence generation which would systematically link the criteria of case

1 The symbol C stands for 'clause'. 26

and agreement to the abstract syntactic form 1 . The inadequacy of characterizing NPa- V-S in terms of syntactic structure alone appears clearly if we compare the basic form of (8)—(41) with the corresponding abstract structure of straightforward personal constructions such as (42)—(45): (42) (43) (44) (45)

Gifpu fronne ne gelyfest fraet he sy Codes Sunu (BH 151,31) Ac fra Cirusgeahsade paet hiene se... cyningpaer secan wolde (Or. 76, 10) forpaem he tiohchode him ma to fultemanne (CP 305,4) paet hie hit acwencean pohton (Or. 200, 16)

The linear form of the syntactic type represented by these sentences is ahoNP-V-S2, and the uncontroversial constituent structure is identical with that established for the subjectless type. (The position of the NP corresponds to that in (8)-(41) and the internal structure of the VP is the same.·) Such structural parallelism is evidently no basis on which to distinguish subjectless from personal construction. It appears that a systematic distinction between these two types has to refer to the nature of the grammatical relation between verb and NP, in particular to determine the nature of the pseudo-subject function of the NPa. The relation in question is determined by the semantic structure of predicates and goes together with a specific semantic role of the NPa. In the following we will turn to the relational aspect.

3.

Grammatical relations and semantic roles

3.1

The syntactic function of the animate-NP

In the type of sentences represented by (8)—(41) — henceforth type S - the grammatical relation which interests us most is that between verb and NPa3 . With regard to the NPa it was found in 2.2 that if functions are syntactically defined, the functional status of the NPa in type S cannot be meaningfully distinguished from that in unmarked personal construction.

The respective syntactic weight of inflection and word order in OE is not easy to assess. In more recent work, word order is given priority, but the exact nature of relative interdependence between the two factors has not yet been established. Cf. Gardner (1971). Although the NP is also [+animate], the symbol NPa is not used because it is defined to characterize the specific grammatical status of the NP associated with syntactically subjectless predicates. With regard to the description of grammatical relations there was much optimism caused by the advent of relational grammar, a branch of generative grammar. See Johnson (1976), Cole & Sadock (1977). To judge from published (and commonly available) work at least, relational grammar has not succeeded in illuminating the issue in a way which would suggest it as a descriptive framework for the present purpose.

27

The present notion of basic structure directly reflects the functional parallelism between the two construction types in terms of surface syntax. Insofar as the pseudo-subject nature of the NPa in type S is concerned, however, a proper characterization cannot be achieved within the descriptive framework of basic structure established mainly on the basis of word order generalizations. With regard to a characterization of functions in terms of surface syntax, there are a number of operational tests which can be applied to our material, but an illustration of this issue will have to wait until other syntactic types involving the verb-NPa relation have been discussed (see section 9.2). With regard to a characterization of subjectless syntax in terms of basic structure, however, the syntactic behaviour of subjectless verbs suggests that we include another construction in our analysis, viz. the type with it in subject position (it ofyyncep him paet...). If our description is to capture the relatedness between the two constructions, the basic structure mentioned above (p. 26) has to be revised. Now the main reason for setting up a uniform basic structure is the constant semantic interpretation of the paraphrase types me rues that... / it rues me that... In the following we want to represent this aspect on the basis of semantic NP-roles.

3.2

Semantic role properties of the animate-NP

The semantic roles taken by the NPa associated with subjectless predicates decisively contribute to a correct account of the complex nature of the grammatical relation between these two elements. And although NP roles are obviously determined by the semantic structure of predicates, there seem to be no operational methods linking, for example, elements of a componential semantic analysis with NP roles in a systematic way. We know the agentive, recipient or instrumental character of an NP without always being able to relate it to some single semantic verb feature or to present valid syntactic paraphrases. This kind of semantic knowledge has so far been given its most explicit expression in the work of Chafe1 and within the framework of case grammar, where the verbs (as lexical entries) are associated with so-called case frames representing the semanticsyntactic valency in terms of the number and kind of semantic roles. Note that apart from the well-determined concept of agenthood there has been a notorious lack of relevant — not to mention crucial — criteria defining semantic roles. Although case grammar at the moment appears to be the most adequate framework to deal with the kind of grammatical information described by semantic roles, it cannot serve as an acceptable descriptive basis in the case of OE, and this for three reasons: first, its theoretical outlook is that of generative grammar, which presupposes native knowledge of the language analysed. Second, and more directly relevant, there is the practical problem (for OE) of adducing paradigmatic semantic and syntactic paraphrase evidence, an important type of evidence for justifying deep cases or semantic roles. Third, an issue until now unresolved by case grammar concerns the link between the semantic level of cases and the level of surface 1 See in particular Chafe (1970). 28

syntax, a crucial lack in view of the present task. We cannot, ultimately, satisfy another demand arising for work within the case model, viz. devise new tests to demonstrate the specific verb-NPa relation assumed to hold for the present set of subjectless predicates. For the purpose of characterizing the role properties of the NPa occurring with subjectless predicates, we shall adopt a practical approach. If we infer, on the basis of the OE semantic verb classes I-V, the general semantic features 'rue', 'please', 'behove', 'be fitting', 'seem' and 'happen', we achieve a semantic framework not subject to change in time. The distinctive NP-roles associated with these meanings are those of recipient and experiencer. If a verb is associated with these NP- roles, it will be used intransitively. A description of the present set of verbs in these terms thus forms the basis for an account of their semantic development. As case grammar has shown, there is no strict correspondence between semantic roles and syntactic functions which could be brought to bear on our description. Despite this, there is a non-accidental correlation between roles like recipient and experiencer and their syntactic expression: they will typically appear as indirect objects, often marked by prepositions. Syntactic paraphrase evidence can throw light on the grammatical indeterminacy of the NPa in type S. If in the it -construction mentioned above (p. 14) the pseudo-subject occurs in indirect object function, there is still no problem with the semantic interpretation. But when the same predicate also enters the third potential variant type, with the NPa as a pure nominative subject (Ί rue, desire, behove'), the question then arises as to how this construction is to be distinguished from type S and whether semantic variation is involved. We will only return to the form and descriptive value of the basic structure in the light of primary OE evidence on these matters. Within this sketchy framework of grammatical relations and syntactic realization possibilities, then, let us now turn to our OE material, dealing with the different semantic subclasses in turn. We will proceed in the following way: first, on the basis of the core meanings, the potential set of possible semantic NP-roles will be assessed. Then the syntactic occurrence properties of individual predicates and their respective productivity in the syntactic variation types offered by OE grammar will be illustrated. In this way we will achieve a comprehensive picture of the lexical valency of the verb classes I-V in terms of their role structure and syntactic occurrence paradigm, both conditions for the final evaluation of the grammatical properties of subjectless verbs in OE.

4. The OE RUE class BT gives the following meanings for the verbs subsumed here under the RUE class: hreowan 'to rue, make sorry, grieve';gehreowan 'to rue, repent, grieve, pity';ofhreowan 'to cause grief or pity; to feel pity'; sceamian 'to feel shame, be ashamed; to cause shame'; 29

forscamian 'to be greatly ashamed; to make greatly ashamed'; ofscamian 'to put to shame'; ascamian 'to be ashamed or make ashamed'; gescamian 'to blush, be ashamed,...; to shame, cause or bring shame to, confound'; eglian 'to trouble, pain, gneve';geeglan 'to trouble, injure';off>yncan 'to cause regret, sorrow; cause displeasure'. The semantic element 'rue' denotes the cause of sorrow as well as the experience of regretting, etc. Different syntactic variants correspondingly reflect that the focus is either on the affected person or the (usually inanimate) cause. This distinction is properly stated in terms of the thematic structure of sentences and does not influence the internal semantic structure of predicates. Here we are not concerned with possible stylistic nuances due to the predominant thematic value of either the [+animate] or [—animate] argument, but with the effect of such use on the grammatical status (in terms of syntactic functions and semantic roles) of the NPa. It appears from the OE sentences above that in the unmarked situation these intransitive verbs take the animate-NP as their pseudo-subject (him rues that...). If on the other hand the [—animate] NP is made the subject of the sentence, 'rue' is used quasi-transitively; the situation is imagined as an event passing from the [—animate] source to the person. However, this type of sentence (his sins rue him) is not grammatically transitive in the strict sense; (there is no accusative NP capable of becoming the subject of a corresponding passive sentence). Similarly, if the person is made the syntactic subject (he rues his sins), the [—animate] object is not transitively affected, but remains the source of sorrow etc. On this level of logical semantics, then, the grammatical relations between the predicate and its two arguments remain constant. Our description of NPa roles and their syntactic realization has to take into account this potential converse way of looking at the semantic situation, paraphrasable by 'NPa rues χ / χ rues NPa'. In particular, we will have to ask in what way the potential three-way variation offered by the syntax affects the interpretation of semantic NPa-roles. Especially within the context of normative development towards the general SVO pattern, the possibility of a quasi-transitive reading of subjectless predicates will have to be kept in mind. In relation to this fundamental semantic framework, then, we will discuss the nature of semantic NP roles occurring with RUE verbs (section 4.1) and the attested syntactic occurrence paradigm (4.2).

4.1

Semantic NP roles associated with R HE verbs

The [—animate] argument of RUE verbs (occurring syntactically in the form of the clausal or nominal complement) would be assigned the label 'neutral' or Objective' by different models of grammars based on deep case relationships.1 The NPa on the other hand takes the roles of 'experiencer' (the animate being affected by sorrow) and 'recipi1 Objective' by FiUmore (1968), 'neutral' by Stockwell et al. (1973). 30

ent' (the animate receiver)1, i.e. generally 'the animate being passively implicated by the happening or state'.2 These roles are inherently associated with the semantic arguments of RUE verbs and are thus not subject to change. The recipient role properties of RUE verbs are found to be correct by speakers of modern English and German3, and they can be supported by grammatical paraphrase evidence representing the semantic and relational structure in a more explicit way, i.e. lexically expressed to a higher degree. Thus the paraphrase 'what I experience is ... (sorrow, grief)', for example, shows the experiencer quality of the NPa with the present verb class. However, speakers of modern English (and German) remain notoriously uncertain if asked about a possible agentive interpretation of the NPa role. This uncertainty seems only to reflect the fact that — on a surface level — this pseudo-subject can be imagined (hypostasized) as taking a quasi-agentive role in the event of regretting etc. Unless the levels of logical-semantic relations and the interpretation of surface syntax are strictly distinguished, grammatical tests will remain inconclusive, as they only mirror this indeterminacy. We should be careful, however, not to evaluate OE syntactic and semantic evidence on NP roles solely from a modern English point of view, mainly because the possibilities offered by the syntactic paradigm are crucially different. While the fundamental possibility of quasi-agenthood of the NPa (not necessarily substituted for, but rather in addition to its recipient nature) must be equally assumed for OE, there is an important difference: the functional and semantic interpretation of OE syntactic structure is not yet governed by the transitive SVO pattern of modern English and its associated generalization of the one-to-one relation between initial position and subjecthood on the one hand and between subjects and agenthood on the other. The question of quasi-transitive use of RUE predicates has to be discussed in view of the complete syntactic paradigm entered by OE subjectless verbs and relative to their statistical occurrence in different constructions.4 Let us now turn to the syntactic behaviour of these verbs.

1 2 3 4

'Dative' in Fillmore's (1968) terms, later changed to 'experiencer'. Quirk et al. (1972, 350). Modern RUE verbs are commonly analysed as being associated with dative/experiencer. That the situation is of gradual rather than clear-cut nature is indicated by facts from modern German. Whereas - contrary to English - the intransitive/quasi-transitive distinction has been lexicalized in the pair reuen/bereuen, the latter (agentive) member is still restricted to occur with most agentive adverbs. In the case of schämen, however, the notion of agenthood in the NPa is absent. Such evidence should not be disregarded in evaluating the respective role properties associated with OE hreowan and sceamian. 31

4.2

Syntactic occurrence properties of R UE verbs

Whereas semantic evidence in historical linguistics must needs remain indirect, the syntactic paradigm is a direct source of information. We have mentioned the potential three-way variation (p. 14) offered by the syntax to express the predicate-argument relation with RUE verbs. The present section first observes the distributional facts concerning actual use of syntactic forms. In cases of syntactic variation we then ask whether the formal variation pattern also reflects semantic differences and in what terms they are adequately described. In the construction type with sentential complements we meet a highly structured situation. With RUE verbs OE almost exclusively uses type S constructions, such as illustrated by (46)-(48): (46) (47) (48)

hie forscamige paet hie eft sua don (CP 151,17) Me ofhreow fjaet hi ne cupon pa godspellican lare (Aelf.Hom. I, 2, 22) pa ofpuhte him pact... (Or. 116,14)

All verbs are attested in this type, the more generally used ones in great number. For those predicates not recorded in type S we have to decide in each case whether the absence is due to semantic reasons or whether it is accidental, i.e. possibly due to the small number of overall occurrences. In terms of use, however, the situation is quite clear: type S was the indigenous syntactic expression of RUE verbs in construction with sentential complements. Now contrary to the productivity of type S, the variant with a nominative subject (he hreowep paet...) is extremely rare with this verb class. In the material at our disposal the personal type is only attested with sceamian1: (49) (50)

Gifwe ponne scomiap paet we to uncupum monnum suelc sprecen (CP 63,5) pu ne pearft sceamian ponne... (Soul & Body I, ASPRII, 145).

Hreowan does occur with nominative subjects, but only in construction without sentential complementation.2 Examples like (49) and (50) - and evidence from the type with nominal complements in chapter II - make it clear that the occurrence of nominative NPa with RUE verbs was not basically constrained. Thus sorgian 'to grieve', hreowsian 'to grieve, repent', maenan 'to moan' and sarettan 'to lament', which all belong to the semantic field RUE, are only found construed in this way: He ... hreowsade paet... (CP 199, 17), he sarette paette ... (CP 153, 9). The point to be noted is that OE RUE verbs occur typically without a nominative NPa. 1 The variant forsceamian regularly occurs with nominative NPa, but in a different construction type, with nominal complements (see chapter II). Hreowsian (not a subjectless verb) likewise is found with nominative subjects only. Both verbs, however, are not met with frequently and in our material do not allow any valid generalizations. 2 See (52) below. 32

The occurrence of the sentences above, however isolated, proves that OE possessed the paraphrase pair (51): (51) a b

us sceamaf) baet... we sceamal) baet...

Is there semantic variation involved between the reading of (51) a and bl Certainly not in the sense that there are two verbs sceamian with distinct semantic structure. Further, to state that the sentences differ in the respective intransitive or transitive use of the verb remains insufficient, because there is no genuine transitivity involved and such a distinction in semantic terms remains elusive.1 There is more OE evidence for the occurrence of the present verb class with nominative noun phrases from a construction type without complementation. Sentences like hreaw him (BT) and hie ne magon ... hreowan (BT), me sceamaf) (BT) and Na ic ne scamode (BT) suggest the presence in OE grammar of variation patterns of the kind (52)o b (53) b

me hreoweb ic hreowe me sceameb ic sceamige

Again there are no reasons to assume that (52) and (53) reflect proper semantic variation. However, we cannot exclude the possibility that these patterns express a change of semantic roles on the part of the animate-NP. In view of the functional status of the OE inflectional system, the variation in (51)—(53) can hardly be morphological only. However, the question as to whether it is the result of a desire to give expression to the distinction between a purely recipient and a quasi-agentive view of the animate-NP has to remain unsettled until there is evidence suggesting unambiguous transitive use. It is certainly the case, on the other hand, that variation pairs like (51)-(53) represent an enriched stylistic pattern in late OE, a new possibility in the syntactic paradigm of these predicates. As a phenomenon in the history of English, the greatly restricted occurrence of nominative subjects with RUE verbs seems to reflect merely the first trace of a 'personalization' tendency gaining force all through the ME period and resulting in the canonical use of nominative subjects in initial position, associated with the establishment of rigid SVO order.

1 Modern German evidence cannot decide the issue, because the intransitive/quasi-transitive distinction has been lexicalized (mich reut, dass ... / ich bereue, dass...) or 'grammaticalized' as in (mich schämt, dass ... / ich schäme mich, dass ...). In the latter case the distinction involved cannot even be described in terms of semantic roles, as a pseudo-agentive interpretation of the personal variant is not possible. German grammars notoriously lack information on this issue, except for the impressionistic statements by Brinkmann (1971), Erben (1972) and the Duden-Grammatik. 33

Returning to the occurrence properties of RUE verbs, we find that comparative evidence from modern German suggests that we look for another variant structure of the type represented by es reut mich, dass..., a perfect semantic paraphrase of mich reut, dass... As it stands there are no sentences of the putative form it hreoweb him pact... in our material. The earliest example is from the \2c.Apoloniusof Tyre: Neofpingb hit begif... (29). Note, however, the variant form with the demonstrative pronoun pact instead of hit, as in (54)

ba ofbuhte baet Mariuse... baet...1 (Or. 236, 3)

That the syntactic form it- V-NP-S is available already in OE appears clearly on the basis of evidence from complex predicates (see (55) and (56) below). For reasons of syntactic structure, then, OE RUE verbs cannot be excluded from entering this construction type; from a semantic viewpoint RUE verbs even invite the associated animate-NP to have the indirect object function and semantic recipient role it has in the construction with it in initial position. In (55) and (56) the complex RUE predicates beon sorhlic / ofbyncend represent the extremely rare iY-construction: (55) (56)

Hit is ... sorhlic... eallum mannum to gehyrenne... (Wulfstan, Polity (Jost) p. 245 § 70, quot. after Vi. 963) hit pa eallum baem senatum ofbyncendum ... baet... (Or. 244,16)

(55) is synonymous with the far more productive type S variant (57)

eallum is sorhlic to gehyrenne ...

Again it seems that the proper account of this situation is in terms of use. We have seen above that the personal variant never became productive in the syntactic paradigm of RUE verbs, because (we assumed) there was as yet no pressure from normative (SVO) syntax nor was there a semantic reason to add agenthood to the status of the NPa. Now in the /f-construction we have a syntactic variant which expressed the recipient role by means of postverbal position (prepositional marking is never found). Like the personal type, however, the it -variant was clearly not productive in OE. Had there been a desire or necessity to express either the recipient role of the NPa more clearly (by syntactic means in addition to morphology) or to give syntactic form to an ongoing re-interpretation of the semantic NPa role towards agenthood, the syntax would have offered the means for it. On the basis of evidence from OE RUE verbs at least the conclusion is that none of these putative conditions applied, and that type S came closest to expressing the double nature of the NPa resulting from the discrepancy between the logical-semantic interpretation and particular language-dependent syntactic form. This conclusion receives support from the facts holding with complex predicates of the type be adjective / noun. Complex predi1 This is just as rare as the it-type. Statements on the //-construction will include the {fact variant, which will also be used for illustration.

34

cates express the whole range of detailed meanings within the semantic framework RUE. Beon wa 'woe', sorh 'care', ege 'fear', ... are all very productive in type S. They are never found with nominative NPa, however (ic earn wa pact...) and only sporadically in the itconstruction. In the light of both the fundamental semantic situation and evidence from the attested OE syntactic paradigm of RUE verbs it is clear that neither the personal nor the it-construction can be excluded as a variation possibility. Neither paraphrase can be shown to have an effect on the semantic interpretation, as an agentive reading of the NPa on this level of structure is impossible. Both variants are syntactically unmarked. In functional terms, however, the if-construction renders the NPa as a pure indirect object, opposed to its pseudo-subject and subject status in S and the personal type respectively. The situation of a potential three-way variation, then, is a central syntactic aspect of OE subjectless grammar, and subsequent development will have to be described with reference to this fact. The assignment of the semantic roles recipient and experiencer to the NPa in type S has so far been based purely on the assumed general semantic features of RUE verbs. It is now necessary to adduce evidence for the correctness of these assumptions. For OE this can be done by means of syntactic tests and paraphrase constructions, an issue taken up in the next section.

4.3

Operational tests and paraphrase evidence for semantic roles

As already mentioned, possible justification of covert categories such as semantic roles by operational test is strictly limited in OE. Furthermore, complementary evidence from modern languages has to be evaluated carefully, because there is as yet no generally accepted set of tests to determine semantic roles. One very direct type of evidence concerns the co-occurrence of predicates with certain adverb classes showing the agentive or recipient character of the associated animate-NP. This test cannot be applied to OE because most of the adverbials used (volitionally, willingly, deliberately on the one hand, and personally on the other) do not exist as lexical items in the vocabulary of the language at this stage. Agent-denoting adverbials are never found in the present semantic context, however. This ties up with the fact that the modern English verbs regret, repent, (rue), be ashamed (as well as please, be fitting, happen and seem) all fail the test for agenthood of the associated NPa. They cannot occur with the class of adverbs represented by deliberately (*John deliberately regretted that ,../.1 This test provides negative evidence only, by showing that with these predicates the NPa cannot have agent quality in a semantic sense, however much the syntax may use transitive form to express the verb-NPa relation. (It suffers, furthermore, from the fact that the range and structure of adverbials crucially associated with the agentive status of the sub1 The asterisk means that the sentence is ungrammatical. It will also be used with examples of ungrammatical sentences in OE.

35

ject is by no means clear.) The adverb test is not suited to demonstrate the recipient nature of the NPa, i.e. to characterize this role positively. The most relevant test available seems to be the one suggested by Postal, who noted that experiencer-NPs (associated with what he calls 'psychological' predicates) alone can take the adverb personally1:1 personally am annoyed with Jack, but *Jack is annoyed with me personally. RUE verbs qualify as experiencer-taking predicates under this test (I personally regret that...). With regard to our OE data, however, the adverb test is not applicable. Another, less direct operational criterion takes the form of relationships between two syntactic constructions entered by the same verb, of which one member explicitly marks the indirect object status of the NPa — the indigenous realization, on the level of functions, of the recipient role — by prepositions, as in / like it j it pleases to me. With modern English RUE predicates such evidence is not available for lexical reasons. Now the most frequently met paraphrase type in OE is of the form NPa be adjective/ noun, in which the NPa appears inflectionally marked by dative case (as in type S): wa bib paem (Beow. 183), waes him micel ... sorh (BH 135, 2), bio him swipe egefull... (CP 262, 15). Only in late OE do we come across sporadic occurrences of prepositional expressions such as paet he mid mannum is ... egeslic ('dreadful, horrible', Metrical Psalms, ASPR V, 98, 3) oiyrre wip me ('angry', Aelfric, quot. by Visser, I, 327).2 The generalization of prepositional marking in early ME is a posteriori justification for the assumption that indirect object function and recipient role also characterized the verb-NPa relation in OE. Modern evidence provides additional support for the experiencer/recipient nature of the NPa occurring with RUE verbs: in the transformational operation of clefting, nouns denoting sorrow, grief etc. are possible complements to the paraphrase type what NPa experiences is ... If the operational verb is do (denoting agentive subjects), on the contrary, the resulting sentences are ungrammatical: *what NPa does is experience grief, etc. It will be seen below (section 9), however that in terms of syntactic behaviour (in particular word order and occurrence in complex sentences) the NPa can be exactly parallel to 'real', agentive subjects. The semantic framework of RUE verbs is thus systematically associated with the experiencer/recipient role of the NPa. Evidence from a whole range of languages (genetically related or not), which use inflectional or other means to mark the NPa as indirect object and semantic recipient shows that the paradigm of type S syntax is not particular to OE, but represents a general aspect of grammar.3 The following section will investigate the grammatical properties of the PLEASE/DESIRE class of predicates.

1 Postal (1971, 39-54). 2 For further examples with PLEASE and BEHOVE verbs see pp. 39,42. 3 Type S constructions are found in all Germanic dialects and in such diverse languages as Russian, Kannada and Polynesian dialects. See Li (1975b). 36

5. The OE PLEASE/DESIRE class This is the semantically least uniform set of the verb classes I—V. BT gives the following meanings: lician, gelician 'to please';gelystan 'to please, cause a desire'; lystan 'to list, cause pleasure or desire'; langian 'to cause longing, desire'. Two further verbs have been added to this class, because they contribute relevant evidence with regard to semantic-syntactic patterning: the purely intransitive verb lustfullian 'to rejoice, be glad' and the optionally transitive gelustfullian 'to be delighted, pleased; rejoice, delight, please'. The range of dictionary meanings again reflects the respective focus on the receiving person or on whatever is the source of pleasure.

5.1

Semantic NP roles associated with PLEASE I DESIRE verbs

With the exception of the transitive use of gelustfullian, the NPa occurring with this verb class can only take the set of role features typically including experiencer and recipient in the relational situation paraphrasable by 'NPa receives pleasure' and 'experiences pleasure or desire' respectively. This semantic context is associated with the recipient function of the NPa as human experiencer. The nature of the semantic NPa role with the present verb class is uncontroversial, as appears from its unambiguous representation as experiencer, recipient or 'dative' in modern studies. The semantic situation is especially clear if two animate-NPs are present. Lician, for example, enters the syntactic type represented by (58): (58)

he me wel licaf} (Aelf. Horn. II, 40, 5)

In sentences of this kind — which cannot be interpreted as transitive structures in OE — the nominative NP has the neutral role discussed earlier, i.e. the role otherwise taken by the 'thing'. There can be no ambiguity as to which NP is the giver or receiver of pleasure (despite word order variation), because the role distribution is determined by the semantic structure of the verb itself. However, as with some RUE verbs, the part taken by the NPa — in particular with the subclass DESIRE - can also be imagined as a quasi-transitive one. It appears again that whereas the relational situation on a proper semantic level remains constant, the abductive nature of surface interpretation may again deviate from the logical pattern by conceiving the NPa as a pseudo-agent. Apart from the occurrence of PLEASE/DESIRE verbs in type S (see next section), the following OE paraphrase constructions with predicates from the same semantic field reflect the nature of the NPa role: first, semantic paraphrases of the type NPa be+adjective in which the NPa is marked as dative, such as waes he eallum his geferum leaf ('dear', BH 213,12), Se is Code wel liciendlic ('pleasing', BT), Heo waes swipe lufigendlic eallum onlociendum ('lovable', BT).

37

Second, paraphrases in which the NPa is prepositionally marked as indiiect object, as in leoftaele mid 'dear, desirable, pleasant, ...' liciendlic on ... 'pleasant, pleasing'.1 In these cases the functional and semantic interpretations rely on the constant distribution of roles in the presence of two animate-NPs. In OE this distribution is governed by the intransitive meaning of PLEASE/DESIRE predicates, which demands that while one NP functions as recipient, the other remains neutral. In the following discussion of the occurrence paradigm we want to investigate in what way the syntax expresses the semantic situation characterized in this section. 5.2

Syntactic occurrence properties of PLEASE I DESIRE verbs

In the construction type with sentential complements — by far the best attested — all predicates mentioned above except lustfullian2 regularly occur in type S. (59)—(61) represent this construction: (59) (60) (61)

licade us efencuman aefter peawe ... (HE 276, 12) ... cwaep paet hine lyste mid him etan & drincan (HE 398, 7) Us gelustfullap gytfurpur to specenne (Aelf.Hom. I, 360, 29)

The syntax is thus in line with the recipient role of the NPa. Within the framework of OE morphology sentences like paem cyninge licep paet... rendered the semantic situation as directly as formally possible, and in addition expressed the thematic importance of the NPa by giving it front position. The personal construction is again extremely rare. There is one single example of personal use with OE lystan: (62)

Manige men hine geornlice lystan geseon (BTs)

(62) shows that it was not impossible for a nominative NPa to occur with this verb, although such structures were clearly avoided. Similarly, there is only one example for lustfullian (the only attested occurrence of this verb with sentential complement): (63)

ic lustfullede [var. lusfulliende waes] ... para gemaenan ... brucan pe ic on paere stowe sceawade (HE 432, 30)

Lustfullian is regularly found with nominative NPa3 and only productive in construction with nominal complementation (see chapter II, section 2.3.2). 1 See Visser (I, 228-30), with further examples for prepositional construction with different semantic verb classes. 2 This verb is not frequently found in OE and only attested with nominative NPa. 3 This distributional fact, however, is not per se sufficient motivation to assume a semantically-based constraint on this syntactic form in OE, although it obviously reflects an important aspect of language use.

38

As with RUE verbs, then, the occurrence of subjectless PLEASE/DESIRE predicates with nominative NPa is not essentially impossible, but it is still a marked exception. Just as the NPa associated with the RUE verbs hreowsian and sorgian takes nominative case, lustfullian only occurs in this way. The same is true for the complex predicates beon lustfull, beon lustbaere ('to be desirous, pleasant'): (64) (65)

ic ... waes swibe... lustbaere (hine) to gehyranne (B. 50, 10) gifhis hwa sie lustfull mare to witanne... (Or. 100, 26)

On the other hand, however, there are predicates whose syntactic behaviour (in the present case non-occurrence in the personal variant) appears to be determined by other factors than usage alone. Langian, for example, is never found in construction with an unambiguous nominative NPa in OE. This situation can clearly not be accounted for in terms of a semantic distinction between langian and lystan. Rather the fact that the former predicate is never, the latter only once attested with nominative subjects at all suggests that the freedom of use could be limited by the close association of distinct semantic roles (here recipient) with a certain verb or verb class. The fact that there is no variant of the form he lysteb, he langeb corresponding to the often found hine lysteb, hine langeb must decide the interpretation of examples like Haeleb langode hwonne ... (BT) or Geat... restan lyste (Beow. 1792), where the case distinction between nominative and accusative is neutralized, in favour of a subjectless reading. The question now is whether the occurrence of syntactic variation pairs only distinguished by the case marking on the NPa also reflects a semantic difference. Again the answer is the same as in the context of RUE verbs: apart from a possible surface interpretation of the NPa as a pseudo-agent, the distinction in pairs like (66) and (67) cannot be stated semantically: (66) a ic licige (BT) b me liceb (BT) (67) a ic gelustfullode (BTs) b me gelustfullode (BTs) As far as the grammatical status of type S is concerned, the fact that none of the PLEASE/ DESIRE verbs occurs in the it -construction is a further indication of its exclusive use. It appears that sentences like it liceb me baet... did not occur in OE, although, as observed already the recipient role of the NPa would thus be adequately expressed.1 The reason for this restricted syntactic paradigm is thus again language use and can therefore not be given more than plausibility support. Type S was preferred over its potential variants because it allowed the most satisfactory expression of the NPa double nature and conformed to the personalization tendency of having initial animate NPs.

In view of the frequent occurrence of liken in our texts (and in the concordances) such an extrapolation seems not unwarranted. Note also that Visser's earliest example for this construction dates from the 13c. (1963,44).

39

In construction with sentential complements lician, langian and gelustfullian exclusively occur in type S. Within this paradigm there can be no question of semantic variation in the use of this verb class. Lystan and lustfullian are attested in the personal variant, but only once each. The isolated occurrence of lystan can hardly be the result of a desire to express semantic variation (i.e. transitive use in contrast to regular intransitive use in type S). Whereas lustfullian never takes dative NPa, langian only occurs in this way (all complement types included). In both cases a systematic semantic constraint could be involved, but as the example of lystan shows, the attested OE situation only allows safe conclusions with regard to use, not abstract semantic structure. The further course taken by ME development will throw additional light on this issue. As in the previous section, then, OE evidence indicates that systematic variation with regard to the NPa was not impossible, but that in the rare examples with nominative NPa there is not the slightest trace of a semantic variation involved. With respect to the semantics of OE PLEASE/DESIRE predicates this means that in no context are these verbs used quasi-transitively. Like the RUE predicates, then, PLEASE/DESIRE verbs show an inherent ('deep semantic') recipient relation between the human experiencer and the verb, regularly expressed by the syntax through type S. However, with RUE as well as PLEASE/DESIRE verbs, the NPa can also occur in the nominative form of 'real' (syntactic) subjects in intransitive construction. Although in terms of use the distributional situation is highly significant — type S being the productive variant — a characterization of OE subjectless grammar has to point out that the potential syntactic occurrence paradigm can give rise to the kind of abductive re-interpretation of the original NPa-role discussed.

6.

The OE BEHOVE class

In BT we find the following meanings for the members of this verb class: gebyrian1 'it pertains to, it is fitting or suitable, it becomes, it behoves'; gedafenian 'to be becoming or fit, to behove'; gerisan 'to behove, become, befit, suit'; behofian 'to have need of, to need, require'. OE behofian is thus not a proper BEHOVE verb, but used as a synonym to the pure necessity predicates burfan and geneodian. The earliest example of behofian in the sense

1 gebyrian, like the simplex byrian, also means 'happen; pertain to, belong to' and will, in the former meaning, be treated with the HAPPEN class. The simple verb risan 'to rise; be fitting, becoming" does occur besides gerisan but is apparently very rarely used, and not found in the dictionaries, in construction with either sentential or nominal complements. 40

of 'moral obligation' dates from the 12c., and it appears that the verb was used in the meaning 'be fitting', etc., in ME only.1 BEHOVE predicates regularly take one [+animate] and one [—animate] NP argument, an important restriction on the relational situation; and again the thematic focus can be on the person in need of something or on the thing needed, a situation reflected by the dictionary meanings.

6.1

Semantic NP roles associated with BEHO VE verbs

The semantic relation between verb and NPa can be paraphrased as '(to) NP behoves that...', 'it is fitting for NP that...'2 Again the role properties of the NPa are uncontroversial, including the recipient (and here, more properly, goal) but not the agent notion. The operational test of clef ting makes the nature of this relation manifest: in 'what is fitting for him is...' the NPa is prepositionally marked as indirect object, associated with the semantic role of goal. Except for the meaning 'need' (behofian), then, an agentive reading of the NPa is impossible with BEHOVE verbs. It will be interesting now to see what effect this fundamental semantic situation has on the syntactic occurrence paradigm entered by these predicates.

6.2

Syntactic occurrence properties of BEHO VE verbs

The OE occurrence pattern of BEHOVE verbs is quite uniform: in construction with sentential complements they are regularly attested in type S, illustrated by (68)—(71): (68) (69) (70) (71)

eow gebyrab baet ge beon eft acennede (Aelf. Horn. Supp. 1,480,19) be gedafenab to lerenne and me to hlistenne (SOL 70,17) Wei be geriseb baet bu heafod sie (Christ. ASPR III, 3) on fuwuer wise us behoued turnen to him (Trin. 63,12)3

For the pure BEHOVE predicates gebyrian, gedafenian andgerisan this is what the semantic situation described in the preceding section would lead us to expect: in association with the meaning 'be fitting', the animate-NP can become the grammatical subject but is restricted to the semantic roles mentioned, expressed by the indirect object function.

1 pan alden bihouab äugende bewas (Lamb. 109, 24). OE becuman in the sense 'become, behove' is attested in one (type S) example only and therefore not included here: Gific seile and becyme mec baetic... (BTs). 2 German has preserved the original syntactic expression of the same relation: ihm gebührt, dass... besides es gebührt ihm, dass... 3 This is the earliest example of behofian with infinitival complement (12c.). 41

There is particularly rich evidence in the present case from complex be +adjective predicates, expressing the range of semantic values on the scale from 'be due, fitting' to 'good', 'necessary' and 'useful'. The most representative ones for the OE semantic field BEHOVE are gerisenlic, gebeorhlic, gedafenlic, behoflic, gebyrelic (all 'fitting');god, riht, sael, gedefe, gelidic ('good, proper'); gecundelic ('natural'); nyt, nytlic, behefe ('useful'); bearf, bearflic, behoflic, nidbehefe, neod, bryce ('necessary'). All of these regularly occur in type S, represented by (72)-(76): (72) (73) (74) (75) (76)

God bib mannum baet... (CP 397,18) baem lytegan bonne is betere baet... (CP 205, 3) ne bib him na gebeorhlic baet... (W 238, 60) us is my eel bearf baet we... teolian (BH 125, 11) mycel is neodbearf manna gehwylcum baet... (W 256, 20)

In (77) and (78) the NPa is again prepositionally marked as indirect object: (77) (78)

he is to freonde god (BT) Ne sint gebyrelico ludea to Samaritaniscum (BT)

With behofian the possibility of a personal variant seems much more viable semantically (Ί have to ...'). Paraphrase evidence shows this assumption to be plausible: the most common necessity predicate burfan occurs in personal construction such as ne bearft bu be ondraedon (BH 191, 18). It should, furthermore, be noted that behofian can be paraphrased by a personal structure of the form habban (agan) plus a necessity adjective or noun as in (79) and (80): (79) (80)

we habbab nedbearfe baet... (BH 23,1) Nu age we bearfe... baet... (W 201,18)

Although in proper semantic terms the NPa remains goal in both cases, in (79) and (80) it is couched in the form of personal, quasitransitive syntax. Even if behofian is never found in personal construction of the form he behofeb baet..., we have to assume that this type was grammatically possible, because of sentences like Ac baes behofab se cyning baet... (Aelf. Horn. Supp. I, 380, 46), or bonne behofab se be her wunab ... baet... (Christ and Satan, ASPR I, 209). (Compare also the presence in OE grammar of the type ic behofige baes, i.e. structures with nominal complements in which the NPa has nominative form1.) Now in contrast to the RUE and PLEASE/DESIRE classes, there is direct evidence for a BEHOVE predicate occurring in the /r-construction, as shown by (81): (81)

hit ne gerist nanum ricum cynincge baet... (Lives 382, 260)

Here the NPa is explicitly marked as indirect object in a straightforward syntactic SVO construction. In this case, then, there is primary evidence for the presence in OE of the variation pair: 1 See chapter II, section 3.2 42

(82) a b

paem cyninge gerisep pact... it gerisep paem cyninge past...

Again it is impossible — and, on the basis of directly corresponding modern German evidence1, inadequate — to account for the difference between the a- and i>-sentences above in semantic terms. (In contrast to the variation pattern with a personal variation, there is here no question of a pseudo-agentive reading of the NPa.) The conclusion is then that (82) should be described as a case of pure syntactic (and stylistic) variation. In terms of use this variation pattern was not integrated into the syntactic paradigm of OE subjectless predicates. Its potential availability, however, is an additional crucial fact on which subsequent linguistic development will build. The verb classes dealt with in the following two sections present a fundamentally simpler semantic context.

7.

TheOE HAPPEN class

7.1

Semantic NP roles associated with HAPPEN verbs

The class of HAPPEN verbs offers an even further restricted semantic framework. The possible grammatical relations holding with (ge)limpan, geweorpan and (ge)byrian (all meaning 'happen')2 are limited by the nature of the semantic situation paraphrased by 'it happens to NPa that...', with the NPa in pure recipient role.

7.2

Syntactic occurrence properties of HAPPEN verbs

OE HAPPEN verbs exclusively occur in type S, represented by (83)-(85): (83) (84) (85)

ponne him gelomp paet... (HE 442, 21) pa gewearp hine paet... (CP 197, 14) oft ponne hwaem gebyrep paet... (CP 39, 6)

With this verb class there is no possible indeterminacy with regard to the semantic role features of the NPa. Also in terms of syntactic function its reading as the indirect object is made clear by the semantic interpretation of the sentences. This situation suggests a systematic (i.e. in this case semantically-based) account of the fact that the personal variant is not found at all. 1 See footnote 2, p. 41. 2 OE becuman is also used in the meaning 'happen', but not in construction with sentential complements.

43

Contrary to the predicates dealt with so far 'happen' allows no paraphrase of the complex form be +adjective, so that there is no additional evidence of this kind. However, the data from the type with nominal complements (chapter II) show that the present verb class never occurs with nominative NPa in OE. The complete absence of HAPPEN verbs in the /Y-construction, on the other hand, cannot be given any systematic reason. This type would have neatly expressed the indirect object status of the NPa, fully preserving the original semantic reading. Note uwtgelimpan is attested in the variant with£aef instead of it: (86)

& paetoft gita manegum mannum gelimpeff... paet... (BH 129, 6)

Again the most coherent account for this situation is in terms of language use, not grammatical necessity.

8. The OE SEEM class 8.1

Semantic NP-roles associated with pyncan

The SEEM class verb pyncan offers again a semantically-based restrictive framework for the investigation of grammatical relations. In the situation paraphrasable by 'it seems to NPa that ...', the NPa is a pure recipient. This semantic fact again correlates with the indirect object status of this element in the functional reading of type S. 8.2

Syntactic occurrence properties of pyncan

The syntactic paradigm of pyncan is simple: the verb only occurs in type S form, such as (87): (87)

& him puhte paet... (HE 362, 32)

There is very rich evidence from the copulative type pyncan+adjective/noun, regularly attested in this construction: (88) (89)

him pa raedlecre gepuhte paet he \νφ opernefrijj gename (Or. 96, 15) pa puhte me hefightime pe to tipienne paes (GEN, 76, 3)

As in the case of the HAPPEN verbs, the fact that the /f-construction is not attested with pyncan throws light on the aspect of syntactic use, but has no systematic status. This construction does in fact occur with the complex predicate (in construction with a temporal clause): 44

(90)

To lang hit him buhte hwaenne hi togaedere garas beron (Battle of Maldon, ASPR VI, 133)

(90) again remains an isolated example besides hundreds of type S sentences. The personal construction involves the discussion of a lexical variant, byncan cannot occur with nominative NPa for semantic reasons. But we should note the presence in the OE lexicon of bencan 'to think', which thus allows proper semantic variation to be expressed, as illustrated by (91): (91) a b

me bynceb baet... ic bence baet...

(91) is the only OE example for the complete and productive lexicalization of variant pairs of the present kind, i.e. for the expression of the intransitive and transitive use by two separate verbs.1

9. Summary 9.1

The complete occurrence paradigm

Let us begin this summary on subjectless verbs occurring in construction with sentential complements by looking at the survey chart listing attested occurrences in the structural types met so far:

1 See the well-documented section about fryncan/pencan in vdG (chapter IV). 45

Type S

Personal

it-constr.

Occurrence with nominative NPa in other construction types1

RUE hreowan sceamian eglian ofbyncan gehreowan ofhreowan

χ χ χ χ χ χ

— χ — — — —

— — — χ — —

χ χ χ — — χ

PLEASE/DESIRE lician gelician lystan langian gelustfullian lustfullian

χ χ χ χ χ —

— — χ — — χ

— — — — — —

χ χ — — χ χ

BEHOVE gebyrian gedafenian gerisan behofian

χ χ χ χ

— — — χ

— — χ —

— — χ

HAPPEN gelimpan geweorpan gebyrian

χ χ χ

— — —

χ — —

— —

SEEM byncan (byncan + Adj.

χ χ

— —

— χ

χ)

The overall attested situation shows that type S is the indigenous syntactic expression for subjectless verbs occurring with an animate-NP and sentential complement. This conclusion generalizes to the wider range of data including complex predicates.2 Lustfullian is the only verb which does not appear in type S. Semantically, the verb-NPa relation entered by this predicate is the same as that entered by the other members of the 1 I.e. with nominal complements (see chapter V) or without complementation at all (type ic licige). 2 See vdG for rich material supporting this claim.

46

PLEASE/DESIRE class. As mentioned above, there are only very few records of this verb,1 so that any generalizations remain unsafe. Evidence from the type with nominal complementation will decide the question how, in our material, lustfullian syntactically expresses the specific pseudo-subject relation. As far as the it -variant is concerned, we observed that no aspect of OE grammar would constrain it from occurring with the present verb classes. The structure is syntactically unmarked, and brings the formal expression in line with semantic interpretation by having the NPa in indirect object place. In all the material at our disposal, however, this type is only found v/iuigerisan and the complex RUE predicates mentioned in section 4.2. There is no 'explanation' for such statistical facts. Our conclusion was that this situation had to be accounted for in terms of use. It will be seen in section 11 below that there are general factors in OE grammar which (we will argue) favour the use of type S rather than the itconstruction. From a formal, morphosyntactic point of view the personal variant is evidently possible. As noted, the semantic structure of RUE and DESIRE verbs (as well as the semantic context of 'necessity') in no way prevents these predicates from occurring with nominative NPa. However, only sceamian, lystan and behofian 'need' are attested both in the S-type and the personal construction. Although in this latter variant the NPa is open to a quasiagentive surface interpretation, there can be no question of genuine transitive use. As far as NPa roles are concerned, the extremely restricted use of the personal variant will not provoke a re-analysis from recipient to agent even where such a re-interpretation is first possible, viz. when the NPa can be receiver as well as giver (cf. lician, hreowan). In the construction type with sentential complements this situation will not arise, but we will have to keep it in mind when dealing with the type where two animate NPs are present (see chapter II). With the BEHOVE, HAPPEN and SEEM verbs, on the other hand, the semantics prevent occurrence with nominative NPa in OE. In the light of the facts available to us, it seems that both the use of the nominative NPa with some verb classes and of the //-construction were purely syntactic innovations in the grammar of OE, potentially available for predicates where there is no semantic constraint, but not integrated into a productive pattern of grammatical variation. One important factor contributing to the predominance of type S in OE is its traditional status. We have to assume the uninterrupted use of this construction from IE. OE still reflects the continued force of the particular verb-NPa relation with the verb classes described here. It is precisely the fact that type S has marked status within OE syntax that makes it fitted for the expression of the original recipient nature of the animate being involved. The status of type S within the framework of general OE syntax will be the main topic of the rest of this chapter. 1 The total number of occurrences in my data is only 4. BT(s) lists 13 examples, but only one shows sentential complementation. The verb is not listed in the/teowu//and ASPR concordances.

47

Now we want to turn to the aspect of syntactic function as an additional means of grammatical characterization.

9.2

The functional aspect

The most characteristic feature of OE subjectless syntax dealt with so far is the grammatical double nature of the NPa. Whereas in semantic (and notional) terms this element is 'the object', some of its syntactic properties are those of subjects. Thus, in type S the NPa has to be analysed as syntactic subject in terms of word order and constituent structure, but the semantic interpretation clearly attributes it (indirect) object status, as shown by the purely syntactic variant in the form of the iY-construction. In the following we are interested in the kind of evidence which the functional aspect can provide for the correct grammatical analysis of OE type S and the double nature of its pseudo-subject. Unless stated otherwise, subject and object functions are syntactically defined (i.e. in terms of distribution, reflected by the configuration of basic structure).1 In terms of syntactic functions, subject (S), verb (V) and object (O), the unmarked order of elements in OE is SVO in the main and SOV in the subordinate clause. OE is thus not a pure SVO language, but shows traces of an original verb-final type. Now one of the various possibilities of re-arranging SVO main clause order is into OVS (by topicalization), as subject and object can be freely moved round the more rigidly fixed element, the verb. In configurational terms of syntactic constituent structure, the NPa in type S is a subject, and the construction conforms to the unmarked SVO pattern.2 Semantically, however, the surface order of elements in type S is OVS. It was this discrepancy between the respective linear arrangement of the same grammatical categories on the semantic and syntactic levels which led traditional linguists to call the NPa the pseudo-subject of type S constructions. Within the context of sentential complementation one important restriction was noted: the r/wi-clause cannot occur initially, in syntactic subject place (poet ... gerisep paem cyninge). In type S, it is thus impossible for the syntactic order to overlap with the semantic reading. On the other hand, the syntactic SVO order of this construction corresponds to the pattern in the unmarked personal type. From the point of view of syntactic structure, then, there is no difference between type S and the personal construction. Consequently any syntactically-based definition of functions will establish a functional parallelism between these two construction types.

1 On the requirements of a complete characterization of subjecthood see Lyons (1968, § 8.1). 2 Because we want to preserve coherent reference with regard to the NPa, the syntactic order is given as SVO rather than XVO, which would be more adequate in a different context of inquiry.

48

Let us now turn to further evidence, provided by the syntactic behaviour of the NPa, an aspect also dependent on the notional interpretation of this element by the language user. We are interested, in particular, in finding out whether there is a difference between the behaviour of the NPa in type S and that of the real subject or indirect object(s), respectively, - i.e. those present in personal construction. The most extensively available material illustrating this issue in OE is provided by the syntactic co-occurrence behaviour of NPs in complex sentences. If the first NP is a nominative subject, the behaviour of the second (coreferential) NP can be of two kinds: either it can occur as a nominative heading the (co-ordinated or subordinated) second clause or it can remain syntactically unexpressed, being, however, 'understood' for the semantic interpretation of the subjectless clause. (92) and (93) illustrate the two possibilities: (92) (93)

baet he bonne ma ne wilnige pact he selflicige his hieremonnum ... (CP 141,16) bes cynges brober gaderode swipe mycel folc & bohte to gewinnaneEngleland... (Chr. 224,13)

Now the behaviour of the oblique NPa in type S is identical to that of nominative subjects. The NPs in (94) are also coreferential: (94)

heo ... & hyre gedafenode ... pact heo maeden waere (Aelf. Horn. Supp. I, 214, 417)

The second NP is functionally parallel to the first, i.e. the subject. That subjecthood is not restricted to nominative NPs in OE is also shown by (95), in which the subject of the subordinate clause (coreferential with that of the main verb cweban 'say') is the oblique NPa in a type S construction: (95)

... cwaef) baet him buhte swylce he eode ofer blostman (Lives I, 140, 381)

In co-ordinated structures with a nominative subject in the first sentence, the omission of the coreferential second NP is a general feature of OE syntax:1 (96)

baet ic gumcystum godne funde beaga bryttan, breac bonne moste (Beow. 1487)

(97)—(100) show that the second occurrence of the subject can also be left out if the preceding NP is the pseudo-subject of type S: (97) (98) (99)

gode ofhreow ba and hrabe cwaeb to bam engle... (Lives I, 300,255) ba wearb him aerest ege... andgefor on Bruttibaet land (Or. 198, 32) Antiochuse ... gebuhte baet the rice genoh naefde and wilnade baet he Parthe begeate (Or. 224,17)

1 Cf. Pogatscher (1900).

49

(100) Aefter paem \vearp Maecedonium αφ paet... andfundon paet... (Or. 146,13) As (100) shows, verb agreement in the plural remains operative in this situation. Evidence of this type suggests that the NPa was treated, on the level of syntactic functions, precisely as a real subject. This is crucial information insofar that real indirect objects do not behave in this way. A second property typically associated with subjecthood is the capacity to 'control the deletion' of subjects of dependent infinitive clauses. Coreferentiality is in this case again an essential condition. In (101) a nominative subject controls the omission (in syntax) of the infinitive subject: (101) pa nyste heora nan his naman to secgenne (BT) The same capacity is shown by the NPa in a type S construction: (102) ponne pam menn ne lyst... nan god don (Lives I, 356, 297) (103) pegedafenap to lerenne... (SOL 70, 17) That this feature is not limited to subjects, however, appears from (104)—(105), where the direct object (in an ACI construction) is coreferential with the unexpressed subject of the infinitive: (104) Drihten het pe secgan pas ping (EX 222, 15) (105) mon ... maeg... paet leoht geseon scinan (BH 129,4) The same situation holds in (106)—(107), where the 'understood' absent subject of the infinitive is coreferential with an indirect object: (106) hie him sealdon attor drincan (BH 229,9) (107) ic sende pison folce flaesc to etene (EX 253,5) The capacity to control the deletion of the subject of a subordinate clause is thus not a crucial property of subjecthood and evidence of this kind remains inconclusive, because it does not show in what way real subjects and indirect objects behave differently from the pseudo-subject in type S. Whereas the NPa is treated like a typical subject in (94)-(96) and (97)-(100), in variation pairs like (82) above (repeated here as (101)) it has indirect object status: (101) a paem cyninge gerisep paet... b hit gerisep paem cyninge paet... Both constituent structure (direct domination by the verb phrase) and word order (regular postverbal position) clearly express the indirect object nature of the NPa in (lOlb).

50

In the if-construction the element inverted to form questions is it — and not the NPa as in type S. From the scant evidence available it appears that whenever it and the NPa occur together, it is in syntactic subject position, whereas the NPa is postverbal. There are no main clause sentences of the form hit fraem cyninge gerisefj paet..., with both elements preverbally. What we see, then, is that the NPa in type S is treated variously as syntactic subject (cooccurrence in complex sentences) or indirect object (paraphrase by means of the //-construction). Such functional behaviour once again reflects the grammatical double nature of the NPa.1 One possible approach to the grammatical analysis of such indeterminacy is to treat the NPa as a 'squishy subject'2, allowing it to take several values on a scale between pure subject and object status, without necessarily reducing it to either. In the light of the inconsistent interpretation of this element in type S on different levels of sentence structure, a description in terms of such a concept of subject/objecthood would appear to be a natural solution. Under the squishy subject hypothesis, the NPa would qualify as subject if it showed 'a clear preponderance' of relevant subject properties,3 and the same would be true for objecthood. However, this concept has not yet received convincing theoretical backing. On the contrary, recent research seems to indicate that there is no universal category subject, and the list of subject properties proposed so far does not cover the subject properties necessary even for one language.4 Furthermore, most of the items on the subject properties list cannot be properly applied to OE. There are at present no systematic parameters along which the relative subject and object properties of the NPa in type S could be evaluated. However, the assumption of an informal concept such as a multifactor view of subjecthood is justified in the present context if we consider it as an essential condition for the further historical development of the construction type. As long as we cannot state precisely the respective status and interdependence of syntax, functional interpretation, semantics - and pragmatics, such a concept offers a plausible descriptive framework, in particular in view of the grammatical reanalysis which characterizes the history of subjectless grammar. After a first look at syntactic structure in section 2, we focused, in sections 3—8, on the semantic aspect of NPa roles and the syntactic occurrence paradigm of subjectless predicates. On the basis of the results summarized in the present section, we may now return 1 Supporting evidence for this situation is provided by an operational test suggested by Glinz (197072, II, 20), where subjects are isolated by a 'transformation' of the verb from finite to infinite form, which has the effect that subjects can no longer occur in the infinitive complex. The test attributes subject status to the NPa in German type S constructions with RUE verbs such as reuen, bereuen and schämen: reuen, dass etwas passiert ist / ihn. Schämen, dass etwas passiert ist / ihn. Bereuen, dass etwas passiert ist / er. 2 This term was coined by Ross (1972) and adopted by Keenan (1975) into his concept of subjecthood. 3 Keenan (1975, 312). 4 Cf. Li(1975b).

51

to the formal aspect of the descriptive framework, viz. the characterization of subjectless syntax in terms of basic structure.

10.

Towards an integrated basic structure

The synonymity of type S and the ίΥ-construction is the main reason for assuming a grammatical relationship between them also in terms of basic structure. If the variation between these two types is purely syntactic in nature, then our method should be made capable of expressing this fact. Two questions arise under this aim: if the correct characterization of the NPa in type S seems to be in terms of the squish concept, how should it be represented in the basic structure? And how should the relation between type S and the /i-construction be formally captured? The only syntactic subject feature of the animate-NP in S is its regular initial position, a fact which must lead any syntactically-based analysis to have the subject position occupied by this element. However, in OE syntax preverbal position is not uniquely associated with subject function, as is the case in strict SVO languages. The first condition to be satisfied is that the verb is not left uncovered in front position, but that some other element precedes it. If the basic structure of type S is to be linked to general OE syntax, it has to take into account the general validity of this structural principle. To let the subject position be occupied by the NPa would not only render impossible any systematic reference to the syntax of the /r-construction, but would also be incompatible with the constant semantic interpretation of both syntactic variants. The alternative is to give the NPa (indirect) object place in basic structure, which would allow the correct semantic interpretation of both type S and //-sentences formed on this pattern. It is thus by giving it semantic content that we can speak of the common basic structure representing these two types. What element, then, should occupy the subject position? Note that in OE the preverbal subject representative, (semantically empty) it, is still optional, i.e. not a necessary element in our concept of basic structure. It follows that the subject position should remain empty at this level of abstraction. The common basic structure of the OE construction types entered by the present verb classes can then be represented by the following tree:

(NP)

it

52

This basic structure is meant to reflect the semantic fact that OE subjectless predicates take two necessary arguments, the animate-NP and the sentential complement. We can thus draw a systematic distinction between the purely formal syntactic element it and the NPa as a semantic argument of the verb. (Although in terms of grammatical relations the complement would qualify as semantic subject, this position is left empty, because the basic structure is not an ordered representation of semantic elements. It is, in the present form, an integrated syntactic-semantic structure, in which the position of elements is justified by criteria from both aspects.) The characterization of the integrated basic structure as a subjectless construction is further supported by the following set of facts. Many of the verbs in question here enter constructions without an NPa, i.e. of the type it-V-S (hit gedafenafr fjaet...). The syntactic subject it is again optional, as it is not a semantic argument of the verb. This type of sentence is very productive and the most frequent form is without it, although usually some other element precedes the verb. (109) is an example of genuine subjectless construction: (109) gelamp fjaet... (BTs) Such structures were obviously complete sentences in OE, semantically as well as syntactically. Besides this type there is the variant with it in subject position: (110) hit gelamp sume daeg baet... (GEN 128, 15) In (109) and (110) the dependent clause is the only argument of the verb. The subject position must thus remain empty in any syntactically-oriented analysis because the clause cannot occupy initial position, a constraint in English grammar up to the 16c. With other predicates there is an (unspecified) experiencer NP which is understood but not expressed syntactically: (111) hyt gedafenode paet... (Or. 38,3) (Notice the type with specified experiencer of the form hit gedafenode fjaem cyninge fjaet ...) Again we should characterize the abstract structure representing the constant syntactic properties of the type without an animate-NP as subjectless like the basic structure for the construction with NPa, viz. by leaving the configurationally defined subject position unoccupied. It will have been noticed that the personal variant has not been included at all in the discussion of the form of basic structure and the function of its elements. This follows from our assumption that the appearance of the personal construction with OE subjectless predicates has to be seen as a purely syntactic phenomenon, viz. as the earliest manifestation of a tendency to give type S the normative SVO form of unmarked declarative sentences. Formally, the basic structure suggested above easily allows the description also of the personal construction as a stylistic variant. This type cannot have its origin in semantic variation; it is not known, with the present set of verbs, in the oldest stages of Germanic dialects. The historical situation thus suggests that originally type S was not a

53

marked variant to the personal construction in any Germanic language, but an indigenous type associated with a specific semantic situation. As mentioned already, it is not our purpose mechanically to relate the present concept of basic structure to the actual form of sentences. In the following section, however, we will try to establish a systematic link between basic structure and regular surface occurrence in terms of general grammatical tendencies.

11.

A dynamic view of OE subjectless grammar

Our basic structure is explicitly not associated with a device to 'generate' surface structures; we cannot, in other words, account for the actual form of sentences by means of movement transformations. Nevertheless, we cannot leave undiscussed the apparent discrepancy between the position of the NPa in the basic structure and its regular occurrence in initial place in type S sentences. We have yet to establish a link from the first step, the establishment of a static syntactic-semantic framework, to the second, a dynamic account of the actual occurrence paradigm of subjectless predicates. The formulation of the question, however, is not that of a transformational approach (which would probably try to motivate the application of NPa-preposing in the case of type S and surface it -introduction in the case of the /i-construction). Under the present analysis, the striking phenomenon of the productivity of type S will be related to two general factors present in the grammar of OE. These have the form of a structural principle and a thematic-oriented tendency. Within the limits set by the formal syntax, the predominant use of type S in OE is first of all a result of the fact that it comes closest to 'what one wants to say'. Its communicative value is clearly a major factor responsible for the attested situation. The first general factor to be mentioned is the so-called verb-second constraint, a very distinct tendency (or, to use a modern term, target) in OE syntax, concerning the position of the verb in structured linear order. There is a definite tendency for the verb to be in second place in unmarked declarative order, a constraint typical of SVO languages, which OE — at least in main clause structure — was becoming. This positional restriction cannot, of course, be made responsible for the attested picture of type S. However, it is a general feature of OE syntax which sets a target to be satisfied by any type of structure, and conformity with this principle is an essential condition for syntactic patterns to become productive. Let us first look at the nature and form of the verb-second target in OE and then discuss how it bears on type S. It was mentioned above that the presence of it is optional in the it -construction, but that there is a definite tendency to have some element (e.g. an adverbial one) in preverbal position. This tendency is a manifestation of the verb-second constraint developed in OE syntax.

54

In subjectless structures this target did not have imminent consequences for the syntactic subject position. Sentences could still remain subjectless, if only there was a preverbal element. Thus, whereas genuine subjectless constructions such as (1 12) and (1 13) do still occur, the effect of the verb-second target is to produce structures of the form as in

(112) (1 13) (114) (115) (1 16) (117) (118)

gelomp f>aet ... (BTs) nis aenig wundor hu ... (Christ, ASPR III, 1016) pa gelamp paet . . . (BH 1 99 , 3) forbaem gelimpp paette ... (CP 259, 23) gifponne aefregebyreb pact ... (SOL 71,17) forpy is betere pact ... (B. 25 , 27) oft eac gebyreb paet ... (CP 105, 19)

The construction with complex predicates offers a simple way to satisfy the verb position target, viz. by topicalizing the predicative adjective or noun, a most productive process in OE prose: (119) cup is wide ... paet ... (Guthlac, ASPR, III., 536) (120) cup is paet ... (BH 33, 7) (121) maniges mannes wise bib paet ... (BH 55,17) Note that in terms of basic structure (1 1 2)— (1 21) are alle characterized as subjectless. The same is true for (122)— (125), in which, however, it serves as a syntactic subject marker: (122) ( 1 23) ( 1 24) (125)

hyt eac wel gedafenode paet ... (Or. 38, 3) hit gewyrp paet . . . (CP 233 , 5) hit bepearfpaet . . . (CP 43 1 , 4) hit is ungecyndelice ofermodgung paet ... (CP 109, 1 1)

The force of the verb-second target is indicated by the fact that in this configuration the putative verb-initial variant gedafenap hit paet ... is not found in our material. If a conjunction or adverb co-occurs with it, we can distinguish two equally productive main clause variants, both satisfying verb-second order: the CONJ/ADV element is initial, with it in postverbal position: (126) (127) (128) (129)

pa gelamp hit. ..paet... (Chr. 148, 15) pagewearb hit ... paet ... (Or. 108, 24) ne gedafenap hit no paet . . . (CP 1 73 , 1 7) nu gewurpe hyt ... paet ... (GEN 130, 10)

or it occupies the subject position, with the other elements pre- or postverbal, according to their respective weight:

55.

(130) (131) (132) (133)

hytgelamp t>a ... paet... (GEN 119,1) hit gelamp sume daeg paet... (GEN 178, 15) hitgewearp aefter ures drihtenes prowunge... paet... (H 86, 11) hit eftgelimpep paet... (Beow. 1753)

In (134)—(136) the prepositional phrase (otherwise hardly found in front position) serves to realize the same target: (134) purh paes cypdo swipost gelomp paet... (HE 230, 15) (135) aefter peosan gewinnegewearp paette... (Or. 98, 29) (136) onpaem dagum gewearp paette... (Or. 100, 28) The generalization suggested on the basis of these data is fully borne out by the OE material considered in this study: the first requirement is for the preverbal position to be occupied. If an element of the class CONJ/ADV is present, it is regularly used to satisfy this condition. The same applies to complex predicates. If there is no such element, it is often added to keep the verb in second place. The development of verb-second order in OE is thus relevant for type S insofar as it gradually restricts the occurrence of verb-initial structures of the kind gelamp him paet... More important, however, is the fact that this target (which is obviously preferred to the tendency to have nominative subjects) allows the oblique animate-NP to take that position in the regular form of type S. It is in this sense that sentences like him gelamp paet... can be said to conform with OE SVO syntax. The verb target alone is still insufficient to account for the productivity of type S in OE. There is, however, another general tendency, which together with the verb target has the effect of producing initial animate-NPs. The nature of this animateness (or personalization) target means that it cannot be formalized in syntactic terms, but assuming it lends substance to an account of the actual OE situation.1 The general tendency to have animate-subjects is a topic-related feature2 especially important in construction with two NPs (cf. chapter II). It is relevant for the occurrence properties of the NPa insofar as it favours structures with preverbal animate-NPs, i.e. typical subjects in unmarked personal syntax. Thus, while the basic structure correctly characterizes the NPa as an indirect object, the association of this abstract structure with the animateness target links its initial occurrence in type S to a general topicalization process. As the verb classes concerned here are inherently associated with animate-NPs, type S falls naturally within the domain of this tendency.

1 Cf. also the notion 'perceptual strategies' used by Bever & Langendoen (1972). 2 Cf. Mathesius (1929) and recent studies such as those collected by Li (1975). It appears, that this target relates to a general, topic-related phenomenon at work across languages. 56

The animateness target is the strongest explanatory factor in accounting for the productivity of type S vis-ä-vis the if-construction. The latter was a perfectly possible grammatical structure in OE, satisfying both the verb-second constraint and syntactic SVO. However, reference to the animateness target yields content to an account in terms of use: animateNPs (even in oblique case) are preferred initially to semantically empty elements such as it. (Or even, as chapter II will show, to non-animate nominative NPs). The basic structure reflects the semantically-determined grammatical relations holding with subjectless predicates in the form of abstract syntactic structure. By associating it with the effect of the two targets mentioned, we achieve a grammatical description which is semantically correct and directly relevant to the actual OE occurrence paradigm. Whereas the basic structure characterizes the present type of sentences as subjectless, verb-second order in actual sentences can be accounted for by reference to different factors. First, if a conjunction or adverb is present, it will precede the verb (pa gelamp him paet ...). Second, with complex predicates the adjective or noun will regularly occur initially (riht is him paet...). Third, in the absence of an element with semantic content, the syntactic subject marker it may be used to keep the verb in second place (hit gelamp him pact ...). Note that the animate-NP remains in post-verbal position in all these processes. Despite the availability of all three grammatical possibilities, attested use shows that only the first two form productive patterns. Structures (such as type S) combining both targets are evidently preferred. Although we cannot assess their respective individual force, the structural demand for verb-second order normally has first priority. If a time adverb is present, for example, there are two variants with markedly different productivity. By far the most frequent type is with the time adverb initially, as in pa gelamp him pact..., leaving the verb in second position. To judge from the present material, the variant him gelamp paet... (a possible main clause order) is less often used in narrative prose. We should not forget, finally, that sentence rhythm is an additional relevant factor determining linear syntactic order. The description of OE subjectless grammar suggested in the preceding sections takes into account information from different levels of linguistic analysis to establish the fixed framework of an integrated basic structure, reflecting the relatedness of the two syntactic variants involved. At the same time, however, the actual use of different variants could be given a rationale by associating basic structure with the dynamic concept of targets. A methodology incorporating this concept allows an account in gradual rather than absolute all-or-nothing terms, which is exactly what the grammar of subjectless verbs demands. In what way the assumption of target structures can give content also to an account of pathways of syntactic development will appear in the second part of this study.

57

CHAPTER II SUBJECTLESS CONSTRUCTIONS WITH NOMINAL COMPLEMENTS

1.

Preliminary remarks

Most members of the verb classes investigated in chapter I also occur with nominal instead of sentential complements, a situation which enables us to establish valid generalizations and to refer from grammatical aspects of one type to those of the other within a coherent semantic framework. In this chapter we need to examine first a type of construction which corresponds to type S but for the complement. In sentences such as Oft pone gepyldegestan scamap paes siges Often the most patient man is ashamed of the victory'(CP 227,19), the NPa is in objective case as in type S, whereas the other NP is marked by the genitive. This is a syntactically subjectless construction whose parallelism to type S suggests an analysis along the parameters established in chapter I. Evidence from the construction with nominal complements promises to be especially important with regard to the aspect of semantic roles. This is due to the fact that whereas the relation between the verb and the sentential complement remained neutral, the case marking on nominal complements will allow more insight into the content of semantic roles. This in turn will throw more light on the grammatical double nature of the animateNP. In section 3 two further syntactic variants involving the animate-NP will be discussed. In the first there is also an NPa in oblique case; the second NP, however, is in the nominative, as in me pin modsefa Heap leng swa wel "The longer I know them, the better your mind and spirit please me' (Beow. 1853). The second is a personal variant of the type with genitive objects mentioned above: in sentences like he sceamep paes the animate-NP occurs as unambiguous subject. The treatment of each variant will follow the methodology developed in the preceding chapter. Whereas the semantic framework remains constant, this chapter will complement the syntactic occurrence paradigm of subjectless predicates and thus try to achieve a correct account of the syntactic-semantic properties of OE subjectless grammar. It follows from the considerably more complex syntactic evidence obtained in this way that the present task can be approached on a significantly stronger empirical basis than hitherto possible. Let us then turn to the kind of syntactically subjectless sentences mentioned at the beginning, henceforth called type N.

59

2. Type N 2.1

The semantic framework

One of the two NPs associated with the verb in type N is regularly [+animate]. With the verb classes concerned here, the animate-NP is inherently experiencer, recipient and goal, in the relational situation which may be paraphrased by 'NPa is affected by (V+object)'. This suggests that the grammatical relation between subjectless predicates and their NPa remains constant in both the sentential and the nominal type. Likewise, the intransitive use of the verb remains the same in types S and N. The type with nominal complementation allows a more explicit definition of the grammatical relation between the verb and the second (genitive) NP. Visser calls it the 'causative relation'1; this certainly captures the essence of the semantic relationship in question, but remains of course both vague in respect to the exact nature of the relation and silent as to the semantic role of the genitive object. Like the verb-NPa relation, the verb-genitiveobject relation is present in IE, and it has proved notoriously difficult to ascertain the exact nature of both types of relation. The problem is particularly felt in English, as subsequent development has rendered awareness of both kinds of grammatical relation wholly obsolete. For the present purpose the exact content of the causative relation does not concern us, as it is sufficiently distinguished from all other types of verb-NP relation dealt with here. We will thus use Visser's term for the sake of clear reference. In line with this semantic context, the genitive NP takes the semantic role of 'cause, source'. The following examples represent OE type N: (1) (2) (3) (4) (5)

him ofhreow paes marines (BT) ponne ofpyncp him paes ilcan pe he aerforbaer (CP 225,19) pe scamode swelces gedwolan (B. 19, 30) hine nanes fringes ne lyste on pisse worulde (B. 102,9) me a langap paes pe ic... (BT)

The complement can also take prepositional form, although examples like (6) and (7) are rare: (6) (7)

menn... scamap nu forgoddaedan (W. Sermo 62, 153) minre sowie pyrste to pan lufigende gode (Wai. 44)2

1 (1963,355). 2 pyrstan is here used as a synonym for langian 'desire'.

60

2.2

The syntax of type N

As in type S, there is the animate-NP in objective case. There is no number agreement in the plural and the verb remains in the 'neutral' third person singular. The second NP is in the genitive case (with a prepositional variant where the valency of the verb allows it). Evidence from word order is also very consistent: the NPa is regularly initial, the genitive NP postverbal. The verb-second constraint is thus satisfied in the unmarked form of type N. I have found no example with initial genitive where this NP is a full noun. Nor is the prepositional phrase ever met in front position. However, topicalization of the postverbal element is possible under certain restricted conditions. In pin me ofhriwp (BT), for example, the genitive-NP appears initially due to the fronting properties of pronominal elements in OE. (In this sentence moreover the animateness target is satisfied). Contrary to type S, declarative verb-first order is not found. It appears that this situation is not accidental, but is due to a general restriction: structures with two postverbal nominal complements (none of which in the nominative) appear not to be present in OE.1 Notice that in such a case it would have been possible to interpret the structure as a question: hreowep him paere daede?, unless the context disambiguated the situation. The place of the second NP in generalized main clause order, then, is postverbal. In contrast to type S, however, preverbal appearance is possible, dependent on rhythmical properties. The NPa, on the other hand, regularly has initial position, subject to variation by the factors established for type S (co-occurrence with conjunctions and adverbs). OE type N sentences, then, have the unmarked order NPa-V-NP. In establishing the basic structure of type N, the arguments used and the result achieved in chapter I apply: the semantic role and functional indirect object status of the NPa in N are parallel to those in S. A basic structure compatible with semantic interpretation again has to represent the NPa in object place, leaving the subject position open; (a deictic it cannot occur in the type with nominal complementation):

1 Such structures are only possible in transitive sentences with dative and accusative objects. See Carlton (1970, chapter 8). 61

The regular process by which verb-second order in actual sentences is satisfied is to front the NPa (under the animateness target). Due to the rhythmical properties of pronouns, also the second NP can occur initially in a marked variant of the type pin me hreowep mentioned above. Let us turn to the aspect of syntactic parallelism between the constructions with nominal or sentential complements; the unmarked syntactic order is as in (8): (8) α b

him hreowep paet... him hreowep paere daede

(9) is a possible variant with both types if an introducer element is present: (9) a b

hreaw him paet... hreaw him paere daede

As for initial position of the complement, there is an absolute constraint on (lOa), whereas the restriction on (lOb) is relative to rhythmical structure: (10) a b

*paet... hreowep him (*) paere daede hreowep him (pin me hreowep)

In contrast there is an absolute constraint on (lib), which corresponds to the sentential it -construction: (11) a b

it hreowep him paet... *it hreowep him paere daede

Sentences of the form (1 Ib) do not occur in our material and can be excluded as possible syntactic variants on the following grounds. In type S it was treated merely as a surface deictic element heralding the extraposed sentential complement. With type N, on the other hand, there is no comparable reason for it to be present. Whereas one of the functions of it in the sentential type was to preserve verb-second order in those cases where the NPa had postverbal position and there was no introducer (it gerisep him paet...), there is no evidence in OE for the co-occurrence of it with two nominal complements. It appears that it is not used, in this configuration, for the sole purpose of establishing syntactic verb-second order, but it does appear that the animateness target regularly helps to produce sentences with initial NPa. No element in Type N is constrained to last position in the same fundamental way as the sentential complement is in type S. And whereas verb-initial structure is possible — if rare — with type S, in the construction with nominal complements it does not occur at all. It appears that if a verb takes two NPs, one of them will be preverbal. 62

Again the structural verb-second principle alone is insufficient to account for this situation. It is the additional presence of the animateness target that gives content to the occurrence paradigm of type N sentences. That the syntactic behaviour of the NPa in complex sentences with two co-referential NPs is parallel to that of real subjects is shown by (12) and (13). In these sentences the second occurrence of the coreferent in the co-ordinated structure is omitted; while in (12) the controller NP is a real nominative subject (se maessepreost), in (13) it is the oblique NPa: (12) (13)

pa se maessepreost paes marines ofhreow, and scof... (Lives II, 142, 262) & him gelicade hire peawas & pancode gode (Chr. 201, 32)

On the other hand, examples like (14) reflect the double role of the NPa in the same complex sentence. In (14)

peah hy paes lyste, peah hy eodon ... (Paris Ps. 40, 6, quoted by Vi. 367)

there is no nominative/accusative distinction in the plural pronoun, but the absence of verb agreement in the first clause ensures the proper pseudo-subject reading associated with the intransitive semantic use of lystan in this construction. Type N is thus characterized as a subjectless structure with two object NPs, one of which (the NPa) regularly occurs as pseudo-subject. The following sections will build on this analysis.

2.3

The occurrence of subjectless predicates in type N

Turning to the occurrence properties of the present verb classes in type N, we are particularly interested in the question of whether the presence or absence of a predicate (or class of predicates) in this construction can be systematically linked to its semantics and in what terms such regularities should be described. Needless to say, it is not always possible to classify negative evidence sharply into arbitrary and principled absence; each individual case has to be discussed in its own terms. With the verb classes in question, type N is attested as follows (x = attested): RUE

PLEASE/ DESIRE

hreowan ofpincan

X χ

eglian sceamian

-

X

X

lician (ge)lustfullian lystan langian

BEHOVE --

behofian gebyrian

--

χ χ

gedafenian gerisan

-

HAPPEN

SEEM

geweorpan -gelimpan -

pyncan

-

1 In the meaning 'pertain, have to do with' gebyrian occurs in subjectless construction of the type Him negebyrap to (jam sceapum (BT). 63

2.3.7 The OE RUE class All RUE verbs except eglian occur freely in type N, represented here by ofbyncan and sceamian: (15) (16)

Maeg baes bonne ofbyncan beodne Heabobeardna (Beow. 2032) Ne bearfbe baes eaforan sceomigan (Genesis, ASPR I, 2329)

The semantic structure of RUE predicates allows the distinct causative relation to the [—animate] NP which is syntactically expressed by the genitive. (Compare also the complex predicate beon wa in bam folce waes aegbres waa 'The people rued either' (Or. 114, 31)). The absence of eglian in this construction is difficult to account for; its dictionary meaning 'to trouble, pain, grieve' does not exclude the object relation expressed in type N, and typical of the semantic subclass in question. This would mean that we have to consider sentences like him egleb baere daede as potentially possible variants not recorded by accident. (Notice that the German cognate ekeln is recorded in the N-construction in Middle High German (DWB)). On the other hand the frequent overall occurrence of this verb in OE and ME suggests a more systematic reason for the gap in its syntactic paradigm. It will be seen in section 3.1.1 below that eglian is regularly used quasi-transitively, in a syntactic type representing the source of grief etc. not as a causative argument (expressed by the genitive) but as the pseudo-agent in nominative case, as in him sculan eglan oberra monna brocu ... 'He must grieve for the troubles of others ...' (CP 75,10). At this stage, then, it cannot be decided whether the restricted occurrence paradigm of eglian is due to a specific semantic constraint. Evidence presented in section 3, however, indicates that this is not the case.

2.3.2 The OE PLEA SEI DESIRE class With regard to the causative relation in type N, we have to distinguish the subclasses lystan / langian and lician/(ge)lustfullian. With the former group the original IE type (with the thing desired in the genitive) is continued. There is ample evidence for this type of grammatical relation in OE with other verbs from this semantic field, as witnessed by verbs like girnan 'to yearn, seek for, require', higian 'to hie, hasten, strive', fundian, gitsian and tilian 'to strive after', all with genitive objects but nominative animate subjects instead of the oblique NPa. (17) and (18) illustrate type N with lystan and langian: (17) (18)

hine baes heardost langode (BH 227,1) begnas... gelystegargewinnes (Judith, ASPR IV, 306)

The complete absence in type N of lician and (ge)lustfullian, on the other hand, calls for two different types of explanation. The semantic structure of lician does not allow it to take genitive objects by entering the causative relation in the manner of lystan. There is

64

no single example in the very richly attested paradigm of this verb showing occurrence with a genitive object. This reflects a matter of grammatical (semantic) principle, not use. We will see in the following sections what syntactic form the 'thing causing pleasure' takes with lician. In contrast to this situation, the gap in the syntactic paradigm ofgelustfullian can only be accidental. The verb is not often recorded, but it does occur with oblique NPa (in type S) as well as genitive objects (see section 3.2), a syntactic situation which rules out a semantically-based account. In the case of lustfullian the attested records are not sufficient to allow a conclusive evaluation of its syntactic occurrence properties. This predicate is regularly used intransitively, with the experiencer-NPa in syntactic subject function. It is only found with nominative-NPa and the object in prepositional, genitive, dative or accusative form, as in (19)-(22): (19) (20) (21) (22)

Swa ic lustfullige on bisum lapwn witum ... (Lives I, 202,116) Waes heo seo cwen sona lustfulliende paere godan foresetnesse (HE 452, 19) pa ongon he lustfullian paes biscopes wordum ... (HE 122, 33) Se cyning... ongon lustfullian baet claenoste lifhaligra... (HE 62, 10)

In the light of this paradigm and the apparent restriction on nominative NPa subjects, we can safely assume that type N was not a grammatical variant in the OE paradigm of this verb. It is possible, in the present case, to link the occurrence in this syntactic type with proper semantic features: the predominant semantic element 'desire' of lystan/langian is easily compatible with the causative relation. With lician, as in other verbs of the same semantic class, however, the element 'experience pleasure' does not enter this relation.

2.3.3 The OE BEHOVE class None of the BEHOVE predicates proper occurs in this construction, and the restrictive attested situation is too constant to be arbitrary. The BEHOVE class offers a clear example of a semantically-based absolute constraint on syntax, here concerning the object relation between the verb and the [-animate] NP. 'What is fitting for somebody' does not stand in a 'causative' relation to the verb, but is again best described, in terms of semantic roles, as neutral or objective as in the case of sentential complements. With the BEHOVE verbs proper, then, the impossibility of occurring in type N is a fundamental semantic fact determining syntactic behaviour. It does not apply, however, to verbs of this class used with a different meaning, such as behofian 'need'. In the same way as the DESIRE verbs lystan and langian, behofian does occur with genitive objects.1 Compare also the occurrence in type N of pure necessity predicates like purfan and geneodian:

1 See section 3.2.1 65

(23) (24)

gif be smaelre candelle geneodige (Wal. 92) bonne be salteres beburfe (Wal. 94)

The fact that behofian is not found in type N can therefore not be given a systematic reason, but is probably due to the small overall number of records.

2.3.4 The OE HAPPEN and SEEM classes In construction with nominal complements the HAPPEN verbs geweorban, (ge)limpan (and gebyrian) do not take genitive objects. Nor are the syntactic type and associated relational situation found in IE.1 There are a few examples in OE of the genitive of the demonstrative pronoun, baes, occurring with HAPPEN verbs: him baes wirs gelamp (Wai. 75), hu gewearb be baes? (BT). However, gelimpan and geweorban are never attested with the genitive of full nouns, which suggests that/wes here should be interpreted as an adverb or conjunction.2 Likewise the SEEM verb byncan cannot occur in type N on semantic grounds, and in sentences like baes me bincb the first pronoun is probably used adverbially.3

2.4

A synopsis of type N

Evidence from their occurrence in type N contributed to our knowledge of the valency of subjectless predicates. In its pure form type N expresses the causative relation between the verb and the complement, the semantic content of which is most clearly present with the RUE and DESIRE verbs. As to the occurrence properties of individual verbs, it was possible to assess the nature of the absence of verbs not recorded in this construction type relative to their attested lexical valency (occurrence in other syntactic constructions). Thus the gaps in the syntactic paradigms of lician, the BEHOVE and SEEM verbs were assumed to be of a systematic kind, due to the semantic structure and inherently associated relational possibilities. With eglian, on the other hand, syntactic behaviour could not be related to semantic structure, but appeared to be the result of specific use. Within the paradigm of constructions with nominal complements, type N is no longer productive in OE. However, although N-sentences are not nearly as numerous as type S constructions, they regularly occur with such verbs as lystan, langian and sceamian, as well as beon bearf.* The presence of this construction type in the lexical valency of sub1 2 3 4

66

Cf. Delbriick/Brugmann (Grundriss, vol. 3, chapter 8). Also DWB geschehen. See Wiilfing( 1894, chapter 2). Similarly Wülfing (1894, 35). Cf. Behaghel (Syntax I, 564 ff., 609 ff.) for a comment about the scarcity of this construction in the Old Germanic dialects.

jectless predicates allows the same semantic situation to be expressed by either sentential or nominal complementation, as illustrated by (25) and (26): (25) a b (26) a b

me hreoweb pact ic dyde pa daed me hreoweb baere daede hine lysteb baet he baet ping haebbe hine lysteb baes binges

In addition to type N, the syntactic occurrence paradigm of OE subjectless verbs involves two further construction types, on the basis of which both their lexical valency and semantic structure can be investigated. The following section will deal with this aspect.

3. Two variant constructions entered by subjectless predicates In the following, two variant constructions complementing the syntactic paradigm in which OE subjectless verbs can occur will be investigated. The question to be answered is what such evidence can contribute towards a characterization of the grammatical status of type N and the relational properties associated with the set of verbs concerned here.

3.1

Variant type I

Important evidence for the analysis of the NPa and the syntactic use of the present verb classes is provided by the construction represented by sentences like be gedafenab ymen (OE Texts, Vesp. 273, 11). In this syntactic pattern — henceforth type I — the verb has associated with it two NPs, one of which is again regularly animate and has the relational and morphological properties of the NPa in types S and N. The second NP, however, corresponds to the object-NP in type N. It is in nominative case and governs verb agreement, a typical syntactic subject capacity. Type I is thus not a syntactically subjectless construction. What relates it to the present study is the fact that it is entered by the same predicates regularly occurring in type S and N and that the verb-NPa relation appears to be identical to the one holding in these types. What remains to be investigated is thus the semantic role of the second NP and the syntactic properties of this construction. Let us deal with the syntax first. There are two main word order variants of type I. Either there is a syntactic nominative subject initially, as in (27)—(31), henceforth subtype la: (27) (28) (29) (30) (31)

ba ofbuhte baet anum baes cyninges geferan (BT) bes sige gewearb Punicum (BT) Aeghwylc man ... burh gode daeda Code lician sceal (BH 129, 33) baet naenige binga ne gedafenode swa aebelum cyninge (HE 130, 31) Hit be wyrs ne maeg on binum hyge hreowan (Genesis, ASPR I, 826)

67

or word order is reversed, with the anirnate-NP in front position, as in (32)—(38), henceforth subtype Ib: (32) (33) (34) (35) (36) (37) (38)

fjylaes... him bonne gehreowe sio aelmesse (CP 325, 7) gifpam gifran ungemetlicu spraec ne eglde (CP 309, 2) hu him se sige gelicade (Or. 156,25) hem behofeb Cristes helpe (Chr. 262. 18) gifhimforsip gebyrige (Ancient Laws, quot. after Vi. 288, s.v. byriari) be gerisep lofsang (BT) him ealle pincg gelumpon swa swa... (Lives I, 268, 104)

(27)-(31) and (32)-(38) are representatives of the same construction. It seems clear that (27)—(31) must determine the basic form of type I, because there is a syntactic subject which is interpreted as the subject of an intransitive sentence. This leaves the NPa in postverbal - indirect object - position. The syntax is here in a one-to-one relationship with the functional reading, viz. SVO. (27)-(31) thus conform to the most basic normative tendencies in OE prose syntax. But how should (32)—(38) be accounted for? Their syntactic word order is OVS, i.e. a marked variant of regular SVO, though a common topicalization order in OE. The oblique NPa is the element focused in this word order variant. In terms of targets this is a strong case for the effect of the animateness target, which here overrides the normative tendency of having nominative subjects in initial position. Structures of the form (32)—(38) are also favoured by the general rhythmical tendency to front light elements (here pronouns), a situation frequently arising because of the association of the present verb classes with pronominal experiencers. Subtype Ib thus has the appearance of a subjectless structure (me hreowep ...), although there is a syntactic subject in postverbal position. It is in particular the verb-NPa relation which links type I to subjectless syntax. This time, however, the co-occurrence with a syntactic subject means that the oblique animate-NP can only function as indirect object, despite regular topicalization under the animateness target. In the following we need to discuss the occurrence situation of subjectless verbs in this construction.

68

3.1.1 Subjectless predicates in type I The occurrence of the present verb classes in type I is the following: PLEASE/ DESIRE

RUE

hreowan ofbincan

eglian sceamian

χ χ

lician gelustfullian lustfullian lystan langian

BEHOVE χ χ

behofian gebyrian

χ χ

HAPPEN1

SEEM

geweorban χ gelimpan χ

byncan

-

gedafenian χ gerisan χ

The construction is very productive with RUE verbs, but there is a striking gap: sceamian is not recorded in type I. Are him sceameb pa word or seo daed sceameb him to be considered ungrammatical? In view of the regular occurrence of both these subtypes with all the other RUE predicates it seems unlikely that the absence of sceamian is accidental. If this assumption is correct, a very plausible reason for the restricted syntactic behaviour of this verb suggests itself. Sceamian is one of the most productive predicates occurring with genitive objects, and it appears that the particularly strong causative notion associated with the source-argument of this verb determined its syntactic paradigm. While the indigenous expression of this relation is through type Ν (and II, see section 3.2), the syntax of type I, with its neutral relation between the verb and the second NP, was avoided. This is also consistent with the fact that sceamian is often used as a reflexive verb. Note furthermore that the pair lystan/langian, both with equally distinct causative objects, are also absent in type I (see below). Again the pathway of development in ME will lend support to the correctness of the semantically-based account suggested here. The two subclasses within the PLEASE/DESIRE class which appeared in the occurrence chart for type Ν are also reflected by type I. With lician the attested syntactic paradigm (present in I, absent in N) can be given content by making it dependent on the semantic structure of this verb, which does not allow the causative relation, but enters the neutral relation to its second NP. Besides the grammatical structure him se sige gelicode, a putative *him baes siges gelicode must be regarded as ungrammatical in OE.2 The same type of explanation applies to the attested situation with gelustfullian.

1 Becuman 'happen' also occurs in type I: Him becomon fela yrmba (BT). 2 Cf. also leofian 'to be dear,pleasant;delight' which occurs in type I but never with genitive objects: Him leofedan londes wynna bold on beorhge (BT), 'the pleasures of the country were dear to him, the house on the hill'.

69

With lystan and langian, on the other hand, the causative object relation seems to be exclusively used, and the restricted syntactic paradigm of these verbs is parallel to that of sceamian discussed above. With lustfullian, the absence in type I has systematic status. As mentioned already, the verb takes its animate-NP in nominative case only, i.e. it does not enter the kind of verbNPa relation typical of the present verb classes. The syntactic behaviour of BEHOVE verbs must be accounted for along similar lines. As is the case with lician, BEHOVE verbs do not enter the causative relation, a fact responsible for their absence in type N. The semantics of these non-agentive verbs regularly involves a recipient/goal animate-NP and a neutral [—animate] NP in intransitive construction, a situation which finds its most adequate expression in type I.1 The neutral/objective role taken by the [—animate] NP in I appears equally clearly with HAPPEN verbs, which are all productive in this construction.2 The SEEM predicate fryncan, on the other hand, does not enter this type. We have so far found evidence for the semantically-governed occurrence paradigm of lician/gelustfullian and the BEHOVE verbs (only in type I) and langian, lystan, sceamian (exclusively in N). All these predicates enter the specific verb-NPa relation characterizing subjectless syntax, but the causative or neutral nature of the relation to the second NP fundamentally restricts their occurrence possibilities. Insofar as the syntactic paradigm formed by N and I is concerned, then, the OE data appear to reflect a systematically governed situation. The occurrence properties of predicates in these constructions can be made dependent on the nature of their semantic relation to the second NP. If the causative relation predominates, type I is not a possible choice. It appears that in this case the 'relational inversion' from causative object in N to subject in I is prevented. Besides further evidence from OE (section 3.2 below) and from ME (chapter III), the rare examples of a variation pair formed by N and I sentences must provide the test for this assumption. The type of variation offered by these constructions is not productive in OE. In the present material only the RUE verbs hreowan and offryncan remain potential candidates for this variation pattern. Both verbs occur quite freely in N and I, and the kind of relational inversion involved can be paraphrased as follows: in a variation pair of the form him hreowep fraes / him hreowef) fraet the causative object of the former (fraes) 'corresponds' to the syntactic subject of the latter (paet). This kind of variation, however, cannot be described in terms of two different semantic readings, as (39) clearly demonstrates:

1 Gebyrian is quite frequent in this construction in the meaning 'pertain' (see BT). 2 Cf. also the use of becuman in the meaning 'happen': Us becymfr swa peak call poet to gode ('it will all happen to us for the good'; Lives I, 378, 208). The rarely attested gesceon 'happen' is also exclusively found in type I. 70

(39) a b

pa ofpuhte paet anum paes cyninges geferan (BT) ofpincep pe ealles pe fju to yfele haefst geworht? (BT)

Unfortunately, evidence on this issue remains extremely scarce.

3.1.2 The type it-V-NPa Let us turn now to a special case of subtype la, with pronominalized subject NP, of the schematic form it-V-NPa (hit licode paem cyninge). The syntax of this construction shows a significant word order constraint in that it is obligatorily in initial position. There are no sentences of the form *paem cyninge hreowep hit, "'him licep hit etc. besides perfectly regular Ib sentences like paem cyninge hreowep seo daed, him licep se sige etc. This rigid word order falls first of all within the domain of the general rhythmical tendency to have light elements preverbally. Moreover it has here the thematic value of connectives, a further fronting property. Even in construction with two pronouns this structure is regularly preserved. It appears that these factors combine to form a word order constraint on OE sentences, overriding the animateness target, which would favour sentences of the form him licep hit. With regard to the occurrence of the present verb classes in the type it-V-NPa, the following has to be noted: sceamian is the only RUE verb not attested in this construction. It should be remembered that we assumed its complete absence in type I to be due to the strong causative role of the source-NP, syntactically expressed by the genitive. In the same way the syntactic behaviour of the PLEASE/DESIRE class corresponds exactly to the pattern for type I: while hit licode/gelustfullode him are possible sentences in OE, the DESIRE verbs lystan and langian do not enter this construction. With other verb classes the occurrence pattern is the same as in type I. Evidence from the subtype it-V-NPa thus fully confirms the results achieved in the previous section. (Note that the constraint on a verb to occur in it-V-NPa form can also have a different reason. Hyngrian and pyrstan are never met in this type. On the basis of attested data the assumption is justified that this situation is due to the aspect of referential!ty on the part of the syntactic subject. In type I proper the subject is a full noun and therefore referential. In the same way its referent of the pronominal subject it is necessarily present in the discourse context. In a putative *hit pyrstep me, *hit hungrep me, however, it could only be interpreted as non-referential in meaning. This is another aspect showing that in OE it is still not used purely to produce verb-second order.)

3.2

Variant type Π

We have seen in the preceding sections that type I retained the specific semantic relation between verb and animate-NP met in Ν and S, but had the second NP in nominative case and syntactic subject function. Its two word order variants allowed a functional represen-

71

tation of the semantic situation as OVS or SVO respectively. There is yet another syntactic variant - type II — into which the present verb classes enter, but this time it is the animate-NP itself which takes nominative form, while the other NP is the genitive object of type N. Type II sentences, then, are the personal variants to subjectless type N; besides him hreowep paes, we also find the form he hreowep paes. (40)—(44) illustrate type II: (40) (41) (42) (43) (44)

pa se maessepreost paes marines ofhreow ... (Lives II, 142, 262) no he paere feoh-gyfte for scotenum scamian porfte (Beow. 1027) pa ongan se bysceop lustfullian paes iungan snytro & his wislicra worda1 (HE 454, 4) seo sawl... lystep godes rices (Waerferth, Dial., quot. after Vi. 367) nu behofap ure freo-dom aefre godes fultumes (Lives 1,382, 266)

In this type the syntactic subject regularly precedes the object-NP. This order is preserved in structures like (42) and (44), where the verb precedes both NPs after certain introducers. Syntactic SVO order, with the animate-NP in subject position, combines in this type with the animateness target to form an apparently absolute constraint on the possibility of having the object-NP initially. Whereas with type N front position of the genitive-NP could occur if it was pronominal and [+animate], this is not the case in type II. The grammar of OE appears not to allow a putative *pin he hreowep besides pin him hreowep.

3.2.1 Subjectless predicates in type II Type II is attested as follows: RUE

PLEASE/ DESIRE

hreowan ofpyncan

χ -

eglian sceamian

χ

lician gelustfullian langian lystan lustfullian

BEHOVE -

behofian gebyrian

χ -

χ χ

gedafenian gerisan

-

HAPPEN

SEEM

geweorpan gelimpan -

pyncan

-

Let us first discuss the systematic or arbitrary nature of the attested situation within the context of syntactic structure and relational variation discussed so far. The occurrence of RUE and PLEASE/DESIRE predicates in type N shows that the causative object relation is well established with this class (eglian excepted). It was also noted, on the other hand, 1 Wulfing (1894, 20) notes that wislicra worda could be parallel to paes iungan, i.e. dependent on snyttro. However, this interpretation seems unlikely, and is also not adopted by Miller's translation (EETS, O. S. 96). There are no other, unambiguous examples of genitive objects in the material consulted.

72

that their occurrence with nominative animate subjects could not be excluded on semantic grounds - as is the case with OE BEHOVE, HAPPEN and SEEM verbs. This situation renders type II an indigenous construction with the RUE class, as reflected by the syntactic behaviour of hreowan and sceamian. Ofbyncan regularly takes genitive objects, but is never attested with nominative animateNP, in whatever construction type. In view of the scant overall occurrence of this verb, we cannot decide whether this situation is due to a semantic constraint, although in the light of the syntactic valency of RUE verbs this seems unlikely. The absence of eglian, on the other hand, must be due to the semantically-based restriction (mentioned in the context of type N) disallowing the causative object relation.1 Similarly, the absence of lician and gelustfullian in this construction can be subsumed under the same reason. With these verbs the nature of the relation to the second NP is never of the causative kind, which accounts for their absence in both N and II. It is thus not the pseudo-subject relation between verb and NPa which is responsible for this restriction. We have seen in chapter I that lician does take nominative animate subjects, in absolute constructions like hu wel hie liciab (CP 231,9) and ic licige. On the other hand, the highly restricted occurrence in type II of lystan and langian cannot be linked to the causative relation, which both verbs freely enter in type N. Their occurrence with nominative animate subjects cannot be excluded, however. Lystan in (43) above remains an isolated example, and langian is never met with nominative subjects at all.2 Although no semantic constraint prevents lystan and langian from occurring in type II, the attested distribution makes clear that this construction was never a productive variant in their syntactic paradigm. It will be interesting to return to this issue in the light of early ME data in the next chapter. The occurrence situation of pure BEHOVE verbs is determined by the same semanticallybased constraint noted in the context of type N: these predicates do not allow the causative relation syntactically expressed by the genitive. Furthermore, none of the proper BEHOVE verbs is ever attested with an animate nominative subject, another general restriction directly following from their core meaning 'be fitting'.3 Behofian has again to be dealt with individually, because its regular meaning 'need' determines a causative/partitive relation to the object. Just as the necessity predicate (be)burfan, behofian freely enters type II: (45) (46)

he bebearfeac micles fultumes (BT) he ne behofab baes huxlican reafes (Aelf. Horn. Supp. I, 434, 1)

1 Further RUE verbs occurring in type II, but never with oblique NPa, are hreowsian and sorgian. 2 Further PLEASE verbs taking genitive objects, but nominative animate subjects, are faegnian 'to rejoice' and gefeon 'to be glad, rejoice'. 3 Note that gebyrian does take nominative subjects if used in the meaning 'concern, belong' as in Ealle pa fre to Codes rice gebyriap (BTs).

73

The OE HAPPEN and SEEM class, finally, cannot occur in type II, because it does not take the recipient-NP in nominative case.1 The occurrence possibilities of OE subjectless verbs in this construction are thus systematically limited by their capacity to enter the causative relation and to occur with nominative human subjects. The former aspect is responsible for the non-occurrence in this type of eglian, lician and gelustfullian, the latter for the absence of HAPPEN and SEEM verbs in this construction. Contrary to such a grammatically-determined account, the restricted occurrence paradigm of lystan and langian must be regarded as a matter of language use. With regard to subsequent development, it is important to note that only the RUE verbs and lystan/langian allow a potential re-interpretation of the NPa if it occurs in nominative case. The BEHOVE, HAPPEN and SEEM verbs are unable to undergo such reanalysis. If type II is the 'personal' variant of N, there arises the question of what effect the syntactic variation involved in pairs of such variants have on the semantic situation. (47)— (49) illustrate the issue: (47)α b (48)α b (49) a b

f>a se maessepreost baes mannes ofhreow (Lives II, 142, 262) him ofhreow baes mannes (Aelf. Horn. I, 192,16) gifbu be ofsceamian wilt pines gedwolan (B. 9,18) bone gebyldegestan scamab baes siges (CP, 227,19) bonne seo sawl... lysteb Codes rices (BT) bonne hine... aetes lysteb (The Whale, ASPR III, 173,51)

As in the case of variation between type S and the personal construction, a proper semantic distinction is not possible. Even in terms of the semantic role features of the NPa the interpretation can be synonymous. In both types the animate-NP can be read as the experiencer and recipient. In the a examples, however, the syntax allows a quasi-agentive interpretation. The distinction could be described by reference to the semantic feature [volitionality], but we are unable to say whether in OE such a re-interpretation actually took place. In the light of otherwise exclusive intransitive use this seems unlikely. An aspect on which a safer stand is possible concerns the respective productivity of N and II. To judge from the attested situation, it is N, i.e. the highly marked variant, which is preferred. If this fact cannot resolve the question of semantic interpretation, it does indicate that the presence in OE of type II is not the consequence of some previous semantic reanalysis of NPa roles, of a desire to give expression to a quasi-agentive interpretation of this element. Rather, it appears to be the result of a general syntactic tendency bringing sentence structure in line with unmarked SVO.

1 With fjyncan+adjective the (reflexive) human subject is sporadically found in the nominative: He fr h t e him selfum suipe unlytel (BT).

74

Within the syntactic paradigm of nominal complementation, type II is of special interest with regard to the ME form of the paradigm. While still an intransitive construction, it may give rise to structural indeterminacy if the case distinction between nominative and accusative is neutralized. This is the case with several classes of nominal inflection.1 Now with predicates potentially allowing a quasi-transitive reanalysis of the NPa role, sentences like he andsware hreoweb (i.e.without a definite article) are open to a pseudoagentive interpretation. Consider (50), where the singular of the feminine noun myltse could be interpreted as an accusative direct object ofbehofian 'need' used transitively: (50)

Ic com eac synful and myltse behofige baes heofenlican dryhtnes (Lives 1,82, 557)

Although the fact that behofian is never used with unambiguous direct objects in OE suggests an interpretation of (50) as type II, this reading is completely dependent on its context. Indeterminacy can also arise in relative clause construction such as/w lore ...be hi ealle behofigen (CP 455, 4). If the semantic conditions for reanalysis are given, such sentences are open to a transitive interpretation. (Note that this situation can arise with a large number of OE 'personal' verbs taking genitive objects).2 Of the present verb classes, however, gelustfullian is the only ditransitive predicate. In contrast to lician it can take animate-NPs as its arguments, as in bu ne gelustfulladest fynd mine (BTs), with an accusative object. (Cf. also Ic wes gelustfullad (BTs)). Except for gelustfullian, then, there is no evidence at all that OE subjectless verbs were used transitively. However, as a result of phonological development there were gradually more syntactic structures which could be interpreted in a quasi-agentive way.

4. Summary On the basis of data from subjectless verbs in construction with nominal complements we obtained much fuller evidence about the range of construction types into which they can enter as well as their individual occurrence properties in the syntactic paradigm. The result is a differentiated picture of syntactic behaviour and the semantic role properties of the NPa. Insofar as the grammatical relation between verb and NPa is concerned, the result achieved in chapter I can be generalized: the specific pseudo-subject status of this element, associated with the experiencer/recipient role, is its indigenous feature also in the type with nominal complementation. It is noticeable that during the whole OE period the marked types (N and Ib) remain productive and typical of the syntactic occurrence of the present set of verbs. With the proper BEHOVE, HAPPEN and SEEM predicates, the 1 Cf. the singular/plural of o-stems (and subclasses), the plural of u-stems, and the masc./fem. singular of the weak n-declension. 2 Cf.Visser (1963, 366 ff.). 75

syntactic expression of the verb-NPa relation is even restricted to the form it has in N and I, as these verbs do not combine with animate-NPs in nominative case. With the RUE and PLEASE/DESIRE classes (and behofian), on the other hand, this restriction does not hold. Intransitive structures like ic licige, ic hreowe are not fundamentally excluded by the semantics as *ic gelimpe, *ic gerise. With these verbs, then, the occurrence with nominative animate subjects (as in type II) will produce structures which fully conform to the unmarked SVO norm. However, there can be no question in OE of transitive use associated with the SVO syntax of la or II. In the material consulted there are only two examples of indeterminacy, baet he us eglan moste (Judith, ASPR IV, 11) zndgelustfullodon be dohtra cyninga (BT). Here the formal sentence structure is ambiguous (in the first example, it could be read as type I or a transitive construction, according to the accusative or dative case of us, but a transitive interpretation is unlikely, because in the numerous examples with full NPs the latter is always marked by the dative). With gelustfullian - the only di-transitive verb (cf. BT) — the external context must decide the correct interpretation: Note that even with necessity predicates like behofian, burfan and geneodian, transitive use is never met. Another sign of general intransitive use is the situation noted with lician: the experiencer-NP is always marked as a dative-NPa, and if an animate nominativeNP is present, this is regularly the source of pleasure. Besides the constant association with a recipient-NP marked by oblique case, the capacity of predicates to enter the causative relation with the second NP is also important in determining their syntactic occurrence paradigm (cf. eglian, lician, gelustfullian). The potential syntactic variation possibilities open to OE subjectless predicates can be demonstrated on the basis of the attested occurrence of hreowan, which is the following: (51)o b c d

me hreoweb baere daede (N) me hreoweb seo daed (I) ic hreowe baere daede (II) me hreoweb baet... (S)

The nature of the variation pattern between (51) a and b concerns the grammatical relation to the second NP, causative in the a-, but neutral in the ό-example. An alternative, neutralizing, as it were, the distinction, is available in the form of (51) d, the variant with sentential complementation. In terms of relative frequency, this type is by far the most productive. (51) a and c, on the other hand, express variation with regard to the NPa, which is a pure recipient in the a-sentence, but open to a pseudo-agentive reading in c. Notice, finally, that there is no variant of the form *ic hreowe ba daed, the putative transitive variant. With the subclasses BEHOVE, HAPPEN and SEEM, the occurrence properties in this potential paradigm are determined by their semantic structure, which allows them only to enter types S and I (regularly in the form Ib, to give emphasis to the NPa). 76

The kind of variation involved with RUE and PLEASE/DESIRE verbs and the possibility for individual predicates to occur in the paradigm represented by (51) can be constrained in two ways. While, for example, hreowan is found in all variants of (51), this is not the case with other verbs. Thus sceamian, lystan and langian have just one type of object relation associated with them, viz. the causative one expressed by N and II. Lician does not enter this relation and is restricted to type I. (The same is true for intransitive gelustfullian). The status of the different variant constructions both within growing SVO syntax and in terms of productivity has been discussed in the individual sections. What emerges clearly in view of the complete syntactic paradigm is the fact that all structures falling within the domain of the animateness target (S, N, Ib) are productive, whereas those types conforming to normative syntax (the personal construction, la, II) are less frequently used. In particular the possibility of type II (or the personal variant) being used to express some semantic distinction has to be excluded. Both structures are the result of a general tendency for sentences to have the form of unmarked SVO sentences. This chapter has led us to the important conclusion that apart from the di-transitive verb gelustfullian, OE subjectless predicates do not enter the transitive construction with direct objects. This does not mean that in some semantic contexts the logically recipient animate-NP cannot be imagined to take a pseudo-agentive role. It does mean, however, that the set of predicates concerned here was never used transitively, i.e. in variation patterns implying genuine semantic change. It is in this sense that a study of OE subjectless grammar can be conducted within a constant semantic framework, a methodological parameter assumed for the present work. With regard to subsequent development in ME, however, a characterization of OE subjectless syntax in the above terms would be incomplete. Within the OE framework outlined here a number of situations can potentially give rise to indeterminacy and thus the possibility of reanalysis - in the present context in the direction of transitivity. Thus we noted in chapter I that in the construction with sentential complements, occurrence in complex sentences is a common source of indeterminacy. (Another, theoretically as possible, but practically unimportant, is the occurrence in type S or I of proper nouns as NPa). In this chapter, then, type II was shown to be a potential transitional structure due to the possibility of case neutralization in the object. In particular with RUE and DESIRE verbs a re-interpretation of the genitive to an accusative object can easily take place. Another, lexical aspect contributing to a transition towards personal use has been mentioned, viz. the presence in the semantic fields RUE and DESIRE of predicates regularly taking nominative animate subjects, such as hreowsian and sorgian. While these verbs normally occur with genitive objects (in type II constructions), hreowsian is also attested in genuine transitive construction: fre fja gefjohtan synna hreowsiafr (CP 417, 32). Note also that predicates close in meaning to 'rue', like maenan and sarettan (both 'to lament, mourn') take accusative objects: hi hit gemaenafr (Or. 136), paet ike sarette se witga (CP 267, 2).

77

Likewise the near synonyms of lystan/langian, willan ('to want') and wilnian ('to wish') occur in transitive constructions such as hwa haefde call jjaet he wolde (BT) or wilniaft hiora woruld aefter pa (B. 53, 16). However, none of those RUE and DESIRE verbs characterized here as subjectless is ever found thus construed. A further, semantic aspect relevant in the present context concerns the occurrence of one or two animate-NPs with the verb. If with the RUE and PLEASE/DESIRE verbs both NPs are animate, both can potentially function as recipient or source/giver. The possibility of keeping these verbs unambiguously distinct depends completely on a functional phonological and morphological marking system. These factors, then, seen in the context of a general (if still weak) tendency towards SVO syntax and gradually more frequent occurrence of structures with nominative NPa, will all contribute to a personal reanalysis of subjectless constructions as soon as there is a decisive leak in the system needed to express their marked form or if grammatical norm no longer allows such structures at all. This was evidently not yet the case in OE, where none of these factors provoked any syntactic or semantic change. On this basis let us now turn to the aspect of development, which will be examined in the next chapter.

78

PART TWO THE DEVELOPMENT OF SUBJECTLESS GRAMMAR

INTRODUCTION 1.

Aim

Our aim in this second part is to describe the history of the grammar of subjectless predicates through and beyond the ME period and to determine the syntactic and semantic properties of the same set of verbs on which the OE investigation was based.

2.

Theoretical aspects

As outlined in the introduction to the first part, there is as yet no theory of historical and diachronic linguistics capable of providing a better explanation of the complex phenomena in question than has been achieved by describing them within the parameters established in the previous two chapters. It would be wrong to conclude from this that there has not been fruitful research and real progress in some areas (such as, for example, the typological work done by W.P. Lehmann). However, as far as syntax (and, even more, semantics) is concerned, generative models for the analysis of diachronic change — from Klima (1964) to Lightfoot (1979) - have not succeeded in integrating the different levels of description, the distinction of which is necessary for the present set of OE data. An account in terms of transformational rule-change models (whether in the form of King (1968), Lakoff (1969), Traugott (1965) or Lightfoot (1979)) is clearly not feasible for the present task, for reasons which will become evident in this chapter. Neither can a case model (such as used by Baron (1974) and Jacobson (1971)) accommodate the range and type of data concerned here, in particular facts from syntactic typology. Apart from the notorious difficulty of defining and justifying deep cases in historical syntax, the interdependence between syntax, grammatical relations and the semantic structure of predicates cannot be embraced by case grammar. Another descriptive framework which holds some potential promise under the present aim is R. Hudson's (1976) non-transformational 'daughter-dependency grammar', due to its extended use of features and special function rules for subject and topic assignment. This model has, however, been exclusively used for synchronic data so far. Valency-based fragments of grammar, finally, offer convenient means for lexical description, but lack a systematic method of referring to overall syntactic development and the targets such as those established in chapter one. If none of these analytical frameworks is adopted here this does not by any means imply their worthlessness for the task. On the contrary, a rule-change model - whether transformational, lexical or diachronic - will adequately account for the set of data to which it is applicable. So will the case model as far as semantic roles, and systemic grammar as far as complex lexical entries are concerned. Similarly, Bever and Langendoen's (1972) hypothesis about perceptually-based strategies and constraint is probably still the richest explanatory theory suggested to date 81

for reanalysis in the diachronic syntax of relative clauses.1 As noted already, the present research situation in diachronic syntax differs markedly from that in phonology, the main reason being that a corresponding neogrammarian heritage is lacking and that methodological problems of this kind were also not on the structuralist programme. The main difficulty lies in the absence of a systematic notion of 'correspondence', a fact which has led some linguists to adopt (implicitly or explicitly) the nebulous framework of a 'diachronic grammar' (as a continuum) for their analyses.2 Unless the concept of correspondence can be narrowly defined in restrictive terms, however, isolated evidence of vaguely parallel sentences cannot be a sufficient basis for the evaluation of change and its direction. In the present study the notion of correspondence is given content by the constrained semantic framework determined by the verb classes I-V, associated with the formal typology represented by our concept of basic structure. The present description does not adopt any of the above-mentioned models, because none promises a satisfactory account under our formulation of the aim and in the light of the empirical results of the study of OE subjectless grammar in chapters I and II. As far as subjectless grammar is concerned, a reductive account (for example in terms of autonomous syntax) cannot embrace those aspects which it is impossible to state properly on this level of structure, first and foremost the role properties of the NPa. The results of the analysis in the previous chapters, then, should determine the parameters also of the diachronic account undertaken in the following. The description of development will thus be based on the occurrence patterns formed by the predicate classes RUE, PLEASE/DESIRE, BEHOVE, HAPPEN and SEEM. Given this semanticaUy-defmed framework, their occurrence in the range of constructions met in OE (and possibly others) lends itself quite naturally to comparison and grammatical evaluation. (The OE sentence me hreowefr pact... belongs to the same construction type as ME me rewep pat... If no semantic change intervenes, the role properties associated with OE hreowan are identical to those of ME rewen.) Syntactic congruity is of course not always so complete, but variations of the OE paradigm can in this sense be discussed within a coherent syntactic and semantic framework. An additional factor which has to be taken into account is the changing structure of the lexicon. The considerable ME change affecting in particular the present verb classes (many OE predicates becoming obsolete3 and new verbs entering the language) is liable to have a crucial effect on a marked subpart of grammar such as subjectless syntax in practical terms of lexical availability and use.

1 The typology-based kind of framework has recently found favour again, having been little estimated in the heyday of TG. Cf. Antilla (1972, 317 f.) and Lehmann (1978). 2 Cf. Lightfoot (1976b). 3 Cf. Visser (1963, 25).

82

One important methodological problem remains to be dealt with: the delimination of chronological 'stages' in which the data can be said to belong to 'the same' grammar, as opposed to preceding and subsequent grammars. The present attempt to link synchronic variation to diachronic change in a systematic way presupposes a notion of 'adjoining grammars'. The problem of grammatical criteria determining the limits between adjoining grammars has not yet been seriously discussed and will remain a thorny issue for some time. Again our approach to this question will be largely practical, in that the assumption of intact communication between three generations will be taken as the only condition. From this aspect the question of what historical stages should be distinguished in our ME material is answered by the data base: if it should allow the sort of generalizations and relative assessment of relational properties and the semantic use of predicates which was possible for OE, the time span of a century suggests itself. This is quite a strict chronological criterion, not normally applied to the description of syntactic change.1 In the present case, however, the external parameter of centuries promises to locate as correctly as possible the catalytic periods of change within the overall development from impersonal to personal construction. The geographical aspect (in particular northern vs. southern dialects) plays no decisive role in this context. Our external framework for comparison will thus embrace the 12c.—17c., complemented by evidence from the 18c. and 19c. where necessary. Before embarking on our analysis, the methodological outlook and results of van der Gaaf s study should be mentioned. He illustrates the OE situation quite cursorily and follows the development up to the 16c. (The bulk of the data is from the 14c.). This leads him to the conclusion that 'until the end of the 13th century punchen was the only type A verb [i.e. subjectless predicate ] that was thusjoined to a personal verb;(...). After 1300 other type A verbs also enter into constructions like those quoted above [i.e. personal, usually in co-ordinated sentences]. This shows that the difference between type A and 'personal' verbs was beginning to be lost sight of; most of the former, in fact, were, at times, already used 'personally" (1904, 34). He tends to illustrate semantic and relational points rather than attempt explanations by way of paraphrase possibilities: 'to contend that in the following quotations loben etc. are 'personal' verbs, would be stretching the point too far' (loc. cit, there follow co-ordinated constructions without explicit subjects in the second clause). After long lists of examples, the whole summary of the development is given in one short passage: 'with a few exceptions all type A verbs began to be used 'personally' in the first half of the 14th century (emphasis vdG). In the case of most of them the original construction continued to exist beside the new until about 1500; in the case of some the old construction kept its ground a century or a century and a half longer, while a few have preserved the A [=personal] construction till the present day.' (1904,142). Van der Gaaf s sole basis of classification is the relation between verb and animate-NP, which makes it l

Cf. Strang (1970, 21f.)

83

impossible to include such crucial aspects as the nominal object in N and its subject counterpart in I. There is a detailed account of the ME use of the following predicates: biren, me is leef, me hadde lever, ic am leef, I am better (...), me is wel, me is wo, rewen, langen, lothen, liken, listen, meten, dremen, bencan, byncan, forbinken, gaynen, happen, lakken, myster, me is nede, neden, (all of which changed to the personal construction) and eilen, greven, plesen, semen, smeten (which preserve the original relation). Van der Gaaf assumes a wholesale ME change from 'impersonal to personal' construction caused by the single factor of case shifting due to morphological coalescence and resulting ambiguity. Both his data base and the explanatory framework will be enlarged in the following account; the analysis of OE subjectless syntax has brought to light the complex nature of variation and the interdependence of individual occurrence properties and general structural tendencies or constraints in the language. We have seen in what way 'personal' use was present in OE, and the traditional account in terms of a wholesale development in ME from 'impersonal to personal construction' cannot characterize the complex nature of the history of OE subjectless syntax. There is, as has been noted on the basis of the situation in OE, a multitude of possibilities, reasons and conditions which can or should be made responsible to the diachronic changes in the individual verbal valency in the lexicon and the associated two possible pathways of development: (1) re-analysis of the animate-NP as a pure subject (he grieves...); (2) interpretation of this NP as an indirect object in postverbal position (it seems to me that...).

84

CHAPTER III THE HISTORY OF SUBJECTLESS PREDICATES UNTIL THE EARLY MODERN ENGLISH PERIOD

1.

The RUE class

1.1

RUE verbs in construction with sentential complements

Let us first turn to the history of the construction types with sentential complements. The ME RUE class comprises two new verbs (greven and forthinken, both entering the language in the 13c.), while the OE representatives are continued.

1.1.1

TypeS

From OE to the 17c. the occurrence properties of these predicates in type S are shown by the following chart:1 TYPES

OE

12c. 13c. 14c. 15c. 16c. 17c. 18c.

rewen ofpincen eilen shatnen greven fort hinken

x χ x χ

χ -

x χ χ χ -

x χ χ χ χ

x

x

-

-

x χ χ

x -

x ~ -

-

(1)—(6) illustrate this construction in ME: (1) (2) (3) (4) (5) (6)

him reowepp patt he nafepp nohht... (Orm 5570) me ofpincp pat... (SEL 337, 308) What alyd the, Lucifer, to fall? (15c. MED) hem shamede to seie pis openli (Wy. 357,15) sorore megreuep pat... (SEL 17, 42) Meforthynkez ful much pat euer I mon made (14c. OED)

The S-construction is attested with all verbs of this class. Specific lexical restrictions apply to ofpincen (obsolete after the 14c.) and eilen, which occurs freely with oblique NPa but very rarely with sentential complements. The external occurrence situation, then, must explain the gap up to the 15c., because type S is attested with this verb in OE. 1 χ means attested, - means not attested. A blank signifies that the predicate in question does not yet (or no longer) belong to the English vocabulary at the period concerned. OE references are added to ME charts where there is a substantial common lexical basis.

85

The situation appears to be a typical case of use, the regular — almost idiomatic — syntactic form in ME being what eileth you? The occurrence situation for the 12c. has to be evaluated carefully for each individual predicate. Documenting 12c. syntax is difficult because of the extreme scarcity of texts. In the present case however, the gaps can all be satisfactorily accounted for: the absence of rewen and shamen must be accidental, as the construction is attested in OE and from the 13c. onwards. The very scanty evidence of eilen in type S has been mentioned; greven and forthinken do not yet belong to the 12c. vocabulary of English. The occurrence in S of both verbs in the 13c. is further proof of the productivity of this type and the associated verb-NPa relation. Up to the 15c. type S is common with all RUE predicates (greven becomes obsolete in this construction already in the 14c.). Only rewen and eilen are still thus used in the 16c. and 17c. respectively. Throughout the whole ME period, then, type S remains a productive construction with the present set of verbs. To determine the relative status of this type within the syntactic paradigm of sentential complementation, however, and to evaluate the pathways of development followed by individual predicates, is only possible in the context of potential variant types, to which we will turn now.

1.1.2 The it-construction The occurrence of RUE verbs in the /r-construction is shown by the following chart: IT-CONSTRUCTION

OE

12c.

13c.

14c.

15c.

16c.

17c.

18c.

19c.

rewen ofbincen eilen shamen greven forthinken

_ -

χ _ -

_ χ x

x _ χ x

x

x

-

-

-

_ χ x

_ x χ x

_ x χ

_ x χ

_ x χ

(7)—(l 1) represent this type in ME: (7) (8) (9) (10) (11)

ne ofbingb hit be gif ic bus wer geceose? (Apol. of T. 29, 34) sore hit me rwes bat... (E.E. Allit. Poems, quot. after Vi. 44) it forthenkith me forgyue to him (Wy. Lk. 17,4 (MED)) yt grevyth me to thynke upon yowr gydeyng (PL III, 201,23) It shameth and irketh me to abide such thyngs as this world doth (16c. OED)

Ofbincen becomes obsolete already in the 13c. and is last attested in the 14c. (in type S). In the light of the present 12c. record, we assume its absence in the ;'f-construction in the 13c. and 14c. to be due to its rare overall occurrence. Rewen is used in this type from the 14c. until the 16c., but is never developed into a productive variant. The direction of development of its syntactic valency will appear more

86

clearly in view of evidence from the personal construction (see 1.1.3 below). As far as the verb-NPa relation is concerned, it is important to note that the great difference in productivity between S and the ίΥ-construction indicates again that the former was the indigenous representative. The semantic and syntactic character of the NPa in S and its nature as pseudo-subject determined its general use. With greven and forthinken, on the other hand, syntactic use shows the it-type to be the regular variant, as opposed to S. How should this situation — resistence of rewen to enter the //-construction, but productivity of greven/forthinken in this type - be accounted for? One possible way would be in terms of grammatical tradition. Whereas the predominant occurrence of rewen in type S continues OE use, the two ME synonyms enter the language at a stage when SVO syntax in the unmarked form of the it -construction has developed into a strong normative tendency. But the respective behaviour of these verbs cannot be made dependent on proper semantic features. Likewise the complete absence of eilen in this type cannot be given any more systematic foundation than the hypothesis suggested in the previous section. Shamen, finally, will be discussed in its proper context below.1 With rewen, greven and forthinken, then, 14c. and 15c. grammar offers the following variation possibility: (12) α b (13) a b

Sore me forthynkkes That euer... (Mal. (Morte A. (1)4,252). MED) me forthynkith hit... that... (Mai. 119,30) wel sore him greuede pat be kyng was angred (14c. MED) Sore hitgreueb me... bat... (CM 10443 Trin.)

The fact that these verbs (and ofbincen in the 12c.) can co-occur in both syntactic construction types, however, is not enough to allow them to be grouped into a distinct semantic subclass. There is no distinctive semantic feature shared by these verbs, but not present in the other RUE predicates. Nor can we systematically distinguish the nature of the verb-NPa relation in S and the /r-construction by reference to roles: the NPa is recipient in both cases. This variation pattern, then, seems to be adequately described in purely stylistic terms. Before turning to the third possibility offered by the syntactic paradigm, the personal construction, let us look at the meanings given for ME RUE verbs by the dictionaries: Rewen 'to affect (sb.) with regret, sorrow. To distress, grieve.' To repent'. (OED). Ofbincen 'to cause regret, sorrow; displease'. (BT). Eilen 'to trouble, afflict, harm'. (MED). Shamen 'feel shame, cause to feel shame'. (OED). Greven 'to injure, harass, oppress; to cause grief. (MED). Forthinken 'to regret, repent; be grieved, displeased; to grieve (sb.)'. (MED). 1 See p. 107.

87

The recurrent elements are 'rue', 'regret', 'repent' and 'grieve', all of which obligatorily require the presence of an animate-NP. In the unmarked case this NP will take the experiencer-recipient role (as in S and the /f-construction). Within the general context of transitivation, however, it can also be imagined as volitionally involved in the action of regretting etc., i.e. it can optionally enter into a pseudo-subject relation with the verb. On the level of syntax the distinction in question is naturally expressed by the nominative case of the NPa, contrary to its dative form in S. We have noted the existence of one general syntactic factor associated with the phenomenon of transitivation1 already in OE, in the form of growing SVO syntax. There is in ME a tendency towards more and more rigid SVO order (interdependent with the loss of case inflection) which will not remain without effect on the subjectless syntax of RUE verbs. Furthermore the changes in the lexicon due to the addition of quasi-transitive synonyms like repent and regret also play an important part. Bearing in mind the changing form of the syntactic and lexical framework, the occurrence and use of the personal variant will now be described within the context of the full ME variation paradigm.

1.1.3 The syntac tic paradigm of ME R UE verbs The following chart shows the occurrence from OE to the 19c. of optionally subjectless RUE verbs in the three possible construction types with sentential complementation: RUE

OE

12c.

13c.

rewen shamen greven forthinken ofpincen eilen

· Δ·

-

· · ·ο ο ·

· ·

·ο _ _

14c.

·ο · Δ ·ΟΔ ·ΟΔ · _ ·

15c.

16c.

17c.

18c.

·ΟΔ ·ΟΔ Δ _ _ · Δ Ο Δ Ο Δ Ο Δ Ο Δ ΟΔ ΟΔ ο ο ο ·ΟΔ ΟΔ _ _ Δ

·

_

_

·: type S ο: /Y-construction Δ: personal construction (14)—(17) represent the personal variant: (14) (15) (16) (17)

Defend shulde shame... to seye pat... (Wy 418,27) What heylyght pi leggy s now to be lame (15c. OED) Full sore for-thynkyng was he That euere he made mankynde (15c. MED) I grieve to see my Manuilus ielousie (17c. vdG 133)

1 For an illustration of the main aspects of transitivation in ME, see Visser (1963, 127 ff.) 88

19c.

_

It appears from the chart that personal use is not met with before the 14c. Even then examples are still rare, and often ambiguous between a type S or personal reading, such as (18): (18)

pys man ... Forthoght fiat he gafso meche hys sone (HS 1142)

In (19) and (20) the personal use of forthinken and rewen has to be inferred on the basis of analogy: (19) (20)

pis Bisschope ... hedde hope pat heo ...,And forpouhte pat... (14c. MED) The layffraturnyt... And rwytfull sar that... (15c. MED)

(19) and (20) are transition structures par excellence, inviting a personal, quasi-agentive interpretation of the second NP parallel to the first. In terms of use, then, the personal construction is not a productive variant before the 15c. The question now is how this relates to the OE situation. There we met shamen in the personal type as well as widespread use of nominative NPs with the set of subjectless predicates (in type II, he hreowep paes). The nature of the 12c. data situation does not allow us to make statements with any degree of certainty about that period, but it is the absence of personal constructions in the 13c. which has to be explained. Withgreven and forthinken the gap could possibly be attributed to the fact that they only enter the lexicon in the 13c., although such an explanation would remain weak unless supported by semantic factors, which are not available. Further evidence will have to decide whether these verbs do not generally take nominative subjects in the 13c. Even more problematic, however, is the case of rewen and shamen, whose 13c. paradigm appears to be restricted to the use of type S. We have seen in the previous chapter that shamen is the only OE subjectless predicate to be recorded in the personal variant. This would clearly suggest uninterrupted use until its next attested occurrence in the 14c. The absence of the personal variant in the 13c. can again not be given a more coherent explanation than the assumed result of a combination of a weak data base and general rarity of the construction type concerned. The following discussion of the ME paradigm of RUE verbs in construction with sentential complements will deal with two complementary aspects: the development of the valency of individual predicates (to be complemented by evidence from the type with nominal complementation) and the development and use of the three syntactic types involved. The most striking facts in the ME paradigm of RUE verbs are the following: the variation pattern S fit -construction from the 14c.—16c. and the three-way pattern including the personal variant in the 15c. and 16c. The generalization of both the it-type in the 14c. and the personal variant in the 15c. we assume to be determined by factors of use. Both structural types are extant in late OE and have remained possible syntactic structures ever since. On the other hand the continued productivity of type S up to the 16c. and 89

17c. indicates the prominent status of this original construction in ME. The occurrence of all three variants during the 15c. and 16c. can throw additional light on the grammatical nature of the animate-NP in type S. The fact that they are found in the same text or used by the same author suggests that the distinction involved was clear enough to be made use of. Whereas the /f-construction stands apart as a syntactic and rhythmical variant, the contrast between S and the personal construction is most properly described by reference to the traditional recipient role of the animate subject in the former, associated with a distinct thematic value. We cannot of course be certain that the personal variant also expressed the same recipient role (in which case we would be dealing with pure syntactic variation). However, in contrast to type S, the personal construction will always allow a quasi-agentive reading of the NPa, a crucial fact in the context of growing overall transitivation in the grammar. It is again to be noted that this syntactic paradigm cannot be given content in proper semantic terms, but remains - apart from the possible change in semantic role properties - largely a stylistic device. We have noted that type S can give rise to ambiguity already in OE, in particular with proper nouns, but also with full noun classes. The co-occurrence of S and the personal variant in the same grammar does not affect all structures equally: it will first render ambiguous the S-reading of sentences with proper nouns (St. Peter reweth that...), then full nouns with articles $e king reweth that...). However, it will not affect pronominal animate-NPs (him reweth that...). This last point is not unimportant in the present context, because RUE verbs - as all experiencer predicates - are inherently associated with first and third person subjects. Pronominal NPs will thus be more productive than (third person) full nouns, especially in the spoken language. With non-pronominal NPs the functionality of the variation pair S/personal construction is wholly dependent on case inflection, and ambiguity can arise easily in syntactically indeterminate sentences like (21): (21) fre fend shulde shame heere to seye pat... (Wy.418,27) Whether (21) should receive a personal or type S interpretation can only be decided on the basis of plausibility arguments from text-internal use. In the present case a personal reading is unlikely, because the construction is found nowhere else with this author. The kind of ambiguity arising through case opacity was not disambiguated by exclusive use of the personal variant. All through the ME period type S- and the specific verb-NPa relation associated with it is most common with RUE verbs. From the 14c. onwards this relation (and the recipient role of the NPa) was given more explicit syntactic form by the /f-variant. It is in this sense that the semantic structure of RUE predicates can be said to have determined the pathways of syntactic development. The obsolescence of type S in the 16c. is certainly due to a considerable degree to morphological neutralization of the nominative/dative distinction in ME full nouns. We 90

should not forget, however, that this distinction remained fully functional with pronouns - and that pronominal NPa are typically associated with RUE verbs. (The most telling example for the correctness of this assumption is the syntactic history of seem and think, dealt with at the end of this chapter.) Moreover, it seems that when the /'/-variant was established, there was less resistance to giving up the traditional type, as the original functional and semantic role properties were continued by the new syntactic expression. The syntactic history ofshamen, greven and forthinken (see below) illustrates just this pathway of development. That the opposite direction of development could also be followed is shown by the history ofrewen before its obsolescence in construction with sentential complements. There is a complete 'personalization' to the exclusion of the it -variant in the 17c. Rewen does not disappear from the lexicon in the 17c. With nominal complements it is attested until the 19c., and the present result will be complemented by this type of evidence, in section 1.2. The case is again different with greven and forthinken. In the 14c. and 15c. they show a similar syntactic paradigm to rewen, with the variation pattern formed by S and the ittype. With both verbs, however, the personal variant is only sporadically attested and (with greven) not found after the 16c. But note the occurrence of the type lam grieved, with sentential complements up to the 18c.: He was grieved to see ... (18c. OED, the type with prepositional complement is also met with in the 19c.). After the 16c., however, the //-construction is the sole variant used with proper sentential complements. In the 17c. and 18c. forthink is found in personal constructions with a relative clause: ... forthink what she had done (18c. OED). This construction is included as evidence for the transitive type (i.e. with nominal complements). With these verbs, then, the personal variant is never integrated into the paradigm of the construction with sentential complements. The 16c. marks the turning point also in the history of the lexical valency of shamen. While S is no longer attested we meet the i'Mype for the first time. But the 16c. variation pattern between the personal and /'/-construction is not preserved, as the former becomes obsolete in the same century. Far from 'merging' into the personal variant, then, the original subjectless type finds its grammatical relations continued in the syntactically acceptable /'/-construction. Again it is not possible to account for this development in terms of individual semantic verb structure. It is clear that the use of RUE verbs in the /'/-construction does not provoke any semantic change, and that syntactic variation does not reflect any semantic restructuring taking place in the lexicon. Whereas the situation can be described in terms of semantic roles (recipient vs. quasi-agent), and syntactic functions (indirect object vs. subject), a causal explanation of the different pathways followed by RUE predicates is not available. As far as general syntactic targets are concerned, both the personal construction and the /'/-variant satisfy the verb-second constraint associated with SVO syntax. We will return to this issue in the light of evidence from the type with nominal complements.

91

To sum up the ME syntax of RUE verbs in construction with sentential complements, we note that of the predicates which occur in this way with any frequency, all enter all three construction types. This is in line with the initial observation about the semantic structure of RUE verbs, which allows the animate-NP to function as experiencer/recipient or quasi-agent, expressed, on the level of syntactic functions, by subject (personal construction), indirect object (/Y-variant) or pseudo-subject (type S). There is a transition period from the 14c.—16c., characterized by the apparently unconstrained possibility for most verbs to occur in all three types side by side. In the case of rewen we have the only example of development towards the personal variant as the sole representative in the syntactic paradigm. In terms of lexical specification, this development is reflected by a change in the nature of the semantic role of the NPa, from pure recipient to quasi-agent. It is also possible that the presence of the transitive lexical variants regret and repent contributed to this development of rewen. That, on the other hand, semantics can determine the opposite pathway of development is seen from the history ofgreven, which discontinues the personal construction, adopting the ;i-variant instead; in this way the original verb-NPa relation can be expressed as closely as possible within SVO syntax. Before dealing with the history (in the type with sentential complements) of the other verb classes in sections 2—5, let us now turn to the ME development of RUE verbs in types N, I and II. Widening the data base in this way allows evidence from the type with nominal complementation to bear on the history of subjectless grammar.

1.2

RUE verbs in construction with nominal complements

1.2.1 Type N Type N was the OE subjectless construction par excellence, with no nominative NP present. Apart from the di-transitive verb eglian (ME eilen} all RUE verbs were productive in this type, repeated here by (22): (22)

Ne fje hreowan frearfealles swa micles swa fru me sealdest (BTs)

Concerning type N in ME, Visser speaks of a radical break between Old and Middle English with regard to the use of this syntactic construction in general, due mainly to the fact that case ambiguity may arise with non-pronominal NPa and that some verbs of the semantic classes involved here take accusative objects, with resulting indeterminacy between genitive singular and accusative plural.1 This view is of course correct in a narrow sense, but not if we assume the original relational situation to have been adequately expressed by a prepositional object instead of the 1 Visser (1963, 358) with examples from Layamon. 92

inflectional 'causative' genitive. There is no reason to doubt that this was the case, as indicated by the variation, already in OE, between synthetic genitive and prepositional objects.1 The following chart lists occurrences in the ME type N variant of RUE verbs: TYPEN

OE

12c.

13c.

rewen ofpincen eilen shamen greven forthinken

χ χ _ χ

_ -

χ — _ χ _ χ

14c.

15c.

χ χ — _ _ χ χ _ _ _ χ χ

16c.

_ _ -

The 12c. gap is accounted for by the external reasons given in the last section. Up to the 15c. rewen, shamen ma forthinken are found in N-constructions: (23) (24) (25)

Himm reowepp off hiss a^henn woh (Orm 5566) penne scheomep me perwip (Marh. 34, 30) me forthynkith ofyoure displeasure (Mai. 66, 11)

With shamen and forthinken type N is only sporadically attested in ME; enough, however, to prove the continued use and grammaticality of this construction, despite growing pressure from normative syntax. The relative productivity of rewen in N shows that the recipient role of the NPa was regularly associated with this verb. Ofpincen, eilen and greven are not recorded in ME type N. For each of these predicates its absence in this construction has to be evaluated individually. OE ofpyncan does occur in N. In view of the continued occurrence of hreowan and scamian in this type, we cannot assume the absence of ofpincen to be due to a syntactic constraint. If its non-occurrence were to have systematic status, the reason would have to be a change in its role structure (the kind of semantic roles which it could take). However, in the light of its attested syntactic valence (see below) this view cannot be upheld. It appears, then, that the gap in the ME paradigm of ofpincen with nominal complements is a result of the weak data base available and cannot be linked to a systematic constraint - of syntactic or semantic nature - affecting this construction type. (Note also that the verb disappears from the lexicon in the 14c.). OE eglian, on the other hand, never enters type N, and its non-occurrence in the ME representative of this construction must be due to the semantically-based reasons given for its absence in OE.2

1 Cf. Oft frone gefryldegestan scamaf) paes siges (CP 227, 19) besides menn ... scamafr nu forgoddaedan... (W. Sermo62, 153). 2 See chapter II, p.64. 93

The example of greven and forthinken, with partially overlapping dictionary meanings, again shows the difficulty of inferring both semantic structure from the syntactic occurrence paradigm and relational valency from apparent synonymity. Greven does not occur in N because it lacks the specific object relation which characterizes this type. Whether the different lexical valency of greven and forthinken can be expressed in terms of proper semantic features will be discussed in the light of evidence from the full syntactic paradigm with nominal complements in section 1.2.5. There, we will also assess the disappearance of type N in the 15c. on the basis of evidence from the other verb classes.

7.2.2 Type! We have mentioned in chapter II that despite two word order variants the relation holding between verb and NPa is fixed in this construction, with the NPa functioning as indirect object, in the semantic role of recipient. If we look first at the subtype with // in subject position, we recall that in OE there was an absolute constraint on word order, with it obligatorily occupying initial position. There was thus complete parallelism between the syntax and the semantic reading, namely SVO. A survey of the it-V-NPa construction in ME shows the following picture: it-V-NPa

12c.

rewen ofpincen eilen shamen greven forthinken

x χ _

13c. x χ x _ χ -

14c.

15c.

16c.

17c.

18c.

19c.

x χ _ _ χ χ

x

x

x

x

x

_ _ χ χ

_ _ χ

_ _ -

_ _ -

_ _ -

(26)-(30) represent this type: (26) (27) (28) (29) (30)

and eure hit hem seal rewen (W 107, 22) pt eft hit him ofpinchep (Lay. 1680) pet hit ne eili me nawt (Marh. 22,14) hit ne greuep ham nawt (HM 9, 77) ... it me forthenketh (Gower, CA III, 630)

The OE word order constraint demanding it in first position remains effective throughout ME, although it is no longer absolute if grammatical relations are explicitly expressed (for example by a prepositional phrase), as in the following example with neden: (31)

to pee it nedip not (Cloud 125,19)

The productivity up to the 19c. of rewen in the type it reweth me shows one potential pathway, viz. continued occurrence in the OE type, which conforms to unmarked syntac94

tic SVO structure and does not involve any semantic change of the original 'subjectless' meaning. Similarly ofpincen and forthinken continue to be thus used, both disappearing from the lexicon in the century for which they are last recorded in this type. The single occurrence of eilen in the 13c. is difficult to assess. Again there is no question of a syntactic constraint on this construction, nor can we ascribe its apparently restricted occurrence to the relation it enters into with the NPa, as it freely occurs with its NPa in indirect object function (see type I below). This situation can thus not be accounted for in terms of use (as in the case of RUE verbs with sentential complements), as the occurrence of the pronoun is discourse-dependent and will only occur once the reason of affliction has been mentioned. Note furthermore that eilen, however semantically adequate in this construction, was only used in a very limited paradigm in ME, most typically in the frequent form what eileth you? The problem is different again with greven, which evidently allows the construction but is no longer used in this way after the 15c. This situation has to be accounted for in the context of the history of type I (see below). Note however the uninterrupted occurrence in the sentential type it greveth me that... until the 19c. Shamen, finally, never entered this type in OE. The specific relational properties holding with this verb have been discussed in chapter II and the issue will be taken up below in the context of its complete lexical valency. Note that the occurrence of shamen in it shameth me that... cannot serve as a relevant basis for comparison in the present case, as it is only the result (first observed in the 16c.) of specific development within the framework of the type with sentential complements.1 In isolation from the full paradigm of nominal complementation, the account of the type it-V-NP again emphasized the idiosyncratic behaviour of individual verbs, whether due to semantic structure (shamen) or external factors (eilen, greven). Contrary to type N, however, there is no pressure on the morpho-syntax, as the structure of it-V-NP fully conforms to the normative order in unmarked declarative sentences. Type I proper has two full nouns. A crucial aspect of this construction concerns the presence of one or two animate-NPs. On the basis of OE evidence it was noted that if there is only one animate-NP it regularly functions as indirect object in intransitive constructions (pact hreowej? me). As observed in chapter II this is a 'personal' construction in terms of syntax by having a nominative subject NP and showing verb agreement. By far the most frequent type is with an inanimate nominative NP (the source of affliction). With regard to word order such a structure will be in a conflict situation caused by the relative effect of the normative tendency favouring syntactically unmarked nominative NPs and the animateness target observed in OE, which favours animate-NPs in initial position. ME type I also occurs in two word order variants: la (32)—(35), with initial nominative NP, and Ib (36)-(38), with inverted order: (32)

and alle wordes him ... greuejj (A 142, 27)

(33)

A thing that myghte theforthynke (Ch. TC 2, 1414)

1 See section 1.2.5.

95

(34) (35) (36) (37) (38)

pis gude lif schamis us (N. Ben. 47, 22) a that sor rewyth me (MK 50,27) te schal... bireowe pat sip (HM 11,88) pan sail pe noght rew pi trauayle (N. Ben. 63, 556) Gief him her ofpinchep his gult (Poema Morale (Trin. 225,166)).

To judge from the data available it appears that both word order variants occur with about equal frequency in ME. Again the fact that the Ib variant is preserved against the gradual establishment of SVO syntax shows that the indigenous nature of the verb-NPa relation continues to be realized during the whole ME period with the present verb class. Whereas in (32)—(35) syntactic order is in line with the functional reading SVO, in (36)— (38) it is inverted to the marked OVS form. However, given the changed facts of ME use, the la order of (32)—(35) can be interpreted as a quasi-transitive structure, with the animate-NP functioning as direct object. As shown by (32) plural inflection on the noun/ verb cannot decide the issue. If the NPa is a pronoun, variation between la and Ib order, as in (39) a alle wordes greuep him b him greuep alle wordes has the following effect. Both examples can receive a reading as type I, with OVS or SVO order respectively, but constant indirect object function of the NPa. (39)b on the other hand also allows a functional interpretation as a quasi transitive SVO structure. With full nouns, however, the situation is different: (40) a pe king greuep alle wordes b alle wordes greuep pe king Due to the state of case morphology, (40)a now is unambiguously a transitive SVO structure ; whereas the order of (40)b can be read as a transitive construction or as type I, with the king as direct or indirect object respectively. The latter interpretation, however, will only be favoured as long as the predicate in question is still used with indirect objects. The issue is thus again largely a matter of use, and individual evaluation will have to take into account the occurrence of prepositionally marked indirect object NPa besides ambivalent cases such as (40)b. The situation of indeterminacy created by case neutralization will not remain without effect on the relational properties in the original type I, but will, within the general framework of growing SVO syntax, be resolved more and more in the direction of a transitive reading. Grammatical indeterminacy even increases if both NPs are [+animate]. In this case the possibility of a type I reading is even more fully dependent on case inflection because 96

prepositional marking appears to be less frequently used with full nouns than with pronominal NPs; in the absence of a functional case system in ME an interpretation as type I becomes gradually more dependent on word order. The la sentence himgreveth the knight will still be read as type I under unmarked conditions (with the transitive reading with topicalized NPa a theoretical possibility). In the Ib variant, however the kinggreveth him, the unmarked interpretation will always be the transitive one. The same is true if two full nouns are present: the king greveth the knight is unambiguously transitive (allowing the marked reading with topicalized object, if forced). A type I interpretation, on the contrary, is only theoretically possible, as it would presuppose the intact use of the NPa as indirect object, a situation clearly not given. Note that text-internal evidence cannot decide the function of the NPa as indirect or direct object respectively. It is in the context of such fluctuation that the 'transition to the personal construction' appears in all its complexity - and historical reality. Even if the situational context will in most cases ensure the desired interpretation, the interplay between the varying factors involved (one or both NPs [+animate], possibility of a quasi-agentive reading, syntactic and functional order, pronominal or full NPa) renders this type a focal point of the change. This situation will have to be borne in mind when dealing with the syntactic history of type I below and in particular its co-occurrence with the genuine transitive construction. Next we need to present the occurrence properties of RUE verbs in type I and to discuss syntactic behaviour on the basis of individual predicates. The ME chart shows the following picture: TYPE I

OE

12c.

13c.

rewen ofpincen eilen shamen greven forthinken

χ χ χ -

χ -

χ χ χ χ χ

14c. χ χ χ χ χ

15c.

16c.

17c.

χ

χ

-

χ χ χ χ

χ χ χ

-

Ofpincen continues OE use until it disappears from the language in the 14c. Shamen highlights a case problem in diachronic syntax. The absolute non-occurrence of this verb in the form it-V-NP and up to the 15c. can hardly be accidental, but appears to be due to a semantically-based constraint on syntactic form. In view of the (single) 15c. occurrence in pis gude lif schamis us (N. Ben. 47, 22) the most plausible account of the historical situation would seem to be in terms of highly restricted use, possibly due to the strong 'causative' relation between the verb and the second NP in this construction type. As will be seen from the complete paradigm below (p. 104) shamen is productive all through ME in just those constructions which give explicit expression to this grammatical relation. In 97

this narrow sense, then, the development of the syntactic valency ofshamen can be said to be partly determined by a systematic semantic constraint. The fact that greven and forthinken are productive in type I from the time they come into the language indicates that the indigenous relation of RUE verbs to their NPa was that of pseudo-subject or indirect object (in the form of Ib and la). Rewen and eilen continue OE usage up to the 16c. (Eilen is regularly met in the type what eileth you?, an almost idiomatic structure last attested in the ISc.)1 The general factors leading to ambiguity and the conditions for reanalysis discussed above apply throughout. The reasons for the obsolescence of type I after the 16c. will be discussed in the context of the full lexical valency of RUE verbs at the end of this section. Let us now turn to the question of to what extent the possible variation pattern offered by the presence of type I determined the history of RUE verbs in construction with nominal complements. The variation possibility allowed by the constructions discussed so far, N and I, is of the kind illustrated by (41): (41) α me ruys of thy hurtys (Mai. 655, 26) b Him ruet alle his rees (15c. OED) In both examples the syntactic expression of the verb-NPa relation is the same, as is the recipient role of the animate-NP. However, what is unambiguously an object NP in N corresponds to the syntactic subject of I. In terms of functions then (41)fl will receive an SVO reading (the initial pseudo-subject is given the subject role because the postverbal NP cannot take this function). (41)6 on the other hand is an OVS structure in terms of syntactic functions. Even if (40)a and b represent close semantic paraphrases, then, there will always be this distinction in terms of functional interpretation. What combines both examples in (41) is the animateness target, which in the b -sentences overrides the demands of unmarked syntactic SVO (offered by la order). Only rewen and forthinken integrate this variation pattern into their paradigm of nominal complementation.

1.2.3 Type II The next syntactic variant to be dealt with is characterized by the presence of an animateNP in nominative case and subject function and the genitive object of type N, as in OE he hreoweb baere daede. The following examples illustrate type II in ME: (42) (43) (44)

For drighten luue, yee reu o me (CM 24564) bai shame more with a foule cloth o baire body ben ... (Rolle, quoted after vdG 160) Forbenkeb nat of ^oure almes^yuyng (HS 6906)

1 Post-16c. records of this form are not included because a type I interpretation is no longer certain. 98

The diachronic situation is represented by the following chart: TYPE II

OE

12c.

13c.

14c.

15c.

16c.

17c.

18c.

19c.

rewen ofpincen eilen shamen greven forthinken

χ _ x

_ -

x _ x _ -

x

x

x

x

x

x

_ x _ χ

_ x _ χ

_ x _ -

_ x _ -

_ _ -

_ _

In OE we only find rewen and shamen in this type, which remains a productive variant until the 19c. and 17c. respectively. The example in particular of rewen shows continued occurrence with nominative subjects and causative objects from OE to the 19c. Also, it appears that inflectional neutralization between genitive/accusative need not bring about a change to the transitive construction if the original causative relation associated with the verb was strong enough to be expressed prepositionally. The use of individual prepositions (generally of, but also with) allowed the continued specific expression of this relation. As mentioned, ofpincen becomes obsolete in the 14c. We noted in chapter II that its absence in this construction in OE could not be assessed with any certainty, although a semantically-based restriction could not be imagined because the verb does take genitive objects. Now in the light of the enlarged syntactic paradigm its non-occurrence in ME type II can be given a plausible reason: it was not the impossibility to occur with nominative NPa — a conclusion suggested by the OE data — which prevented ofpincen to enter II; rather it appears that the specific causative relation to its object-NP was lost - in favour, we assume, of direct objects (see 1.2.4 below).1 The extent to which the development of the simplex thinken (OE pencan and pyncan) contributed to this change in valency is impossible to assess. The complete absence of eilen in this type is probably due to the reason mentioned for its non-occurrence in N, viz. the lacking causative element. It appears quite clearly from the survey chart below (p. 104 ) that type I was the indigenous ME representative in the syntactic paradigm of eilen; in this construction the source/cause is in nominative case. Forthinken also takes nominative animate-subjects since its first occurrence in the 13c. Its use in type II is restricted, but the fact that the causative object relation is independently present, from the 13c.-15c. (in type N) indicates that the thinly attested evidence cannot be due to a systematic reason. The absolute non-occurrence of greven in this construction, however, again suggests a more systematic account. The reason for its absence cannot be due to a constraint on its capacity to occur with nominative subject NTs, but seems to be due to the semantic structure, which does not allow the explicit 'causative' object relation associated with 1 A further possible reason is that the double occurrence of of was avoided.

99

type II. Again the apparent semantic overlap (in terms of dictionary meaning) with rewen does not warrant inference as to identical syntactic valency. Type II combines the demands of SVO form with the animateness target in intransitive constructions. That this construction, despite its syntactically unmarked form, was not generalized as a productive variant with the RUE class is due to the fact that only the semantic structure of rewen, shamen and forthinken favoured the explicit syntactic expression of the original causative object relation. The presence of type II in the paradigm of constructions with nominal complements now offers a variation pattern concerning the role features of the NPa (whereas variation between N and I involved the second NP). We have thus side by side type N and II sentences such as (42) a he reweth of the deed b him reweth of the deed As ME can only express this distinction with a limited set of NPa (the class of pronouns), the functionality of the variation between II and N would seem to be weak. However, we have already had occasion to note that the predicates involved here can determine syntactic behaviour which violates general grammatical tendencies, such as the development of functional SVO and the transitive use of ditransitive verbs. (42) does not involve any distinct semantic change, as both verbs receive an intransitive reading. Likewise an analysis of syntactic functions leads to identical results: clear SVO in (42)a, and SVO with initial pseudo-subject in (42)b. In terms of semantic NPa roles, however, the distinction can be appropriately made: whereas in (42)b the NPa is recipient, it is imagined as a quasi-agent in (42)a. Those RUE verbs which enter into the causative relation with the NP, viz. rewen, shamen and forthinken generally show the variation pattern between II and N, which is further support for the distinctive value of the verb-NPa relation in the latter type. (43) is an example from the 15c.: (43) a me forthynkith of your e displeasure (Mai. 66, 11). b I forbynk sore of hir dede (15c. MED) As far as the NPa is concerned the kind of variation in (43) corresponds exactly to that observed with the syntactic pair S and personal construction, illustrated by another 15c. example: (44) a me shamed ...to haue more adoo with you (15c. OED) b Schamys thow nocht that thow neuirgeit didgud? (15c. OED) It is true that the opposition in (44) is often neutralized with non-pronominal NPs in ME. But it has to be noted again that the verb classes concerned here are inherently associated with pronominal NPa, a factor which contributes to the continued use of N side by side 100

with II until the 15c. The structure of type II offers the natural transition to the transitive construction, dealt with in the next section.

1.2.4 The transitive construction The only apparent evidence for the transitive use of a RUE verb in OE was eglian. baet he us eglan moste (Judith, ASPRIV, 11) is indeterminate between the transitive construction and type I. But even here a transitive interpretation is unlikely, as the OE verb is 'chiefly used impersonally' (BT). There is no unambiguous case of an accusative object with eglian. On the contrary, where inflectional endings are distinct, the person affected is regularly in the dative. This suggests that we interpret us in the sentence above as an indirect object. The meaning of OE hreowan was 'to rue, grieve, affect with sorrow'. It is thus not the dictionary meaning per se which disallowed RUE verbs to be used transitively, but the regular association with oblique animate-NPs. The proper semantic reading is thus not affected by the kind of relational indeterminacy arising from a neutralization of the distinction between intransitive type I and the transitive construction. Shamen and rewen do occur with nominative animate-NPs in OE, but only in intransitive construction, such as hie ne magon ... hreowan (BT). There is no genuine evidence in OE of transitive use as in a putative *he hreoweb ba daed. It was noted, however, that already in OE neutralization of the genitive/accusative distinction may arise, although it will have no decisive effect (yet) on the relational interpretation of the object within the context of uniform subjectless use. Clearly we have to assume that co-ordinated structures of the form met earlier,1 Ic earn eac synful and myltse behofige ... (Lives I, 82, 557) would also lead to ambiguity with RUE verbs. Imagine hreowan in a putative sentence Ic earn synful... and pa cwala hreowe. It was mainly for this reason that the OE situation was characterized as potentially allowing the transitive use of RUE verbs. Actual attested use however showed that OE did not realize this possibility, but expressed the object-source by the 'causative' genitive. In the few 12c. texts we come across more widespread ambiguity between intransitive and transitive use caused by advanced inflectional decay. Relative clause constructions such as (45) and (46) from the Old English Homilies illustrate a typical source of indeterminacy: (45) (46)

Armheorted is be man be swibere reoweb his sinne (Trin. 95, 29) Mildheorted beb be man jje reoub his nehgebures unselbe... (Trin. 95, 33)

In these examples rewen can receive an interpretation either as intransitive type I or as a transitive construction with his sinne and his unselbe as direct objects. Whereas similar cases of formal indeterminacy in OE can be assigned the correct reading on grounds of general semantic use ('RUE verbs do not take direct objects'), this is no longer possible in the 12c. because in a general climate of transitivation such structures will be open to a 1 Seep. 75.

101

reanalysis in the transitive direction. Even if there are no unambiguous examples for the transitive use of rewen (or any other RUE predicate) in the same text, we have to assume this possibility as equally likely as the regular OE interpretation. A further argument speaking for the possibility (if not probability) of transitive use in the above 12c. sentences is a posteriori, but nevertheless obviously valid. Already the earliest 13c. texts show transitive use fully established and generally used as a variant alongside the traditional types. This can certainly not be explained by assuming radical change between the 12c. and 13c., but as a result of gradual development, determined by such factors as the loss of case distinctions and the general context of transitivation, which gave rise to more and more sources of ambiguity between intransitive and transitive use. There is a second possible transition from an OE type into the ME transitive construction: if not prepositionally marked, the genitive object of II will often coincide with the accusative form of direct objects1. Within the overall form of syntactic development in ME the most likely way to disambiguate such structures is again by reanalysis as transitive ones. Whatever the actual 1 Ic. and 12c. use was, the transitive construction in the 13c. is fully productive with RUE predicates, as appears from the survey chart below. The occurrence properties of RUE verbs in unambiguous transitive construction of the type he reweth his sins are the following: TRANSITIVE

OE

12c.

13c.

14c.

15c.

16c.

17c.

18c.

19c.

rewen ofpincen eilen1 shamen greven forthinken

— -

-

X

X

X

X

X

X

X

X

X

X

X

X

(χ)

(χ)

M

(χ)

-

X

X

X

X

X

X

X

X

X

X

X

X

X

X

X

X

X

X

X



The transitive type is represented by the following examples: (47) (48) (49) (50) (51) (52)

do ase def) pe pettican. ofpunche hit swupe sone (AR 51, 37) gifpou forpenke a mannys prowe (HS 3931) Greui nel ich f>u no^t (SEL 216, 69) Wit pis word scorned (CM 17429) gifhe... at st nde & halt on to eili pi flesch (HM 67,711) That thai sail rew thar barganyng (Bruce, XII, 253)

In the light of general transitive use of RUE in the 13c. we can plausibly infer that the 12c. was a transitional period, characterized by the kind of indeterminacy present in (45) and (46) or arising through neutralization of the case distinction between II and the transitive construction. In the following summary the development of the different construction types will be discussed in context. 1 See Visser (1963, 356f.) 2 Post-16c. entries stand for the type what ails you? 102

7.2.5 The development of the ME paradigm with nominal complements In this section we will first illustrate and discuss the complete range and the nature of variation possibilities offered by the syntactic paradigm with nominal complementation. Two variation patterns have to be first distinguished, (both are continued from OE). With regard to the animate-NP the constructions involved are N and II, with respect to the second NP types N and I. In the 14c. and 15c. the variation pair N and II allows the occurrence side by side of sentence pairs such as (53) and (54): (53) a b (54) a b

of Havelok rewede him ful sore (Hav. 503) ο me he suld reu (CM 18419) me rewyth of thy hurte (Mal. 665, 26) thow oghtist rewe on it (15c. vdG 63)

Both syntactic types can express the same factual situation, the only difference being the grammatical relation between verb and animate-NP: the NPa is recipient in the a-, but quasi-agent in the o-example. Now the availability from the 13c. onwards of the transitive construction offers a new pattern of relational variation to be realized. In the pair formed by type I and the transitive construction, the nature of the variation concerns the NPa roles as in (53) and (54) above, but now there is also a change in syntactic functions: the α-examples receive an OVS, the -sentences an SVO reading: (55) a b (56) a b

him reuys his sinnes sare (CM 28676) thou shalt it rewe (15c. vdG 63) pan sail be noght rew bi trauayle (N.Ben. 63, 556) yee shul reewe it deere (15c. vdG 63)

In pairs formed by type II and the transitive construction, on the other hand, it is the object relation which is concerned. In (57)a and (58)a the causative relation is expressed by a prepositional phrase: (57) a b (58) a b

o me he suld reu (CM 18419) He may that werre sore rewe (Gower, CA, Prol. 1004) Thow oghtist reewe on it (15c. vdG 63) thou shalt it nat reewe (15c. vdG 63)

As appears from the chart for type II above, this construction, though syntactically possible, is only productive with verbs with a strong causative relation to the second NP. Note that although II is a likely source for transitive reanalysis, there is no change in any sense from this type into the transitive one in ME, as the genitive can be rendered prepositionally if necessary. 103

The following survey chart shows the syntactic valency of OE and ME RUE verbs: OE

12c. 13c. 14c.

rewen

ο®·

·

ΟΘ·Δ Ο®·Δ ΟΦ*Δ Θ·Δ

shamen

ο®



οθ

· -

ΟΔ ·Δ ·Δ ·Δ

forthinken greven eilen ofpincen ο: ·: ©: Δ:

· ο·

ΟΦΔ

15c. 16c. 17c. 18c. 19c. ΟΘ·Δ ΦΔ

ΟΦ·Δ ΟΦ·Δ ·Δ ·Δ ·Δ ·Δ ·Δ ·Δ ·Δ ·Δ

®Δ ®Δ ΘΔ ΘΔ

Δ

Δ

Δ Δ

Δ Δ

Δ

Ν (him reweth of NP) I (him reweth NP) II (he reweth of NP) transitive (he reweth NP)

The most prominent feature of the syntactic paradigm of RUE verbs is the complete absence of type II with greven, eilen and ofpincen, due to the individual reasons discussed in section 1.2.3. Als already noted in the individual occurrence charts, there is a large gap in the 12c. column. This is entirely due to the data situation, which renders any entry quite arbitrary. Nevertheless, the scant 12c. evidence which is available cannot be excluded because of completeness and because it indicates the first traces of structural indeterminacy. Now from the 13c. the variation paradigm is complemented by a further possibility, the transitive construction. In the 13c. and 14c. all RUE verbs are thus used, a fact which may again indicate how readily a re-interpretation from recipient to quasi-agent status of the NPa was adopted by the present set of predicates. As has been noted above, it is not only the inflectional neutralization between genitive/accusative (i.e. between type II and the transitive construction) which can be made responsible for this development, but also the general syntactic and semantic factors of SVO order and transitivation. A first step had been the use of RUE verbs with nominative animate-NP, which opened the way to a reanalysis in the agentive direction. Another crucial aspect is the pathway of change via the type la (NP reweth him, see below). With the restrictions discussed in the accounts of individual verbs above, RUE predicates thus show the complete variation pattern with regard to both associated NPs from the 13c.-15c. The second major break occurs after the 15c., when type N disappears. This construction is under normative syntactic pressure during the whole ME period, but is preserved by rewen, shamen and forthinken until the 15c. As appears from the chart, the consequence of this formal demand for type N is not a change in the semantic roles of either NP, ie. a merger with either I or II, but more and more restricted use in highly marked contexts. Furthermore, the fact that the verbs in question tend to occur with first person pronominal subjects (or third person pronouns into whose bearer's mind the speaker/author can enter freely) helps to keep this type distinct as a syntactic construction. The continued 104

use of type II shows that it is not the obsolescence of the causative object relation which causes N to disappear, but the fact that this is the only construction without a nominative NP present. The general conditions for the re-interpretation of the initial animate-NP, mentioned in the context of type I, also apply in the present case. Although the clear break after the 15c. suggests that the structural requirement to have nominative subjects changed its status from an optional tendency to a much stronger constraint in the 16c., it is not the case that nominative subjects are obligatory, as the type with sentential complements shows. It is true that coalescence between N and II is possible from the 12c. with full nouns (the king reweth of NP), but the historical situation does not allow for an account in terms of N 'changing into' or 'merging with', II. Both constructions co-exist for centuries, and the loss of N cannot be directly linked to II, but is due to the general form of diachronic syntactic development mentioned as well as to more and more restricted use of the original pseudo-subject relation. To assess the full weight of this apparent break at the end of the 15c., we have to wait for further evidence from the other semantic verb classes. Even in the 16c. the direction of development remains unclear, but in the 17c. the third major break becomes visible: the loss of type I, with the result that only the transitive construction (and type II with rewen and shameri) continues the rich syntactic paradigm of ME. The obsolescence of I during the 16c. cannot be explained on syntactic grounds alone, as there is always a possibility to satisfy unmarked SVO order (in word order type la). The ME representative of the regular OE form (word order type Ib) is first of all up against the syntactic tendency to have nominative subjects initially and the animateness target. If the nominative NP is [—animate] inversion is of course possible, because relational confusion cannot arise: (59) (60)

& that forthinket h me (Mand. 220,13) ... hoot that myght hem greue (15c. MED)

Structures like (59) and (60) however serve as a bridge for a transitive re-analysis of grammatical functions, the NPa being open to an interpretation as direct object. One immediate consequence of such a change is that the occurrence properties of the NPa in initial position are severly restricted, i.e. word order variation for purely stylistic reasons as observed in subtypes la and Ib is no longer possible. We have to assume that in late OE and early ME an interpretation as type I was still ensured, as both word order variants are regularly met. However, subtype Ib (him forthinketh that) is an OVS structure and thus has no place within SVO syntax except as a highly marked variant. At a time when the transitive interpretation has been generalized (in the late 16c.) the original Ib variant will be reanalysed as a transitive SVO structure with topicalized object (him greveth his arm). Before that time, however, we can never be sure that a transitive reading is intended, because the accusative case of the object cannot be expressed unambiguously. 105

It would seem that the first predicates to invite such a «analysis will be those which (optionally) take accusative NPa in OE (such as scamian, lystan, langiari). If the type I word order of sentences like mec longade, hine lyste, hy scamajj (plus postverbal nominative NP) is inverted, the structure becomes indistinguishable from a transitive one. Subtype la, on the other hand, (his arm greveth him), becomes positionally and morphologically identical with a transitive structure and will give rise to more and more cases of syntactic indeterminacy. Already in the 15c., and certainly in the 16c. it appears that in ambivalent constructions the original type I interpretation only occurred if forced by the situational context. Otherwise the original indirect object will be reanalysed as the direct object of a transitive construction. The loss of type I during the 16c. thus has to be seen as the consequence of the interaction of multiple causes, morphological, syntactic, semantic, and communicative. If the subject is [—animate], it will receive a quasi-transitive (still non-agentive) interpretation. If both NPs are [+animate], word order type Ib (him greveth the^king) will hardly arise if the unmarked reading is the transitive one. A putative sentence in la form, however, (the king greveth him), will be re-interpreted also in terms of semantic roles, by means of a change of the recipient to agent status of the king.. To show that this development is inherently associated with the transitive use of RUE verbs - not accompanied by a change in their semantic structure - is our aim in the next section. After the 16c., RUE verbs only occur in syntactic structures of the quasi-transitive kind with non-human pseudo-agents (NP grieves him) and the transitive he gieves NP. What is preserved is the variation with regard to the object, with type II still used for the explicit causative relation. The overall result of syntactic development was thus a re-interpretation of grammatical functions and semantic roles: the original recipient pseudo-subject is lost and the agentive nominative subject substituted for it. The development towards a transitive paradigm can be well illustrated on the basis of the development of rewen and shamen. The lexical valency of rewen from OE to the 19c. is the following: REWEN

OE

12c.

Type N Type I Typell transitive

x x x -

x -

13c. x x x x

14c. x x x x

15c. x x x x

16c. x x x

17c. x x

18c. x x

19c. x x

With the disappearance of N in the 15c., the 16c. preserves the variation pattern for both NPs (pairs I/II, transitive with regard to the NPa, and I/II/transitive with regard to the second NP). After the loss of type I, however, the NPa can no longer appear as recipient. What is optionally preserved is the presence of the causative object relation. With shamen the individual pathway of development is slightly different, but the factors involved remain the same: 106

SHAMEN

OE

12c.

Type Ν Type I I transitive

χ — x - -

13c.

14c.

15c.

16c.

17c.

18c.

xxx x —

x x

x x

— x x

— x x

— x

19c. — x

Shamen only enters the pseudo-subject relation to the NPa in the form of type N. When this construction becomes obsolete in the 15c., the animate-NP is generally a nominative subject and the variation pattern is limited to the object relation: it can express the 'causative' variant by type II (up to the 17c.) or the direct object relation by means of the accusative. Contrary to rewen, then, the relational contrast involved does not appear to have been made full use of with shamen, with the result that II is abandoned after the 17c.

2.

The PLEASE/DESIRE class

The ME predicates of this class form a semantically less coherent set than the RUE verbs. This situation suggests a different arrangement from the one followed in the preceding section. It seems better to deal with the ME history of liken, langen and listen in turn 1 , discussing both complement types together.

2.1

liken

2.1.1 In construction with sentential complements OE (ge) lician 'please' exclusively occurred in type S, repeated here by (61) (61)

sibfran gelicade eallum foleum baet... (Or. 106)

This usage is attested until the 16c. The first occurrence of the //-construction (absent in OE) is met in the 13c. text of Layamon's Brut: (62)

hit be likede wel pat fru us adun laeidest (Lay. 4362)

(62) remains the only example of this type in my 13c. material. Nevertheless we have to assume that from the 13c.—16c. liken enters the variation pattern represented by S and the it -construction: (63) a b (64) a b

hem lyketh to be dene (Ch. WB 97) It liketh him at wrastlyngfor to be (Ch. PF 165) mee likes... go see the hoped heaven (Tottel's Misc., quot. after Vi. 28) it liked him ... to bring... (More 307)

1 (ge)lustfullian is not recorded after the OE period. 107

The occurrence of these constructions over such a long period again indicates that both structures represented indigenous syntactic types, with distinct stylistic value. In the semantic context of this verb, it is especially important to note that the NPa remains a pure recipient in both cases; the difference is not describable in terms of semantic variation. In so far as the grammatical status of type S is concerned, it unambiguously continues the traditional construction if the NPa is a pronoun. With full nouns, however, this structure can again give rise to indeterminacy: (65) (66)

God liketh nat that Raby men us calle (14c. MED) His grace liketh not that... (More 27)

The syntactic form of (65)—(66) is ambiguous between S and the personal construction. But to what extent the construction is open to a semantically personal reading (with NPa subject) can only be assessed in the light of the 14c. and 15c. use of liken. In contrast to RUE verbs, liken only sporadically occurs with nominative NPa before the 14c., typically in absolutive constructions such as (67): (67)

Amidde pe redunge hwen pe heorte likep kimep up a deuotiun (AW 148, 16)

From this time the verb is used in the meaning 'wish,choose;like', besides 'please' (MED). The ground is thus also prepared semantically for a personal interpretation of the type represented by (65)—(66), although it is often impossible to be sure about the intended meaning. In the 14c. and 15c. verb agreement often disambiguates otherwise indeterminate structures: (68) (69)

^e pat liken in love swiche fringes to here (14c. vdG 68) Men lykyn lestis for to here (15c. MED)

Only by the 16c., when S becomes obsolete, can we be more confident about the personal interpretation of ambiguous constructions such as (70)

Our Saveour lyked to slepe (16c. MED)

The survey chart for liken shows its occurrence properties from OE to the 19c.: LIKEN

OE

12c.

Type S It Personal

x — —

x —

13c. x X —

14c. x X x

15c. x X x

16c. x X x

17c.

18c. 19c.

— x

— — x

— x

From the 14c. liken can potentially occur in the three-way variation pattern illustrated in the previous section for RUE. The crucial difference is now that the use of the pair it108

and personal construction allows a maximally distinct relational and semantic variation to be expressed, viz. the alternation between recipient/agent roles of the NPa and the dictionary meanings 'please/like (wish)'. Up to the 16c. type S is the regular representative of the construction with sentential complements. The disappearance of this type after the 16c. is exactly parallel to the situation met with RUE verbs and is due to the same factors: the general syntactic development towards the unmarked SVO form of sentences, in which the subject is expressed by the nominative case. It has to be stressed again that there is no evidence at all that semantic use contributed to the loss of this construction, i.e. that semantic change preceded syntactic change. The /Y-construction, on the other hand, is never frequently used in ME. There can clearly be no syntactic reason for its obsolescence, and its non-occurrence after the 16c. has to be accounted for specifically on the basis of the lexical valency and the semantic structure of liken. As long as the verb is used with indirect object NPa in the meaning 'it pleases NPa that' ... respectively 'NPa is pleased that...', the ff-construction is an adequate syntactic expression. It seems, however, that when S was given up for syntactic reasons, the further pathway of development was not towards the ^-construction, which was never a productive variant, but that the transitive use and the meaning 'like' were generalized. Whereas syntax was responsible for the obsolescence of type S, the semantic change associated with this development prevented the /r-construction to 'take over' the expression of the original relational situation. Again in contrast to the case of RUE verbs, the disappearance at the same time of both variants in which the NPa is not the subject is intimately linked to the semantic restructuring of ME liken, from the dictionary meaning 'please' to 'like'. The history of liken receives more content if seen in the lexical context ofplesen. From the 14c. this verb is used as a synonym for liken 'please' in all constructions with an indirect object NPa. In addition to the external reasons mentioned above, the fact that plesen could take over these functions makes the semantic change undergone by liken easier to understand. Compare the occurrence of the two verbs side by side in construction with sentential complements:

TypeS It Personal

14c.

15c.

16c.

17c.

18c.

19c.

X

X

X







xo

xo xo

xo xo

0

xo

ο xo

ο xo

X

x: liken; o: plesen Plesen is also used transitively from the 15c., in the meaning 'wish, desire'. The crucial fact in the present context is that from the 14c., when the transitive use of liken sets in, the syntactic paradigm ofplesen offers the /f-construction to be continued:

109

(71) (72)

it plesyd owr Lord to visityn hir (MK 140, 34) yf it shall please hym thus to loke on us (Fisher 121,4)

This is the first case, then, where the original relational situation and the intransitive meaning of an OE subjectless verb is preserved by a new verb in the English lexicon. The development of liken with nominal complements will offer crucial evidence with regard to the lexical history of this verb.

2.1.2 In construction with nominal complements The sole representative of this construction type in OE was type I, repeated here by (73) and (74): (73) (74)

Gif...pe heora peawas liciap (B. 16, 3) [Dauid]... Code gelicode op his lifes ende (Lives I, 386, 32)

OE lician, then, does not enter the syntactically subjectless construction (N) in this complement type. The relational situation associated with type I, however, is the one met in genuine subjectless syntax, viz. 4x pleases to NPa', with the latter in the recipient role. As has already been noted, the grammatical structure of type I with two full NPs is under the influence of various conflicting factors, belonging to different levels of description. Liken continues type I unchanged all through ME. It has been mentioned that there was only one possible word order for the subtype with a pronominal it subject, namely it-V-NPa. This formal constraint serves as a valid parameter along which to describe syntactic development. The ME data show that /'/ liketh him is continued as the unmarked order, thus ensuring the desired (intransitive) reading: (75) (76) (77)

it him on sume wise likep (Trin. 103,9) hit mei lutel like godd (AR 121,18) for pat it misse-likip hem (Wy. 338, 24)

From the 13c. onwards, however, the NPa can precede it, with the result that a type I interpretation is only unambiguously clear with pronominal NPa, as in (78)

swa pat him bape hit wel likede (VV 119,17)

The postverbal NP is often marked as the indirect object by pre-positional form, as in (79) (80)

alsuo hit ne likep... to pe dyeule (A 187, 25) yifit like unto the (Ch. Bo. 4 p. 7, 1570-5)

Now if the NPa is given initial position in the sentence and case distinction is neutralized, the result is syntactic indeterminacy, as in 110

(81) (82)

Carrais ne likede hit naming (Lay. 272) if fortune it Iyke(l5c. MED)

Structures like (81) and (82) obviously give rise to relational reanalysis in the way observed for RUE: the original recipient NPa can be interpreted as the (pseudo) agent of a transitive construction. But whereas with RUE verbs this change did not cause radical semantic restructuring, the case is different with liken: if the inherent relational situation associated with the meaning 'please' is no longer possible, the dictionary meaning itself has to change. Its direction has been made clear in the context of sentential complements, viz. the transitivation from 'please' to 'like'. Type I with two full NPs continues to occur in both word order variants. (83)—(85) illustrate the construction with only one animate-NP, in which the intransitive reading is ensured through the presence of a prepositionally marked pronominal NPa: (83) (84) (85)

hem likede here löbliche sinnes (Trin. 7,11) bo him hise wise wel ne like (13c. MED) He doth all that hym lyketh (Merlin 1, 8)

Indeterminacy arises in the following cases, which can receive an OVS (type Ib) or SVO (transitive) interpretation: (86) (87)

Oonan, pe hinges neueu, ne likede nogt pis... (14c. MED) Roulond likete pat wordfol wel (14c. MED)

Another early source of indeterminacy arises from co-ordinated and relative clause construction: (88) (89) (90)

Mildheorted hep pe man pe reoup his nehgebures unselpe and likep here alre selpe (Trin. 95, 33) ... pe gie michel... luuiep and likep (VV 69,15) This is my loued sone that lyketh me (15c. OED)

The example from the 12c. Trinity Homilies (88) is the earliest case of ambiguity with liken. As long as we cannot be certain about a transitive interpretation, however, the original type I reading is more likely. In (88) the intransitive or transitive use of liken is also partly determined by the use of rewen, which could not be conclusively decided but appeared to be intransitive.1 Furthermore, the fact that liken is not attested in unambiguous transitive construction before the late 14c. (also in the type with sentential complements) speaks against such early transitive use. Whereas (90) is disambiguated by common external knowledge of the facts, sentences like (89) from the 14c., remain ambivalent. The important thing to note is that from the 12c., the syntax of type I will potentially allow a reanalysis of NP-roles also with liken. Type I in its indigenous form (Ib order) 1 Cf. example (46) p. 101. Ill

remains productive up to the 19c. (see chart below), due to preserved case distinction with pronouns or the possibility of prepositional full nouns and the semantic change which clearly takes place between the two readings. As the following 17c. and 19c. examples show, it is usually the case that the situational context makes the desired reading clear: (91) (92)

How that way may like you (17c. OED) a certain path that liked me not (19c. OED)

Turning to the word order type la (with the [—animate] nominative subject initially), we recall that the use of this form was apparently restricted in OE. Even in the 12c. we normally find this word order in the syntactic form of relative constructions only: (94) (93)

& gode peawes heo healt pe Code lichigep & mannen (E.E. Horn. 13, 8) pu hafast gecoren pone wer pe me wel Heap (Apol. of. T. 62)

Under the effect of the gradually generalized syntactic SVO pattern, la order becomes more and more common during the ME period. With only one animate-NP present, the intransitive type I interpretation is ensured: (95) (96) (97)

off all pan Drihhtin likepp (Orm 305) as oft as pe song likip me more pan ... (Wy. 191, 32) pi seruyse lykyth me wel (MK 18,32)

The productivity of this type is underlined by the occurrence, from the 14c. onwards, of the NPa explicitly marked by a prepositional phrase: (98) (99)

ase merci likep to god, alsuo hit ne likep ... tope dyeule (A 187,25) pe biddyng... ne likep not to God (Bk. of W 233, 23)

The crucial case, however, arises again in the presence of two animate-NPs, both of which can take the semantic roles of recipient and agent (receiver and giver). If both NPs are [+animate], an unambiguous functional interpretation with the NPa as indirect object is secured as long as the case distinction can be expressed. Otherwise the general conditions for reanalysis are given. Syntactically we again meet both word order variants, him I like and / like him. The latter form is an obvious pathway of development because it combines both the animateness target and structural SVO. We should not forget the pragmatic fact, however, that the present verb classes, and liken in particular, are typically associated with pronominal NPs. This will contribute to the continued use of Ib order and the original type I relational situation throughout the ME period: (100) ... pone wer pe me wel Heap (Apol. of T. 62) (101) gif me likes stalewurpe lefmon (Lamb. 273,11) (102) whanne hem likip... men (14c. MED) 112

The following chart represents the syntactic valency of liken in construction with nominal complements.1 LIKEN

OE

12c.

13c.

14c.

15c.

16c.

17c.

18c.

Type I transitive

x -

x —

x -

x x

x x

x x

x x

x x

19c. x x

The new lexical situation arising with the transitive use of liken in the 14c. has to be complemented by mentioning another fact: the addition of p lesen to the ME lexicon in the 14c. This verb, in the intransitive meaning 'to be agreeable to', occurs in type I, as liken does: (103) pis vice displeseb mache God (Bk. of W 47, 14) (104) And thes knyghtes... pleasith me more than ... (Mai. 98, 21) (105) puplesystme(MK32,23) (106) thow pleses Hym noghte (RR 29, 23) In contrast to liken, plesen appears to be attested only in word order type Ib, i.e. putative sentences of the form *him pleseth the song, with the NPa in subject position, seem to be heavily constrained. On the other hand, there is the variant with the indirect object function of the NPa explicitly marked by a preposition: (107) no ping plesip to a fool but folie (Wy. 268, 7) (108) yifthis mauere plese to ony worthi man (Mand. 94, 28) The same word order restriction holds for the type with it in subject position: (109) & pe betir it plesip hym (Cloud 89, 20) (110) yfit pleasid God (Cax. 23,13) This construction also occurs with prepositional marking of the NPa: (111) gifit plese more to god (Wy. 321, 8) (112) as it pleseth to our Lorde (Cax. 117,8) That the use of plesen corresponds directly to the intransitive use of liken is shown by the following examples from two different texts of the Wyclif Bible: as hit likep tope / / as hit plesith be (MED). In the 14c. and 15c. liken and plesen thus occur side by side in type I, a fact which highlights the close association of the semantice structure of liken with the indirect object function of its animate-NP in ME.

1 Liken does enter type II, but only in the meaning 'approve of (cf. OED like of). 113

2.2

listen (OE lystan)

2.2.1 In construction with sentential complements With sentential complements OE lystan 'cause pleasure, desire' exclusively occurs in type S. This construction remains the main representative also with listen 'desire, wish, choose; please' during the whole ME period: (113) (114) (115) (116)

us luste leornie hwu... (Trin. 199, 33) himm lisste pa wel etenn off an appell (Orm 282) ow ne luste nowper speoken ... (AW 43, 14) butyifhim liste to go ... (Mand. 173, 35)

The //-construction is first met in the 15c. only: (117) ... hit lyste me to com myselffoute (Mal. 1187, 20) (118) it liste hym ... to departe (Cax. 8, 8) This is a late date for the earliest occurrence of the syntactically unmarked type, and we assume the historical situation to reflect a systematic semantic constraint on the syntactic occurrence of this verb for the following reasons. On the one hand, general syntactic development makes clear that the it -construction was available as a syntactic possibility much earlier. The fact that it is not once attested in the rich 13c. and 14c. literature investigated1 can hardly be accidental, but appears to be due to specific use, which did not favour the pure recipient role and explicit indirect object relation to the NPa. It seems quite plausible to assume that the regular use in the meaning 'desire, wish; choose' readily allowed an additional quasi-agentive role to be taken by the NPa. Such a semantic double nature could certainly be associated with type S, but not with the itconstruction. A further indication for the plausibility of this assumption is the fact that the variant *it listeth to him, with the NPa in prepositional form, does not occur with this verb. The following 14c. and 15c. examples show the pronoun it as direct object: (119) Til pe Lorde ofpe lyfte hit abate (MED) (120) per was no man pat list it withseyn (MED) These semantic and syntactic aspects combined are enough to account for the restricted ME paradigm of listen. Even in the 14c. and 15c. the /Y-construction was never a productive type besides S.

1 Nor in the dictionaries and correspondences. 114

There is another, syntactic aspect which could have helped to restrict the occurrence of listen with it as subject. Note that this construction is syntactically congruent with a marked word order variant of S. Listen is a ditransitive 'equi'1 verb, which obligatorily takes infinitive complements. If used in the ^-construction the result can become structurally ambiguous in that the initial it may be interpreted as a proposed object of the complement clause: it list him to sing would be a viable word order variant of type S him list to sing it (i.e. the song). While it is true that the situation is underdetermined, it still indicates that although listen and liken are given as synonyms by MED, attested use shows them to behave differently with regard to the ME paradigm of sentential complementation. Now also in the 14c. we come across the first personal constructions: (121) bo makedon hi hope ioye & blis fiat luste (MS variant: lusteri) er bet to grone (SEL 578, 198) (122) whan we... liste to drinke (14c. vdG) (123) and thou lyst to fyght (Mai. 687,28) (124) if ye lyst of me to make... (Ch. RR 1967) Again personal use is prepared by the fact that listen takes nominative NPa already in OE: (125) bonne seo sawlbyrsteb and lysteb Codes rices (BTs) A 13c. example representing intransitive use is (126): (126) her he him losteb (A 246,14) The presence of an intransitive variant with the NPa in subject function makes a personal reading of the following syntactically ambiguous sentences equally possible as a type S interpretation: (127) gifa man list for tose... (Cloud 56,3) (128) many men ne list not to yeue credence... (Mand. 229,9) The same is true for the following 14c. centences with non-pronominal NPs: (129) whan that fortune list to flee (MED) (130) Of the myracle... Seint A mbrose... list to seye (MED) (131) Na creatur sal ban list to plai (CM 22601) That despite this potential source of ambiguity the semantic structure of listen freely allows the relational variation involved between type S and the personal construction, i.e. between the pure recipient and quasi-agent roles of the NPa, is shown by their occurrence side by side from the 14c.—19c.: 1 Verbs which govern so-called equi-NP-deletion obligatorily take infinitive complements by deleting the second occurrence of a coreferential NP. 115

LISTEN

OE

12c.

Type S i t personal

x -

x —

13c. x -

14c. x x

15c. x X x

16c. x x

17c. x X x

18c.

19c.

x x

x x

Contrary to the case of liken, the appearance of the personal construction in the 14c. does not follow a previous semantic change. The resulting variation pattern between S and the personal construction (him listeth that... /he listet h that...) was clearly distinctive enough also in semantic terms to be continued up to the 19c. The dictionary meaning 'wish, desire' can be expressed by both syntactic variants, but in type S the verb can additionally mean 'please'.

2.2.2 In construction with nominal complements The OE paradigm entered by lystan is N/II, characterized by the constant causative object relation and the variation between the recipient and pseudo-agent roles of the NPa. The exclusive occurrence of this relation was assumed to have been due to the same semantic reason responsible for the non-occurrence of type I in OE (*hine lystep ealu).A An important aspect determined by the semantic structure is the fact that lystan regularly takes one [+animate] and one [—animate] argument — the pseudo-subject and the object. The potential relational situation is thus essentially different ('simpler') than the one met with lician/liken. This semantically-based explanation of the restricted syntactic pattern of lystan/listen is supported by the ME facts. In ME, type N is continued, with the object-NP in prepositional form: (132) him lystep Codes andweardnysse (E.E. Horn. 99, 30) (133) A33 lisste himm äfften mare (Orm 10200) (134) ... to ping p us luste efter (AW 141, 15) Type II, on the contrary, is nowhere as productive as N, and rare before the 14c.: (135) sothlik til offrandes lustes pow noghte (14c. vdG) (136) One pe lawnde righte per pay lay... And liste no thynge of play (15c. OED) The scarcity of II also means that neutralization between N and II through lack of distinctive case marking (respectively verb agreement in the plural) did not put pressure on a possible reanalysis. In the context of the attested 12c. (and 13c.) paradigm, then, (137) will be unambiguously interpreted as a type N structure:

1 See chapter II, section 2.3.2 116

(137) pis hys god ta pan mann pe hure metes ne lyst (12c. MED) Type I, which is absent in OE, is first found in the 14c., although only in relative clauses such as al that hir list (MED) or in the type what him list (see below). There are no examples with full nouns in ME, possibly for the semantic reasons mentioned in chapter II 1 . Structures with it as subject are met at about the same time:2 (138) For me lyst hyt noght (Ch. HF 1564) (139) Whil it you list (Ch. TC 5, 1330) Clearly the word order in dependent clauses like (138) can give rise to a transitive reanalysis with /'/ as direct object. The same is true of the non-pronominal variant of what him list (eg. ... to do what him //si)3, a type quite frequently met in ME, besides as him list, when him list etc. The earliest examples of what him list are found at the end of the 14c. although the type is OE (ealpaet hine lysteh): (140) Tomorwe wall seye thee what me leste (14c. MED) (141) And what him liste he tok (Gower, CA I, 1922) (142) pe tale pat me lyste (14c. MED) Because of the unambiguous role distribution due to the presence of just one [+animate] NP, a type I interpretation (with the NPa as goal) is ensured in these cases. The same syntactic construction will of course become ambivalent if full nouns are substituted for the pronoun, as in a putative what the king listeth. 15c. evidence4 shows clearly that variation between type I and the form with nominative NP could arise: (143) offer or leeue, wheper pe lyst [var. thu list] (15c. MED) Note also that from the 13c. there is a variation pattern of the following kind: 14c. 15c. 16c. 17c.

when pam lyst (RR 45, 1)as pou lest (Rolle, OED) whar hir //sf(N. Ben. 85, 1351)as thei lyst (PL I, 238, 17) as me listeth (Tyndale OED) where he lysteth (Tyndale, OED) as they list (Life of M. 5 5, 14)

It is a fact that after the ME period listen is hardly ever used with nominal complements, although quite productive with sentential ones. The patterns continued are what/as/when him list . This situation makes it practically impossible (and uninteresting) to investigate the question of a possible re-interpretation towards transitive use. Significantly, however, there are no examples before the 16c. of unambiguous transitive constructions such as ""he listeth the mete. This cannot be explained in terms of the semantic structure of 1 2 3 4

Section 2.3.2 15c. copy of 14c. text. MED 'quasi-impersonal'; OED 'impersonal-transitive'. Possibly even earlier, because MED adds c. ? Ί 300'. 117

listen, but of specific use within its lexical field: the transitive verbs desire and wish were used in their particular meanings since the 13c. and OE respectively. Transitive use is most rare, and usually found in relative clause construction. The earliest example is from the 16c. and the construction is not found after the 18c.: (144) and bailed whom I list with my sword... (Kyd, Ard. 1936) (145) what they list (18c. OED) If we look at the survey chart below we see that the strongest characteristic feature of listen — its association with 'causative' objects — is given up in the 15c. for syntactic reasons. The syntactically unmarked variant to N, type II, which is met in OE already, is only sporadically attested, and not developed into the main representative, because the whole construction type becomes obsolete in the 16c., possibly due to the loss of the specific object relation. After the ME period, then, listen is only used in the set of rather stereotype phrases mentioned above. LISTEN

OE

12c.

13c.

Type N Type I Type I I 'transitive' ('what h e list')

x — x

x —

x x

-

-

14c. x x x

-

-

15c.

16c.

17c.

18c.

x x x

— x x

-

- - — -

-

x

x

19c.

x

-

The development of listen is thus not characterized by semantic change, but by a change in its syntactic valency after the 16c., when only the construction with sentential complements is used as a productive type.

2.3

langen (OE langian)

2.3.1 In construction with sentential complements The OE paradigm of langian 'to cause longing, desire' is restricted to type N with nominal and type S with sentential complements: (146) hine baes heardost langode... (BH 227,1) (147) Haeleb langode hwonne hie ...(Genesis ASPR 1,431) ME longen continues to occur in S: (148) hure longede wib hure broker speke (SEL 59, 12) (149) sore hym longed to wife how... (Merlin 233, 16)

118

Again syntactically ambiguous structures can arise from the 12c. onwards. However, the difference between a type S or a personal interpretation of (150) cannot be described in terms of dictionary meanings, but as the opposition between pseudo-agentive or recipient roles of the NPa met above: (150) pe cwen ... longede for to seon pis meiden (Kath. 1556) From a semantic point of view, then, such structures were under no immediate pressure of reanalysis. In the 13c. we find paraphrase constructions with nominative NPa: (151) he wes sory pe of-longed to his fader blisse (13c. vdG 64) The unambiguous personal construction is attested from the 14c.: (152) Thanne longen folk to goon on pilgrimages (Ch. Prol. 12) (153) scho ... langit for to heir ofhir weilfair (14c. vdG 65) Strikingly, the it -construction does not occur at all. Although sentences like it longeth him to go must be considered possible grammatical structures, they were not used, possibly for the same reason suggested for the parallel behaviour of listen. The syntactic history of longen with sentential complements is summed up by the following chart: LONGEN

OE

12c. 13c.

14c.

15c.

16c.

17c.

18c. 19c.

Type S personal

x —

x —

x x

x x

x x

— x

— x

x —

— x

Between OE and the 14c. type S is the sole representative. As with liken, the possibility for the animate-NP to be in non-subject relation disappears in the 16c.

2.3.2 in construction with nominal complements In construction with nominal complementation, longen continues the only representative of the OE paradigm, type N, up to the 15c.: (154) him wile sone lange par after (Trin. 148, 19) (155) Us longep to him (Trin. 131, 14) (156) Me longed after nouelrie (15c. OED) In the 13c. and 14c. the structure with nominative animate-NP is first met with, so that there is the variation possibility between subject and pseudo-subject function of the NPa, represented by types II and N:

119

(157)α pa longede swipe Luces pon kinge... of pan lauerd Criste (Lay. 5049) b Ich langy so swipe after Gor lays his wifue (Lay. 18918 Otho) (158^7 Us langis eftire a thynge ofpe world (14c. OED) b Heo longeden in here mod after oure Lord echon (SEL Prol. CV 36 (Eg.)) (159) shows how indeterminacy can arise between these constructions through lack of case distinction in the NPa: (159) pe soule langep perafter sore (SEL 476, 377) That the characteristic feature associated with this verb is the explicit (prepositional) causative object relation can be seen from the fact that, after the loss of N, type II is continued as the only representative of the nominal paradigm (see chart below). There is no transitive reanalysis of longen. Type I is only sporadically used in the 14c. and 15c.: (160) Horn longep tramtris pe trewe, for heled was his wound (14c. MED) (161) ... to seen his sustyr that her longeth so (15c. MED) It is not developed as a productive variant however, becoming obsolete before the 16c. for two possible reasons: first, the characteristic object relation is not expressed by type I; second, the nominative NPa in the role of agent is generalized from the 15c., which leaves no place in the paradigm for this construction. The development of the syntactic valency of longen is thus again semantically determined: the grammatically viable variant type I is discontinued for semantic reasons, whereas nontransitive type II — an SVO structure — remains in use. The survey chart reflects the skeleton of attested history: LONGEN

OE

12c.

Type I I Type N Type I

x -

x -

13c. x x -

14c. x x x

15c. x x x

16c. x — -

17c. x — -

18c. x — -

19c. x -

To sum up the history of these three verbs during and after ME, then, we note that in the nominal type the post-16c. development is towards the exclusive subject role of the animate-NP, in transitive construction with liken, in construction with prepositional objects with listen and longen. In the type with sentential complements the subject function is generalized after the 16c. with liken and longen, but the pseudo-subject nature of the NPa in type S is preserved with listen. Contrary to the paradigm of RUE verbs, then, the /Y-construction and the corresponding pure indirect object function of the NPa is not developed with these three verbs, a fact clearly due to semantic structure, which in all cases brought about a change towards agentive subjects. Only with liken did the development cause radical semantic change. Listen and longen, with the associated relational context of (pseudo-)agent/object 120

of desire were under no semantic pressure to change. That with listen the traditional verbNPa relation is preserved in a closely-knit syntactic paradigm (what/when/as him listeth), is another striking example of the effect on syntactic development by idiosyncratic use.

3.

The BEHOVE class

3.1

BEHO VE verbs in construction with sentential complements

It is necessary first to mention the lexical changes affecting this semantic class in ME. The OE verb (ge)dafenian no longer belongs to the ME lexicon after the 12c. OE gerisan occurs in the form birisen and (ge)byrian is continued by biren, behofian by bihoven and the contracted northern form bus. Additionally, booten 'to be of use, avail', bicumen1 'to be fitting ...', langen 'to be appropriate, fitting' and fallen 'to be fitting, proper' are ME representatives of the BEHOVE class. In type S these predicates continue to be used until the 16c., as shown by the survey chart and illustrated by the examples below: TYPES

OE

12c.

13c.

14c.

15c.

16c. 17c.

bihoven biren bus bicumen booten langen fallen dafenian birisen

χ χ

χ χ

χ χ χ χ

χ χ χ χ χ

χ χ χ χ χ χ χ

χ χ χ χ

(162) (163) (164) (165) (166) (167) (168) (169) (170) (171) (172) (173)

χ

χ χ χ

-

-

χ χ

us bihoueb bat we don alse (Trin. 71,8) swa biriseb ban Sungan bet... (Lamb. 109, 24) unnc birrb babe bannkenn Crist (Orm Ded. 27) Mikil dowte birde bam haue (N. Ben. 37,23) and therfore behoueb us to joyne to the erthe (Cax. 51, 28) thame behufit be swa (Bruce IX, 725) To rekkenynge buse us rise (16c. OED) be bus do als Crist counseiles (OE Legends 27 (Horstmann) quot. after Vi. 27) me botis not barly your biddyng with stonde (Destr. Troy, quot. after Vi. 27) ne bicomb be to werche wanne... (SEL 119, 2) ... as longeth to a maiden for to have (15c. MED) Me idafenaeb to wyrecenne his weorc (Bod. Horn. 58, 27)

As with RUE and PLEASE verbs, type S can again become a source of ambiguity with full nouns: 1 OE becuman 'become, happen' and 'come, enter'. MED sense 6. The following verb langen (OED v. 2) is recorded in the present meaning from the 13c.-16c. 121

(174) f)t mann birrjj spellen to be folk of... (Orm. Ded. 35) (175) pan birds men casten hem ... (Hav. 2101) (176) be king burd send ober to ask (14c. MED) That this situation need not necessarily lead to personalization will become clear below. In the //-construction the same class of verbs is attested as follows: IT-CONSTRUCTION bihoven biren bus bicumen booten longen fallen dafenian birisen

OE

12c. 13c. 14c.

_

_ x

_ χ

χ

x _ x χ χ

15c. 16c. 17c. 18c. 19c. x _ χ -

x _ x χ χ -

x

x

x

x χ

x χ

x χ

χ

The following examples illustrate the /'/-type in ME: (177) (178) (179) (180) (181)

hit birisaeb ure drihten baet... (Bod. Horn. 78, 13) nu bicumeb hit... to uwilche cristene monne ...to haligen... (Lamb. 45, 19) perfore it behoveth him bat... (Mand. 87, 17) it bated him not to aske why... (More 18) To God it belongeth to gouern the tonge (More (Wks. 1534) 1207, 84, quot. after Vi. 45) (182) // ne valb no^t to be wip ^onge children pleie (SEL 119, 20) From the 13c.—17c. seem (semen) is also used in the meaning 'to suit, be fitting'. As a member of this semantic verb class it enters the //-construction: (183) it semyth hym to sittyn on a white cuschyn (MK 211,9) This raises the question of how to account for the gaps in the above chart. A syntactic explanation has to be excluded. BEHOVE verbs are quite well attested during and after the ME period. Despite scant evidence before the 14c., there is no reason to assume that the occurrence of this verb class in the it-construction was in any way syntactically constrained. Neither is it possible to describe the gaps in the occurrence paradigm of BEHOVE verbs in terms of their semantic structure. Evidence on this aspect is provided by their co-occurrence properties in the three syntactic variation types mentioned. In the 12c., birisen and bicumen co-occur in S and the //-construction. In the 13c. and 14c., the same behaviour is attested wiih fallen, and during the period from the 14c.—16c. bihoven, bicumen, booten and longen show the same occurrence properties, illustrated by (184)— (187): 122

(184)z swa birisep pan jungan pet... (Lamb. 109, 24) b hit birisaep ureDrihten baet... (Bod. Horn. 78, 13) (185)fl & welbisemep be to beon & bikimep to beo ... (Jul. 55,15) b Hit bicomep for a kyng... To^iue meede to men (14c. MED) (186)a hym behoueth to come doun (Cax. 50, 7) b perfore it behoveth him pat he... (Mand. 87,17) (187)a Him booteth not resist (Spenser, FQ 1.3, 20) b it booteth no man to speke to them (More 18) As in previous examples of this variation pattern, the distinction involved cannot be described in terms of semantic verb structure. The aspect which can be captured by the descriptive means at our disposal concerns again the relational situation expressed by the semantic roles of the NPa: pseudo-agent in the a-, recipient/goal in the b -examples. This again means that the attested history of BEHOVE verbs in the //-construction cannot be evaluated on the basis of proper semantic structure. The variation expressed by the above pattern is most adequately described as one of style. However, the ME behaviour of this verb class shows distinct preferences of selection. If we look at the attested situation in terms of what use was made of the grammatical variation offered by type S and the //-construction, there emerges a striking pattern, formed by the syntactic use of biren/bus and bicumen/booten respectively. ME biren (according to MED 'a synonym of bihoven') is almost exclusively at home in the north Midlands and the North.1 The northern contracted form ofbihoven is bus. Biren and bus are found in type S until the 15c./16c., but they never enter the //-construction (nor do they ever occur with prepositional NPa as indirect object). This fact alone does not allow any semantic inferences. However, both verbs are met in the personal construction from the 13c. (biren) and the 14c. (bus), a situation which will give rise to the following variation patterns: (188)a & uss birrp hemm purrhsekenn (Orm 242) b pat alle... Weilgeorne birrdenn clennsenn hemm (Orm 4029) (189)* of a womman bos himm be born (CM 9870) b We erflesshlie men & bus nedelyngis eatt (15c. MED) Again we have to assume that the relational distinction in this pattern was functional, i.e. that the role features of the NPa in S were systematically distinguished from those associated with subject function in the personal type. From the 14c.—16c. we also find bihoven in the same variation pattern. This suggests that — at least in personal construction — biren and bus were used in the meaning 'need', not 'be fitting'. It is impossible to say whether the syntactic variation pair was constantly used to express the semantic distinction between these two meanings, although an investigation of the individual contexts might throw more light on the semantic aspect. In view of this situation, the complete absence of biren and bus in the //-construction must again 1 Orm uses biren and bihoven. with obvious preference for the former.

123

be described as a matter of usage rather than grammatical principle. (In the case of the contracted form bus, a rhythmical constraint could be imagined, because in a putative it bus him that... there would be two (or more) light elements initially). We cannot follow up the history of biren beyond the 16c.,because the verb disappears from the lexicon after the ME period, (most probably because it is a regional northern item). Neither does bus occur in our post-16c. material. Notice, however, that OED lists bus as a Scottish dialect word, with 18c. and 19c. examples of the personal construction. (The meaning of the modern Scottish verb is 'must'). EDD does not list it. Now quite a different type of syntactic behaviour is shown by bicumen and booten: both co-occur in S and the /r-construction, but are not used in the personal variant. This situation can be accounted for in semantic terms: both verbs are proper 'be fitting' predicates, i.e. they are never used in the semantic context 'necessity'. Insofar as the semantic role properties of the NPa are concerned, they can best be negatively defined, as never being of the agentive kind. With these verbs there are two further sources of evidence: their pathway of development after the 16c. and their behaviour in the type with nominal complements. Let us turn to the former aspect first. The occurrence chart for BEHOVE verbs in construction with sentential complements shows the following picture: OE

bihoven biren bus bicumen

· ·

booten langen fallen

12c.

· · ·ο

13c.

14c.

15c.

•0

• ΟΔ

•Δ

•Δ •Δ

•Δ

•O

•O

• •



·ο

16c.

17c.

18c.

19c

• ΟΔ

•ΟΔ

ΟΔ

ΟΔ

ΟΔ

•Δ

Δ Δ

Δ



• O

0

0

0

•0

0

0

0

0

•O

O

·ο

·

·

·: Type S o; /r-construction Δ: personal construction As with all the predicates met so far, type S is discontinued after the 16c. With bicumen/ booten, however, there is no personal reanalysis of S. When this type is abandoned, the /'r-construction is continued as the only variant until the 19c. Contrary to biren/bus these verbs never take a nominative NPa, a fact which helps to disambiguate any indeterminate type S construction, (the king booteth ...) in the direction of the traditional interpretation. This development has been described as following from their constant meaning 'be fitting'. That the direction of development is not an either/or matter (as might be suggested by the respective syntactic valency of biren/bus and bicumen/booten) is shown by the history of bihoven. 124

In the 14c.-16c. this verb occurs in all three construction types, thus realizing all relational possibilities to its animate-NP. This fact again suggests an account in semantic terms: bihoven can be used as a pure 'be fitting' predicate or in the meaning 'need'. Whereas in the former meaning it enters type S and the /f-construction, it can occur in all three syntactic variants if used as a necessity predicate. The kind of variation involved is illustrated by the following examples: (190)0 theym behoved for no good that... (Mai. 588, 26) b For hit behovyth the now to chose (Mai. 257, 29) c he behoued to lete Bedyuer ly stille (Merlin 661, 31) (191)7 Perforce behuifit him than fall doun (Lyndesay (16c.) quot. after Vi. 27) b it behoveth him to dye (More 18) c he behoved nedes dye (16c. MED) When type S becomes obsolete in the 16c., bihoven continues both syntactically viable variants until the 19c.

3.2

BEHOVE verbs in construction with nominal complements

We saw in chapter II that in the meaning 'be fitting, proper' OE BEHOVE verbs exclusively occur in type I. There is a constant relational pattern associated with this construction, as it regularly involves a [+animate] and a [-animate] NP. An account of the grammatical behaviour of this verb class in ME involves reference to different levels of description. Thus there are the two word order variants la and Ib, focusing the 'person' or 'thing' respectively; and again we will have to find out whether the syntactic variation is systematically associated with the expression of the semantic distinction between 'need' and 'be fitting'. Although in many contexts both meanings are possible and can therefore not be made dependent on syntactic form, the semantic variation involved is relevant with regard to the possibility of transitivation. In the meaning 'be fitting' type I remains the only possible syntactic expression. As there can be no reanalysis towards a transitive reading for semantic reasons, neither word order type will create indeterminacy, as the relational situation remains constant. In situational contexts in which both meanings can occur, word order cannot determine the desired interpretation. From the 12c. we meet syntactically ambiguous examples such as: (192) swa bihoueb be saulefode (Lamb. 63, 153) (193) lef child bihoveb lore (13c. MED) (194) lewed men bos anober lere (15 c. MED)

125

The syntactic structure of (198)—(200) allows a reading as intransitive type I or as a transitive construction, corresponding with the recipient or quasi-agentive role of the humanNP and the semantic use as 'be fitting' or 'need' respectively. In the meaning 'need', structures like (192)—(194) are open to a transitive interpretation according to the force of SVO order in the respective synchronic grammar and text-internal use. However, Ib order (the king behoveth NP), which easily allows a transitive re-interpretation of the original grammatical relations expressed by type I, is not frequently met. (195)-(201) illustrate ME type I, (195)-(198) in la, (199)-(201) in Ib order: (195) (196) (197) (198) (199) (200) (201)

mid al pet him bi-heovede (Lay. 3276) death thee behoves (15c. (?) MED) Whiche thynges... bihouen to penitence... (Ch. Pars. I, 83) wysdom behouys the (\Sc. MED) ban alden bihoueb äugende pewas (Lamb. 109,23) him behofde bob ofblod (AR 146, 12) me bihoueb his help (HM 37, 384)

In relative clause construction the NPa clearly has indirect object function in (202)— (203): (202) albat him be-houed (CM 1209) (203) and what behoved him (17c. OED) However, the grammatical relations in (204) are indeterminate: (204) Somme seknes in the legge whiche behoueth a medycyne (15c. OED) It has to be noted that the indirect object status of the NPa can always be expressed by prepositions: (205) thys shelde behovith unto no man but unto sir Galahad (Mai. 878, 27) There is thus no syntactically unambiguous evidence for the occurrence ofbihoven with direct objects in ME. If the verb does take nominative NPa (as in type II), the second NP is a genitive object. Biren is similarly used, but only found in texts from the north Midlands and the North. As a necessity predicate OE behofian also entered type II (he behofeb metes). This use is continued until the 14c.: (206) he nanes binges ne bihofaeb (Bod. Horn. 82,26) (207) Ouber understondinge bihoueb of be ryueres... (14c. OED)1

1 In this example the NPa is unexpressed, though understood. 126

In ME the causative (respectively partitive) relation is not often met with this verb. That it was kept alive with necessity verbs, however, appears from the core predicate neden, which preserves this relation until the loc.1 The diachronic lexical valency of bihoven with nominal complements thus shows the following picture: BIHOVEN

OE

12c. 13c. 14c.

15c. 16c.

17c. 18c. 19c.

Type I Type I I transitive

x x _

x x _

x — _

x — _

x x _

x x _

x — _

x _

x __ _

The continued productivity of type I again reflects the force of semantic use in determining the pathways of syntactic development. Even against the general tendency towards transitivation and the availability of the syntactically unmarked type II, I is preserved as the sole representative of the nominal paradigm. The essence of this development can best be characterized informally: the 'thing' comes to be expressed by the nominative NP, the 'person' by the dative. In terms of semantics, this correlates with the post-16c. generalization of the meaning 'be fitting'.2 Bicumen 'be fitting' is first attested in this meaning in the 13c. It almost exclusively occurs in word order variant la, thus preserving the fit between syntactic and functional levels: (208) Gretunge... walde wel bicumen be (Kath. II, 207) (209) be kirtel bicom him swibe wel (14c. OED) The prominent status of the indirect object relation with only one animate-NP present is further shown by the productivity of type I with /r-subjects: (210) hu wel hit heom bicome (Lay. 23018) (211) hit becom me wurse (AR 112, 21) As mentioned above (p. 126), Ib order can lead to relational indeterminacy. This may be the reason for the apparent constraint on the type him becom be kirtel, which is not found in our material. Contrary to bihoven, bicumen is used quasi-transitively, in the meaning 'suit, agree', from the 14c.-19c.:

1 Cf. OED v. 2, III.

2 The last type I record for behoven 'need' in OED is from the 17c.

127

(212) and that bicam me wel (Ch. WB 603)1 (213) A book would become his hands better (18c. OED) With the semantic structure of this verb the relational situation in the transitive construction is just the inverse of that in type I, i.e. there is no semantic change involved. This fact may be responsible for the continued occurrence side by side of both construction types, in the variation pair NP bicumef) him / he bicumejj NP. The survey chart for bicumen represents its syntactic occurrence from the 12c. to the 19c.: BICUMEN

12c.

13c.

14c.

15c.

16c.

17c.

18c. 19c.

Type Ι transitive

χ —

χ χ

χ χ

χ χ

χ χ

χ χ

χ χ

χ ('be fitting') χ (suit, agree')

Beseem (besemen 'be fitting') is attested in construction with nominal complements from the 13c.-19c.: (214) A poynt... bat uche mon may euel by seme (14c. MED) (215) A prison may best beseem his holiness (Marlowe, E2, 207) Besides bihoven, bicumen, besemen (and booten, which is only rarely found in construction with nominal complements), there are a number of verbs used in the meaning 'be fitting' for a limited period. All of them occur in type I. Thus birisen is last recorded in the 13c.: (216) wisdom birisaef) weran (Lamb. I l l , 16) (217) bingges be biriseb to aelche kinge (Lay. 4897) Longen is a member of the present lexical class from the 14c.—16c.: (218) Actyflyf... longeb to worldly men (14c. MED) (219) ... all the blastis that longed to all maner of game (Mai. 682, 33) Similarly, fallen 'be fitting, pertain' (and its antonym misfalleri) are thus used from the 13c.-16c.: (220) Mani miracle bar feol α-day (13c. MED) (221) ... bet hit misualle to him (A 193,19) All these verbs preserve the original pseudo-subject or indirect object relation to their animate-NP, in the syntactic form of Ib or la order respectively. As they do not occur in

1 In my 14c. material I have found no unambiguous accusative object in this construction. 128

the semantic context Necessity', relational indeterminacy cannot arise. With this class, then, the semantic structure determines the possible pathways of syntactic development.

4. The HAPPEN class

4.1

HAPPEN verbs in construction with sentential complements

The OE representatives (ge)limpan and (ge)weorpan exclusively occur in type S. In ME the members of this class are limpen, bicumen1, fallen (from the I3c.), fortunen and happen (from the 14c.),2 all occuring freely in type S: (222) (223) (224) (225)

hemfallib... to haue lordchip (W. 412, 15) him ilomp... pat... (Lay. 1391) Betere him become, boren bat he ne were (13c. MED) And hoped me... we take a schippe (PL I, 429, 12)

That the semantic structure of HAPPEN verbs easily accepts the indirect object status of the NPa is reflected by the fact that from the 12c. they form a variation pattern with the /f-construct! on: (226) Feawe mannen gelimpp on byssen dagen pat... (E.E. Horn. 58, 7) b Ah bah seib sum pat hit limpep to euch wummon ... forte werien (AW 215,21) (227)7 andforbi him bicumeb bat... (Trin. 45, 20) b Nu bicumeb hit... to cristene monne ...to haligen ... benne dei (Lamb. 45, 19) The same variation possibility is also adopted by happen and fortunen, entering the lexicon after the 13c.: (228)a at the laste... hym happynd... to com to a ... courtelage (Mai. 272, 37) b hit happed hym to come to the röche (Mal. 132, 18) (229)a A solom cite me fortuned to fynde (15c. OED) b whom it fortunyd to metyn wyth ... (MK 127, 5) In the 14c. the personal construction is first met with bicumen, fortunen and happen: (230) be feile bor bicam to come (MED)3 1 OE becuman with sentential complements is not attested in the meaning 'happen', but 'become, behove'. The earliest use of'happen' is met in the 12c. Bodleian Homilies: he cwaep hit bicymefr & iwurp (Bod. Horn. 16,3). 2 The ME HAPPEN verbs chauncen and lukken are not included because attested evidence does not allow any valid generalizations. 3 Cf. the manuscript variant gan to come (MED). 129

(231) thu may hapin to sla sum dere (CM 3602 Gott.) (232) frei happyd to corny n under... (MK 240, 31) (233) This maner Ende that controuersie or stryfe fortunyd to take (15c. MED) The distinction between S and the personal construction is again neutralized with full NPs: (234) a godly lady ... happed to plonge... in a depe pitte (15c. vdG) With HAPPEN predicates, as with proper BEHOVE verbs, there can be no transitivation in a semantic sense; the change in the occurrence paradigm of these verbs consists in a restriction — after the 17c. — to the unmarked SVO variant obeying the animateness target. The survey chart shows the following historical situation:

limpen bicumen fallen fortunen happen

12c.

13c.

14c.

15c.

16c.

17c.

· ·ο

·ο ·ο ·

· Δ ο ·ΟΔ

_ ·ΟΔ ·ΟΔ ·ΟΔ

_ _ ΟΔ ΟΔ

_ _ ΟΔ ΟΔ

18c.

_ Δ Δ

19c.

Δ

·: Type S ο; it -construct! on Δ: personal construction Limpen , which disappears from the lexicon in the 15c., never occurs with nominative NPa. The detailed semantic development of bicumen has not been studied yet. Its main meanings in MED are 'to become; happen; suit'. In construction with sentential complements the verb is not used in the meaning 'happen' after the 14c. (It continues this meaning with nominal complements until the 19c.) The most striking aspect of the history of HAPPEN verbs concerns their lexical valency in the 15c. Fallen, fortunen and happen occur in all three variants: (244)a At be latiste hym happed He bouhte suche a bargayn (15c. MED) b as it happyd hir to be herberwyd (MK 78, 22) c ban happyd sehe to fyndyn per... (MK 129, 18) (245)zat... befyllehym... that... (Mai. 1017,3) b hit befelle hym that... (Mai. 982, 19) c In bedde yf bou falle herberet to be (15c. MED) Whereas semantically the S- and if-constructions are well motivated (the NPa is not a subject and cannot take the agent role), the personal variant must be accounted for on a syntactic basis, viz. as a formal adaptation to SVO.

130

Fallen 'happen' is not found in construction with sentential complements after the 15c., but the type without an experiencer-NP (it fell that...) is continued. In any kind of syntactic variation the relational situation remains constant: the NPa is the recipient/goal, the complement remains neutral. This fact may be responsible for the disappearance of type S a full century before it becomes obsolete with the other verb classes. In the 16c. and 17c., the two natural syntactic representatives of the grammatical relations associated with 'happen' — type S and the //-construction — are continued, although the latter is no longer found in the 18c. Again the personalization tendency (in the form of the animateness target and syntactic SVO) is satisfied without any semantic pressure.1 There appears to be no more coherent reason for the obsolescence of the //-construction than the fact of the animateness target observed from OE onwards. There is a further aspect to the structural history of the //-construction. If the NPa is not distinguished for nominative/dative case, as in (246) // happyd be creatur... to ben per present (MK 56, 15) (247) hyt happened sir Gawayne at Astolot to lodge (Mal. 1077, 24) (248) // may also fortune a man to be sory for his synne (Fisher 32, 11) there is an alternative interpretation, as a type it-verb-sentence (of which the NPa is the subject). // in this case is present purely to produce a verb-second structure, without having the deictic function it has in the //-construction proper. In examples like (246)—(248), with infinitive complements, the NPa can of course only be said to be subject in a semantic sense; (in a formal syntactic account, the subject of the complement clause would be 'deleted' under identity with the object of the main clause). However, the effect of this kind of surface semantics on the hearer/reader - and therefore its liability for surface reanalysis - should not be underrated, as a 14c. example from Mandeville shows: (249) // wolde semen him to dye for drede (Mand. 203, 21) In particular, if the NPa is not a real animate noun, a semantic interpretation as a structure of the type it-verb-sentence is likely: (250) it fortuned... a great subsedie to be demanded (Life of M. 40, 16) The same interpretation must be assumed if the complement is a that-clause: (251) once it happened the Duke ofNorffolk

... found him (Life of M. 51, 15)

The last records in OED and Vi. are from the 16c. The present 17c. example is from the Life of More: If it happened ante learned man to resort to him (56, 5). 131

4.2

HAPPEN verbs in construction with nominal complements

With nominal complementation the syntactic paradigm entered by HAPPEN verbs is again semantically determined. The relational situation is constant: the animate-NP is experiencer, the second NP takes a neutral semantic role. The grammatical relations associated with 'x happens to NPa' find their syntactic expression in type I, with the [—animte] NP in subject function. The only variation possibility thus concerns the two order types of this construction. Until the 13c., the HAPPEN class is represented by limpen, bicumen znd fallen: (252) (253) (254) (255)

... alle god be limpet) to sawle (AW 155, 21) pis gelamp pan fulhtere Johanne (E.E. Horn. 56, 3) Wei michel sorge is me becumen (GEN & EX 2227) & moni ping ham fallep to (AW 179,4)

In the 14c. happen joins this paradigm, continuing the construction up to the 19c.: (256) Now wol I teilen forth what happed me (Ch. WB 563) Fortunen* is only sporadically found during the 15c. and 16c.: (257) And this not onely of hym fortuned..., but... (Sk. 338, 29) (258) If any thynge fortuned well to us... (16c. OED) As appears from (252)—(257), the regular word order is la, with a nominative subject and optional prepositional marking of the NPa. The word order variant Ib (with initial animate-NP, as in Hym selfe, noble kyngArthour, Hathforteynd syche a chans (15c. OED)) is hardly found at all, although it does not create any conflict between linear syntax and the OVS interpretation of grammatical relations. In the nominal paradigm, then, HAPPEN predicates are attested as follows: TYPE I

12c.

13c.

14c.

15c.

16c.

17c.

18c.

limpen bicumen fallen happen fortunen

χ χ

χ χ χ

χ — χ χ -

χ — χ χ χ

χ χ χ

χ -

χ -

19c.

χ χ -

Limpen disappears from the lexicon after the 15c. and bicumen is discontinued in the meaning 'happen' after the 17c. Although the occurrence of HAPPEN verbs in absolute constructions is rare, it is sporadically met until the 19c.: 1 This verb is used transitively, in the meaning 'to control somebody's destiny', from the 14c.-17c. (OED). It also occurs with nominative animate subjects in intransitive construction of the type he fortuned wel, from the 14c.-18c.

132

(259) iffrou limpes (15c. MED) (260) pey he beo odde yffalle (14c. MED) (261) if ye happe therupon (Gower, CA 5.2336) The syntactic history of these verbs in construction with a nominal complement is the clearest example of a semantically-determined paradigm. There is the obligatory presence of the NPa in recipient role — syntactically expressed as indirect object - and the event which happens, in syntactic subject function. No semantic change is possible, and the order of elements will not cause any indeterminacy even if conforming to SVO syntax (la). This constant semantic situation is reflected by the continued use of type I with happen ana fallen.

5. The SEEM class The OE representative is pyncan, which continues to be used in the meaning 'seem' in its ME form thinken.! In OE, type S is the only syntactic possibility. Only in the 13c./14c. do we meet rare occurrences of the if-construct! on:2 (262) And tah hit punche opre men pat... (HM 9, 76) (263) Vor hit him pingp pet... (A 135, 19) (264) Thanne is it wysdom, as it thynketh me to maken vertu of necessitee (Ch. Kn. 2183) From the 12c. there is only one iormpincen, meaning 'seem' in type S and 'think' in the personal construction: (265) Hym thoughte pat his herte wolde breke (Ch. Kn. 100) (266) Thanne thoughte they it was the beste reed (Ch. Sum. 322) In the present case the reasons for the continued productivity of type S lie outside linguistics proper: the close association of first person experiencer with this syntactic form ensures that methinks (often indeed conceived of as one form) remains the indigenous rendering of OE me pyncep. This is the second example (besides listen) which shows that the general syntactic development leading to the loss of type S is not without its exceptions. In the 13c. semen enters the language. Semantically, a generic experiencer is of course present, but it is not expressed syntactically: (267) hisgrisliche tep semden of swart irn (Marh. 20, 24) 1 The same form also continues OE pencan 'think'. 2 Note, however, that the type it thinks him ADJ/NP is productive throughout the ME period.

133

'His horrible teeth seemed to be of black iron'. Here the syntactic subject corresponds to the head of a subordinate clause of the form it seemed that his teeth were... There is thus a fundamental difference between this syntactically personal construction and the personal variant met so far. Within the semantic framework of SEEM verbs a nominative NPa never corresponds to the experiencer NPa of OE me bynceb baet..., but represents the 'raised' subject of a subordinate clause.1 There can thus be no personalization of type S for semantic reasons. Specified experience« have the regular dative form of the NPa: (268) hym semeth the nombre IX so holy... (Mand. 164, 22) In the 14c. and 15c. we also meet the (personal) type represented by (267) with a dative pseudo-subject as in the following example: (269) Hir semed na wight to be wilde (CM 3284) Note, however, that the initial animate-NP in this sentence does not correspond to the experiencer NPa of regular type S constructions, but appears to be a pure surface marking of animate-NPs occurring with semen. The history of the syntactic valency of semen and thinken shows the following picture:

semen thinken

12c.

13c.

14c.

15c.

16c.

17c.

18c.

19c.

Δ

-2 ·ΟΔ

·ΟΔ ·ΟΔ

·ΟΔ ·Δ

·ΟΔ ·Δ

·ΟΔ ·Δ

·ΟΔ ·Δ

·ΟΔ ·Δ

·

·: Type S ο; //-construction Δ: personal construction The most striking aspect is the three-way variation pattern into which seem enters from the 14c.-19c.: (270)z what semeth the to be the resoun of this...? (Ch. Bo. 4 p. 5, 1425-30) b it sholde seme to som folk that... (Ch. Bo. 4 p. 2, 1240-5) c the wikkide men semen to be bareyne (Ch. Bo. 4 p. 2, 1230—5) The continued productivity of type S after the 16c. has to be accounted for in a similar way to that of methinks above: there is a particularly close relation to the first person speaker/writer, which is reflected by the synthetic form meseems, methinks. This is in line with the fact that S-type sentences like (271) Certes me semeth ... that... (Ch. Bo. 3, p. 9, 800-5) 1 Cf. Paul M. Postal, On Raising, 1974. 2 13c. records are all of the type represented by (267) above, i.e. without a specified experiencer.

134

are most productive with first person pronouns, whereas with second and third person the it -construct! on is more general: (272) it ne seme not to men tlwt... (Ch. Bo. 5 p. 5, 1840-5) However, the construction is always possible as a syntactic variant also with pronominal NPa in the first person: (273) And righte as it semethe to us that thei ben ... (14c. OED) (274) It seems to me... that... (19c. OED)

6.

Summary

The historical development of subjectless grammar involves different linguistic levels and demands, as the present study shows, a description which takes this fact into account, even at the apparent cost of a lack of systernaticity. The syntactic history of the present verb classes was treated within the parameters determined by their semantics. RUE, PLEASE/DESIRE, BEHOVE, HAPPEN and SEEM predicates regularly involve a 'person' and a thing/state/event. In terms of semantic roles, the relational situation associated with this framework is such that the animate being is experiencer/recipient, the element passively involved, while the 'thing' is seen as source/cause, or in a neutral role. Now the syntactic occurrence of subjectless predicates in OE neatly reflects the form which the adequate expression of this semantic situation took: if the 'thing' is given the form of a sentential clause, the syntactic type S (me hreowefr fraet...) is used almost exclusively. The fact that the perfect semantic paraphrase of the form it hreowefr me pact ... is hardly met, despite its availability in the syntax, indicates that the relational properties associated with the animate-NP are a decisive factor for the productivity of type S. The NPa is indeed the focal point of the whole problem. Its main syntactic feature, besides oblique case marking, is its regular initial position in all OE constructions except the inverted subtype Ib (paet hreowefr me). Several independent factors were assumed to contribute to this important fact: in terms of syntax this order was congruent with the SVO form which unmarked OE sentences began to adopt quite generally; with regard to thematic structure the initial NPa reflects the priority given to animate-NPs in terms of information value; as the NPa furthermore predominantly occurs in pronominal form, the rhythmical tendency to front light elements also works towards the same result. In SVO languages, initial position is associated with nominative subjects. The unmarked form of the OE main clause was SVO, and we saw that from the time of the earliest texts the present verb classes could take nominative NPa, in absolutive construction (ic scamode) or type II (ic scamode fraes). OE also showed the first traces of personal use in construction 135

with sentential complements. Likewise, we mentioned the personal paraphrase possibility of the type ic haebbe bearf baet ... Within the framework of constant intransitive use, however, there was no question of the animate-NP ever taking an agent role. There is not a single example of the transitive use of any RUE or PLEASE/DESIRE predicate, a fact which means that syntactic variation did not involve semantic change even in those cases where a re-interpretation would have been first possible. Apart from the systematic non-occurrence - on semantic grounds — of the BEHOVE, HAPPEN and SEEM verbs in personal construction, the limits of syntactic variation in OE seemed to be wholly due to the factor of use. Under this aspect the productivity of all types expressing the specific verb-NPa relation (S, N, Ib) clearly indicates that the attested occurrence paradigm is due to the fact that the grammar of subjectless predicates was conceived of as a special coherent subpart of the grammar of OE. The pathways of development in ME were described within the limits set by general syntactic and semantic parameters. In syntax, the tendency towards SVO order with nominative subjects gradually becomes more rigid; it naturally favours structures which conform to this pattern (such as the personal and ^-construction, types la, II and the transitive construction). Associated with this formal aspect, two general semantic processes determine the occurrence properties of subjectless verbs. Within an overall context of transitivation, indeterminate structures will be re-interpreted as transitive ones, first by substituting a (pseudo-) agent reading for the original recipient role of the animate-NP. As the example of RUE verbs shows, this process need not lead to proper semantic change, i.e. to a restructuring in the lexicon. The effect of such general semantic and syntactic factors is of course disproportionate. However, they gradually limit the ME realization possibilities of the verb-NPa relation, because they combine in constraining nonnominative initial animate-NPs. The original semantic situation will thus appear more and more in the new garb of unmarked syntactic form. Of course the animate-NP remained a recipient in a purely semantic sense, as the core meanings RUE, BEHOVE, PLEASE, HAPPEN and SEEM are inherently associated with this semantic role; but the process of surface abduction will give rise to a quasi-agentive re-interpretation, with subsequent semantic change as in the case of liken. The first major change affecting the ME grammar of subjectless predicates is associated with the three processes of syntactic SVO, personalization and transitivation, all of which have their roots in late OE. Up to the 14c. the occurrence of these verbs with animate nominative subjects was restricted to type II (he reweth of...). Now from about the midHe, this use is generalized to the construction with sentential complements. Also, subjectless verbs are construed with direct objects, i.e. in a purely transitive way. (A notable exception is booten, whose semantic structure prevents it from taking the 'person' as subject.) That individual pathways of development may only satisfy one of the above-mentioned tendencies is shown by the example ofgreven ana forthinken, which conform to the syntactic norm by entering the /f-construction, but violate the demand for initial animate nouns. 136

That the new possibility of personal and transitive construction in the 14c. did not provoke a general reanalysis of grammatical relations is shown by the fact that all syntactic variants which express the original verb-NPa relation (except type N) are continued for a further 200 years. From a purely formal viewpoint, type S is no longer a viable (because ambiguous) structure from the 13c. onwards if the NPa is a full noun. Despite the presence of the unmarked //-variant, type S resisted change during all this time. Indeterminacy could thus arise, but it appears that the situational context usually made the nature and directionality of grammatical relations clear; the ambiguous examples presented in this chapter were not uttered/written in isolation. Since the work of van der Gaaf (1904) the view has been repeated that the ME syntax of the verb classes concerned here is in a state of confusion until the 16c., which marks the end of the period of syntactic and semantic change. In recent work 16c. syntax in particular is said to undergo a series of radical changes amounting to a wholesale reanalysis of 'impersonal' as 'personal' constructions.1 It is certainly true that the 16c. marks a turning point in the history of subjectless grammar, as shown by the last occurrences of types S and I with a number of verbs. But to describe the development in absolute terms is inadequate. The empirical results achieved in this study do not support a radical break theory, whether syntactically or semantically-based. We have seen that the 'break' is prepared in the 14c. and becomes effective in the 15c. (loss of N), and the gradual nature of the development necessarily demands that we follow it up into the 17c. in order to assess the situation in the 16c. more relevantly. The major event in the 16c. history in the present context is the almost complete obsolescence of the OE verb-NPa relation also in syntactic type S (him shameth that ...), which in the 17c. is only preserved by list, seem and think. There can be no inherent semantic reason for this fact, but the particularly close association with pronominal first person subjects probably contributed to the development followed by these verbs. Now the effect of the disappearance of S from the paradigm with sentential complements again highlights the idiosyncratic nature of development. Of the verbs which give up S in the 16c., eilen discontinues sentential complementation altogether. How questionable it is to say that S 'changed into' another construction is shown by the example of those verbs which continue only one syntactic variant. By the 17c. rewen, liken and langen have adopted the personal type as their sole possibility in construction with sentential complementation. Shamen, greven, bicumen and booten, on the other hand, continue the /^-construction. This situation (especially in view of the practical synonymity of rewen and greven) forbids an account in terms of semantic structure, so that reference to the vague notion of 'predominant use' must suffice as a description. The occurrence paradigm of bihoven, happen and fortunen, on the other hand, shows that the maximal variation possibility — between the personal and /f-construction — has been adopted. Syntactic realization can now take place within the limits determined by strict SVO order. 1 Cf. Lightfoot (1976, 1978) 137

In the case of semen and think, finally, the uninterrupted use of type S allows the stylistic three-way (respectively two-way) variation to be expressed until the 19c. In the construction with nominal complementation the pseudo-subject relation is expressed by types N and Ib. N is no longer found in the 16c., which means that strictly speaking there are no more syntactically subjectless constructions after the 15c. As in the case of the constructions with sentential complements above, the course of development followed by individual predicates cannot be given any single systematic reason, but a stricter account in semantic terms is now possible. The typical pseudo-subject relation is given up by RUE verbs and listen/longen in the course of the 16c. With this set of verbs, then, 17c. occurrences of type I (in la order) have to be read as transitive constructions, because after the loss in their syntactic valency of S, N and Ib, the original verb-NPa relation can no longer be expressed. It is in this sense only that type I can be said to 'change into' the transitive construction by the process of reanalysis. Liken (in the meaning 'please'), as well as the HAPPEN and BEHOVE verbs preserve type I. In the case of liken the conditions assumed above for continued use of S by seem and think apply. The occurrence of HAPPEN and proper BEHOVE predicates in type I, on the other hand, is semantically determined. If in the 16c. the various factors contributing to the personalization tendency bring about a reanalysis of the NPa as the grammatical subject, the simultaneous process of transitivation is the major force in bringing about, through the more widespread transitive use of verbs, a reanalysis of the original causative NP (whether syntactically expressed by la (NP reweth him) or II (he reweth of NP)) as a direct object. Again the seemingly causal account of this development as a change of la and II into the transitive construction possesses no explanatory content. All types co-exist for centuries, and the sole survival of the transitive one is due to the general factors mentioned, independently present in the language. Whereas the syntactic parameters determining the limits of possible development can be clearly established, the part played by semantic verb structure is less easily determined. In no single case could a proper semantic feature (say of a componential analysis) be linked to a specific course of development of syntactic verb valency. At the same time 'semantics' in terms of associated NP role properties obviously determined such idiosyncracies as the absence of type I with shamen and the absolute non-occurrence of causative objects with greven and eilen. Listen gives up this grammatical relation in the 16c., but the synonym langen preserves it until the 19c. It was this situation which suggested that individual syntactic behaviour was described with reference to semantic roles rather than abstract semantic structure. There can be no question of a wholesale semantic change associated with the development of syntax in the 15c.-17c. The HAPPEN, SEEM and BEHOVE classes do not undergo any semantic change even if they adapt their lexical valency to the normative paradigm. With RUE verbs and liken (listen, langen), on the other hand, the situation is fundamentally different because these predicates may take two animate-NPs (or the

138

[—animate] NP may be conceived of as the agentive originator). The process of role inversion - the original receiver (the NPa) becoming the pseudo-agent - will cause semantic restructuring in the case of RUE verbs by adding 'regret, repent, ...' to 'feel contrition' and radical lexical change in the case of liken, from 'please' to 'like'.

139

CHAPTER IV LINGUISTIC CHANGE WITHIN A RESTRICTIVE FRAMEWORK

1.

Introduction

The aim of this chapter is to focus on one main issue discussed in the previous chapter from the most restrictive aspect possible, viz. of strictly synchronic grammar. Chapter III showed the co-occurrence of syntactic variants relative to the chronological criterion of whole centuries and evaluated the nature of attested variation in the light of the development of formal syntactic types and with regard to the semantic structure and use of individual predicates, within a framework not explicitly defining a coherent dialect. In the present context we intend to deal with the range of actual variation internal to single texts/authors, seen against the syntactic possibilities available in the respective century. Also, under the aspect of use, the occurrence of different syntactic types will be discussed in quantitative terms. Within a restrictive dialect-internal account, the upper limits of variation will be determined by the generalizations and constraints characterizing the syntactic valency of predicates within the framework of a century. A text-internal approach in particular allows a much stricter stand on morphology, a decisive factor in any context of structural indeterminacy. Also, on the basis of individual texts, or groups of texts written by the same author, it is possible to make valid claims concerning the semantics of predicates, although this issue must remain inconclusive, as will be seen. These aspects together form the basis for a correct analysis of ambiguity. Under a dialect-internal approach, syntactic and relational indeterminacy of the kind met in chapter III can be treated more systematically. In chapter III ambiguous constructions necessarily had the same status as unambiguous ones, i.e. they had to be assumed as potentially possible structures. This is clearly not the case in a synchronic context, with communication-based limits on admissible ambiguity. In synchronic grammar we have to assume a tolerance level defining the limits of syntactic indeterminacy in quite a practical way. By describing the syntactic behaviour of subjectless verbs within the systematically restrictive framework of single texts, we hope further to characterize the factors which determine the relation between potential variation possibilities and actual use. As mentioned in the introduction to chapter III, we cannot choose systematically the points for which we would like to have the synchronic situation comprehensively attested. Our procedure will be to select, for the period between the 12c.-16c., texts which amply illustrate the aspect of syntactic variation and which allow founded statements concerning the representativity of attested variation. It seems in order, at this point, to stress the methodological importance of the marked difference between phonological and syntactic variation in different manuscript copies of the same source. What can clearly be attributed to different dialects on the basis of phonological criteria does not generally warrant such a classification from the point of

141

view of syntax. Manuscript differences of course include lexical changes or departure from the original syntactic construction altogether. But evidence for the systematic use of one variant in contrast to others, or a preference for certain paraphrase types will be shown to be scarce, though not completely lacking.

2.

The 12th century

The most extensive prose documents of the 12c. are the Lambeth and Trinity Homilies, which will form the data base representing 12c. syntax.1 This material will be complemented by roughly contemporary data from the Early English Homilies.2 With respect to the occurrence of verbs in the syntactic construction types concerned here, the situation is the following: with sentential complements type S and the /f-construction, but not the personal type, are attested. With nominal complements, on the other hand, the paradigm includes types N, I and II, but not the transitive construction. The most regular type with sentential complements is S, illustrated here on the basis of the BEHOVE verbs: (1) (2) (3)

us bihoueb pat we don alse... (Trin. 71,8) swa biriseb ban jungan bet he abbe ihersumnesse (Lamb. 109, 24) And forbi him bicumeb bat... (Trin. 45, 20)

There is only a single occurrence of the /i-construction in the two texts: (4)

nu bicumeb hit berfore to uwilche cristene monne ... to haligen ... benne dei (Lamb. 45, 19)

The relational situation between verb and NPa as used in these texts is typical of OE. Although there is clearly no syntactic constraint on the it-type, it is not productive with any of the five verb classes. While the variation possibility is there, it is not made use of. The generalization is that in 12c. prose the NPa is not given the syntactic form of indirect objects (as in the tf-construction), but is regularly preserved in the pseudo-subject role, thus continuing OE use.

1 EETS 29, 34 and 53. On the basis of phonological criteria the Trinity text (originally southern/ south-western) shows south-east Midland features, due, probably, to the copyist's dialect (Morris, EETS 53, p. XII). The Lambeth MS, on the other hand, is said to belong to the Middlesex or west Midland region by R.M. Wilson (Early Middle English Literature, p. 116). The 12c. Life of Chad, the earliest example of non west-saxon homiletic prose (Wilson, op. cit. p. 115), is too underdetermined with respect to subjectless syntax to be included here. 2 EETS 152 (MS Vesp. D XIV) ed. Warner. 142

Nor does the personal variant seem to be used. Unambiguous structures of the form he behofeb gon are not found in the Homilies. Compare, however, the following sentence from the Lambeth text: (5)

Nu bihoueb be forwunded wreche bet... (83, 26)

Within the less restrictive framework of 12c. syntax (5) is indeterminate between S and the personal construction. In the present case, however, text-internal evidence can disambiguate this construction. The word order criterion is inconclusive, as postverbal position of subjects, especially after the class of introducers met in chapter I, is still freely possible in 12c. syntax. Morphology, however, offers the crucial test. The indicative element as to the case properties of the NPa is the definite article. In the Lambeth text the inflectional distinction between nominative and dative form of the definite article is regularly preserved.1 (5), then, has to be read as a personal construction. Note, however, that while in this particular example morphological evidence is decisive, it is irrelevant in other contexts, for example, with proper nouns. Furthermore, verb inflection cannot disambiguate the construction by verb agreement in the plural. Type S and the personal construction could thus only be kept distinctly apart if the NPa was a pronoun or a full NP as in (5). Otherwise there was always the possibility of indeterminacy with regard to the functional reading of the NPa and — depending on the verb - semantic variation. This situation leaves open two possible accounts. It either means that in the synchronic grammar the relational and semantic variation involved by a personal or type S interpretation was not strong enough to force syntactic disambiguation, or clear enough to guarantee the intended reading despite the indeterminate syntactic form. In the light of text-internal evidence the second analysis seems correct. Syntactic disambiguation in the direction of the it-type is of course always possible, but it involves a definite change in the relational properties of the NPa, from pseudo-subject to indirect object and away from the quasi-agentive interpretation of this element. On the other hand, resolving the indeterminacy in the direction of the personal construction could only be achieved by generalizing the occurrence of these verbs with nominative animateNPs. This is clearly not the case in the 12c. The unmarked form of the construction with sentential complements was type S, as in the following examples from the Early English Homilies: (6) (7)

Neusne gedafeneb bat... (55,2) forbygedafenabbanbearnenbat... (75, 18)

Morphology also plays a crucial role in the next example, which illustrates inter-textual variation: 1 The paradigm is pam, pan, pa for the dative masc. against pe for the nominative, pe is also used as objective case after preposition+accusative.

143

(8) (9)

wa is mine saule pet... (Lamb. 157,19) wuo is mi soule pat... (Trin. 149,8)

Whereas saule/soule can be nominative or dative1, (8) is disambiguated in the direction of type S by the dative form of mine. (9) on the other hand is indeterminate between a type S or personal reading. Mi(n) is no certain sign of nominative case, as the final -e is omitted in other cases too.2 Word order suggests an S reading for both sentences. There are several external arguments speaking against a personal analysis of (9). First, there is only minimal inter-textual variation of phrases like (8) and (9) in the two series of the Homilies. The situation can well be illustrated on the basis of the three parallel versions of the Poema Morale, the Lambeth, Trinity and Egerton manuscripts respectively. For the range of constructions concerned here, there is complete parallelism in the following cases:3 Lambeth

pe me ofpinchet nupe (10) al pet me likede (\\) for him ne scamep (165)

Trinity

pe me ofpinkep nupe (10) al pat me likede (11) for hem ne samep (167)

Egerton

pat me ofpinchet nupe (10) al pat me likede (13) for heom ne scamet (167)

Cf. also lines 164 and 203 (Lambeth). There are few cases of change in syntactic construction; the following examples illustrate the typical kind of variation: Lambeth

hit schal rowen (19) a hwilce time se eure man ofpinchp his misdede (130) lutel hit punchep moni mon ac muchel wes pa sunne (201)

Trinity

hit sal him rewen (21) elche time salpe man ofpunche his misdade (132)

Egerton

hit hym scael new en (21) a wulche time so eure pe man ofpinchet his misdede (132) lutel ipenchp mani man hu muchel wes pe synne (203)

There is no variation in the sense relevant here, i.e. between type S, the rr-and personal constructions or the different types with nominal complements. 1 Morris, EETS 29, 34, p. XXVII. 2 Morris, EETS 5 3, p. XVIII. 3 The numbers in brackets refer to lines in the text. 144

Now (8) and (9) above are not from parallel texts, but it seems clear that we have to assume that the form of (9) was an S construction in the source from which the Trinity MS was copied. Second, external syntactic history shows that the first instances of the use of nominative NPa (/ am wo) are found in the first halb of the He.1 This observation is supported by our material, in which the only type used in the 13c. is of the form me is wo i..e. with the NPa in non-subject relation, as in (10) (11) (12)

nu me is wa pat... (AR 22, 29) wel is me pat... (Kath. 1896) wa /5me(SawleaWarde253,21)

However, the syntactic expression of this relation will always be underdetermined in relativ clause construction, as in (13)

wa is pan (MS: pet) Mon pet wa is (Lamb. 145, 8)

a situation which can only be tolerated if the semantic use is clearly understood. Note furthermore that there is a paraphrase construction to (9) of the form & wa ha habbap (Sawles Warde 253, 22). In view of the presence of this syntactically unmarked type, the productivity of the variant me is wo... is further evidence both for the indigenous nature, and use of the verb-NPa relation in the latter construction. Potential ambiguity of the kind represented by (8) and (9) concerns the functional reading of the NPa, but does not involve a change in the semantic structure of the verb. With a predicate such as liken, however, the consequences of syntactic indeterminacy for the semantics are more serious. Consider (14), from the Trinity MS: (14)

perefore likede ure drihten swo wel pat... (167, 20)

Within the context of 12c. prose neither morphology nor word order can disambiguate this sentence. The safest basis on which to evaluate (14) is to infer from the syntactic valency of liken its semantic structure and use. We have seen in chapter I that OE lician is attested with nominative NPa only in slavish translation, in non-narrative contexts.2 As in OE, liken is regularly used with non-nominative NPa in the Homilies, in the intransitive meaning 'please'. Compare, however, the following syntactically ambiguous example, which would equally well accept a personal reading: (15)

Mildheorted bep pe man pe reoup ... and likep here alre selpe (Trin. 95, 33)

1 Cf. vdG59. 145

In the sections of 12c. prose studied there is only one other example, with nominal complement, which not only allows a subject interpretation of the NPa (flesch), but also transitive use of liken, in the meaning 'please' or like': (16)

hwi nis me bitter al fret mi flesch likeb (Lamb. 201, 32)

Overall synchronic semantic evidence, however, does not suggest a personal reading for either (15) or (16). We can also make use of diachronic evidence on this issue. The regular use of liken in the 13c. texts grouped with the Homilies is with the traditional verb-NPa relation, as in & hit te wule eft likin (AW 123, 3) and the first unambiguous examples of the construction type as he wel liked are only attested from the mid-He.1 Contrary to the variant construction with habban+adjective, the paraphrase type lefben, present already in OE, allows no change of the relational situation. Compare the following examples from Layamon: (17) (18) (19)

heo him leofe wear en (996) hu leofaem ich be (1480) ban king heo wes deore (2995)

The transitive variation possibility of (19), realized in the Otho MS, κ be king hine lofuede. The status of originally subjectless syntax as it appears in the \2c.Homiiies, then, is still largely the OE one. Syntactic ambiguity can arise, but the functional reading is generally decided by semantic context and use. The situation will, however, be liable to reanalysis along the unmarked functional SVO pattern if either of these factors changes.

3.

The 13th century

If we represent the 13c. west Midland dialects by the texts of the Ancrene Riwle, Ancrene Wisse, Hali Meidenhad, The Katherine Group, Layamon and the South English Legendary, the syntactic paradigm into which subjectless verbs enter differs in one important aspect from that of the 12c. Before the mid-13c. we first meet the transitive construction with nominal complements, although it is still used as a much less productive variant besides the other types. The occurrence of the transitive construction besides type II means that originally subjectless verbs are now used more and more with nominative subjects and that the conditions for a re-interpretation of semantic roles are now given, as there is a direct object. It is thus in the 13c. that the gradual change towards personalization starts. S is still the most frequent type with sentential complements, although with the 1 Cf. also vdG 68f. 146

predicates entering the language in the 13c., the it-type seems to be used with about equal frequency. Strikingly, however, there is no unambiguous example of personal construction (he reweb bat...), despite the fact that there is now also semantic support for this reading, as will be seen below. On the basis of 13c. West Midland evidence some aspects of development emerge clearly. Consider first the sort of variation occurring in parallel texts in the two versions of Ancrene Riwle preserved in the Titus and Nero manuscripts respectively.1 Besides syntactic parallelism throughout, there are only a couple of paraphrase examples which are meaningful in the present context: Titus:

forbi'fhim reowe of ow (10,21) hwer to nedes ow? (90,23 )

Nero

uor biifhim areowe ou (29, 12) hwarofhabbe neode (117, 26)

The personal paraphrase type habben+adjective (examples (22)—(23)) is also met in Ancrene Wisse besides the traditional verb-NPa relation associated with neden and ben neod: (20) (21) (22) (23)

...-frhim to neodeb (178, 13) & for bi as ow is muche neod (220, 22) an bing-irich hefde neode to (134, 8) Two manere men habbeb neode to ... (134, 19)

The complex relational pattern involved by the semantic context 'need' is reflected by the additional paraphrase type ben nedful, which is only used with nominative NPa: (24)

of mete he wes se neodful (133, 1)

In the south-western2 (Titus and Bodleian) manuscripts of Mali Meidenhad there are no syntactic differences of the kind concerned here. In the south-western manuscripts of Layamon (Caligula and Otho), on the other hand, there is a category of noteworthy changes in the Otho text: Caligula

berfore him ofte scomede (6868) eb him wes on heorten (1117) feirest bat heom bohte (655)

Otho

barfore he ofte samede glad he was on heorte fairest bat hiifunde

The examples of personal construction in the Otho MS represent a weak tendency, though significant, in the direction of the use of the NPa in subject function. 1 For the Nero MS see EETS 225. The MSS are 'very close1 according to Mack (EETS 252, p. XVII). 2 Cf. MED, Introduction p. 12.

147

A similar case of personal paraphrase is found in the Trinity manuscript of Cursor Mundi: Cotton

A schippe behoues pe to dight (1665) berfor behoues now man eteflesse (1998)

Trinity

A shippe most pou nedes di?t per fore man mot now eteflessche

The increasingly widespread use of personal paraphrase possibilities can also be noticed in the following examples from the South English Legendary: ·* (25) homofssamede sore (323,42) (26) sore was pe peofofssamed pat... (86, 23) (27) Sorihewas... (159,13) In this context the analysis of ambiguous constructions becomes naturally more difficult, especially if word order corresponds to the unmarked SVO pattern, as in (28) (29)

pe soule longeb berafter sore (476, 377) pe erche bissop... longede after him sore (74, 86)

(28) and (29) are indeterminate between N and II. Of these variants, SEL only uses N: (30)

hym longede after reste (30, 174)

In the absence of a single safe example of type II with pronominal NPa, (28) and (29) should probably be read as type N structures. Consider the Layamon manuscript Caligula, where the inflection of the article disambiguates a structure corresponding to (28) and (29): (31) (32)

pa longede swube Luces bon kinge... of pan Lauerd Criste (5049) Goffar pan king... nes hit noht iqueme bat... (705)

The next set of examples with ofpincen illustrates the complex syntactic and semantic variation pattern that has to be taken into account in dealing with ambiguous constructions. (33) is a sentence with nominal complementation which is also indeterminate with regard to the verb-NPa relation. The two possible readings are as type I or the transitive construction: (33)

be godeman sone ofpohte is bo-jt (122)

Now in SEL ofpincen is regularly used with non-nominative NPa, as in: (34) (35) (36)

148

me ofbincb bat... (337,308) it ne ssel ^ou ofbenche no^t (138,44) For beose miracles hure ofbo^te... (116,189)

However, the verb also occurs in co-ordination with the personal construction: (37)

we beob anyd and ofbencheb sore (642, 1005)

The morphological singular/plural distinction is neutralized, and we cannot decide on either of the two intransitive readings. Compare, however, that a personal paraphrase is also available: (38)

of be uuel... neuere eft igreued he nas (278, 122)

The two situations illustrated above are representative of the fact that in the 13c. the syntactic expression of the pseudo-subject relation holding in S (me reweth that...), N (me reweth of...) and I (me reweth the ded) did not necessarily become indeterminate with full nouns or in co-ordinated (or relative clause) construction. This situation again speaks against the radical change hypothesis. The regular OE types are also the productive variants in most ME texts studied - a fact which reflects that the semantic situation was still the original one. While the transitive and personal constructions are available in the grammar of 13c. English, they are still only sporadically used. A marked exception in my material is the Otho manuscript of Layamon, where the use of a nominative NPa in both the transitive and intransitive construction is evidently the rule. Apart from this single case, however, it appears that the proper subject relation was more often expressed in paraphrase constructions. Compare the following examples from Ancrene Wisse: (39) (40) (41) (42) (43)

Ich am ofdred leste... (176,7) Ich am offearet sare pet... (176,11) Habbep bah to ower bihoue bis... (175, 5) Sorhful ich am and sari bat... (47, 1) Two manere men habbeb neode to eote wel (134,19)

Constructions of the form (39)—(43) did not replace the traditional use with simple predicates, but they show that syntactic ambiguity as represented above could easily be avoided if the desired semantic interpretation was not ensured.

4.

The 14th century

The main change between 13c. and 14c. syntax concerns the use of the verbs in question with nominative NPa to form the personal construction with sentential complements. At the same time, we can observe the beginning of a tendency to use direct objects besides the original (genitive or prepositional) 'causative' one. The result is in both cases a transitive SVO structure conforming to the general pattern of declarative sentences.

149

Within the century-framework relevant for chapter HI both the personal and transitive types represent possible syntactic variants in the 14c. paradigm of subjectless verbs. In the present context, however, the potential occurrence properties of verbs in the 14c. will be evaluated relative to their actual use text-intemally. This method will allow us to assess the relationship between the abstract representation of diachronic development (as in chapter III) and the nature of actual syntactic and semantic variation within a single dialect. It has already been noted that there appears to be a tendency in the Trinity manuscript of Cursor towards the construction with nominative NPa. The preferred variant of constructions with sentential complements is thus the personal one, as represented, in (44), by the necessity verb owen: (44)

men owe to ponke him (3692)

The other manuscripts have type S or syntactically indeterminate constructions such as men aght to thane him (3692 Cotton). Longen occurs with subject-NPa in the Cotton text: He wil noght lang pat i be her (20306), where the other manuscripts have different syntactic paraphrases. Similarly, thu may hapin to sla sum dere (3602 Gott.) is rendered by altogether different constructions in the other texts.1 These examples are representative of the fact that genuine paraphrase variation is used rather than variation on the basis of the construction types which form the syntactic paradigm of subjectless verbs. In all manuscripts, however, the traditional construction with the pseudo-subject role of the NPa is still quite regularly used. That the occurrence with nominative subjects in intransitive construction (types I, II) and the personal type had not yet led to a change in the regular use of originally subjectless predicates is clearly shown by the scarcity of the transitive construction. Only fort hinken is attested in this type — a verb distinct from the other members of the same set by its association with transitively used pincen. Compare (45)

sco hir sin forthoght (25792)

But the regular occurrence of the NPa in subject function is still found in type II: (46) (47) (48)

of (other MSS:on) him pou reu (14173) ο me he suld reu (18419) ... pin husband par, pat pou has longed efter sare (20306)

A more restricted situation is found in the Ayenbite. The verbs concerned here are still regularly used with a non-nominative NPa, either in the syntactically unmarked it-construction, as in (49)

Vor hit him pingp pet... (135,19)

1 t>ou mai drep me sum dere (Cotton), pou may gete sum wilde dere (Bod.), pou may sie sum dere (Trin.).

150

(50) (51) (52)

Hit fringf) to pefole pet... (184, 34) ...pet him hit likep to zuerie (63,23) hit him behouep pet... (211,14)

or in the regular S variant: (52) (53)

him behouep paye (137,34) ase him pinep to done wel... (141, 32)

Likewise, there is a tendency for the indirect object relation to be prepositionally expressed, as in the following examples: (54) (55) (56) (57) (58)

ase behouep to pe dyeule (56,23) More behouep to one kuene panne... (216, 25) ... pet hit misualle to him (193,19) ase merci likep to god alsuo hit ne likep... to pe dyeule (187, 25) No chastete lykep to god (207,7)

In this text again the personal construction is not at all a productive variant, although the type must clearly be assumed to be part of the 14c. syntactic paradigm of the verbs occurring in (49)—(58). The personal construction is only met in the semantic context 'necessity'; the complex predicates of this class thus allow the relational three-way variation to be expressed: (59) (60) (61)

... huet ham is nyed (208, 23) hit is ous nyeduol (112,18) We habbep niede oftuo ping (209, 23)

The grammatical use of subjectless verbs in the Ayenbite does not reflect the process of continuing personalization and transitivation; it remains quite conservative, possibly due to the particular history of the tradition of this text. Both Chaucer and Wyclif are conservative in their syntactic expression of the verb-NPa relation. With sentential complements Chaucer uses bihoven only once with nominative NPa, whereas liken and shamen occur in the form of S and the if-type, but apparently not the personal construction. Whereas in the text ofBoethius we come across the variation pair (62) a the byhoveth discovre thy wounde (Bo. 1 p. 4,95-100) b it behoueth the to shewen it (Bo. 2 p. 3, 325-30) there is only a single occurrence of bihoven used as in (63): (63)

/, servant of God, bihove not to chyde (Pars. 630-5) 151

The following syntactically indeterminate example is disambiguated towards the intransitive reading by the semantic context: (64)

that hir housbonde liked for to seye (Mel. 2250-5)

Transitive use of liken with nominative NPa, as in (65)

she lyked him the bette (LGW 1076)

is still very rare. The only other example appears to be she gan encline To like him first (TC 2, 675), but compare the intransitive use in we! was him that coude best devise To liken her (TC 3,613). With listen, personal construction, as in (66)

if ye lyst of me to make... (RR 1967)

is most rare. I have found only two further instances, both in relative clause constructions. A type S reading is not possible here, as the verb agrees with its plural subject: (67) (68)

Ye sustren nyne eek, that... listen for to abyde (TC 3,1810) he that precheth to hem that listen not heren his wordes (Mel. 2234)

If the NPa is a proper noun or a relative pronoun, structures with sentential complements remain indeterminate: (69) (70)

Thogh Juno lyst not at the feste bee (LGW 2249) Who list to havejoie... (RR 5028)

The 'necessity' predicate owen is also used in the variation pattern formed by S and the personal type: (71) a ... god... to whiche him oghte to ben obeisaunt (Pars. 675—80) b wel oghten they thanne have desdayn (Pars. 150—5) It is impossible to decide, however, whether the marked word order in the following examples was enough to disambiguate their functional interpretation: (72) (73)

wel oghte a man have desdayn (Pars. 140-5) muchel oghte a man to drede... (Pars. 160—5)

In construction with nominal complements the transitive type is only attested with neden, a verb of the 'necessity' class used with nominative subjects already in OE. Compare in this context us lyketh yow (Cl. 106), which combines the general use of pseudo-

152

subjects with liken with the direct object form of the complement NP. Examples like this are usually adduced to illustrate the state of 'confusion' in ME syntax. In the light of 14c. evidence met so far, we can only accept a much weaker version of this statement, viz. that this shows that the transitive construction was still not the regular variant in the present lexical context. As in Cursor, type I is the indigenous representative here. The situation is quite similar in the English works of Wyclif. In construction with sentential complements the it-type is very productive. The variation pattern between S and the personal construction is used with shamen: (74) a hem shamede to seie pis openli (357,15) b & jit frei schäme sumdel to seie... (266,16) with the result of syntactic ambiguity as in (75)a pe fend shulde shame... to seye pat... (418,27) b why shulde a prelat shame to lyue pus... (414, 7) Besides SVO word order there is also text-internal evidence which favours a personal reading of (75): the NPa is often prepositionally marked, even in initial position: (76) (77)

To me, seip god, it longip to do vengance (362,20) To pise sectis ceessip not pe iugement pat... (303,12)

We have to assume that, if the desired reading of (75) had been other than personal, the sentences could have been easily disambiguated syntactically. Again the only predicate used transitively by Wyclif is neden 'need'. The results achieved on the basis of the above texts are representative of the 14c. situation. (In Gower, for example, listen is used with nominative NPa besides the regular Sconstruction, and the occurrence of rewen in the transitive construction He may that werre sore rewe (CA, Prol. 1004) is still exceptional.) Type II (he reweth of...) is the productive variant in construction with nominal complements. Otherwise the use of verbs with the NPa as subject is practically limited to intransitive constructions of the type as ye list and as he wel liked. Synchronic 14c. evidence thus adds a realistic corrective to the diachronic account in the previous chapter. As a possible grammatical variant, the construction with nominative NPa is available, and the semantic structure of originally subjectless predicates does not prevent them from occurring in this way. This fact has to be stressed in view of the occurrence (in some cases since OE) of some of these verbs with nominative NPa in intransitive construction without complements. The important aspect in the present context is thus not the syntax but the fact that the relational variation involved in a pair ic licige/me licep does not express a semantic change. This situation establishes the two 153

main factors determining the 14c. paradigm: the semantic structure of individual verbs and their use by different authors. It will have become clear that there is no question of a 'change from impersonal to personal construction' in this general sense. It is a significant fact about language use that in the 14c. there is a marked tendency to use the if-type — and not the personal construction - as a variant besides S. In terms of semantics this means that there is no semantic change involved. Both constructions are intransitive, and the unambiguous syntactic expression of the indirect object relation is possible at all periods. In variation with nominative subjects, however, there is a change in the role properties of the NPa, from recipient to (quasi-) agent. In this context, then, it does not seem to be accidental that only shamen is used with NPa subjects, as the semantic variation here is minimal.

5.

The 15th century

In abstract terms of 15c. as against 14c. syntax there is no change. In terms of use, however, there is a growing tendency for originally subjectless verbs to occur with nominative NPa. We shall see, however, that this trend is not followed by all authors. The main feature of the 15c. situation is the difference between the more conservative or progressive character of texts. Normative syntactic pressure was obviously not yet decisive, nor had the specific pseudo-subject relation become obsolescent. In line with this situation our method in the following will be on the one hand to deal with one text/author in detail and on the other to discuss the use of some representative verbs and paraphrase constructions in different authors. The work of Malory offers a rich data base for the first part of this task. The most productive variant of the construction with sentential complements is the //-type: (78) (79) (80)

hit behovyth the now to chose (257,29) Now hit befallyth me... to stirre me (263, 3) hyt happynd hym that... (1066,13)

Listen does not occur in this way, but in the variation pattern formed by S and personal construction: (81) a Me lyste not to teile, seyde Balyne (81,9) b yfye lyste to fyght so (111, 22) The simple predicate shamen is used with a subject-NPa; but compare the paraphrase type, corresponding to the /'f-variant, with complex predicates: (82) (83) 154

hit were shame for me to se three knyghtes... (273,13) hit were shame for us... to sette upon ... (1113, 16)

(82) and (83) represent the productive variant in Malory. The most frequent occurrence with nominative NPa is still in intransitive construction, as in the type represented by do asyelyst (536). With sentential complements liken is not used with NPa subjects, but with nominal complements there is the variation pattern between type I and the transitive construction: (84) a every knyght toke the way that hym lyked beste (872, 29) b I trow ye lyke nat youre mete (956,27) Word order and the unmarked interpretation of co-ordinated structures disambiguate the following sentences in the direction of the transitive SVO reading: (85) (86)

/ lyke bettir the swerde (54, 8) kynge Arthure loked on the swerde and lyked it... well (54, 5)

Despite the availability of the transitive construction, liken is still regularly used in the intransitive type represented by (87)-(88): (87) (88)

as hit lykith you (1079, 36) // hit lyke you I woll speke (177,3)

The range of syntactic paraphrase constructions and the semantic and relational variation pattern associated with them can be illustrated on the basis of the semantic context 'shame'. Besides the construction with subject-NPa as in (89), there is the productive syntactic variant (90) and (91): (89) (90) (91)

/ shamed to se so many good knyghtes ayenste hym (1114, 14) it wasgrete shame unto them ... to be overgovernyd (15, 23) Hit were shame to us (160, 6)

Compare also the further possibility with a complex predicate: & yet he thought shame to sie hem (170, 22). Quite a similar situation holds in the semantic context 'necessity' with the main verb neden. Two aspects of the above text-internal survey characterize the syntactic behaviour of subjectless verbs in the 15c. First, most predicates can now be optionally used with a nominative NPa, but this pattern is still regularly realized by the intransitive construction of types I or II, or by an intransitive paraphrase of the type shown above. Personalization — the occurrence with nominative animate-NPs — does not necessarily involve transitivation at the same time. With sentential complements the unambiguous subject role of the NPa is far less common than the indirect object role in the it-type. There is an obvious preference for this construction in the work of Malory, Margery Kempe and Mandeville. Type S likewise continues as a productive variant, despite the fact that with non-pronominal NPa the force of SVO word order will often suggest a personal re-interpretation. 155

The occurrence paradigm of liken allows all three relational variants to be illustrated: (92) a b (93) a b c (94) a b c d

it lykede Him to takeflesch (Mand. 1,6) yifyou lyke to knowe the vertues... (Mand. 116,21) & ban lykyd hym ful wel bat... (MK 144,3) whan it likyth me to spekyn (MK 183, 7) bei... lyked not be menys... (MK 60,9) it liked him to deliver him out of prison (PL I, 97, 27) and that ...ye lyke to sende me redes lettres (PL I, 20, 21) as hir should lyke the Kyng (PL I, 377, 28) he lekyd the contre ryght wel (PL 1,179,20)

Second, the use of paraphrase types is considerably developed in order to disambiguate the verb-NPa relation in either of two possible directions illustrated by examples from Mandeville: (95) (96)

yee shulle ben nedy ofallegodes (108, 7) it is nedefulle to him that... (117,22)

The traditional pseudo-subject relation is replaced by a proper subject in (95) and by an indirect object in (96). In contrast, the main necessity predicate neden is only used with the NPa role as in (97): (97)

of alle thing that hem nedeth ...(185,12)

Not only is it still possible for the syntax to express the original verb-NPa relation where desired; but there are verbs (like neden and liken) which are regularly used in this way. This situation leads to a considerable difference in use among individual authors. Whereas Lydgate, for example, frequently uses listen with a nominative NPa, this does not seem to be the case at all with Hoccleve. It is quite difficult to give an account in terms of gradual change in productivity of the different syntactic variation patterns. In the early 14c. texts of Mandeville the personal construction occurs only scarcely except in paraphrase. For works of the second half of the 15c., however, it is impossible to note a gradual change in syntactic use, although the preference for the it-type — with certain verbs — is a prominent feature. The history of types S and I, however, can only be adequately assessed in the light of 16c. evidence.

6.

The 16th century

The most extensive linguistic study of a 16c. text is Visser's work on St. Thomas More1, on which the first part of this section will be based. With regard to the use of originally subjectless verbs, Visser notes that the language was still 'unsettled' at the time (p. 19). 1 Visser (1946). All references are to pages in part I. 156

Although these verbs now form only a very small lexical group, they still validly represent subjectless grammar. The traditional verb-NPa relation continues to be expressed by types S and Ib, but there are marked differences between the occurrence properties of individual predicates. Become, behove, grieve and happen no longer occur in type S. That the pseudo-subject relation of the NPa to behove and grieve, at least, had not yet become obsolescent is shown by their occurrence in type I: (98) (99)

... to geate that thee... behoueth (26) the thyng whych greueth this blessed bretherhed... (27)

Rue is the only member of its class attested in type S: (100) me rueth so long to haue be thrall (24) However, seem, list and think are frequently found in this construction. These predicates show the variation pattern between S and the personal type, illustrated in (101) by list: (lOl^z me lyst not frendly on them lake (23) b ... as she list to put unto them (27) As (102) shows, the semantic context cannot always disambiguate structurally indeterminate sentences: (102) as soon as Fortune list to laugh (23) With the other verbs there is a general tendency to occur in the it-type: (103) // behoueth hym to dye (18) (104) it becommeth us... to make frendes (17) Boot, long 'suit', grieve, fortune, happen, like and seem also enter this construction. Seem is thus used in the three-way variation paradigm (105)« him semed that... (24) b whiche thinge sith ye seme to impugne (24) c It semeth me... that... (23) There is no evidence that More preferred any of these three syntactic possibilities. A slightly more restricted situation appears to hold in the works of Kyd and Marlowe. Whereas think is frequently used with a pseudo-subject, there are only single occurrences of seem and behove (Kyd) in this construction. The regular type entered by behove and boot is the //-construction, represented by examples from Marlowe: (106) the torment that it boots me not reveal (D 1022) (107) it behoves us to be circumspect (Ατά. 2256) 157

Shame, like, happen, list and long only occur in the personal type, such as (108) I shame to hear thee speak (Marlowe, 3H6 240) (109) / long to hear it at full (Marlowe, 2H6 904) It is noticeable that langen is often used in this way by Marlowe. In contrast to Marlowe and Kyd, 16c. writers like Spenser and Sir Thomas Elyot retain the more conservative use of the S-variant.1 In the works of Shakespeare we find general 17c. use reflected. Type S is preserved v/iui behove, seem and think, whereas grieve, long and shame enter the personal variant, represented by (110)—(112): (110) I grieve to hear what... (Henry VI, 1.4.57) (111) I do shame to think of... (Cymb. 4.3.23) (112) / long to know the truth (Com. Err. 4.4.143) The only predicate still used in all three constructions is again seem: (113)a me seemeth then it is no policy (Henry VI (2), 3.1.23) b it seems to me that yet we sleep (Mid. Night's Dream 4.1.193) c though I seem to drown her remembrance again (Twelfth N. 2.1.31) In order to sum up the result of this investigation, let us look at the survey charts representing the constructions with sentential and nominal complements from the 15c. onwards. They reflect the processes of personalization and transitivation respectively and show the transition periods.

l Cf. Ammarm, R.E. Die Verbalsyntax in Sir Thomas Elyots 'Governor' mit vergleichenden Beispielen aus Roger A schams 'Schoolemaster'. Diss. Bern (1961).

158

Originally subjectless verbs in construction with sentential complements

15c.

16c.

17c.

18c.

19c.

rue shame ail grieve forthink like list

·°Δ ·Δ ·Δ ΟΔ ·ο ·ΟΔ ·ΟΔ

·°Α ΟΔ _ ΟΔ ο ·ΟΔ ·Δ

Δ ΟΔ _ ο

_ ΟΔ ο

_ ΟΔ _ ο

Δ ·ΟΔ

Δ ΦΔ

Δ ·Δ

long

·ΟΔ

·ΟΔ

Δ

Δ

Δ

Δ

·ΟΔ · ο ·ο ΟΔ ΟΔ ·ΟΔ ·Δ

ΟΔ

ΟΔ Δ ο ο Δ Δ ·ΟΔ ·Δ

behove bus become boot happen fortune seem think

·° ·Δ ·ο · ·ΟΔ ·ΟΔ ·ΟΔ ·Δ

ο ο ΟΔ ΟΔ ·ΟΔ ·Δ

.

ΟΔ Δ ο ο Δ ·ΟΔ ·Δ

·: TypeS ο.· /Y-construction Δ: Personal The survey chart shows that type S does not suddenly disappear in the 16c., although it is discontinued with individual RUE verbs and the HAPPEN class. If we take into account the fact that from the 12c. onwards type S structures are liable to give rise to indeterminacy, the longevity and productivity of this construction only go to show how strong the desire for the indigenous expression of the original verb-NPa relation was. In the 17c. the construction becomes obsolete with the RUE and BEHOVE predicates. List, seem and think, however, continue to enter this type until the 19c. Note that only in the case of like and long is it possible to speak of complete personalization of the verbNPa relation. The HAPPEN verbs, although occurring in a syntactically personal construction, cannot undergo the role change to agentive subjects. The development of the syntactic valency of grieve, boot and become proceeds in the other direction: with these verbs the traditional recipient role of the NPa is generalized. The kind of paraphrase construction met in the 16c. also reflects these two pathways of change. That the verb-NPa relation expressed by complex predicates corresponds to that met with simple verbs is shown by the following examples: (114) and to begge lama shamed (NT Mat. 21, 3. All versions) (115) Hercules... was enshamed this myry wose... to carye (Sk. 380, 5) (116) but unto theym it was more greuous than ony dedely woo (Sk. 107,2) 159

(117) it shall bee to greevous... for thee to systayne (Fisher 427,16) (118) it isgreuous... unto them (Fisher 241,1) (119) are not you griev'd that... (Sh. King John 3.4.123) (120)—(123) represent paraphrase constructions within the semantic framework 'behove', with the animate-NP prepositionally marked: (120) (121) (122) (123)

behoefful unto mannes lyf (Sk. 40,15) as it is sittynge and metely unto euery man (Sk. 6, 7) it is... accordyngefor synners to wayle (Fisher 31, 28) benefycyall unto hym (Fisher 3,20)

Originally subjectless verbs in construction with nominal complements

15c.

16c.

17c.

18c.

19c.

rue shame grieve forthink ail like list long

ΟΦ·Δ ΟΘ·Δ ·Δ ΟΦ·Δ ·Δ ·Δ ΟΘ· ο·®

Φ·Δ ®Δ ·Δ ·Δ ·Δ ·Δ ®· ®

ΦΔ ®Δ Δ Δ _ ·Δ _ ®

®Δ Δ Δ Δ

ΦΔ Δ Δ

·Δ _ ®

·Δ _ θ

behove become happen fortune

· · · ·

· · · ·

· · · -

· · · -

· · · -

ο: Type Ν ·: Type I ®: Type II Δ: transitive In the 16c. type I is still quite productive, but from the 17c. onwards it is only preserved in those cases where the semantic situation makes clear that the NPa functions as experiencer/recipient, (like, behove, become, happen). With like in particular the mutual exchangeability of semantic roles leads to a productive variation pattern between I and the transitive construction up to the 19c. The example of rue (and shame in the 17c.) shows that there was no complete transitivation in this semantic context, in which the 'causative' object relation continues to be expressed by type II.

160

Apart from the RUE verbs and like, however, originally subjectless verbs do not undergo the general process of transitivation. Also, the fact that the capacity to enter into the different relations with the NPa is not the same in both complementation types does not invite sweeping generalizations. We would expect that if a verb enters the pseudo-subject relation in type I, it will also enter type S and vice versa. This is not true: forthink does not take subject-NPa with sentential complements, but enters the transitive construction with direct objects. With long and behove, on the other hand, the situation is just the opposite. Subjectless predicates formed only a small lexical group in OE, as in the other Germanic languages. They have an especially prominent status within the grammar because they express basic situations and events with which human beings are confronted. This semantic context finds its indigenous expression in the syntax of types S, Ib and N. The constructions survived despite general case opacity in ME, because of the regular association of subjectless verbs with pronominal animate-NPs. Even the gradually generalized syntactic constraint on initial non-nominative NPs could not completely oust the traditional expression of the pseudo-subject relation, as the example of list, seem and think shows. It is certainly true that the 16c. marks the turning point in the history of originally subjectless verbs, as shown by the last occurrences of types S and I with a number of predicates. But to describe the development in absolute terms is inadequate. As has been amply demonstrated in this and the previous chapter, there can be no question of a wholesale semantic change associated with the development of syntax in the 15c.-17c. After the 16c. RUE verbs continue the variation pattern between recipient and agentive NPa role. Such usage had always been in the language, as OE personal paraphrases of the type He ... bib suibe sorig (CP 227, 8) show. The change after the 16c. consists in a reduced syntactic paradigm. Besides the quasitransitive RUE verbs regret, repent, grieve, the intransitive use is preserved by constructions of the type / am sorry ..., I feel remorse, ... Nor does it make sense to speak of semantic change within the framework of 16c. PLEASE/DESIRE predicates. Like is used in the meaning 'please/like' already in the 13c. (and probably even earlier). In the type with nominal complementation, the continued variation pattern between I and the transitive construction neatly preserves this semantic distinction. That the pure DESIRE verb long should adopt the personal variant as its only syntactic occurrence possibility is semantically understandable. List, on the other hand, again reflects the possible dual view with regard to semantic roles by preserving the impersonal and personal variants side by side. The post-16c. behaviour of BEHOVE verbs is semantically determined: become and boot adopt the /f-construction, but behove — often used as a necessity predicate — enters the variation pattern between a recipient and agentive reading of its NPa. From the point of view of semantics we would expect the HAPPEN verbs to take the same course of development as the pure BEHOVE verbs. This is indeed the case in the 16c. and 17c., but afterwards they only enter the personal construction (in the variant with 'raised' subjects: He happened to come). As mentioned above, there can be no reanalysis of semantic roles in this context. The productivity of the variation which is con161

tinued with seem and think, finally, highlights the attractivity of this kind of stylistic device. It will have become clear that the description of OE subjectless grammar cannot be based on a single factor such as the establishment of syntactic SVO. The development of unmarked declarative syntax must necessarily serve as the potential descriptive framework of any account. Within these formal limits and those of general factors like the targets mentioned in chapter I, however, the syntactic expression of the verb-NPa relation remains a matter of individual development. A study of the present kind necessarily remains a pilot work in many ways. Hopefully, it will have contributed, besides suggesting new parameters for the description of historical syntax, to the independent characterization of the development of Old and Middle English subjectless grammar.

162

Appendix: Translation of OE examples used in the introduction and chapters I and II. (Translations taken from text editions are given in inverted commas). Introduction (4) Now it seems tome that... (5) 'and yet (he) is not willing to do his will...' (6) Often the most patient man is ashamed of the victory...' (7) He longed for this. Chapter I (8) I regret (in my mind) that they should possess heaven. (9) 'diciples...; it repents them that they are of the Jewish race.' (10) 'art thou not ashamed to cut off that which...?' (11) 'he was annoyed at his being better than himself. (12) 'Then it seemed good to her... to take up the bones...' (13) He also desired to see how... (14) The men longed for the time when they might step from durance.' (15) We are pleased yet further to speak... (16) 'now he had good cause to creep into his capacious bag of tricks...' (17) But to them it behoves to bury the people. (18) Ά bishop should be blameless.' (19) 'that it became no man better to prove the work...' (20) 'It sometimes happened to the abstinent that they...' (21) 'and it was thereupon agreed amongst them all to find a new sword'. (22) 'it then happened to him suddenly, by the grace ... that he was healed.' (23) Now it seems to me that... (24) We are ashamed to tell all the shameful sorcery. (25) 'and (they) feel ashamed of them (= sins), that they may not do so again.' (26) They were not eager to ask. (27) 'It behoves him to think...' (28) '... it happened (to him) that he hardly knew whether he was a man at all.' (29) It seems to me that... (30) And every man desires that he might rest. (31) Truly, it is fitting for the preacher to... (32) But it is fitting (for them) to be obedient towards their bishop. (33) 'that very frequently to many a man it suddenly befalleth that He cuts him off.' (34) "Then she is sorry that she did not fulfil her vow.' (35) Afterwards it pleased to all the people that... (36) "Then it seemed good to her... to ...' (37) 'Then it happened that he died.' (38) Then it happened (to her) that she neglected them. (39) 'and (they) take delight in doing what is unlawful as it were lawfully' (40) 'It happened in his youth that he committed...' (41) He greatly rued that... (42) 'But if thou believest not that he is the son of God...' 163

(43) (44) (45) (46) (47) (48) (49) (50) (55) (56) (58) (59) (60) (61) (62) (63)

(64) (65) (68) (69) (70) (71) (72) (73) (74) (75) (76) (77) (78) (79) (80) (81) (88) (89) (90) (92) (93) (94) (95) (96) (97) (98) (99) 164

Then Cirus demanded that the ... king should meet him there. 'because he wished to help him more.' ... that they intended to extinguish it. 'and feel ashamed of them (= the sins), that they may not do so again'. It grieved me that they did not know the Gospel. Then it rued him that... 'If we are ashamed to speak so to strangers'. You need not be ashamed when... It is pitiful... for all to hear ... that... All the Senate and the consuls then rued that he wanted to break their old laws. I am pleased with him. It pleased us to agree... He said that he desired to eat and drink with him. We delight in speaking yet further... Many desired very much to see him. Ί found pleasure in the sweetness and beauty of the place, which I had seen there, and at the same time in the enjoyment of the society and bliss of those I saw in the place.' I was very desirous to hear him. If anybody should like to know more about it... It is fitting for you to be born yet again. It is necessary for you to learn and listen to me. It is rightly fitting (for you) to be king. In four ways it is necessary for us to turn to him. 'It is good for men that...' 'It is better for the cunning to be...' It is most fitting for him... It is most necessary for us to strive after... It is most necessary for everybody to... He is good to his friends. 'Non countur ludaei Samaritarns'. We need... We need... 'It becometh not any rich king ...' It seemed to him more advisable to make peace with the others. It seemed tedious to me to grant you this. It seemed to them too long until the time when they should again carry spears. 'that... he may not desire to please his subjects rather than God'. The king's brother gathered many people and intended to conquer England. She ..., and it was fitting for her ... to be a virgin. '... said that it seemed to him as if he were walking over flowers'. '... that I found a generous distributor of rings, enjoyed him while I could.' 'God had pity then, and quickly said to the angel: ...' Then was he first afraid ... and went to the land of the Bruti. It seemed to Antiochus ... that his kingdom was not enough, and he wanted to take the (people of the) Parthe.

(100) (101) (102) (103) (104) (105) (106) (107) (108) (109) (110) (111) (112) (113) (114) (115) (116) (117) (118) (119) (120) (121) (122) (123) (124) (125) (126) (127) (128) (129) (130) (131) (132) (133) (134) (135) (136)

After this the Macedonians heard that... and found None of them could tell his name. 'when a man careth not to do any good in his life*. It is necessary for you to learn ... The Lord ordered you to say this... 'the light may be seen shining'. 'They gave him poison to drink.' I will send this people meat to eat. It is fitting for the king that... (It) happened that... It happended some day that... It was fitting that... (It) happened that... It is no wonder... Then (it) happened that... Therefore it happens that... Ifitever happens that... Therefore it is better that... Often it also happens that... It is widely known that... (It is) known that... 'It is the practise of many a man...' It was also quite fitting that... It happens that... It is necessary that... 'it is unnatural presumption ... to...' Then it happened that... Then it happened that... It is not fitting that... Then it happened that... Then it happened that... It happened some day that... After the Lord's suffering ... it happened that... Then it happened that... 'It was chiefly through his information that...' After this battle it happened that... In those days it happened that...

Chapter II (1) 'He was sorry for man'. (2) '... he is indignant at what he formerly passed over.' (3) You were ashamed of such an error. (4) He desired nothing in this world. (5) I always desire that which... (6) Man is ashamed of good deeds. (7) My soul desires loving God.

(8) (12) (13) (14) (15) (16) (17) (18) (19) (20) (21) (22) (23) (24) (27) (28) (29) (30) (31) (32) (33) (34) (35) (36) (37) (38) (39) (40) (41) (42) (43) (44) (45) (46) (47) (48) (49) (50)

166

He rues that... / the deed. Then the priest was sorry for the man and... And he was pleased with their customs and thanked God... Although they desired it, they ... It may displease the prince of the Heathobards. You should not be ashamed of your offspring. This he desired most. The warriors were eager to win... 'So greatly I rejoice in these painful torments.' "The queen at once rejoiced in the good intent.' "Then he began to rejoice at the bishop's words.' The king began to take pleasure in the pure life of holy men. If you need a smaller candel. When you need the Psalter. Then one of the king's followers regretted this... This victory the Punici won. Everybody should please God with good deeds. That there nothing was fitting for such a noble king... It may not grieve you worse to think... 'lest (we) repent of our alms'. 'If the greedy were not afflicted with loquacity...' How he was pleased with the victory. They need Christ's help. If he should happen to die. Songs of praise are due to you. 'Everything might befall him as...' a Then one of the king's followers was sorry for this. b Do you regret all the evil you have done? Then the priest was sorry for the man and... 'he needed not feel shame at the precious gift'. Then the bishop began to be delighted by the wisdom of the young man. The soul... desires the Lord's kingdom. 'Now our freedom ever needeth God's assistance'. He also needs much help. He does not need this unbecoming garment. a Then the priest was sorry for the man and... b He was sorry for the man. a If you will repent of your mistake b Often the most patient man is ashamed of the victory.' a When the soul... desires God's kingdom b When he ... desires food Ί also am sinful and need mercy of the heavenly Lord'.

References Admoni, V.G. (1974) An attempt at classifying the grammatical theories in contemporary linguistics, Linguistics 117, 5-20. Aitchison, J. (1976) The Articulate Mammal. London:. Hutchinson. Akmajian, A. and F. Heny (1975) An Introduction to the principles of transformational syntax. Cambridge (Mass.) and London: The MIT Press. Andersen, H. (1973) Abductive and deductive change, Language 49,765—91. Anderson, J.M. and C. Jones (eds) (1974) Historical Linguistics: Proceedings of the First International Conference on Historical Linguistics, Edinburgh 2—7 September 1973,2 vols. Amsterdam: North-Holland. Anderson, J.M. (1971) The grammar of case. Cambridge: Cambridge University Press. Antilla, R. (1972) An introduction to historical and comparative linguistics. New York: Macmillan. Bach, R. (1970) Is Amharic an SOV language? Journal of Ethiopian Studies 8,9-20. Bach, E. (1971) Syntax since Aspects. In: 22nd Annual Round Table Meeting, Monograph Series on Languages and Linguistics (ed. R.J.O. Brien), Georgetown Univ. School of Languages and Linguistics. Bach, E. (1975) Order in base strucutres, in Li (ed.) Word order and word order change. Austin: University of Texas Press. Bacquet, P. (1962) La structure de la Phrase verbale ä l'opoque Alfre"dienne. Paris: Les Belles Lettres. Bäiley, C.-J. (1971) Trying to talk in the new paradigm, Papers in Linguistics 4, 312-38. Bailey, C.-J. and R.W. Shuy (eds) (1973) New Ways of Analyzing Variation in English. Washington: Georgetown Univ. Press. Baron, D.E. (1974) Case grammar and diachronic English syntax. The Hague: Mouton. Baugh, A.C. (1973) A history of the English language. London: Kegan Paul. Baumgärtner, K. (1969) Synchronie und Diachronie in der Sprachstruktur - Faktum oder Idealisierung? In: Sprache, Gegenwart und Geschichte. Bazell, C.E. (1958) Linguistic typology. London: School of Oriental and African Studies. Behaghel, O. (1923-32) Deutsche Syntax: eine geschichtliche Darstellung. 4 vols. Heidelberg: Winter. Bertschinger, M. (1941) To Want: An Essay in Semantics. Diss. Zurich Bever, T.G. and D.T. Langendoen (1972) The interaction of speech perception and grammatical structure in the evolution of language. In: Linguistic Change and Generative Theory, RJ*. Stockwell and R.K.S. Macaulay (eds). Bloomington: Indiana University Press. 167

Bloomfield, L. (1935) Language. London: George Allen and Unwin. Botha, R.P. (1973) The Justification of Linguistic Hypotheses. The Hague: Mouton. Bresnan, J. (1970) On complementizers: toward a syntactic theory of complement types, Foundations of Language 6,297-321. Brinkmann, H. (1971) Die deutsche Sprache. Gestaltung und Leistung. Düsseldorf: Schwann Brugmann, K. (1904) Kurze vergleichende Grammatik der indogermanischen Sprachen. Strassburg: Trübner Bynon, Th. (1977) Historical Linguistics. Cambridge: Cambridge University Press. Callaway,M.(1913)The infinitive in Anglo-Saxon. Washington D.C.: Carnegie Institution. Carlton, C. (1970) Descriptive syntax of the Old English Charters. The Hague: Mouton. Chafe, W.L. (1970) Meaning and the structure of language. Chicago University Press. Chambers, R.W. (1932) On the Continuity of English Prose from Alfred to More and his School. Early English Text Society, Original Series 191 . Cherubim, D. (ed.) (1975) Sprachwandel. Reader zur diachronischen Sprachwissenschaft. Berlin: de Gruyter. Chomsky, N. (1965) Aspects of the Theory of Syntax. Cambridge (Mass.) and London: MIT Press. Chomsky, N. (1976) Reflections on Language. New York: Pantheon; London: Temple Smith. Clement, D. and W. Thiimmel (1976) Plädoyer für eine stärkere strukturierung bei der tiefensyntaktischen beschreibung natürlicher sprachen, Linguistische Berichte 41,1-14. Cole, P. and J.M. Sadock (eds) (1977) Syntax & semantics (vol. 8), Grammatical relations. New York: Academic Press. Coseriu, E. (1974) Synchronie, Diachronie und Geschichte. München: Wilhelm Fink Verlag. Delbrück, B. (1893-1900) Vergleichende Syntax der indogermanischen Sprachen (3 parts). (= Brugmann, K. and B. Delbrück, Grundriss der vergleichenden Grammatik der indogermanischen Sprachen (vols 3—5). Strassburg: Trübner) Diachronie Syntax, Parasession to the 12th annual Chicago Linguistic Society meeting (1976). Dougherty, R.C. (1976) Argument invention: The linguist's'feel' for science. Unpublished. Dougherty, R.C. (1973) A Survey of Linguistic Methods and Arguments, Foundations of Language 10,423-90. 168

Ebert, R.P. (1976) Infinitival complement constructions in Early New High German. Tübingen: Niemeyer. Ebneter, Th. (1973) Strukturalismus and Transformationalismus. List-Taschenbuch 1423. Ebneter, Th. (1977) Impersonals äs modal operators, Italian Liguistics 1. Eggenberger, J. (1961) Das Subjektpronomen im Althochdeutschen. Diss. Zürich. Emonds, J.E. (1970) Root and Structure Preserving Transformations. Doctoral dissertation, MIT. Indiana University Linguistics Club. Erben, J. (1972), Deutsche Grammatik. München: Max Hueber Verlag. Esau, H. (1973) Order of the elements in the German verb constellation, Linguistics 98, 20-40. Fillmore, Ch. (1968) The case for case, in E. Bach and R.T. Harms (eds) Universals in Linguistic Theory. Fourquet, J. (1938) L'Ordre des Elements de la Phrase en Germanique ancienne. Paris; Les Belles Lettres. Fries, C.C. (1940) On the development of the structural use of word-order in modern English, Language 16,199-208. Fritz, G. (1974) Bedeutungswandel im Deutschen. Neuere Methoden der diachronen Semantik. Tübingen: Niemeyer. Gaaf, W. van der (1904) The Transition from the Impersonal to the Personal Construction in Middle English. Heidelberg: Carl Winter's Universitätsbuchhandlung. Gardner, F.F. (1971) An analysis of syntactic patterns of Old English. The Hague: Mouton. Glinz, H. (1970-72) Deutsche Grammatik. 3 parts. Bad Homburg: Athenäum Verlag. Green, G.M. (1970) How Abstract is Surface Structure? Chicago Linguistic Society 6, 270-81. Greenberg, J.H. (ed.) (1966) Universals of Language. Cambridge (Mass.): MIT Press. Haiman, J. (1974) Targets and syntactic change. The Hague: Mouton. Haiman, J. (1976) Agentiess sentences,Foundations of Language 14, 19-53. Halliday, M.A.K. (1967) Notes on transitivity and theme in English, Journal of Linguistics 3,199-244. Held, K. (1903) Das verbum ohne pronominales subject im Althochdeutschen, Palästra 31. Hudson, R.A. (1976) Arguments for a Non-transformational Grammar. Chicago and London: The University of Chicago Press. Hutchins, W.J. (1975) Subjects, Themes and Case Grammar, Lingua 36, 101 ff. 169

Jacobson, R. (1976) The London Dialect of the late Fourteenth Century. The Hague: Mouton. Jespersen, O. (1909—49) A Modern English Grammar on Historical Principles. 7 Parts. London: Allen and Unwin (repr. 1970—74). Johnson, D.E. (1976) Toward a theory of relationally-based grammar. Bloomington: Indiana University Linguistics Club. Isenberg, H. (1975) Diachronische Syntax und die logische Struktur einer Theorie des Sprachwandels, in Cherubim (ed.) Sprachwandel. Itkonen, E. (1976) Linguistics and Empiricalness: Answers to Criticism. University of Helsinki, Department of General Linguistics. Kanngiesser, S. (1972) Aspekte der synchronen und diachronen Linguistik. Tübingen: Niemeyer. Kiparski, P. (1970) Historical Linguistics, in J. Lyons (ed.) New Horizons in Linguistics. Klima, E.S. (1964) Relatedness between grammatical systems, Language 40,1-20. Kooij, J.G. (1971) Ambiguity in natural language. Amsterdam: North-Holland Publishing Co. Koster, J. (1975) Dutch as an SVO language. Linguistics Analysis 1,111 ff. Lakoff, R. (1968) Abstract Syntax and Latin Complementation. Cambridge (Mass.): MIT Press. Leech, G.N. (1969) Towards a Semantic Description of English. London: Longman. Lehmann, W.P. (1973) A structural principle of language and its implications, Language 49,47-66. Lehmann, WP. (1974)Proto—Indo-European Syntax. Austin: University of Texas Press. Lehmann, WJ>. (1978) Syntactic typology. Sussex: The Harvester Press. Lehman, WP. and Y. Malkiel (eds) (1968) Directions for historical linguistics. Austin: University of Texas Press. Leisi, E. (1971) Der Wortinhalt. Seine Struktur im Deutschen und Englischen. Heidelberg: Quelle & Meyer. Li, C.N. (1975a) Word order and word order change. Austin: University of Texas Press. Li, C.N. (ed.) (1975b) Subject and Topic. New York and London: Academic Press. Li, C.N. and S.A. Thompson (1976) On the issue of word order in a synchronic grammar: a case against 'movement transformations', Lingua 39,169—81. Liefrink, F. (1973) Semantico-Syntax. London: Longman. Lightfoot, D.W. (1974) Indeterminacy in Syntax, Canadian Journal of Linguistics 19, 150-165. 170

Lightfoot, D.W. (1976a) Diachronie syntax: Extraposition and deep structure re-analyses, Folia Linguistica 9,197-209. Lightfoot, D.W. (1976b) Syntactic change and the autonomy thesis, Paper presented at the Santa Barbara Symposium on the Mechanisms of Syntactic Change. Lightfoot, D.W. (1979) Principles of Diachronie Syntax. Cambridge: Cambridge University Press. Lockwood, W.B. (1968) Historical German Syntax. Oxford: Clarendon Press. Lyons, J. (1968) Introduction to Theoretical Linguistics. London and New York: Cambridge University Press. Lyons, J. (1977) Semantics. 2 vols. Cambridge: Cambridge University Press. Malsh, D.L. (1976) Clauses and quasi-clauses: VO order in Old English, Glossa 10, 28-39. Marchand, H. (1951) The syntactical change from inflectional to word order system and some effects of this change on the relation 'verb/object' in English, Anglia 70, 70-89. Martinet, A. (1962) A Functional View of Language. Oxford: Clarendon Press. Mathesius, V. (1929) Zur Satzperspektive im modernen English, Archiv für die neueren Sprachen 155, 202-10. McCawley, N. (1976) From OE/ME 'Impersonal' to 'Personal' Constructions: What is a 'subject-less' S? In: Diachronie Syntax, Chicago Linguistic Society. Menzel, P. (1975) Semantics and Syntax in Complementation. The Hague: Mouton. Miller, D.G. (1974) On the history of infinitive complementation in Latin and Greek, Journal of Indo-European Studies 2, 223-245. Mosse, F. (1938) Histoire de la forme periphrastique ETRE+PARTICJPE PRESENT en anglais de 1200 ä nos jours. Paris: Klincksieck. Mustanoja, T. (1960) A Middle English Syntax. Part I. Helsinki: Societe Neophilologique. Nehls, D. (1974)Synchron-diachrone Untersuchungen zur Expanded Form im Englischen. München: Hueber. Nickel, G. (1966) Die Expanded Form im Altenglischen. Neumünster: Karl Wachholtz Verlag. Paul, H. (1968) Prinzipien der Sprachgeschichte. Tübingen: Niemeyer. Penhallurick, J.M. (1975) Old English Case and Grammatical Theory, Lingua 36, 1-29. Peters, S. (1970) Why There Are Many 'Universal' Bases, Papers in Linguistics 2,27-43. Peters, S. and R.W. Ritchie (1969) A Note on the Universal Base Hypothesis, Journal of Linguistics 5,150—2.

171

Piirainen, I.T. (1970) Generative Modelle in der Diachronie, Folia Linguistica 4,32—37. Pogatscher, A. (1900) Unausgedrücktes Subjekt im Altenglischen, Anglia 23, 262-301. Postal, P.M. (1971) Crossover phenomena. New York etc.: Holt, Rinehart & Winston. Poutsma, H. (1914) A grammar of late modern English. Groningen: Noordhoff Quirk, R. (1965) Descriptive statement and serial relationship, Language 41, 205 ff. Quirk, R., S. Greenbaum, G. Leech & J. Svartvik (1972) A grammar of contemporary English. London: Longman. Quirk, R. and J. Svartvik, (1970) Types and Uses of Non-finite Clauses in Chaucer, English Studies 51,393-411. Ries, J. (1907) Die Wortstellung im Beowulf. Halle a.S.: Max Niemeyer Verlag. Ross, J.R. (1972) The Category Squish, Papers from the 8th Regional Meeting, Chicago Linguistic Society. Ross, J.R. (1973) The Penthouse Principle and the Order of Constituents, Papers from the Comparative Syntax Festival, Chicago Linguistic Society. Roth, W. (1914) Die Wortstellung im Aussagehauptsatz angelsächsischer Originalprosa. Diss. Berlin. Sampson, G. (1975) The Form of Language. London: Weidenfeld and Nicolson. Samuels, M.L. (1972) Linguistic evolution. Cambridge: Cambridge University Press. Sapir, E. (1921) Language. New York: Harcourt. Scriven, M. (1959) Explanation and Prediction in Evolutionary Theory, Science 130, 477-82. Sigurd, B. (1966) Generative Grammar and Historical Linguistics, AI Hafn, 10, 35—48. Sprache, Gegenwart und Geschichte (1969) Probleme der Synchronie und Diachronie. Jahrbuch 1968 des Instituts für deutsche Sprache, Düsseldorf 1969. Staal, J.F. (1967) Word order in Sanskrit and universal grammar. Dordrecht-Holland: D. Reidel Publishing Company. Stockwell, R.P., P. Schachter & B.H. Partee (1973) The major syntactic structures of English. New York: Holt, Rinehart & Winston. Stockwell, R.P. and R.K.S. Macauly (eds) (1972) Linguistic Change and Generative Theory. Bloomington and London: Indiana University Press. Strang, B.M.H. (1966) Some features of S-V concord in present-day English, English Studies Today (Proceedings of the Sixth I.A.U.P.E. Congress) Rome. Strang, B.M.H. (1970) A History of English. London: Methuen. 172

Sunden, C.F. (1918) Essay, 2: A category of predicational change in English. In: Upsala Universitets Arsskrift 1918. Thompson, S.A. (1973) On subjectless gerunds in English, Foundations of Language 9, 374-83. Traugott, E. (Closs) (1965) Diachronie syntax and generative grammar, Language 41, 402-15. Traugott, E. (1969) Toward a theory of syntactic change, Lingua 23, 1-27. Traugott, E. (1972) On the notion 'restructuring' in historical syntax, Stanford Occasional Papers in Linguistics 2. Ullmann, S. (1951) The Principles of Semantics. Glasgow: Jackson, Son & Co., London: Basil Blackwell (repr. 1967) Untermann, J. (ed.) (1973) Theorie, Methode und Didaktik der historisch-vergleichenden Sprachwissenschaft. Wiesbaden: Dr. Ludwig Reichert Verlag Ureland, S. (1973) Verb complementation in Swedish and other Germanic languages, Umea. Vachek, J. (1968) A Note on Future Prospects of Diachronistic Language Research, Lingua 21,483-93. Vater, H. (1975) Toward a generative dependency grammar, Lingua 36,121—45. van de Velde, R.G. (1976) Zur deskriptiven Adäquatheit der Linguistik älterer Sprachstufen, Folia Linguistica 4,125—34. Vennemann, T. (1972) Analogy in generative grammar: the origin of word order, Paper read at the International Congress of Linguists, Bologna. Vennemann, T. (1974) Topics, subjects, and word order: From SXV to SVX via TVX, in Anderson and Jones (eds) Historical Linguistics I. Visser, F. Th. (1946) A syntax of the English language of St. Thomas More, part I. Louvvain: Librairie Universitaire. Visser, F. Th. (1963-73) An historical syntax of the English language. 4 vols. Leiden: E.J. Brill. Wagner, K.-H. (1969) Generative studies in the Old English language. Heidelberg: Julius Groos. Wahl6n, N. (1925) The Old English Impersonalia, part I. Göteborg. Weinreich, U. and W. Labov and M.I. Herzog (1968) Empirical Foundations for a Theory of Language Change, in W.P. Lehmann and Y. Malkiel (eds) (1968). Wilson, R.M. (1968) Early Middle English Literature. London: Methuen.

173

Wittmann, H. (1971) Recent developments in diachronic linguistics, Linguistics 65,90101. Wiilfing, J.E. (1894-97) Die Syntax in den Werken Alfreds des Grossen. 2. vols. Bonn: P. Hausteins Verlag.

174

Index of Old and Middle English verbs occurring in text examples becuman/bicumen 69,70,212,123,127-129,132,142,157 behofian/bihoven22,41,68,72,73,75,101,121, 123, 125,126, 142,143,148,151, 154,157 beseem 123,128 beburfan 55,66,73 birisen 121-123,128, 142 booten 121-123,157 bus 121,123, 125 byrian/biren 121-123 gebyrian 22,24,25,41,43, 55,63,68,73,74 gedafenian 22,24,41,50,53, 55,67,121,143 eglian/eüen 21,68,76,85,88,101,102 fallen 122,128-130,132,133 befallen 130,154 misfallenl28,151 forthinken 8S-89, 100,102,105,150 fortunen 129-132 gerisan41,42, 68 greven 85-88,102,105,157,158 happen 129-133, 154 hreowan/rewen 3,21,25,67,85,86,89,101-103,144, 145,147,150,157 gehreowan 68 ofhreowan 32,49, 60, 63, 72, 74 hreowsian 32, 77 langian/longen 8, 22,60,64,118-121,128,148,150,153,158 leofian 69 Iician/liken22,25,37,38,49,67,107, 108,110-113, 144-146,151,152,155, 156 gelician25,63,68,107, 110 limpan/limpen 129, 132,133 gelimpan 22, 24, 25,43,44,53, 55, 66, 68,129 175

lustfullian 38,65,72 gelustfuUian22,38,75,76 lystan/listen 8,22,24,38, 50, 60,63,72,74,114-116,117,118,152,154,157 gelystan 64 gemaenan 77 neden 147,156 geneodian 66 ofpyncan/ofbincen 21, 32, 34, 60, 64, 67, 71, 85,86,102,144, 148, 149 owen (oughte) 103,150,152 pinen 151 plesen 110,113 displesen 113 sarettan 32,77 sceamian/shamen 8, 21, 24, 32,60,64, 72, 74, 85,86, 88, 90,100,102, 144, 149,153, 155,158 forscamian 32 ofscamian 74,148 semen 122,131,133-135,157,158 Jiencan 49 pyncan/thinken 8, 22,24,44,45,49, 74,133,147,150,154 pyrstanoO, 115 geweorpan 22,43, 55, 56,66,67 wUlan78 wilnian 78

be+adjective/noun accordynge 160 ashamed 159 behofful 160 benefycyal 160 betere 55 176

αφ 50, 55 de ore 146 ege49 enshamed 159 gebeorhlic 42

god 42 greuous 159,160 grieved 149,160 iqueme 148 leof 146 lustbaere 39 lustfuU 39 metely 160 nedeful 147,151,156 nedy 156 neod!47,151 neodbearf 42 shame 154, 155 sittinge 160 sorhful 149 sort 148 bearf42 wa 144,145 habban nedbearf 42 hebben neod 147, 149, 151 a^an bearf 42

177